SOIL MOISTURE CONTROLS ON SPATIAL AND TEMPORAL PATTERNS OF

CARBON DIOXIDE FLUXES IN DRYLANDS

by

Andrew Neal

_____________________

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF HYDROLOGY AND WATER RESOURCES

In Partial Fulfillment of the Requirements For the Degree of

DOCTOR OF PHILOSOPHY

WITH A MAJOR IN HYDROLOGY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2012

2

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Andrew Neal entitled Soil Moisture Controls on Spatial and Temporal Patterns of Carbon Dioxide Fluxes in Drylands and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy _______________________________________________________________________ Date: 11/23/2011

Paul Brooks _______________________________________________________________________ Date: 11/23/2011

Shirley Papuga _______________________________________________________________________ Date: 11/23/2011

Hoshin Gupta _______________________________________________________________________ Date: 11/23/2011

Willem van Leeuwen _______________________________________________________________________ Date: 11/23/2011

Steven Archer Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College. I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement. ________________________________________________ Date: Dissertation Director: Shirley Papuga ________________________________________________ Date: Dissertation Director: Paul Brooks

3

STATEMENT BY AUTHOR This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library. Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author. SIGNED: Andrew L. Neal

4

ACKNOWLEDGEMENTS

This dissertation would not have been possible without the assistance of my committee, Steve Archer, Hoshin Gupta and Wim van Leeuwen, and particularly the guidance of my co-advisors, Paul Brooks and Shirley Papuga. Editorial, analytical, and moral support from Daniel Bunting, Jon Martin, Krystine Nelson, and Zulia Sanchez Mejia has been invaluable. Kyle Hartfield and the students at the Arizona Remote Sensing Center helped me through technical computing issues with remote sensing data and spatial analysis. Rafael Rosolem and Jeff Gawad were very accommodating to an out-of-town student coming back to finish his dissertation. Funding for the research was provided by the Semi-Arid Hydrology and Riparian Areas (SAHRA) NSF Science and Technology Center (NSF agreement EAR-9876800). Data for this study were provided by the Ameriflux network and were used in accordance with their fair-use policy. Data were also provided by the USDA-ARS Southwest Watershed Research Center. Funding for these datasets was provided by the United States Department of Agriculture, Agricultural Research Service. Thank you to my parents, Donna and Rick, my brothers, Rich and Tom and my wife, Hayley, who were my safety net throughout this process.

5

5

DEDICATION

To Richard, who planted the seed of inspiration and inquisitiveness

in a young boy,

and

To Hayley, who has nurtured that seed in a grown man.

6

6

TABLE OF CONTENTS

LIST OF FIGURES ............................................................................................................ 8

ABSTRACT........................................................................................................................ 9

ABSTRACT........................................................................................................................ 9

1. INTRODUCTION .................................................................................................... 10

1.1. Problem Statement ............................................................................................ 10

1.2. Background Information............................................................................... 11 1.3. Dissertation Format........................................................................................... 18

2. METHODS ............................................................................................................... 20

3. PRESENT STUDY................................................................................................... 22

3.1. Appendix A: ENVIRONMENTAL CONTROLS ON ECOSYSTEM

RESPIRATION IN DRYLANDS: THE ROLE OF THE VERTICAL DISTRIBUTION

OF SOIL MOISTURE .................................................................................................. 22

3.2. Appendix B: OVERCOMING LIMITATIONS TO REMOTE SENSING-

BASED UPSCALING OF CARBON FLUXES IN SEMIARID SYSTEMS ............. 23

3.3. Appendix C: HYDROLOGIC REDISTRIBUTION SUBSIDIZES CARBON

FLUX IN SEMIARID LANDSCAPES........................................................................ 24

4. FUTURE RESEARCH OPPORTUNITIES ............................................................. 26

5. REFERENCES ......................................................................................................... 30

APPENDIX A: SOIL MOISTURE AND VEGETATION CONTROLS ON

ECOSYSTEM RESPIRATION IN DRYLANDS............................................................ 41

APPENDIX B: SOIL MOISTURE CONTROLS REMOTE SENSING-BASED

UPSCALING OF CARBON FLUXES IN A SEMIARID SHRUBLAND ..................... 89

7

7

APPENDIX C: LATERAL HYDROLOGIC REDISTRIBUTION SUBSIDIZES

CARBON FLUX IN SEMIARID LANDSCAPES ........................................................ 131

APPENDIX D: USING LANDSAT ETM+ NDVI TO CLASSIFY VEGETATION

COVER........................................................................................................................... 170

8

8

LIST OF FIGURES

Figure 1. Drydown of shallow soil moisture (θ, 0-20 cm depth) following a storm. Points

indicate mean soil moisture, error bars are one standard deviation.................................. 15

Figure 2. Identification of transition points between negative (uptake) and positive

(release) NEE to define the deep soil moisture threshold (θ*deep). Gray bars indicate

individual instances of prolonged carbon uptake.............................................................. 16

9

9

ABSTRACT

Dryland ecosystems provide a unique opportunity to study the effects of water limitation

on ecosystem activity. The sensitivity of these systems to small inputs of moisture is

well-documented, but the expression of water limitation in terms of carbon dioxide flux

between the ecosystem and atmosphere remains unclear. Applying a simple conceptual

approach to soil moisture dynamics, patterns in carbon flux become clear. Release of

carbon dioxide via respiration is primarily driven by moisture in the shallow soil, and

differences in respiration rates among plant functional types are only evident after

controlling for soil moisture. Alternatively, carbon uptake by a semiarid shrubs

ecosystem is largely driven by the availability of deep soil moisture. This link to deep soil

moisture improves spatial scaling of gross and net carbon uptake using remote sensing

data. Lateral redistribution of moisture on the landscape connects readily observed

physical features, namely topography, to ecosystem function, but redistribution is

generally not considered in carbon models. A simple runoff scheme coupled to a

conceptual model for carbon flux demonstrates the high degree of spatial heterogeneity in

carbon dioxide flux resulting from moisture redistribution. The importance of

redistribution in carbon modeling is highlighted by interannual variability in modeled

carbon fluxes under different rainfall characteristics (event size, event duration,

interstorm duration). The links between hydrology and ecology across spatial scales

become clearer when topographically-based moisture distribution is used as an

organizing variable. In all, this research identifies new avenues for research where

moisture dynamics are of central interest in dryland ecohydrology.

10

10

1. INTRODUCTION

1.1. Problem Statement

This study examines the response of carbon dioxide fluxes in semiarid ecosystems

based on moisture availability. The prevalence of water-limitation in semiarid

environments emphasizes the importance of hydrologic processes (e.g. runoff,

infiltration, percolation) on ecological function. Semiarid systems exhibit different

carbon flux behavior in response to shallow and deep soil moisture (Kurc and Small

2007) Runoff redistributes moisture laterally which augments local moisture from

rainfall. Using specific patterns in the vertical profile of moisture and with simple runoff

models, we can better understand the dynamics of carbon flux. This framework allows us

to more clearly identify source/sink patterns of carbon dioxide exchange between the land

and atmosphere.

This study addresses three main sets of questions:

How does soil moisture at different depths drive ecosystem respiration? Is that

difference more significant than the differences among vegetation functional types?

Can soil moisture information be used to identify a relationship between net carbon

flux (and photosynthetic carbon uptake) and remote sensing data? How can this

relationship be applied spatially?

Does the lateral redistribution of moisture as runoff subsidize local soil moisture to

extend periods of increased photosynthesis and respiration? What is the effect of

hydrologic redistribution on watershed-scale carbon flux?

11

11

Addressing these questions from an empirical standpoint, the importance of vertical

(i.e. shallow versus deep) soil moisture is evident from previous studies as a control for

carbon dioxide uptake (Kurc and Small 2007). Further, moisture patterns may be

influenced by hydrologic redistribution, altering the spatial patterns of carbon flux.

Altogether, these results point toward new, spatially-distributed methods to quantify

carbon flux based on soil moisture as a direct driver of ecosystem processes.

1.2. Background Information

Semiarid environments are a major component of the terrestrial earth surface,

comprising approximately 40% of the land area (Reynolds et al. 2007). These areas are

important to global carbon balance, and may act as substantial sources or sinks for carbon

at a level that is globally significant (Wohlfahrt et al. 2008). Carbon cycling in semi-arid

systems may be altered under potential land use change, species invasion, climate change

and desertification (Conant and Paustian 2002; Conant et al. 2001; Lal 2004).

Quantifying the landscape-level carbon balance and understanding the drivers of carbon

flux will improve our ability to assess potential changes based on management and

climatic changes.

The net ecosystem exchange of carbon dioxide (NEE) is comprised of two main

components: the carbon dioxide released during all ecosystem respiration processes (Re)

and the gross primary production (GPP) associated with photosynthesis. In dryland

12

12

environments, carbon fluxes have shown a close link to moisture availability, which has

led to the concept of precipitation pulses as drivers of ecosystem activity (Noy-Meir

1973). Precipitation pulses provide a simple tool for analyzing carbon fluxes (Huxman et

al. 2004), particularly in a modeling context (Ogle and Reynolds 2004; Reynolds et al.

2004). However, precipitation does not capture differences in moisture within the soil

column, which more closely influences ecosystem activity in vegetation and soil

microbes (Fernandez 2007).

Several studies have demonstrated the importance of soil moisture in controlling

carbon fluxes in semiarid environments (Bell et al. 2008; Kurc and Small 2007; Scott et

al. 2008; Thomas et al. 2009). The water-limitation inherent to these systems provides an

opportunity to explore the mechanisms of different soil moisture conditions on ecosystem

activity. For example, small storms may only wet the shallow soil (~20 cm deep).

Moisture in shallow soils is often short-lived due to evaporative demand (Wythers et al.

1999). This brief, shallow moisture favors Re (Barron-Gafford et al. 2011; Scott et al.

2006) over GPP, which requires more extensive moisture redistribution within plants.

Alternatively, large storms or a series of storms have been shown to yield more moisture

deeper in the soil profile (>20 cm deep) (Kurc and Small 2007). This moisture may be

preferentially used by plants, thus corresponding to increased GPP (Cavanaugh et al.

2011; Kurc and Benton 2010; Scott et al. 2006).

13

13

Further, ecosystems respond dynamically to available moisture. Vegetation develops

complex root systems to access moisture from different locations in the soil (Cavanaugh

et al. 2011; Donovan and Ehleringer 1994; Scott et al. 2006). Plants may grow or

abandon roots to ensure moisture availability (Chaves et al. 2002) and often have

developed mechanisms to cope with water stress (Porporato et al. 2001). Plant

communities may also directly influence local soil moisture by trapping overland flow

(Moran et al. 2010; Pugnaire et al. 2004; Villegas et al. 2010). As a result, a simple

“bucket” conceptualization of root zone soil moisture may not adequately represent the

ability of a plant to access water.

As moisture moves down through the soil, we anticipate different responses in the

flux of carbon dioxide. Kurc and Small (2007) demonstrated a link between NEE and

evapotranspiration when deep soils were wet and shallow soils were dry. Extrapolating

this result, a conceptual model for soil moisture is used to analyze the patterns of carbon

dioxide flux. This conceptual model divides the soil into two zones: an upper, shallow

zone (0-20 cm) and lower, deep zone (20-60 cm). These depths were selected based on

the maximum depth of direct bare soil evaporation, at approximately 20 cm (Wythers et

al. 1999), and the maximum depth of available soil moisture data at the sites analyzed.

Data from soil moisture reflectometers at each site were averaged within the two zones.

Each individual instrument was assumed to measure an equal distance above and below

the depth of installation.

14

14

The conceptual model identifies wet and dry periods for the two soil zones based on

different thresholds. The shallow soil zone threshold is determined at each site based on

exponential drydown curves associated with daily rainfall greater than 8 mm (Kurc and

Small 2007). Shallow soil moisture data for each rainfall event were averaged and an

exponential curve was fit to that average (Figure 1). The moisture threshold was defined

as the asymptote of the exponential drydown. Deep soil moisture thresholds were

determined from the transition between net carbon dioxide uptake and net release.

Transitions were identified when more than five days of carbon uptake were followed by

five or more days of net carbon release. The transition occurs on the first day of net

carbon release. The values of deep soil moisture for each transition were averaged to

produce a single threshold value for each site (Figure 2).

15

15

Figure 1. Drydown of shallow soil moisture (θ, 0-20 cm depth) following a storm. Points

indicate mean soil moisture, error bars are one standard deviation.

Moisture conditions at any point in the landscape may also be influenced by lateral

redistribution of moisture (Guntner and Bronstert 2004; Ravi et al. 2010). Large storm

events in semiarid environments often produce infiltration-excess runoff (Goodrich et al.

1994), which may later re-infiltrate as runon when trapped by vegetation or having

(Ludwig et al. 2005; Wilcox et al. 2003). This runon moisture has been linked to

vegetation patterns on the landscape (Bhark and Small 2003; Ludwig et al. 2005) and

produces highly heterogeneous patterns in lateral soil moisture (Breshears et al. 2009;

16

16

Caylor et al. 2006). These lateral moisture effects, in concert with the vertical moisture

effects described above, drive the patterns of activity in semiarid ecosystems (Belnap et

al. 2005).

Figure 2. Identification of transition points between negative (uptake) and positive

(release) NEE to define the deep soil moisture threshold (θ*deep). Gray bars indicate

individual instances of prolonged carbon uptake.

Understanding the function of coupled hydrologic (runoff/runon, infiltration) and

ecological (root hydraulic properties, plant physiological behavior, microbial activity)

systems is critical to predict carbon fluxes. This is particularly true in systems that

experience much of their rainfall during the summer growing season, as in the

southwestern United States (Gochis et al. 2006; Higgins and Shi 2000).

17

17

From this study, we find that using soil moisture states can more clearly identify

expected patterns in carbon dioxide fluxes in semiarid landscapes. Simple models can be

used to link precipitation to soil moisture states (Huxman et al. 2004). Through these

models, the effect of seasonal precipitation on rates of GPP in the growing season

(Emmerich and Verdugo 2008; Potts et al. 2006) can be directly linked to landscape

hydrology. In the context of altered climate, the southwestern US may experience

reduced total annual rainfall (Diffenbaugh et al. 2005) although this may include either

increased or decreased seasonal rainfall (Fatichi et al. 2011; McAfee and Russell 2008;

Trenberth et al. 2003). Applying simulated precipitation data from a climate change

scenario with daily resolution, a simple soil moisture characterization could identify the

relevant soil moisture states to estimate both Re and GPP according to the results

presented here.

This study focuses on the ecohydrological response resulting from the distribution of

soil moisture on the landscape, focusing on the distribution of moisture both vertically

and laterally. The effect of the vertical moisture distribution is assessed with respect to

carbon dioxide flux in terms of Re (Appendix A) and NEE and GPP (Appendix B).

Lateral redistribution of moisture may enhance both Re and GPP fluxes by subsidizing

both shallow and deep soil moisture. The effect of hydrologic subsidy on carbon fluxes

and a discussion of relevant spatial scales are included (Appendix C). A method for

delineating vegetation cover types using satellite data is also included (Appendix D).

18

18

1.3. Dissertation Format

A brief treatment of some of the analytical techniques used in this study is given in

the “Methods” section. This offers an overview of some of the technical aspects of eddy

covariance and remote sensing data collection and processing, which may not be included

in the subsequent chapters.

Following the methods is the “Present Study” which provides a summary of the

analyses and conclusions of three papers prepared for peer-reviewed publication. The

three articles are included as appendices, namely:

Appendix A: “Environmental Controls on Ecosystem Respiration in Drylands: the

Role of the Vertical Distribution of Soil Moisture”. In this paper I performed all analyses

interpreted results and drafted the manuscript, which is submitted at Journal of

Geophysical Research-Biogeosciences;

Appendix B: “Overcoming Limitations to Remote Sensing-Based Upscaling of

Carbon Fluxes in Semiarid Systems”. In this paper I performed all analyses, interpreted

results and drafted the manuscript, which will be submitted to Agricultural and Forest

Meteorology;

Appendix C: “Hydrologic Redistribution Subsidizes Carbon Flux in Semiarid

Landscapes”. In this paper I performed all analyses, interpreted results and drafted the

manuscript, which will be submitted to Advances in Water Resources.

19

19

Altogether these three papers discuss findings that will enhance our understanding of

the mechanism of water limitation on ecosystem carbon dynamics. Also included is an

appendix describing work in development as part of the study in Appendix B. Appendix

D: “Using Landsat ETM+ NDVI to Classify Vegetation Cover” discusses an approach to

delineate parts of the landscape that contain specific vegetation types, specifically

creosote shrubs. This analysis was developed in conjunction with the analyses in

Appendix B, and a modified version of the method is in development to extend the results

found in Appendix B.

The dissertation concludes with a section on Future Research Opportunities, which

describe several potential studies that may extend the results presented here. These

studies delve further into the ecohydrological behavior of dryland ecosystems.

20

20

2. METHODS

Direct, landscape-scale measurements of NEE are often made using the eddy

covariance method (Baldocchi et al. 1988). The eddy covariance method gives an

estimate of ecosystem carbon dioxide, water vapor and sensible heat flux based on high-

frequency records of turbulent exchange in the atmosphere. Data are collected via a three-

dimensional wind speed and infrared gas analyzer for carbon dioxide and water

(Moncrieff et al. 1997). Associated micrometeorological measurements, e.g. relative

humidity, temperature, and net radiation are also collected to corroborate the flux data.

Data used in this study were accessed from the Oak Ridge National Laboratory

Ameriflux archive (http://public.ornl.gov/ameriflux/) at Level 2 quality control.

Various approaches have been developed to partition tower-measured NEE into GPP

and Re (Desai et al. 2008). They include correlation structures in the raw, high-frequency

data (Scanlon and Kustas 2010) or model approaches based on the 30-minute flux data.

These models may predict GPP, often via a light use efficiency (LUE) scheme (Lasslop

et al. 2010) or Re, through either biophysical models (e.g. BIOME-BGC, (Reichstein et

al. 2007). Often one model is used for either GPP or Re and the other is calculated as the

residual of observed NEE and the model output.

In order to scale tower-measured carbon fluxes, researchers rely on satellite

remote sensing data to provide information about the land surface over a broad spatial

area (Wylie et al. 2003). Because satellites only detect surface conditions, scaling models

21

21

based on remote sensing data predict GPP as a function of vegetation conditions

(Mahadevan et al. 2008; Sims et al. 2005; Sims et al. 2008; Xiao et al. 2010). These

models use reflectance data that are sensitive to vegetation as inputs, often in the form of

a vegetation index such as the normalized difference vegetation index (NDVI, (Tucker

1979) or the enhanced vegetation index (EVI, (Huete et al. 2002). Vegetation indices

respond to the presence of photosynthetic leaf surfaces (Verhoef 1984), thus GPP models

that use vegetation indices are explicitly or implicitly LUE models (Turner et al. 2002;

Turner et al. 2005). In fact, more recent satellite missions have included data products for

absorbed photosynthetic radiation (Knyazikhin et al. 1999; Myneni et al. 2002) and other

indices related to photosynthetic absorption, e.g. the photosynthetic radiation index (PRI,

(Garbulsky et al. 2010) in an attempt to better characterize plant response to radiation.

A major drawback of LUE models is the coupling of water limitation with

photosynthetic efficiency (Garbulsky et al. 2010). From a remote sensing standpoint,

detection of moisture characteristics is difficult, because satellites only detect surface

reflectance across limited spectra (Fensholt and Sandholt 2003). In order to overcome

this, new indices, such as the normalized difference water index (NDWI, (Fensholt and

Sandholt 2003)), based on the absorption of infrared radiation or land surface temperature

(LST, (Sims et al. 2008)) have been used to represent the moisture characteristics of the

landscape. Other studies have used ground-based estimates of water vapor flux to

represent moisture conditions on the ground (Sjöström et al. 2011).

22

22

3. PRESENT STUDY

The methods, results, and conclusions of this study are presented in the papers

appended to this dissertation. The following is a summary of the most important findings

in these documents.

3.1. Appendix A: ENVIRONMENTAL CONTROLS ON ECOSYSTEM

RESPIRATION IN DRYLANDS: THE ROLE OF THE VERTICAL

DISTRIBUTION OF SOIL MOISTURE

This study examined the influence of vertical (shallow versus deep) soil moisture on

Re. The findings of this analysis include the poor predictive relationships for Re based on

soil temperature (Tsoil) and volumetric soil moisture, individually. When soil moisture

was divided into wet and dry states in shallow and deep soil zones, distinctions between

Re fluxes emerged. Predictive relationships based on Tsoil in the four soil moisture cases

indicated higher rates of Re when shallow soils were wet. These relationships held across

vegetation type, and differences in respiration rates associated with vegetation type

(grass, shrub and savannah) became clear only in the context of the wet/dry soil moisture

cases.

The results of this study indicate the importance of soil moisture as a driver of Re in

semiarid ecosystems. However, Re does not respond as a continuous function of soil

moisture, but rather based on a binary wet-dry classification. Other variables—Tsoil,

vegetation type—only yield meaningful relationships with Re after the wet-dry

23

23

classification is performed. We point toward the potential of the soil moisture cases to

identify particular sources of carbon dioxide (e.g. autotrophic maintenance and growth,

heterotrophic metabolism) at the landscape scale.

3.2. Appendix B: OVERCOMING LIMITATIONS TO REMOTE SENSING-

BASED UPSCALING OF CARBON FLUXES IN SEMIARID SYSTEMS

This study explored empirical scaling relationships between remote sensing data and

tower-based NEE and GPP measurements. The goal was to develop a simple function

linking the temporally-detailed tower data to spatially-extensive satellite remote sensing.

The results indicated that the relationship between tower-measured carbon fluxes and

vegetation index data for the tower source area was sensitive to soil moisture. The wet-

dry soil moisture cases were used to explore the relationships between EVI and NEE and

EVI and GPP based on moisture availability. Correlations between EVI and carbon

fluxes improve near the tower when deep soils are wet.

The connection between carbon flux, especially carbon uptake, and EVI suggests that

deep soil moisture is critical to relax the water stress on semiarid vegetation. Reducing

water stress leads to a stronger relationship between vegetation indices, representing the

solar radiation reflected by plants, and carbon flux. This method offers an approach to

scaling carbon fluxes over wider areas in water-limited systems, given knowledge of soil

moisture conditions.

24

24

3.3. Appendix C: HYDROLOGIC REDISTRIBUTION SUBSIDIZES CARBON

FLUX IN SEMIARID LANDSCAPES

This study considers the potential for soil moisture at any point in the landscape to be

influenced by lateral flow within the watershed. Redistribution of moisture due to

topographic effects can influence the ecosystem carbon flux in semiarid landscapes

depending on runoff/runon processes. Using a simple model for runoff and moisture

routing, the relationship between topography, hydrology and ecosystem dynamics

emerges. Runoff that collects downslope supports longer periods of carbon uptake lower

in the watershed.

Comparing carbon flux with moisture redistribution to carbon flux with no runoff

indicates that the ability of semiarid ecosystems to withstand dry conditions is due in part

to hydrologic features. The results suggest that topography can be used as a predictor of

carbon flux resulting from the redistribution of moisture toward areas that collect runoff

from upslope areas. Spatial heterogeneity in carbon flux resulting from this redistribution

is critical to improve knowledge of ecohydrological processes at larger spatial scales in

drylands.

3.4. APPENDIX D: USING LANDSAT ETM+ NDVI TO CLASSIFY

VEGETATION COVER

This study tested an approach to land cover classification using vegetation index data

and knowledge of vegetation phenology. The goal was to identify parts of the landscape

that are dominated by creosote bush, an evergreen desert shrub. Anticipating minimal

25

25

change in NDVI between growing and non-growing seasons, the method classified the

landscape based on seasonal change in NDVI values. The method gives an approximation

for creosote distribution, but ignores potential variability in herbaceous plant cover,

which may be more or less prevalent depending on a variety of factors, including

seasonal moisture conditions. A new method, corroborating Landsat data with high

resolution (1 meter) remote sensing data, is in development as a result of this study.

26

26

4. FUTURE RESEARCH OPPORTUNITIES

The results presented in this dissertation identify the importance of moisture

distribution, both vertically within the soil and laterally via runoff, on carbon dioxide

exchange in dryland ecosystems. They also suggest further study that will enhance the

understanding of semiarid ecosystems both under current conditions and as dynamic

systems. Ongoing research in dryland ecohydrology continues to develop new

understanding of coupled vegetation and hydrologic behavior. The research proposed

here can extend that frontier across spatial and temporal scales.

Relationships between shallow and deep soil moisture and carbon dioxide fluxes

presented in this dissertation are based on sites where the majority of rainfall occurs

during the summer months—coincident with the typical growing season. Increased

respiration and carbon uptake associated with soil moisture is thus inherently tied to

seasonal climate and ecosystem phenology. Left unresolved is the question of how

ecosystems that receive most of their annual precipitation in winter (e.g. Mediterranean

or Great Basin systems) respond to shallow moisture versus deep moisture. Time lags

between winter moisture inputs and spring and summer carbon flux response become the

basis to explore the mechanism for plant and ecosystem responses. Synthesizing such

results with those presented here can help unify understanding of dryland response to

moisture, whether they receive precipitation in summer or winter.

27

27

The response of ecosystem respiration to shallow soil moisture is demonstrated here

based on respiration over large areas and with limited soil moisture measurement. But

spatial heterogeneity in soil moisture due to variable infiltration rates (Bhark and Small

2003), canopy shading (Breshears et al. 1997), and litter (Villegas et al. 2010) may

introduce variability between canopy and bare soil carbon fluxes. A reasonable question,

then, is: does the vertical moisture profile influence carbon fluxes over small spatial (and

possibly temporal) scales? Detailed analysis of carbon fluxes in bare and canopy areas,

potentially including isotopic analysis to partition flux sources, would describe the effect

of spatially varying soil moisture on respiration.

Isotopic analysis of carbon dioxide fluxes could also be used to determine if the four

moisture cases favor autotrophic or heterotrophic respiration. A working hypothesis that

emerged from this study was that respiration under fully dry soils represented background

levels of respiration for both autotrophs and heterotrophs; that shallow soil moisture was

the driver for heterotrophic respiration; and that deep soil moisture was associated with

growth respiration in vegetation. Relationships between soil moisture and respiration

components will explain the processes and pathways to better understand carbon

dynamics in the future.

This study indicates the importance of incorporating runoff in carbon dioxide flux

estimates at watershed scales. However, the experiment performed here used a limited

precipitation data set over a relatively small area. Rainfall is known to be highly variable

28

28

in space and time, including variability within a single storm event, between seasons and

between years. Rainfall variability can be characterized by intensity, duration, and

frequency. From the findings here, reduced storm frequency and intensity would be

expected to reduce runoff and thus carbon dioxide uptake.

Results presented in Appendices A and B suggest different sensitivities of ecosystem

properties to shallow and deep soil moisture, i.e. ecosystem respiration responds to

shallow moisture while carbon uptake is sensitive to deep soil moisture. The connection

between shallow and deep soil moisture and ecosystem behavior is an emergent property

of dryland ecosystems. Exploring the correlation structure and frequency patterns

between shallow and deep moisture and partitioned carbon and water fluxes would

describe the sensitivity of ecosystems to potential change and identify positive feedbacks

between vegetation and soil moisture. Further, detailed frequency analysis would identify

time scales associated with stress thresholds on vegetation dynamics. As more research

sites continue to maintain long-term data records, this research will be easier to facilitate

and more fruitful in potential outcomes.

The remote sensing data applied in this study was at a relatively coarse spatial scale

relative to the processes of interest. Identifying subgrid vegetation patterns and temporal

landscape change are complicated by the lack of resolution. New data products derived

from multiple satellite sensors may enable improvements in both spatial and temporal

resolution via a data fusion approach. Applying these new tools will enable consistent

29

29

analysis of long-term trends in dryland ecosystem dynamics at the landscape scale by

linking data from current sensors to legacy systems.

30

30

5. REFERENCES

Baldocchi, D.D., Hicks, B.B., & Meyers, T.P. (1988). Measuring biosphere-atmosphere

exchanges of biologically related gases with micrometeorological methods.

Ecology, 69, 1331-1340

Barron-Gafford, G.A., Scott, R.L., Jenerette, G.D., & Huxman, T.E. (2011). The relative

controls of temperature, soil moisture, and plant functional group on soil CO2

efflux at diel, seasonal, and annual scales. J. Geophys. Res., 116, G01023

Bell, C., McIntyre, N., Cox, S., Tissue, D., & Zak, J. (2008). Soil microbial responses to

temporal variations of moisture and temperature in a Chihuahuan Desert

Grassland. Microbial Ecology, 56, 153-167

Belnap, J., Welter, J.R., Grimm, N.B., Barger, N., & Ludwig, J.A. (2005). Linkages

between Microbial and Hydrologic Processes in Arid and Semiarid Watersheds.

Ecology, 86, 298-307

Bhark, E.W., & Small, E.E. (2003). Association between plant canopies and the spatial

patterns of infiltration in shrubland and grassland of the Chihuahuan Desert, New

Mexico. Ecosystems, 6, 185-196

Breshears, D.D., Myers, O.B., & Barnes, F.J. (2009). Horizontal heterogeneity in the

frequency of plant-available water with woodland intercanopy-canopy vegetation

patch type rivals that occuring vertically by soil depth. Ecohydrology, 2, 503-519

Breshears, D.D., Rich, P.M., Barnes, F.J., & Campbell, K. (1997). Overstory-Imposed

Heterogeneity in Solar Radiation and Soil Moisture in a Semiarid Woodland.

Ecological Applications, 7, 1201-1215

31

31

Cavanaugh, M.L., Kurc, S.A., & Scott, R.L. (2011). Evapotranspiration partitioning in

semiarid shrubland ecosystems: a two-site evaluation of soil moisture control on

transpiration. Ecohydrology, 4, 671-681

Caylor, K.K., D'Odorico, P., & Rodriguez-Iturbe, I. (2006). On the ecohydrology of

structurally heterogeneous semiarid landscapes. Water Resour. Res., 42, W07424

Chaves, M.M., Pereira, J.S., Maroco, J., Rodrigues, M.L., Ricardo, C.P.P., OsÓRio,

M.L., Carvalho, I., Faria, T., & Pinheiro, C. (2002). How Plants Cope with Water

Stress in the Field? Photosynthesis and Growth. Annals of Botany, 89, 907-916

Conant, R.T., & Paustian, K. (2002). Potential soil carbon sequestration in overgrazed

grassland ecosystems. Global Biogeochemical Cycles, 16

Conant, R.T., Paustian, K., & Elliott, E.T. (2001). Grassland management and conversion

into grassland: Effects on soil carbon. Ecological Applications, 11, 343-355

Desai, A.R., Richardson, A.D., Moffat, A.M., Kattge, J., Hollinger, D.Y., Barr, A., Falge,

E., Noormets, A., Papale, D., Reichstein, M., & Stauch, V.J. (2008). Cross-site

evaluation of eddy covariance GPP and RE decomposition techniques.

Agricultural and Forest Meteorology, 148, 821-838

Diffenbaugh, N.S., Pal, J.S., Trapp, R.J., & Giorgi, F. (2005). Fine-scale processes

regulate the response of extreme events to global climate change. Proceedings of

the National Academy of Sciences of the United States of America, 102, 15774-

15778

Donovan, L.A., & Ehleringer, J.R. (1994). Water Stress and Use of Summer Precipitation

in a Great Basin Shrub Community. Functional Ecology, 8, 289-297

32

32

Emmerich, W.E., & Verdugo, C.L. (2008). Precipitation thresholds for CO2 uptake in

grass and shrub plant communities on Walnut Gulch Experimental Watershed.

Water Resources Research, 44, 9

Fatichi, S., Ivanov, V.Y., & Caporali, E. (2011). Simulation of future climate scenarios

with a weather generator. Advances in Water Resources, 34, 448-467

Fensholt, R., & Sandholt, I. (2003). Derivation of a shortwave infrared water stress index

from MODIS near- and shortwave infrared data in a semiarid environment.

Remote Sensing of Environment, 87, 111-121

Fernandez, R.J. (2007). On the frequent lack of response of plants to rainfall events in

arid areas. Journal of Arid Environments, 68, 688-691

Garbulsky, M.F., Peñuelas, J., Papale, D., Ardö, J., Goulden, M.L., Kiely, G.,

Richardson, A.D., Rotenberg, E., Veenendaal, E.M., & Filella, I. (2010). Patterns

and controls of the variability of radiation use efficiency and primary productivity

across terrestrial ecosystems. Global Ecology and Biogeography, 19, 253-267

Gochis, D.J., Brito-Castillo, L., & Shuttleworth, W.J. (2006). Hydroclimatology of the

North American Monsoon region in northwest Mexico. Journal of Hydrology,

316, 53-70

Goodrich, D.C., Schmugge, T.J., Jackson, T.J., Unkrich, C.L., Keefer, T.O., Parry, R.,

Bach, L.B., & Amer, S.A. (1994). Runoff simulation sensitivity to remotely

sensed initial soil water content. Water Resour. Res., 30, 1393-1405

33

33

Guntner, A., & Bronstert, A. (2004). Representation of landscape variability and lateral

redistribution processes for large-scale hydrological modelling in semi-arid areas.

Journal of Hydrology, 297, 136-161

Higgins, R.W., & Shi, W. (2000). Dominant factors responsible for interannual

variability of the summer monsoon in the southwestern United States. Journal of

Climate, 13, 759-776

Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., & Ferreira, L.G. (2002).

Overview of the radiometric and biophysical performance of the MODIS

vegetation indices. Remote Sensing of Environment, 83, 195-213

Huxman, T.E., Snyder, K.A., Tissue, D., Leffler, A.J., Ogle, K., Pockman, W.T.,

Sandquist, D.R., Potts, D.L., Schwinning, S., & Yu (2004). Precipitation pulses

and carbon fluxes in semiarid and arid ecosystems. Oecologia, 141, 254-268

Knyazikhin, Y., Glassy, J., Privette, J.L., Tian, Y., Lotsch, A., Zhang, Y., Wang, Y.,

Morisette, J.T., Votava, P., Myneni, R.B., Nemani, R.R., & Running, S.W.

(1999). MODIS Leaf Area Index (LAI) and Fraction of Photosynthetically Active

RadiationAbsorbed by Vegetation (FPAR) Product (MOD15) Algorithm

Theoretical BasisDocument. In

Kurc, S.A., & Benton, L.M. (2010). Digital image-derived greenness links deep soil

moisture to carbon uptake in a creosotebush-dominated shrubland. Journal of

Arid Environments

34

34

Kurc, S.A., & Small, E.E. (2007). Soil moisture variations and ecosystem-scale fluxes of

water and carbon in semiarid grassland and shrubland. Water Resources

Research, 43

Lal, R. (2004). Carbon sequestration in dryland ecosystems. Environmental Management,

33, 528-544

Lasslop, G., Reichstein, M., Papale, D., Richardson, A.D., Arneth, A., Barr, A., Stoy, P.,

& Wohlfahrt, G. (2010). Separation of net ecosystem exchange into assimilation

and respiration using a light response curve approach: critical issues and global

evaluation. Global Change Biology, 16, 187-208

Ludwig, J.A., Wilcox, B.P., Breshears, D.D., Tongway, D.J., & Anton, C.I. (2005).

Vegetation Patches and Runoff-Erosion as Interacting Ecohydrological Processes

in Semiarid Landscapes. Ecology, 86, 288-297

Mahadevan, P., Wofsy, S.C., Matross, D.M., Xiao, X., Dunn, A.L., Lin, J.C., Gerbig, C.,

Munger, J.W., Chow, V.Y., & Gottlieb, E.W. (2008). A satellite-based biosphere

parameterization for net ecosystem CO2 exchange: Vegetation Photosynthesis

and Respiration Model (VPRM). Global Biogeochem. Cycles, 22, GB2005

Masek, J.G., Vermote, E.F., Saleous, N., Wolfe, R., Hall, F.G., Huemmrich, F., Gao, F.,

Kutler, J., & Lim, T.K. (2006). A Landsat surface reflectance data set for North

America, 1990-2000. Geoscience and Remote Sensing Letters, 3, 68-72

McAfee, S.A., & Russell, J.L. (2008). Northern Annular Mode impact on spring climate

in the western United States. Geophys. Res. Lett., 35, L17701

35

35

Moncrieff, J.B., Massheder, J.M., deBruin, H., Elbers, J., Friborg, T., Heusinkveld, B.,

Kabat, P., Scott, S., Soegaard, H., & Verhoef, A. (1997). A system to measure

surface fluxes of momentum, sensible heat, water vapour and carbon dioxide.

Journal of Hydrology, 189, 589-611

Moran, M.S., Hamerlynck, E.P., Scott, R.L., Stone, J.J., Holifield Collins, C.D., Keefer,

T.O., Bryant, R., DeYoung, L., Nearing, G.S., Sugg, Z., & Hymer, D.C. (2010).

Hydrologic response to precipitation pulses under and between shrubs in the

Chihuahuan Desert, Arizona. Water Resour. Res., 46, W10509

Myneni, R.B., Hoffman, S., Knyazikhin, Y., Privette, J.L., Glassy, J., Tian, Y., Wang, Y.,

Song, X., Zhang, Y., Smith, G.R., Lotsch, A., Friedl, M., Morisette, J.T., Votava,

P., Nemani, R.R., & Running, S.W. (2002). Global products of vegetation leaf

area and fraction absorbed PAR from year one of MODIS data. Remote Sensing of

Environment, 83, 214-231

Noy-Meir, I. (1973). Desert Ecosystems: Environment and Producers. Annual Review of

Ecology and Systematics, 4, 25-51

Ogle, K., & Reynolds, J.F. (2004). Plant responses to precipitation in desert ecosystems:

integrating functional types, pulses, thresholds, and delays. Oecologia, 141, 282-

294

Pennington, D., & Collins, S. (2007). Response of an aridland ecosystem to interannual

climate variability and prolonged drought. Landscape Ecology, 22, 897-910

36

36

Peters, A.J., Eve, M.D., Holt, E.H., & Whitford, W.G. (1997). Analysis of Desert Plant

Community Growth Patterns with High Temporal Resolution Satellite Spectra.

Journal of Applied Ecology, 34, 418-432

Porporato, A., Laio, F., Ridolfi, L., & Rodriguez-Iturbe, I. (2001). Plants in water-

controlled ecosystems: active role in hydrologic processes and response to water

stress - III. Vegetation water stress. Advances in Water Resources, 24, 725-744

Potts, D.L., Huxman, T.E., Cable, J.M., English, N.B., Ignace, D.D., Eilts, J.A., Mason,

M.J., Weltzin, J.F., & Williams, D.G. (2006). Antecedent moisture and seasonal

precipitation influence the response of canopy-scale carbon and water exchange to

rainfall pulses in a semi-arid grassland. New Phytologist, 170, 849-860

Pugnaire, F.I., Armas, C., & Valladares, F. (2004). Soil as a mediator in plant-plant

interactions in a semi-arid community. Journal of Vegetation Science, 15, 85-92

Qi, J., Chehbouni, A., Huete, A.R., Kerr, Y.H., & Sorooshian, S. (1994). A modified soil

adjusted vegetation index. Remote Sensing of Environment, 48, 119-126

Ravi, S., Breshears, D.D., Huxman, T.E., & D'Odorico, P. (2010). Land degradation in

drylands: Interactions among hydrologic-aeolian erosion and vegetation

dynamics. Geomorphology, 116, 236-245

Reichstein, M., Papale, D., Valentini, R., Aubinet, M., Bernhofer, C., Knohl, A., Laurila,

T., Lindroth, A., Moors, E., Pilegaard, K., & Seufert, G. (2007). Determinants of

terrestrial ecosystem carbon balance inferred from European eddy covariance flux

sites. Geophys. Res. Lett., 34, L01402

37

37

Reynolds, J.F., Kemp, P.R., Ogle, K., & Fernandez, R.J. (2004). Modifying the 'pulse-

reserve' paradigm for deserts of North America: precipitation pulses, soil water,

and plant responses. Oecologia, 141, 194-210

Reynolds, J.F., Smith, D.M.S., Lambin, E.F., Turner, B.L., Mortimore, M., Batterbury,

S.P.J., Downing, T.E., Dowlatabadi, H., Fernández, R.J., Herrick, J.E., Huber-

Sannwald, E., Jiang, H., Leemans, R., Lynam, T., Maestre, F.T., Ayarza, M., &

Walker, B. (2007). Global Desertification: Building a Science for Dryland

Development. Science, 316, 847-851

Scanlon, T.M., & Kustas, W.P. (2010). Partitioning carbon dioxide and water vapor

fluxes using correlation analysis. Agricultural and Forest Meteorology, 150, 89-

99

Scott, R.L., Cable, W.L., & Hultine, K.R. (2008). The ecohydrologic significance of

hydraulic redistribution in a semiarid savanna. Water Resources Research, 44

Scott, R.L., Huxman, T.E., Cable, W.L., & Emmerich, W.E. (2006). Partitioning of

evapotranspiration and its relation to carbon dioxide exchange in a Chihuahuan

Desert shrubland. Hydrological Processes, 20, 3227-3243

Sims, D.A., Rahman, A.F., Cordova, V.D., Baldocchi, D.D., Flanagan, L.B., Goldstein,

A.H., Hollinger, D.Y., Misson, L., Monson, R.K., Schmid, H.P., Wofsy, S.C., &

Xu, L.K. (2005). Midday values of gross CO2 flux and light use efficiency during

satellite overpasses can be used to directly estimate eight-day mean flux.

Agricultural and Forest Meteorology, 131, 1-12

38

38

Sims, D.A., Rahman, A.F., Cordova, V.D., El-Masri, B.Z., Baldocchi, D.D., Bolstad,

P.V., Flanagan, L.B., Goldstein, A.H., Hollinger, D.Y., Misson, L., Monson,

R.K., Oechel, W.C., Schmid, H.P., Wofsy, S.C., & Xu, L. (2008). A new model

of gross primary productivity for North American ecosystems based solely on the

enhanced vegetation index and land surface temperature from MODIS. Remote

Sensing of Environment, 112, 1633-1646

Sjöström, M., Ardö, J., Arneth, A., Boulain, N., Cappelaere, B., Eklundh, L., de

Grandcourt, A., Kutsch, W.L., Merbold, L., Nouvellon, Y., Scholes, R.J.,

Schubert, P., Seaquist, J., & Veenendaal, E.M. (2011). Exploring the potential of

MODIS EVI for modeling gross primary production across African ecosystems.

Remote Sensing of Environment, 115, 1081-1089

Thomas, C.K., Law, B.E., Irvine, J., Martin, J.G., Pettijohn, J.C., & Davis, K.J. (2009).

Seasonal hydrology explains interannual and seasonal variation in carbon and

water exchange in a semiarid mature ponderosa pine forest in central Oregon.

Journal of Geophysical Research-Biogeosciences, 114

Trenberth, K.E., Dai, A., Rasmussen, R.M., & Parsons, D.B. (2003). The Changing

Character of Precipitation. Bulletin of the American Meteorological Society, 84,

1205-1217

Tucker, C.J. (1979). Red and photographic infrared linear combinations for monitoring

vegetation. Remote Sensing of Environment, 8, 127-150

39

39

Turner, D.P., Gower, S.T., Cohen, W.B., Gregory, M., & Maiersperger, T.K. (2002).

Effects of spatial variability in light use efficiency on satellite-based NPP

monitoring. Remote Sensing of Environment, 80, 397-405

Turner, D.P., Ritts, W.D., Cohen, W.B., Maeirsperger, T.K., Gower, S.T., Kirschbaum,

A.A., Running, S.W., Zhao, M.S., Wofsy, S.C., Dunn, A.L., Law, B.E.,

Campbell, J.L., Oechel, W.C., Kwon, H.J., Meyers, T.P., Small, E.E., Kurc, S.A.,

& Gamon, J.A. (2005). Site-level evaluation of satellite-based global terrestrial

gross primary production and net primary production monitoring. Global Change

Biology, 11, 666-684

Verhoef, W. (1984). Light scattering by leaf layers with application to canopy reflectance

modeling: The SAIL model. Remote Sensing of Environment, 16, 125-141

Villegas, J.C., Breshears, D.D., Zou, C.B., & Law, D.J. (2010). Ecohydrological controls

of soil evaporation in deciduous drylands: How the hierarchical effects of litter,

patch and vegetation mosaic cover interact with phenology and season. Journal of

Arid Environments, 74, 595-602

Wilcox, B.P., Breshears, D.D., & Allen, C.D. (2003). Ecohydrology of a Resource-

Conserving Semiarid Woodland: Effects of Scale and Disturbance. Ecological

Monographs, 73, 223-239

Wohlfahrt, G., Fenstermaker, L.F., & Arnone, J.A. (2008). Large annual net ecosystem

CO2 uptake of a Mojave Desert ecosystem. Global Change Biology, 14, 1475-

1487

40

40

Wylie, B.K., Johnson, D.A., Laca, E., Saliendra, N.Z., Gilmanov, T.G., Reed, B.C.,

Tieszen, L.L., & Worstell, B.B. (2003). Calibration of remotely sensed, coarse

resolution NDVI to CO2 fluxes in a sagebrush-steppe ecosystem. Remote Sensing

of Environment, 85, 243-255

Wythers, K.R., Lauenroth, W.K., & Paruelo, J.M. (1999). Bare-soil evaporation under

semiarid field conditions. Soil Science Society of America Journal, 63, 1341-1349

Xiao, J.F., Zhuang, Q.L., Law, B.E., Chen, J.Q., Baldocchi, D.D., Cook, D.R., Oren, R.,

Richardson, A.D., Wharton, S., Ma, S.Y., Martini, T.A., Verma, S.B., Suyker,

A.E., Scott, R.L., Monson, R.K., Litvak, M., Hollinger, D.Y., Sun, G., Davis,

K.J., Bolstad, P.V., Burns, S.P., Curtis, P.S., Drake, B.G., Falk, M., Fischer, M.L.,

Foster, D.R., Gu, L.H., Hadley, J.L., Katul, G.G., Roser, Y., McNulty, S., Meyers,

T.P., Munger, J.W., Noormets, A., Oechel, W.C., Paw, K.T., Schmid, H.P., Starr,

G., Torn, M.S., & Wofsy, S.C. (2010). A continuous measure of gross primary

production for the conterminous United States derived from MODIS and

AmeriFlux data. Remote Sensing of Environment, 114, 576-591

41

41

APPENDIX A: SOIL MOISTURE AND VEGETATION CONTROLS ON

ECOSYSTEM RESPIRATION IN DRYLANDS

Andrew L. Neal1, Shirley A. Kurc1, 2, Paul D. Brooks1, Russell L. Scott3

A.L. Neal. Department of Hydrology and Water Resources, University of Arizona, 1133

James E. Rogers Way, Tucson, AZ 85721. (aneal@email.arizona.edu)

S.A. Kurc. School of Natural Resources and the Environment, University of Arizona, 

P.O. Box 210043, Tucson, AZ 85721 (kurc@email.arizona.edu)

P.D. Brooks. Department of Hydrology and Water Resources, University of Arizona,

1133 James E. Rogers Way, Tucson, AZ 85721. (brooks@email.arizona.edu)

R.L. Scott. Southwest Watershed Research Center, United States Department of

Agriculture, Agricultural Research Service, 2000 E. Allen Road, Tucson, AZ 85719

(russ.scott@ars.usda.gov)

1 Department of Hydrology and Water Resources, University of Arizona.

2 School of Natural Resources and the Environment, University of Arizona.

3 Southwest Watershed Research Center, United States Department of Agriculture,

Agricultural Research Service.

42

42

Abstract

Terrestrial ecosystem respiration (Reco) is a major component of the global carbon system

and is sensitive to a variety of physical and biological factors, including temperature,

moisture and vegetation. Describing the sensitivity of Reco to these factors is particularly

important to quantify how expected changes to temperature, rainfall, and vegetation will

affect Reco and net carbon balance. Here, we examine the relationships between several

ecosystem characteristics (soil temperature, vegetation type, and soil moisture) and

nighttime Reco from eddy flux towers throughout the southwestern USA. Soil moisture

data provided the best prediction of Reco and was most useful in the context of a wet/dry

classification scheme based on shallow and deep soil zones. Soil moisture thresholds for

the classification scheme were determined using drydown curves (shallow soil, < 20 cm)

and net ecosystem carbon flux (deep soil, 20-60cm). In contrast, we found that soil

temperature alone was a poor predictor for Reco and that including information about

vegetation type did little to improve that prediction. The two zone wet/dry soil moisture

classification also indicates different sensitivity to moisture among vegetation types. The

ability of a threshold-based soil moisture-based approach to capture widespread patterns

in Reco underscores the importance of water-limitation for both soil microbial and plant

respiration in drylands, and provides a useful tool for future analyses of carbon cycling

under climate or vegetation change in these landscapes.

Keywords: ecohydrology; semiarid environments; thresholds; water limitation; carbon

flux; shrubland; grassland

43

43

1. Introduction

Drylands are critical to the global carbon cycle because they store approximately 15% of

soil organic carbon [Jenerette and Lal, 2005; Lal, 2004; Scott et al., 2009] and constitute

about 40% of the terrestrial earth surface [Reynolds et al., 2007]. The net carbon flux in

drylands can vary between source and sink behavior on seasonal to interannual scales [cf.

Scott et al., 2009; Wohlfahrt et al., 2008], because it are largely controlled by moisture

pulses characteristic of drylands [Noy-Meir, 1973]. Carbon fluxes may also vary under

conditions of vegetation change, e.g. woody plant encroachment [Huxman et al., 2005;

Scott et al., 2006a]. Due to their expansive land area, any change in the carbon dynamics

of drylands is likely to have a large impact on the global carbon system [Lal, 2004]. The

sensitivity of these ecosystems to moisture inputs (both frequency and intensity)

underscores the need for improved understanding of the mechanisms driving carbon

fluxes under potentially altered climate [Vargas et al., 2012]. This is particularly true for

understanding the carbon release via ecosystem respiration (Reco) [Raich and Schlesinger,

1992; Schlesinger and Andrews, 2000].

Ecosystem respiration is the release of carbon dioxide both aboveground and below.

Aboveground respiration is predominantly autotrophic plant growth and maintenance

respiration, while belowground soil respiration is composed of both heterotrophic

(microbial) and autotrophic (plant and mycorrhizal) sources [Hanson et al., 2000]. Over

long periods of time, soil respiration is dominant in Reco except during intense vegetation

growth [Franzluebbers et al., 2002, Law et al. 1999]. Although soil temperature is

44

44

generally considered the primary driver of Reco [Bahn et al., 2010; Cable et al., 2011;

Fang and Moncrieff, 2001], in dryland ecosystems, Reco is also driven by soil moisture

[Almagro et al., 2009; Conant et al., 2004; Grünzweig et al., 2009; Liu et al., 2009;

Sponseller, 2007; Xu et al., 2004]. The temperature and moisture effect in respiration is

linked directly to biochemical processes [Davidson et al., 2012] at multiple time scales

[Vargas et al., 2012].

To characterize the influence of soil moisture on Reco in dryland ecosystems, conceptual

methods have used rainfall as a proxy [Huxman et al., 2004; Ogle and Reynolds, 2004;

Reynolds et al., 2004, Vargas et al., 2012]. However, rainfall must infiltrate the soil

before it becomes useful to plants and soil microbial communities [Bhark and Small,

2003; Kurc and Small, 2004; Moran et al., 2010; Ng and Zhan, 2007; Wainwright et al.,

2000] resulting in a temporal lag between rainfall and Reco response [Vargas et al.,

2010b]. Therefore, soil moisture (θ) is more appropriate for describing the influence of

changes in water availability controlling the biological processes involved in Reco [Bell et

al., 2008; Kurc and Small, 2007; Potts et al., 2006; Scott et al., 2008; Thomas et al.,

2009]. Soil moisture is particularly important between rainfall events as persistent

moisture sustains continued ecosystem activity [Kurc and Small, 2007]. Following a

storm, θ can vary vertically in the soil as a function of the magnitude and time since that

storm [Laio et al., 2001], soil hydraulic properties [Porporato et al., 2002] and

topographic and plant hydraulic redistribution [Scott et al., 2008]. This vertical change in

45

45

θ directly influences ecosystem function [S. A. Kurc and Small, 2007; Yaseef et al., 2010]

such as net primary productivity and soil respiration [Thomey et al., 2011].

At the onset of the wet season, when the shallow soil is wet but the deeper soil remains

dry, microbial activity can be high [Barron-Gafford et al., 2011], while plant

transpiration [Cavanaugh et al., 2010; Scott et al., 2006b] and carbon flux [Kurc and

Benton, 2010] are typically low. However, when the surface is dry and deep soil is wet,

such as occurs several days following a storm, plants transpire and photosynthesize

[Cavanaugh et al., 2010; Kurc and Benton, 2010], which has been linked to autotrophic

respiration [Vargas et al., 2011] but may inhibit heterotrophic respiration [Barron-

Gafford et al., 2011; Tang and Baldocchi, 2005]. We note here the lack of a widely

agreed upon description of autotrophic and heterotrophic respiration response to event-

based moisture inputs [Carbone et al., 2008]. The influence of shallow and deep soil

moisture on the components of Reco derives from differences in the temporal scales of

water availability [Loik et al., 2004]. Shallow soil moisture (< 20 cm soil depth) is

relatively short-lived due to the high evaporative demand of arid ecosystems [Wythers et

al., 1999]. Alternatively, deep soil moisture (> 20 cm below the surface) is more

persistent with minimal impact from direct atmospheric effects; thus it is available for

uptake via plant root systems [Caldwell and Richards, 1989; Williams and Albertson,

2004]. Because of its influence on physical and biological processes, the vertical θ

distribution and the timing of moisture availability at different depths is critical for Reco in

dryland ecosystems.

46

46

The objective of this study was to examine the importance of shallow versus deep θ on

Reco across dryland ecosystems in the southwestern USA. We seek a unifying framework

to describe Reco in dryland environments. To accomplish this, we synthesize data from

five sites, which span vegetation types typical of the region. We focus on the influence of

θ in the context of other expected environmental controls: soil temperature (Tsoil) [Fang

and Moncrieff, 2001; Luo and Zhou, 2006], vegetation type [Barron-Gafford et al., 2011;

Cable et al., 2011]. To simplify θ variability in the vertical dimension, we used a

conceptual framework based on two soil zones shallow and deep soil moisture described

above. We applied a classification scheme that divides soil moisture into two states (i.e.,

wet and dry) using threshold soil moisture values for the two zones, resulting in four

specific wet/dry cases. We evaluate the distribution of Reco rates based on the four soil

moisture cases and vegetation type.

We hypothesize that the dynamics of Reco will depend on the presence of moisture in the

soil. Specifically, we expect that Reco is controlled by temperature only when moisture is

present, particularly in the shallow soil. We expect the highest Reco rates to be associated

with wet conditions in shallow soil because of the concentration of labile carbon

available for respiration in the upper soil [Jobbágy and Jackson, 2000]. Reco will also be

elevated when moisture is present in the deep zone, whether or not there is moisture in

the shallow zone, because of root water uptake stimulating autotrophic activity.

Vegetation type is also expected to have an influence Reco [Barron-Gafford et al., 2011;

47

47

Jin et al., 2010] based on the physiological differences in plant functional types. Any

trend in Reco based on vegetation is expected to follow from the pattern in soil moisture,

based on the strength of soil moisture controls across plant functional types.

2. Methods

2.1. Sites

A total of five dryland sites are used in this study (Figure 1). These semiarid sites were

selected because they include a vertical profile of continuous soil moisture measurements

to at least 50 cm depth and all have similar soil textures and climates. The sites also span

a range of grass and shrub vegetation types that characterize semiarid environments [S. A.

Kurc and Small, 2007; S.A. Kurc and Benton, 2010; Scott et al., 2006]. See Table 1 for

details on site characteristics. First, we included two creosotebush-dominated shrublands,

one in the Chihuahuan desert of New Mexico at the Sevilleta Long Term Ecological

Research (LTER) site (Sev Shrub) and one in the Sonoran desert at the Santa Rita

Experimental Range (SRER, SRER-C). These sites are similar climatically, with mean

annual temperatures of 17.5 °C and 21 °C and mean annual rainfall of 230 and 260 mm,

respectively. Second, we also included two semidesert grasslands, one at the Sevilleta

LTER (Sev Grass) and one at the transition zone between the Chihuahuan and Sonoran

deserts in the Walnut Gulch Experimental Watershed (Kendall). The Sev Grass site

experiences the same climate as the Sev Shrub site. Kendall has a mean annual

temperature of 17 °C and mean annual rainfall of 351 mm. Third, we included a fifth site

representing a mesquite savannah, also located at the SRER (SRER-M). Climate at

48

48

SRER-M is slightly wetter and cooler than at SRER-C, with mean annual temperature of

19 °C and mean rainfall of 310 mm. All sites receive the majority of annual rainfall

during the summer months (June-September).

Instrumentation used at the individual sites includes standard open-path eddy covariance

equipment, i.e. three-dimensional anemometers, infra-red gas analyzers and other

micrometeorological equipment [see Baldocchi et al., 2001, and Table 1]. Additional

instrumentation includes soil temperature thermistors within the upper 10cm of soil depth

at all sites and soil moisture sensors at several depths, at least as far as 50cm below the

surface (Table 1). Multi-year datasets in this region reveal that infiltration is usually

limited to the upper ~50 cm of soil in the hot, rainy season and ~100 cm during the cool

and occasionally wet winters [Scott et al., 2000].

2.2. Ecosystem Respiration Data

Data at the five sites were collected from 2002-2009 (see Table 1 for exact range at each

site) and include years representing relatively wet, dry and normal precipitation patterns

at each site, as reported by site investigators (see Table 1 for references). The data used

here were processed at Level 3 according to Ameriflux processing standards and were

used according to the Ameriflux fair use policy (http://public.ornl.gov/ameriflux). This

level of processing applies a standard set of protocols to quality control net ecosystem

exchange of CO2 (NEE) and calculation of Reco [Reichstein et al., 2005].

49

49

Nighttime NEE was used to represent Reco [Lavigne et al., 1997]. To further constrain

NEE to measure only respiration, this study used data collected between 00h00 and

05h00 local time at all sites. We note that these late-night periods are often characterized

by a well-formed stable boundary layer and thus decreased mixing in the atmosphere;

during these time periods, assumptions of eddy covariance methods are likely violated

[Stull, 1988]. To remedy this, a friction velocity (u*) filter was identified for each site

[Goulden et al., 1996]. Filter values were either based on those reported by site

investigators or, in the case of SRER-C, derived for that site individually. A u* filter of

0.2 ms-1 was applied for SRER-C; the filter was identified by the minimum filter value

where annual net carbon flux converges to a stable value [Goulden et al., 1996]. In total,

the filters reduced the available Reco data by 70% of the total recorded data between

00h00 and 05h00. Similar rates of nighttime data loss are reported elsewhere in studies

using eddy covariance systems [Goulden et al. 1996, Lavigne et al., 1997], and nighttime

NEE has widely been use to study Reco [Mahecha et al., 2010].

After filtering, daily average flux rates were calculated. For the remaining data, daily

average values were only determined if at least three non-filtered data points remained.

These daily-level fluxes (units: mg m-2 s-1) were analyzed in conjunction with other daily-

level site data of the forcing variables (i.e., soil moisture and soil temperature).

Anticipating an exponential relationship between Tsoil and Reco [Lloyd and Taylor, 1994],

we use a linearized form of an exponential model [Luo and Zhou, 2006], of the form:

50

50

soileco bTaR += ln (1).

The parameters of equation 1 are also used to derive an apparent Q10 value, to describe

the rate of change in Reco based on a 10 °C change in Tsoil [Xu and Qi, 2001]. The

apparent Q10 is calculated as:

beQ 1010 = (2).

2.3. Ancillary Data and Soil Moisture Cases

Soil moisture, soil temperature and rainfall data were collected with micrometeorological

measurements as noted in section 2.1. Information about vegetation type at each site was

drawn from the Ameriflux website. Soil moisture data were averaged, and rainfall

summed, to daily values, which ensured that any moisture inputs prior to nighttime Reco

periods were included. Soil temperature data from the same nighttime period as the Reco

data were averaged temporally to produce a daily Tsoil value which corresponded to the

Reco data. Tsoil varies by approximately 3°C across all sites during the night.

Soil moisture data from different depths were combined to yield two soil zones, one

shallow and one deep, for each of the five sites. The shallow soil zone includes the upper

20cm of the soil and the deep soil zone ranges from 20-60cm. Assuming that each sensor

measured an equal distance above and below the sensor, weighted average values of daily

(24 hour) soil moisture were determined at each site for the two soil zones for the entire

day. Soil temperatures were analyzed in connection with soil moisture conditions on the

immediate preceding day, i.e. soil moisture conditions for the day correspond with Tsoil

and Reco for the following night. A wet/dry classification was developed from the two-

51

51

zone soil moisture record [Kurc and Small, 2007]. The wet/dry classifications yielded

four soil moisture cases: dry shallow and dry deep zones (Case 1); wet shallow and dry

deep zones (Case 2); dry shallow and wet deep zones (Case 3); and wet shallow and wet

deep zones (Case 4).

Vegetation type and soil moisture case are used as categorical variables in this study. To

determine the relative importance of each categorical variable tested, the exponential

relationship (Equation 1) of each variable was assessed by the goodness-of-fit, measured

as the coefficient of determination (r2). An individual variable (e.g. grasslands) that exerts

stronger control over Reco should produce a better exponential fit compared to a weaker

category variable (e.g. shrublands). Categories (e.g. vegetation type), which produced

good individual model fits exert a stronger control on respiration than those with weaker

fits.

We used an analysis of covariance model (ANCOVA) to compare Reco models (Equation

1) among moisture cases and vegetation types. This ANCOVA test yields individual

models for each categorical variable, and allows those parameters to be compared

directly. Because the Reco distributions are not Gaussian, differences in Reco among

categories were assessed by the nonparametric Kruskal-Wallis test [Milton and Arnold,

2003]. Median Reco values were compared using the Bonferroni test to identify

differences in distributions of Reco rates among on moisture cases and vegetation types.

This test was chosen because it is conservative under conditions of unequal sample size

52

52

[Benjamini and Hochberg, 1995]. Model parameters from the ANCOVA were also

compared using the Bonferroni test.

3. Results

The four soil moisture cases follow a similar pattern in time for all five sites, driven by

the seasonality of rainfall (e.g. Figure 2). Fully dry soils characterized by Case 1 occur

for much of the year, from the end of the summer wet season in September through to the

next wet season in June. Occasional low intensity winter rains wet the soil enough to

produce Case 3 and 4 events that may persist for several days before the soils return to

Case 1. In general, the first 3-5 days at the start of the wet season fall into Case 2,

establishing wet conditions in the shallow soil. Cases 3 and 4 occur intermittently

throughout the wet season among the five sites, totaling 50 (± 29) and 94 (± 48) days,

respectively. Cases 1 and 4 each account for approximately 35% of soil moisture states in

all sites, while Case 3 represents 15% and Case 2 approximately 5% of the soil moisture

time series (these do not total 100% due to instances of soil moisture probe malfunction;

not shown).

The weakness of Tsoil alone as a predictor of Reco is demonstrated by a poor exponential

model fit (r2 = 0.0877, p <0.001; Figure 3a). Many of the Reco values are clustered around

zero and peak Reco is reached around 25 °C, with declining respiration at higher

temperatures. The exponential fit does not capture the variability in the data, indicating

that variables other than Tsoil influence respiration rates in dryland environments (Figure

53

53

3a). Similarly, shallow soil moisture (θs) is a poor predictor of Reco (Figure3b). Low Reco

rates occur even when soils are relatively wet, and high Reco rates are possible even with

relatively little water in surface soil. Ecosystem respiration peaks at intermediate levels of

θs (~ 0.10 m3 m-3) when temperatures are high (Tsoil > 20 °C). These conditions generally

correspond to moisture cases 2 and 4.

When Reco data are separated based on the wet/dry soil moisture cases, distinct patterns of

Reco variability become evident (Figure 2). Respiration is higher in Cases 2 and 4 when

soil moisture is present at the surface (0.06 and 0.07 mg C m-2 s-1 respectively) than in

Cases 1 and 3 when soil moisture is absent at the surface (0.03 and 0.03 mg C m-2 s-1, see

Table 2). The trends in (nighttime) Reco contrast with the pattern in NEE (Figure 2),

which is closely linked to the availability of deep soil moisture. When deep soil moisture

is available (Cases 3 and 4), NEE is generally less than zero (mean NEE is -0.03 and -

0.10 mg C m-2 s-1, respectively) indicating net carbon uptake (Figure 2, day of year 200-

275). Alternatively, dry deep soils (Cases 1 and 2) are more likely to correspond to net

carbon release (mean NEE values of 0.02 and 0.03, respectively) (Figure 2, day of year 1-

200).

When the data are divided based on their soil moisture case and then Reco is predicted

based on the typical Tsoil exponential model, the soil moisture control on respiration is

further reinforced (Figure 4). For instance, the best relationship between Reco and Tsoil (r2

= 0.44, p < 0.001) occurs in Case 2 when soil moisture is high in the surface, but absent

54

54

from the deep layer (Figure 4b). A strong exponential relationship with Tsoil was also

found in Case 4 (r2 = 0.41, p < 0.001; Figure 4d), when both shallow and deep soil zones

are wet. When moisture is only present in the deep soil (Case 3) a relationship exists

although not as strong as in Cases 2 and 4 (r2 = 0.21, p < 0.001; Figure 4c). In Case 1

when the soil is dry, Reco is generally near-zero, even at high temperatures (Figure 4a).

Reco is more responsive to increasing temperature when moisture is present in the soil,

especially near the surface (Figures 4b, c and d). The largest Reco fluxes occur in Case 4,

(Figure 4d), but Reco is also high when only the shallow soil is wet (Figure 4b). By

contrast, when only the deep soil is wet, the response of Reco to Tsoil is diminished (Figure

4c). Expressed as an apparent Q10 value (Equation 2), the dry soil conditions of Case 1

have a Q10 of 1.5 (± 0.09), while Cases 2 and 3 have a Q10 of 2.5 (± 0.28 and ±0.22,

respectively) and Case 4 has a value of 4.0 (±0.23).

In contrast to expectations [Barron-Gafford et al., 2011; Cable et al., 2012], little

difference was found in Reco rates among the three vegetation types (Table 2). The

grassland sites have average Reco rates of 0.04 ± 0.02 mg m-2 s-1 (error range is the

standard error of the mean), while rates in the shrublands average 0.045 ± 0.03 mg m-2 s-1

and the savannah average 0.06 ± 0.02 mg m-2 s-1 (Table 2). The difference between the

grass and shrub Reco rates is not distinguishable at 95% confidence in the Bonferroni test

and the distribution of Reco over the three vegetation types is similar (not shown).

Marginal means for all three vegetation types are within the 95% confidence in the

55

55

Bonferroni test under an ANCOVA model which includes Tsoil with vegetation type, i.e.

with the effect of Tsoil removed.

Differences in Reco trends among the three vegetation types become clear when we

separate the data based on soil moisture cases and then vegetation type, (Figure 5). As

expected, in Case 1, when moisture is absent from the soil profile, all three vegetation

types have similarly distributed and low Reco values. However, in Cases 2 and 4 when

soil moisture is present at the surface, savannah Reco is significantly greater than Reco in

both shrub and grass. The distribution of Reco rates in Case 4 includes larger fluxes at all

sites compared to Case 2. In Case 3, when moisture is only available deep in the soil, the

savannah and shrub sites have mean Reco values greater than the grassland (Table 2).

4. Discussion

At the daily level, temperature alone is insufficient to explain Reco as indicated by the

weak model as a function of Tsoil (Figure 3a). When Reco is analyzed as a function of θ, a

threshold response becomes apparent (Figure 3b). The Reco-θs pattern was the impetus for

exploring a threshold-based classification system for soil moisture in a two-zone soil

column. The threshold response indicates that a small amount of moisture can activate

Reco, but that Reco is not a direct function of soil moisture (Figure 3b). Further, Reco

responds differently to wet conditions in the shallow and deep soil zones (Figure 4, Table

3). Only after soils become wet (Cases 2-4) are other factors limiting for Reco [Austin et

al., 2004].

56

56

The soil moisture dynamic is conceptually consistent with the prevailing “bucket” model

of soil moisture [Knapp et al., 2008] but delineates between the response to shallow and

deep moisture. As this study demonstrates, the location of moisture in the soil may

influence the rate of Reco (Figure 4). Shallow and deep moisture stores may also facilitate

different components of Reco, i.e. autotrophic and heterotrophic respiration, depending on

the level of photosynthetic activity in the vegetation (Figures 3, 4).

Shallow and deep soil moisture vary at different rates. Shallow moisture responds rapidly

based on the frequency and intensity of storm events and evaporative demand [Kurc and

Small, 2004; Sala and Lauenroth, 1982] while deep soil moisture varies over longer

times (seasonal to annual scales; [Amenu et al., 2005]) and is dependent on soil hydraulic

properties [Porporato et al., 2002] and plant hydraulic redistribution [Scott et al., 2008]

among other factors. As a result, the responses in Reco identified here will vary under two

thresholds with two frequency patterns (Figure 6). Understanding future behavior of these

systems will depend on the ability to forecast shallow and deep moisture variability under

changing climate [Cayan et al., 2010].

4.1. The Effect of Soil Moisture on Reco

Our results demonstrate that soil moisture is a strong control on carbon flux in drylands

and drives different responses based on the depth of moisture. The influence of soil

moisture on plant water uptake indicates that deep soil moisture triggers vegetation

57

57

activity, including autotrophic respiration [Borgogno et al., 2010; Porporato et al., 2001].

Other studies have suggested threshold-type responses for soil respiration based on

available soil moisture or storm size [Potts et al., 2006; Sponseller, 2007]. Thresholds in

prior studies were defined for precipitation (either at the event or antecedent rainfall),

rather than soil moisture or matric potential. This study points toward soil moisture

thresholds as a more appropriate paradigm for ecosystem analyses.

The clearest pattern in Reco emerges when analyzed based on the vertical distribution of

soil moisture. The cases identify distinct ranges of Reco flux values. When soil moisture is

present at the surface (Cases 2 and 4) Reco rates are significantly higher (Figure 5)

compared to when moisture is absent at the surface (Case 1 and 3). Moisture deeper in

the profile (Case 3) slightly elevates Reco (Figure 5) compared to the fully dry conditions,

although not significantly so. By comparison, the 24-hour trend in NEE indicates more

net respiration under Cases 1 and 2, while Cases 3 and 4 are more frequently associated

with net carbon uptake. Taken altogether, this suggests the importance of shallow

moisture to Reco (both day and night) which is likely linked to soil microbial respiration.

Similarly, the correspondence with net carbon uptake in Cases 3 and 4 suggests that Reco

in these periods may be dominated by plant respiration.

Strong exponential Reco-Tsoil relationships emerge under the soil moisture case analysis

compared. The best exponential fits across all categorical variables occur in Cases 2 and

4 (Table 4). When moisture is present in the soil column as in Cases 2-4, Reco is nearly an

58

58

order of magnitude more sensitive to Tsoil compared to when the soil is dry. However,

when soils are dry (Case 1), ecosystems are moisture-limited, and respiration is minimal

across a range of temperatures (Figure 5a). In this case, the exponential model is weak.

Soil moisture cases provide a tool to identify times when moisture limitations are relieved

and other variables, such as Tsoil, may influence Reco. The characteristic response of Reco

to temperature varies among these cases, based on the location of moisture in the soil.

The different responses suggest different sensitivities at the ecosystem scale that emerge

from autotrophic and heterotrophic respiration responses, in contrast to a single,

ecosystem-level response [Cable et al., 2011; Chen et al., 2009].

The differences among Reco rates in the three vegetation types are closely linked to a Tsoil

effect. Our analysis shows that those differences are small and may be sensitive to the test

applied (no difference was found in the marginal means in the ANCOVA analysis, not

shown). This clarifies a movement in the literature pointing toward vegetation controls on

respiration processes via canopy-induced microclimates, microbial species assemblages

and root-driven soil moisture manipulation [Barron-Gafford et al., 2011; Breshears et al.,

1997; Metcalfe et al., 2011; Scott et al., 2008]. Differences among the three vegetation

types are only found after analyzing for the soil moisture distribution (Figure 5). For

example, the Reco rates at the savannah site and shrub sites are generally greater than at

the grassland sites, in Cases 2 and 3, although they are similar when considered outside

the moisture cases. The differences in Reco associated with community and plant function

are secondary to the effect of soil moisture.

59

59

Other respiration studies have focused on smaller spatial and temporal scales, and have

identified sub-daily hysteresis resulting from temperature [Barron-Gafford et al., 2011],

and plant and microbial community differences [Bell et al., 2008; White et al., 2009].

Here, we focused on large-scale landscape Reco at daily time steps rather than focusing on

short-term, plot or point measurements of soil respiration. Our approach integrates the

variability resulting from small scale processes, e.g. spatial heterogeneity of vegetation

and microbial communities. At these scales, the differences in respiration rates between

vegetation communities are minimal (Table 2) and only become apparent in the context

of soil moisture distribution (Figure 5).

4.2. Implications of Moisture Cases

The four moisture cases described here provide a useful approach to analyze ecosystem

dynamics in drylands. Changes in carbon fluxes in response to rainfall event size [Knapp

et al., 2008; Munson et al., 2010] suggest the importance of soil hydrologic

characteristics and this study identifies the different roles shallow and deep water play in

driving ecosystem activity. Water limitation and moisture pulse response have been

identified in a variety of other semi-arid landscapes in different environmental conditions

based both on carbon fluxes and metabolic rates [Jenerette et al., 2008; Sponseller,

2007]. Moisture patterns at different scales, e.g. seasonal and interannual, have been

identified in the response of Reco to moisture [Bowling et al., 2010]. The two-zone

method described here provides direct, explicit approach to identifying water limitation.

60

60

The two-zone moisture approach was applied here at a whole ecosystem scale. Research

at microsites near SRER-M pointed out the importance of satisfying moisture limitations

to stimulate soil respiration and the subsequent role of vegetation in respiration processes

[Barron-Gafford et al., 2011]. Differences among the three vegetation types in the

present study become clear when analyzed on the basis of the four soil moisture cases.

The greatest difference in mean Reco rates occurs in Case 2, when moisture is only present

at the surface. During those periods, the mean respiration rate in the savannah is over 0.1

mg m-2 s-1, compared to 0.025 mg m-2 s-1 at the shrub sites and 0.01 mg m-2 s-1 at the

grasslands. When moisture limitations are met in the deep soil as well, the savannah

ecosystem generally respires at a rate greater than that of the grassland and shrubland

sites. This may be due to greater diversity in the savannah relative to the grass and shrub

environments [Zak et al., 2003]. A higher respiration rate in savannahs is consistent with

expected results under woody plant encroachment along a riparian corridor [Scott et al.,

2006a].

The importance of soil moisture cases on Reco may depend on seasonal characteristics of

rainfall patterns and vegetation phenology. Different rainfall regimes may produce

different moisture thresholds and alter the pattern of moisture cases. A similar

dependence on the seasonality of moisture has been found in other semi-arid systems, e.g.

on grasslands adjacent to the Sev Grass site [Thomey et al., 2011, Carbone et al., 2011].

Alternatively, some communities may be less sensitive to the timing of moisture inputs

61

61

based on their ability to manipulate soil moisture, such as sagebrush [Bates et al., 2006].

Overall, the results shown here identify soil moisture as the primary factor determining

subsequent effects on respiration [Cable et al., 2011]. For future studies, the soil moisture

cases described here enable simple analysis in the context of water immediately available

to plants and soil microbes [Fernandez, 2007].

The differences in Q10 values in the present study compared to that of Cable et al. [2011]

suggest the dynamic response of Reco to moisture is important for understanding Reco in

drylands. The large Q10 values in Cases 2, 3 and 4 indicate the potential to dramatically

increase Reco when the ecosystem is wet, particularly in Case 4 when the entire soil

column is wet and the sensitivity to Tsoil is greatest. Plant physiological characteristics

and climate factors vary across dryland environments and may alter the influence of the

soil moisture distribution on Reco. Comparing the Reco responses of the three vegetation

types, we find more frequent high Reco rates at the savannah site compared to the

grassland and shrubland sites (Table 2). In landscapes undergoing a shift from grass- to

shrub-dominance, we may expect a change in the sensitivity of Reco to moisture inputs,

with particularly large Reco fluxes during the transition, in a mixed landscape. This pattern

of increased fluxes during vegetation transition has been described for riparian areas

[Scott et al., 2006a] and at smaller spatial scales in terms of soil respiration [Cable et al.,

2012]; we show that it is consistent in uplands as well.

62

62

In Cases 1 and 2, when deep soils are dry, Reco and daytime NEE are both positive, while

Reco and nee have opposite signs (i.e. NEE is negative) when deep soils are wet. As a

result, we may expect a large part of that Reco is associated with heterotrophic respiration,

while the vegetation is less active due to moisture stress. Similarly, when the deep soil

zone is wet, water is more readily accessed by vegetation, which may promote

autotrophic respiration. The potential relationship between the components of Reco can be

explored in the context of moisture availability at wider spatial scales than in previous

studies, which have relied on upscaling results from soil collar respiration data

[Richardson et al., 2006]. Such an approach could help to bridge gaps in understanding

respiration across wider spatial and temporal scales [Vargas et al., 2010a]. This research

points toward scaling relationships that consider the vertical distribution of moisture as a

potential driver of autotrophic and heterotrophic processes when combined with NEE

data or experimental approaches.

5. Conclusions

Among the variables tested, soil moisture was a strong predictor of Reco in the context of

a two-zone wet/dry classification scheme. Meeting the small moisture requirement in the

two soil zones enables a wide range of Reco flux values. This classification scheme

underscores the importance of the water-limitation mechanism in dryland environments.

Our results further highlight the location of that moisture within the soil column as an

important factor in ecosystem behavior. Moisture in the surface layer leads to higher Reco

63

63

rates than moisture deeper in the soil and Reco responds to Tsoil at different rates based on

moisture in the two soil zones.

The occurrence of different soil moisture cases is largely controlled by the magnitude of

and time between storms and soil hydraulic properties. As a result of potential changes in

precipitation patterns, the timing and frequency of soil moisture cases may also change.

Applying forecasts of precipitation under altered climate is among the potential

applications of the soil moisture cases identified here. The response of soil moisture to

variable rainfall inputs, and the ability to satisfy threshold values of soil moisture,

provides a clear approach to assessing the complexities associated with carbon flux in

drylands.

Appendix A: Defining Soil Moisture Thresholds

Wet/dry thresholds for the upper soil zone were determined based on exponential

drydown curves following rainfall events large enough to yield substantial changes in soil

moisture, i.e. > 8 mm d-1 [ Kurc and Small, 2004]. The terminal soil moisture value at the

end of the mean drydown curve for each site was used as the shallow soil moisture

threshold. The lower soil zone threshold was determined from the soil moisture value on

the first of at least five consecutive days of positive net carbon flux (release) after a

minimum five-day period of negative (uptake) net carbon flux. This threshold was chosen

to identify the value of θ in the deep soil when θ becomes a limiting factor on

photosynthetic uptake. Several transition periods occur in each site record and an average

64

64

soil moisture value was used based on all uptake/release transitions. The wet/dry

classification was performed independently for each site (see Table 2 for threshold

values).

The shallow soil zone threshold is determined based on exponential drydown curves for

each site. The curves are generated from the mean shallow soil moisture values following

daily rainfall events greater than 8 mm. This rainfall event size was selected to identify

storms that yield enough precipitation to increase soil moisture over several days [Kurc

and Small 2007] Shallow soil moisture data for each 8 mm rainfall event in the site

record were averaged and an exponential curve was fit to that average, using the Matlab

curve fitting toolbox (Figure A1). The moisture threshold was defined as the asymptote

of the exponential drydown.

Deep soil moisture thresholds were determined from the transition between net carbon

dioxide uptake and net release, i.e. from NEE < 0 to NEE > 0. Transitions were identified

when more than five consecutive days of carbon uptake were followed by five or more

days of net carbon release. The deep soil moisture transition occurs on the first day of net

carbon release in such a sequence. The values of deep soil moisture for each transition

were averaged for each site record to produce a single threshold value for each site

(Figure A2). Threshold values for the shallow and deep soil zones are given in Table A1.

Acknowledgements

65

65

Funding for this research was provided by the SAHRA (Sustainability of semi-Arid

Hydrology and Riparian Areas) Science and Technology Center of the National Science

Foundation (NSF agreement EAR-9876800) and the University of Arizona Water

Sustainability Program. The authors would like to thank the Ameriflux network for

providing public access to the data used in this study. All data were used according to the

Ameriflux fair-use policy. G. Barron-Gafford, D. Bunting, Z. Guido, H. Neal, K. Nelson,

Z. Sanchez-Mejia, and R. Vargas made helpful contributions to earlier versions of this

manuscript.

66

66

References Almagro, M., J. Lopez, J. I. Querejeta, and M. Martinez-Mena (2009), Temperature dependence of soil CO2 efflux is strongly modulated by seasonal patterns of moisture availability in a Mediterranean ecosystem, Soil Biology &amp; Biochemistry, 41(3), 594-605. Amenu, G. G., P. Kumar, and X. Z. Liang (2005), Interannual variability of deep-layer hydrologic memory and mechanisms of its influence on surface energy fluxes, Journal of Climate, 18(23), 5024-5045. Austin, A. T., L. Yahdjian, J. M. Stark, J. Belnap, A. Porporato, U. Norton, D. A. Ravetta, and S. M. Schaeffer (2004), Water pulses and biogeochemical cycles in arid and semiarid ecosystems, Oecologia, 141(2), 221-235. Bahn, M., et al. (2010), Soil respiration at mean annual temperature predicts annual total across vegetation types and biomes, Biogeosciences, 7(7), 2147-2157. Baldocchi, D., et al. (2001), FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities, Bulletin of the American Meteorological Society, 82(11), 2415-2434. Barron-Gafford, G. A., R. L. Scott, G. D. Jenerette, and T. E. Huxman (2011), The relative controls of temperature, soil moisture, and plant functional group on soil CO2 efflux at diel, seasonal, and annual scales, J. Geophys. Res., 116(G1), G01023. Bates, J. D., T. Svejcar, R. F. Miller, and R. A. Angell (2006), The effects of precipitation timing on sagebrush steppe vegetation, Journal of Arid Environments, 64(4), 670-697. Bell, C., N. McIntyre, S. Cox, D. Tissue, and J. Zak (2008), Soil microbial responses to temporal variations of moisture and temperature in a Chihuahuan Desert Grassland, Microbial Ecology, 56(1), 153-167. Benjamini, Y., and Y. Hochberg (1995), Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing, Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289-300. Bhark, E. W., and E. E. Small (2003), Association between plant canopies and the spatial patterns of infiltration in shrubland and grassland of the Chihuahuan Desert, New Mexico, Ecosystems, 6(2), 185-196. Borgogno, F., P. D'Odorico, F. Laio, and L. Ridolfi (2010), A stochastic model for vegetation water stress, Ecohydrology, 3(177–188), doi: 10.1002/eco.120

67

67

Bowling, D. R., S. Bethers-Marchetti, C. K. Lunch, E. E. Grote, and J. Belnap (2010), Carbon, water, and energy fluxes in a semiarid cold desert grassland during and following multiyear drought, J. Geophys. Res., 115(G4), G04026. Breshears, D. D., P. M. Rich, F. J. Barnes, and K. Campbell (1997), Overstory-Imposed Heterogeneity in Solar Radiation and Soil Moisture in a Semiarid Woodland, Ecological Applications, 7(4), 1201-1215. Cable, J., et al. (2011), The temperature responses of soil respiration in deserts: a seven desert synthesis, Biogeochemistry, 103(1), 71-90-90. Cable, J. M., G. A. Barron-Gafford, K. Ogle, M. Pavao-Zuckerman, R. L. Scott, D. G. Williams, and T. E. Huxman (2012), Shrub encroachment alters sensitivity of soil respiration to temperature and moisture, J. Geophys. Res., 117(G1), G01001. Caldwell, M. M., and J. H. Richards (1989), Hydraulic lift: water efflux from upper roots improves effectiveness of water uptake by deep roots, Oecologia, 79(1), 1-5. Carbone, M., C. Still, A. Ambrose, T. Dawson, A. Williams, C. Boot, S. Schaeffer, and J. Schimel (2011), Seasonal and episodic moisture controls on plant and microbial contributions to soil respiration, Oecologia, 167(1), 265-278. Carbone, M. S., G. C. Winston, and S. E. Trumbore (2008), Soil respiration in perennial grass and shrub ecosystems: Linking environmental controls with plant and microbial sources on seasonal and diel timescales, J. Geophys. Res., 113(G2), G02022. Cavanaugh, M. L., S. A. Kurc, and R. L. Scott (2010), Evapotranspiration partitioning in semiarid shrubland ecosystems: a two-site evaluation of soil moisture control on transpiration, Ecohydrology, doi: 10.1002/eco.157 Cayan, D. R., T. Das, D. W. Pierce, T. P. Barnett, M. Tyree, and A. Gershunov (2010), Future dryness in the southwest US and the hydrology of the early 21st century drought, Proceedings of the National Academy of Sciences, 107(50), 21271-21276. Chen, S. P., G. H. Lin, J. H. Huang, and G. D. Jenerette (2009), Dependence of carbon sequestration on the differential responses of ecosystem photosynthesis and respiration to rain pulses in a semiarid steppe, Global Change Biology, 15(10), 2450-2461. Conant, R. T., P. Dalla-Betta, C. C. Klopatek, and J. A. Klopatek (2004), Controls on soil respiration in semiarid soils, Soil Biology &amp; Biochemistry, 36(6), 945-951.

68

68

Davidson, E. A., S. Samanta, S. S. Caramori, and K. Savage (2012), The Dual Arrhenius and Michaelis–Menten kinetics model for decomposition of soil organic matter at hourly to seasonal time scales, Global Change Biology, 18(1), 371-384. Fang, C., and J. B. Moncrieff (2001), The dependence of soil CO2 efflux on temperature, Soil Biology and Biochemistry, 33(2), 155-165. Fernandez, R. J. (2007), On the frequent lack of response of plants to rainfall events in arid areas, Journal of Arid Environments, 68(4), 688-691. Franzluebbers, K., A. J. Franzluebbers, and M. D. Jawson (2002), Environmental Controls on Soil and Whole-ecosystem Respiration from a Tallgrass Prairie, Soil Science Society of America Journal, 66(1), 254-262. Goulden, M. L., J. W. Munger, S. M. Fan, B. C. Daube, and S. C. Wofsy (1996), Measurements of carbon sequestration by long-term eddy covariance: Methods and a critical evaluation of accuracy, Global Change Biology, 2(3), 169-182. Grünzweig, J. M., D. Hemming, K. Maseyk, T. Lin, E. Rotenberg, N. Raz-Yaseef, P. D. Falloon, and D. Yakir (2009), Water limitation to soil CO2 efflux in a pine forest at the semiarid “timberline”, J. Geophys. Res., 114(G3), G03008. Hanson, P. J., N. T. Edwards, C. T. Garten, and J. A. Andrews (2000), Separating root and soil microbial contributions to soil respiration: A review of methods and observations, Biogeochemistry, 48(1), 115-146. Huxman, T. E., B. P. Wilcox, D. D. Breshears, R. L. Scott, K. A. Snyder, E. E. Small, K. Hultine, W. T. Pockman, and R. B. Jackson (2005), Ecohydrological implications of woody plant encroachment, Ecology, 86(2), 308-319. Huxman, T. E., K. A. Snyder, D. Tissue, A. J. Leffler, K. Ogle, W. T. Pockman, D. R. Sandquist, D. L. Potts, S. Schwinning, and Yu (2004), Precipitation pulses and carbon fluxes in semiarid and arid ecosystems, Oecologia, 141(2), 254-268. Jenerette, D. G., and R. Lal (2005), Hydrologic sources of carbon cycling uncertainty throughout the terrestrial–aquatic continuum, Global Change Biology, 11(11), 1873-1882. Jenerette, G. D., R. L. Scott, and T. E. Huxman (2008), Whole ecosystem metabolic pulses following precipitation events, Functional Ecology, 22(5), 924-930. Jin, Z., Y. S. Dong, Y. C. Qi, and Z. S. An (2010), Soil respiration and net primary productivity in perennial grass and desert shrub ecosystems at the Ordos Plateau of Inner Mongolia, China, Journal of Arid Environments, 74(10), 1248-1256.

69

69

Jobbágy, E. G., and R. B. Jackson (2000), THE VERTICAL DISTRIBUTION OF SOIL ORGANIC CARBON AND ITS RELATION TO CLIMATE AND VEGETATION, Ecological Applications, 10(2), 423-436. Knapp, A. K., et al. (2008), Consequences of More Extreme Precipitation Regimes for Terrestrial Ecosystems, BioScience, 58(9), 811-821. Kurc, S. A., and E. E. Small (2004), Dynamics of evapotranspiration in semiarid grassland and shrubland ecosystems during the summer monsoon season, central New Mexico, Water Resources Research, 40(9), 15. Kurc, S. A., and E. E. Small (2007), Soil moisture variations and ecosystem-scale fluxes of water and carbon in semiarid grassland and shrubland, Water Resources Research, 43(6). Kurc, S. A., and L. M. Benton (2010), Digital image-derived greenness links deep soil moisture to carbon uptake in a creosotebush-dominated shrubland, Journal of Arid Environments. doi:10.1016/j.jaridenv.2009.10.003 Laio, F., A. Porporato, L. Ridolfi, and I. Rodriguez-Iturbe (2001), Plants in water-controlled ecosystems: active role in hydrologic processes and response to water stress: II. Probabilistic soil moisture dynamics, Advances in Water Resources, 24(7), 707-723. Lal, R. (2004), Carbon sequestration in dryland ecosystems, Environmental Management, 33(4), 528-544. Lavigne, M. B., et al. (1997), Comparing nocturnal eddy covariance measurements to estimates of ecosystem respiration made by scaling chamber measurements at six coniferous boreal sites, Journal of Geophysical Research-Atmospheres, 102(D24), 28977-28985. Law, B. E., M. G. Ryan, and P. M. Anthoni (1999), Seasonal and annual respiration of a ponderosa pine ecosystem, Global Change Biology, 5(2), 169-182. Liu, W. X., Z. Zhang, and S. Q. Wan (2009), Predominant role of water in regulating soil and microbial respiration and their responses to climate change in a semiarid grassland, Global Change Biology, 15(1), 184-195. Lloyd, J., and J. A. Taylor (1994), On the Teperature-Dependence of Soil Respiration, Functional Ecology, 8(3), 315-323.

70

70

Loik, M. E., D. D. Breshears, W. K. Lauenroth, and J. Belnap (2004), A multi-scale perspective of water pulses in dryland ecosystems: climatology and ecohydrology of the western USA, Oecologia, 141(2), 269-281. Luo, Y., and X. Zhou (2006), Controlling factors, in Soil Respiration and the Environment, edited, pp. 79-105, Academic Press, Burlington. Mahecha, M. D., et al. (2010), Global Convergence in the Temperature Sensitivity of Respiration at Ecosystem Level, Science, 329(5993), 838-840. Metcalfe, D. B., R. A. Fisher, and D. A. Wardle (2011), Plant communities as drivers of soil respiration: pathways, mechanisms, and significance for global change, Biogeosciences, 8(8), 2047-2061. Milton, J. S., and J. C. Arnold (2003), Introduction to Probability and Statistics, 4th ed., McGraw-Hill. Moffat, A. M., et al. (2007), Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes, Agricultural and Forest Meteorology, 147(3-4), 209-232. Moran, M. S., R. L. Scott, E. P. Hamerlynck, K. N. Green, W. E. Emmerich, and C. D. H. Collins (2009), Soil evaporation response to Lehmann lovegrass (Eragrostis lehmanniana) invasion in a semiarid watershed, Agricultural and Forest Meteorology, 149(12), 2133-2142. Moran, M. S., et al. (2010), Hydrologic response to precipitation pulses under and between shrubs in the Chihuahuan Desert, Arizona, Water Resour. Res., 46(10), W10509. Munson, S., T. Benton, W. Lauenroth, and I. Burke (2010), Soil carbon flux following pulse precipitation events in the shortgrass steppe, Ecological Research, 25(1), 205-211. Ng, C. W. W., and L. T. Zhan (2007), Comparative study of rainfall infiltration into a bare and a grassed unsaturated expansive soil slope, Soils and Foundations, 47(2), 207-217. Noy-Meir, I. (1973), Desert Ecosystems: Environment and Producers, Annual Review of Ecology and Systematics, 4(1), 25-51. Ogle, K., and J. F. Reynolds (2004), Plant responses to precipitation in desert ecosystems: integrating functional types, pulses, thresholds, and delays, Oecologia, 141(2), 282-294.

71

71

Porporato, A., F. Laio, L. Ridolfi, and I. Rodriguez-Iturbe (2001), Plants in water-controlled ecosystems: active role in hydrologic processes and response to water stress - III. Vegetation water stress, Advances in Water Resources, 24(7), 725-744. Porporato, A., P. D'Odorico, F. Laio, L. Ridolfi, and I. Rodriguez-Iturbe (2002), Ecohydrology of water-controlled ecosystems, Advances in Water Resources, 25(8-12), 1335-1348. Potts, D. L., T. E. Huxman, J. M. Cable, N. B. English, D. D. Ignace, J. A. Eilts, M. J. Mason, J. F. Weltzin, and D. G. Williams (2006), Antecedent moisture and seasonal precipitation influence the response of canopy-scale carbon and water exchange to rainfall pulses in a semi-arid grassland, New Phytologist, 170(4), 849-860. Raich, J. W., and W. H. Schlesinger (1992), The Global Carbon-Dioxide Flux in Soil Respiration and its Relationship to Vegetation and Climate, Tellus Series B-Chemical and Physical Meteorology, 44(2), 81-99. Reichstein, M., et al. (2005), On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm, Global Change Biology, 11(9), 1424-1439. Reynolds, J. F., P. R. Kemp, K. Ogle, and R. J. Fernandez (2004), Modifying the 'pulse-reserve' paradigm for deserts of North America: precipitation pulses, soil water, and plant responses, Oecologia, 141(2), 194-210. Reynolds, J. F., et al. (2007), Global Desertification: Building a Science for Dryland Development, Science, 316(5826), 847-851. Richardson, A. D., et al. (2006), Comparing simple respiration models for eddy flux and dynamic chamber data, Agricultural and Forest Meteorology, 141(2-4), 219-234. Sala, O. E., and W. K. Lauenroth (1982), Small Rainfall Events: An Ecological Role in Semi-Arid Regions, Oecologia, 53(3), 301-304. Schlesinger, W. H., and J. A. Andrews (2000), Soil respiration and the global carbon cycle, Biogeochemistry, 48(1), 7-20. Scott, R. L., W. L. Cable, and K. R. Hultine (2008), The ecohydrologic significance of hydraulic redistribution in a semiarid savanna, Water Resources Research, 44(2). Scott, R. L., W. J. Shuttleworth, T. O. Keefer, and A. W. Warrick (2000), Modeling multiyear observations of soil moisture recharge in the semiarid American Southwest, Water Resour. Res., 36(8), 2233-2247.

72

72

Scott, R. L., T. E. Huxman, D. G. Williams, and D. C. Goodrich (2006a), Ecohydrological impacts of woody-plant encroachment: seasonal patterns of water and carbon dioxide exchange within a semiarid riparian environment, Global Change Biology, 12(2), 311-324. Scott, R. L., T. E. Huxman, W. L. Cable, and W. E. Emmerich (2006b), Partitioning of evapotranspiration and its relation to carbon dioxide exchange in a Chihuahuan Desert shrubland, Hydrological Processes, 20(15), 3227-3243. Scott, R. L., G. D. Jenerette, D. L. Potts, and T. E. Huxman (2009), Effects of seasonal drought on net carbon dioxide exchange from a woody-plant-encroached semiarid grassland, Journal of Geophysical Research-Biogeosciences, 114. Sponseller, R. A. (2007), Precipitation pulses and soil CO2 flux in a Sonoran Desert ecosystem, Global Change Biology, 13(2), 426-436. Stull, R. B. (1988), An introduction to boundary layer meteorology, 670 pp., Kluwer Academic Publishers, Dordrecht, The Netherlands. Tang, J. W., and D. D. Baldocchi (2005), Spatial-temporal variation in soil respiration in an oak-grass savanna ecosystem in California and its partitioning into autotrophic and heterotrophic components, Biogeochemistry, 73(1), 183-207. Thomas, C. K., B. E. Law, J. Irvine, J. G. Martin, J. C. Pettijohn, and K. J. Davis (2009), Seasonal hydrology explains interannual and seasonal variation in carbon and water exchange in a semiarid mature ponderosa pine forest in central Oregon, Journal of Geophysical Research-Biogeosciences, 114. Thomey, M. L., S. L. Collins, R. Vargas, J. E. Johnson, R. F. Brown, D. O. Natvig, and M. T. Friggens (2011), Effect of precipitation variability on net primary production and soil respiration in a Chihuahuan Desert grassland, Global Change Biology, 17(4), 1505-1515. Vargas, R., D. D. Baldocchi, M. Bahn, P. J. Hanson, K. P. Hosman, L. Kulmala, J. Pumpanen, and B. Yang (2011), On the multi-temporal correlation between photosynthesis and soil CO2 efflux: reconciling lags and observations, New Phytologist, 191(4), 1006-1017. Vargas, R., M. Carbone, M. Reichstein, and D. Baldocchi (2010b), Frontiers and challenges in soil respiration research: from measurements to model-data integration, Biogeochemistry, 1-13. Vargas, R., Collins, S. L., Thomey, M. L., Johnson, J. E., Brown, R. F., Natvig, D. O. and Friggens, M. T. (2012), Precipitation variability and fire influence the temporal dynamics

73

73

of soil CO2 efflux in an arid grassland. Global Change Biology. doi: 10.1111/j.1365-2486.2011.02628.x Vargas, R., M. Detto, D. D. Baldocchi, and M. F. Allen (2010a), Multiscale analysis of temporal variability of soil CO2 production as influenced by weather and vegetation, Global Change Biology, 16(5), 1589-1605. Wainwright, J., A. J. Parsons, and A. D. Abrahams (2000), Plot-scale studies of vegetation, overland flow and erosion interactions: case studies from Arizona and New Mexico, Hydrological Processes, 14(16-17), 2921-2943. White, D. A., A. Welty-Bernard, C. Rasmussen, and E. Schwartz (2009), Vegetation controls on soil organic carbon dynamics in an arid, hyperthermic ecosystem, Geoderma, 150(1-2), 214-223. Williams, C. A., and J. D. Albertson (2004), Soil moisture controls on canopy-scale water and carbon fluxes in an African savanna, Water Resources Research, 40(9), 14. Wohlfahrt, G., L. F. Fenstermaker, and J. A. Arnone (2008), Large annual net ecosystem CO2 uptake of a Mojave Desert ecosystem, Global Change Biology, 14(7), 1475-1487. Wythers, K. R., W. K. Lauenroth, and J. M. Paruelo (1999), Bare-soil evaporation under semiarid field conditions, Soil Science Society of America Journal, 63(5), 1341-1349. Xu, L. K., D. D. Baldocchi, and J. W. Tang (2004), How soil moisture, rain pulses, and growth alter the response of ecosystem respiration to temperature, Global Biogeochemical Cycles, 18(4), 10. Xu, M., and Y. Qi (2001), Budgets of soil erosion and deposition for sediments and sedimentary organic carbon across the conterminous United States, Global Biogeochem. Cycles, 15(3), 687-696. Yaseef, N. R., D. Yakir, E. Rotenberg, G. Schiller, and S. Cohen (2010), Ecohydrology of a semi-arid forest: partitioning among water balance components and its implications for predicted precipitation changes, Ecohydrology, 3(2), 143-154. Zak, D. R., W. E. Holmes, D. C. White, A. D. Peacock, and D. Tilman (2003), Plant Diversity, Soil Microbial Communities, and Ecosystem Function: Are There Any Links?, Ecology, 84(8), 2042-2050.

74

74

Table Captions

Table 1. Characteristics for the five Ameriflux sites used in this study. Length of data and

years of record describe publicly available data from the Ameriflux website.

Table 2. Mean (± standard error) and standard deviation of R eco rates based on secondary

controls. Results of the Bonferroni comparison tests are also given. Comparison tests

were performed uniquely for each grouping, i.e. moisture cases, vegetation type, and

combined moisture cases and vegetation type.

Table 3. Regression coefficients and coefficient of determination (r2) for exponential

models based on different categorizations. Letters indicate parameter groups based on a

Bonferroni test. Parameters for vegetation type models all belong to the same group.

Table A1. Threshold values for the shallow and deep soil zones for each of the five sites.

Method for determining thresholds are described in section 2.3

75

75

Figure Captions

Figure 1. Map showing study sites. Contours indicate annual rainfall (National

Oceanographic and Atmospheric Administration (NOAA) Climate Maps of the US,

http://hurricane.ncdc.noaa.gov/cgi-bin/climaps/climaps.pl).

Figure 2. Time series of precipitation (P), NEE and Reco at the SRER-M site. Colors

indicate the soil moisture conditions, which transition from predominantly Case 1 in the

dry (winter) season through Cases 2, 3 and 4 in the wet (summer) season.

Figure 3. (a) Respiration as a function of soil temperature at all sites. Fit is an exponential

curve, derived from the linearized form in equation 1. (b) Respiration as a function of

shallow soil moisture (θs). Note the increase in Reco variability when soil moisture is

above approximately 0.03 m3 m-3. Fit is piecewise linear with a breakpoint at 0.28 m3 m-

3.

Figure 4. Reco-Tsoil regressions categorized by soil moisture cases: (a) Case 1, (b) Case 2,

(c) Case 3, and (d) Case 4.

Figure 5. Boxplot of Reco distributions categorized by soil moisture case and vegetation

type.

Figure 6. Revised soil moisture model, after the “bucket” model of Knapp et al [2008].

The solid black line represents average soil moisture while the undulating line indicates

variation around the average. Red dashed lines indicate the upper and lower thresholds

for moisture-controlled ecosystem activity.

Figure A1. Drydown of shallow soil moisture (θ) at the Sev Shrub site. Points indicate

mean soil moisture, error bars are one standard deviation.

76

76

Figure A2. Identification of transition points between negative (uptake) and positive

(release) NEE to define the deep soil moisture threshold (θ*deep) at SRER-M. Gray bars

indicate individual instances of prolonged carbon uptake.

77

77

Site Reference Vegetation Type % Cover Root depth Veg. height (cm) (m) Sevilleta Grass [Kurc and Small, 2007] Grassland (Bouteloua eripoda) 50 100 0.3 Sevilleta Shrub [Kurc and Small, 2007] Open Shrubland (Larrea tridentata) 30 100 0.75 Kendall Grass [Moran et al., 2009] Grassland (B. eripoda, Eragrostis lehmanniana) 40 n/r n/r SRER Mesquite [ Scott et al., 2008] Woody Savanna (Prosopis velutina) 35 n/r 4 SRER Creosote [ Kurc and Benton, 2010] Open Shrubland (L. tridentata) 24 n/r 1.7

n/r--not reported

1

Soil Type MAP MAT Soil Moisture Depths Soil Moisture

Sensor Length of Record (mm) (°C) (cm) (years)

loamy sand 230 17.5 2.5, 12.5, 22.5, 37.5, 52.5 CS500 3 (2002-2004) sandy loam 230 17.5 2.5, 12.5, 22.5, 37.5, 52.5 CS500 3 (2002-2004) sandy loam 357 17 5,15,30,50 TDR100 4 (2004-2007) sandy loam 310 19 5,10,20,30,50 CS616 4 (2004-2007) sandy loam 260 21 2.5, 12.5, 22.5, 37.5, 52.5 CS616 2 (2008-2009)

78

78

N Mean (mg m2 s-1) St. Dev. (mg m2 s-1) All 2422 0.0498 (±0.001) 0.0673 Vegetation Type Grass 924 0.0370 (±0.002) 0.0461 a Shrub 423 0.0457 (±0.003) 0.0631 b Savannah 1075 0.0625 (±0.002) 0.0806 b Soil Moisture Case Case 1 701 0.0255 (±0.003) 0.0347 a Case 2 108 0.0635 (±0.002) 0.0742 b Case 3 499 0.0292 (±0.006) 0.0328 a Case 4 1046 0.0737 (±0.003) 0.0829 b Moisture Case-Vegetation

Case 1 Grass 184 0.0263 (± 0.004) 0.0337 a Shrub 165 0.0251 (± 0.005) 0.0382 a Savannah 352 0.0252 (± 0.003) 0.0336 a Case 2 Grass 39 0.0214 (± 0.010) 0.0277 a,b Shrub 25 0.0344 (± 0.012) 0.0277 a,b,c,d Savannah 44 0.1174 (± 0.009) 0.0867 c,d Case 3 Grass 255 0.0205 (± 0.004) 0.0244 a Shrub 53 0.0329 (± 0.008) 0.0291 a,b Savannah 191 0.0398 (± 0.004) 0.0397 b Case 4 Grass 413 0.0547 (± 0.003) 0.0566 b,c Shrub 145 0.0622 (± 0.005) 0.0711 b,c Savannah 488 0.0933 (±0.003) 0.0990 c,d

79

79

A b r2 All 0.0189 0.0465 0.0877 Vegetation Type Grass 0.0189 0.0383 0.0591 Shrub 0.0241 0.0325 0.0453 Savannah 0.0209 0.0436 0.0678 Soil Moisture Case Case 1 0.0046 (a) 0.0431 (a) 0.0508 Case 2 0.0055 (a) 0.0942 (b) 0.4423 Case 3 0.0024 (b) 0.0933 (b) 0.2115 Case 4 0.0026 (b) 0.1375 (c) 0.4087 All fits are significant at p < 0.001

80

80

Kendall SRER-M SRER-C Sev Grass Sev Shrub Shallow 0.07 0.04 0.09 0.11 0.09

Deep 0.13 0.06 0.1 0.11 0.10

81

81

Figure 1.

82

82

Figure 2.

83

83

Figure 3.

84

84

Figure 4.

85

85

Figure 5.

86

86

Figure 6.

“Bucket” soil zone

Shallow soil zone

Deep soil zone

87

87

Figure A1.

88

88

Figure A2.

89

89

APPENDIX B: SOIL MOISTURE CONTROLS REMOTE SENSING-BASED UPSCALING OF CARBON FLUXES IN A SEMIARID SHRUBLAND

Andrew L. Neal1, Shirley A. Kurc1,2, Willem J.D. van Leeuwen2,3, Paul D. Brooks1

for submission to Agricultural and Forest Meteorology

A.L. Neal. Department of Hydrology and Water Resources, University of Arizona, 1133

James E. Rogers Way, Tucson, AZ 85721. (aneal@email.arizona.edu)

S.A. Kurc. School of Natural Resources and the Environment, University of Arizona, 

P.O. Box 210043, Tucson, AZ 85721 (kurc@email.arizona.edu)

W.J.D. van Leeuwen. School of Natural Resources and Environment, University of

Arizona, 1955 E. 6th Street Tucson, AZ 85721 US

P.D. Brooks. Department of Hydrology and Water Resources, University of Arizona,

1133 James E. Rogers Way, Tucson, AZ 85721. (brooks@email.arizona.edu)

1 Department of Hydrology and Water Resources, University of Arizona.

2 School of Natural Resources and the Environment, University of Arizona.

3 School of Geography and Development, University of Arizona

90

90

Abstract

The application of spatially-distributed models of carbon dioxide flux using remote

sensing data is a rapidly developing line of research. Validation of such models requires

linking remote sensing data to ground measurements of carbon dioxide exchange,

generally collected by eddy covariance towers. In semiarid shrublands, this link is

complicated by sparse vegetation cover and temporal variability in vegetation activity in

response to soil moisture. We find that remotely sensed vegetation greenness near the

tower is not strongly correlated to tower CO2 fluxes during dry soil moisture conditions.

We determine that increased deep soil moisture identifies a relationship between

greenness and carbon uptake via photosynthesis (r2 = 0.37 for net carbon flux, r2 = 0.75

for gross primary productivity). The spatial vegetation distribution around the tower is

shown to have little effect on the link between greenness and carbon fluxes. Together the

two analyses enable improved spatially-distributed modeling of carbon flux from remote

sensing data.

Keywords: net ecosystem exchange (NEE), gross primary productivity (GPP), Moderate

Resolution Imaging Spectroradiometer (MODIS), enhanced vegetation index (EVI),

Sevilleta Long Term Ecological Research (LTER) site, ecohydrology

91

91

1. Introduction

Carbon dioxide fluxes, particularly terrestrial fluxes, may substantially influence

atmospheric and biogeochemical cycles (Friedlingstein et al. 2006; IPCC 2007). Spatially

explicit carbon flux estimates can help quantify the net ecosystem exchange of carbon

(NEE) between the land and the atmosphere (Gilmanov et al. 2005; Wylie et al. 2003),

often generated using models of NEE based on satellite remote sensing data, such as

vegetation indices, temperature and leaf area index (Goerner et al. 2011; King et al. 2011;

Wu et al. 2010; Xiao et al. 2010; Zhang et al. 2011). Several approaches to link ground

measurements of NEE to satellite data have produced reasonable carbon flux estimates at

regional to continental scales (Mahadevan et al. 2008; Wylie et al. 2003; Xiao et al. 2011;

Xiao et al. 2010). These models depend on the link between vegetation greenup as

observed by satellites and the subsequent uptake of carbon by photosynthesis (or gross

primary productivity, GPP) (Sims et al. 2006a; Sims et al. 2006b; Wu et al. 2010).

The link between satellite data and NEE is strongest in humid regions (Sims et al. 2008;

Tang et al. 2011), where the predominant limitation to vegetation activity is energy from

solar radiation (Badeck et al. 2004; Jolly et al. 2005). In these humid regions, vegetation

phenology, e.g. leaf-out, is largely determined by seasonal energy availability, and occurs

in synchrony with plant carbon uptake via photosynthesis (Sims et al. 2006b; Tang et al.

2011); Figure 1a). The synchrony of greenup and carbon uptake in humid ecosystems

reduces the complexity of modeling carbon flux based on remote sensing observations.

Further, the relative uniformity of vegetation cover, as opposed to bare soil (Asrar et al.

92

92

1992; Gao et al. 2000), has further simplified the application of simple regression models

linking satellite data to NEE in humid regions (Sims et al. 2006b; Tang et al. 2011).

However, not all ecosystems are energy-limited. For instance, in water-limited semiarid

ecosystems, plant carbon uptake is strongly linked to intermittent moisture pulses

(Huxman et al. 2004; Noy-Meir 1973). Moisture pulses can have very different effects on

soil moisture, and thus on carbon flux, depending on the magnitude of a rainfall event.

For example, small storms, which occur relatively frequently (Sala and Lauenroth 1982)

generally wet the shallow soil and are associated with a net release of carbon dioxide

(Kurc and Benton 2010; Kurc and Small 2007; Scott et al. 2006) Larger, infrequent

storms are required to wet the deeper soil layers (Kurc and Small 2007), which leads to

net carbon uptake (Cavanaugh et al. 2011; Kurc and Benton 2010; Notaro et al. 2010).

While carbon fluxes are closely linked to moisture pulses, vegetation greenness may not

correlate strongly with carbon flux. Desert shrubs under water stress may retain their

green pigment (Hamerlynck and Huxman 2009), but will restrict their photosynthetic

activity in order to minimize transpiration losses (Ogle and Reynolds 2002). Many desert

shrubs, including creosotebush (Larrea tridentata), retain leaves despite lengthy dry

conditions (Lajtha and Whitford 1989) and can respond quickly when moisture is again

available (Franco et al. 1994). Because of these physiological adaptations, NEE and GPP

are often poorly linked to remotely sensed vegetation greenness (Sims et al. 2006a).

Satellite data may not capture vegetation response to moisture over short time scales

(Sjöström et al. 2009), due to the sensitivity of semiarid ecosystems to brief, localized

93

93

storms that yield small amounts of soil moisture (Huxman et al. 2004; Sala and

Lauenroth 1982).

Further complicating the relationship between satellite-observed vegetation greenness

and NEE is that semiarid ecosystems tend to be sparsely vegetated when compared to

more mesic ecosystems (Smith et al. 1990; Xiao and Moody 2005). While the seasonal

pattern of vegetation change in sparse ecosystems has been described (Fensholt and

Sandholt 2003; Xiao and Moody 2005), this sparseness may limit the ability to precisely

resolve ecosystem carbon dynamics following rainfall events. The limitation is likely due

to the complex mixtures and contributions of vegetation and soil spectral reflectance

values, as well as the effects of moisture limitations on vegetation (Sims et al. 2006a;

Turner et al. 2005). A high percent vegetative cover simplifies the interpretation of

satellite vegetation data (Asrar et al. 1992; Gao et al. 2000), because changes in

vegetation index value can be ascribed to vegetation alone. However, the link between

satellite-observed greenness and carbon flux still depends on the density of vegetation,

based on the connection between leaf area and vegetation indices (Fensholt et al. 2004).

The main objective of this study is to improve the utility of remotely sensed data to

upscale carbon data in semiarid ecosystems. We contend that the weak link between EVI

and NEE in semiarid shrublands is due to a disconnect between observable features

(greenness) and plant physiology (GPP, NEE) based on water stress (Figure 1). We

identify when satellite-observed vegetation greenness is more closely related to carbon

flux, as a function of soil moisture. The relationship with deep soil moisture will yield

94

94

reasonable estimates of carbon uptake for the region on the landscape that is consistent

with the tower area.

Specifically, we address the following hypotheses for this Creosote dominated area. First,

we hypothesize that the relationship between greenness and carbon flux (and specifically

between greenness and carbon uptake) will be strongest when moisture is available deep

in the soil, because it is available to vegetation over long periods of time and is not lost to

direct evaporation. Second, we hypothesize that the relationship between EVI and GPP

will identify when satellite-observed changes in vegetation directly relate to carbon

uptake, which we anticipate to occur with deep soil moisture.

To address our hypotheses, we examine the link between satellite data and carbon flux in

the context of a soil moisture classification scheme that defines wet and dry conditions

suitable for carbon flux activity. Wet and dry periods are defined on the basis of physical

processes in the shallow soil (the so called “drydown” period) and ecosystem water stress

in the deep soil (Neal 2012). Using this classification scheme, we isolate periods when

water limitation is relaxed so that observed reflectance predicts carbon uptake, e.g.

following large storms during the wet season. We test these characteristics at a semiarid

upland shrub-dominated site using NEE and GPP data collected via eddy covariance

(Kurc and Small 2007) and EVI data from the Moderate Resolution Imaging

Spectroradiometer (MODIS).

2. Data

2.1. Site Description

95

95

Our flux tower study site is located within the semiarid southwestern United States (US).

The Sevilleta Shrubland (34.3349 °N, 106.7442 °W) is part of the Ameriflux network and

is located in the Sevilleta National Wildlife Refuge in central New Mexico. Mean annual

precipitation at the site is 230 mm with > 50 % of the annual rainfall occurring July-

September; mean annual temperature is 17.5 °C (Kurc and Small 2007). The site is

predominantly creosotebush (L. tridentata) and percent vegetative cover is approximately

30%. Vegetation is generally uniform and the terrain is flat for at least 500 m upwind of

the tower, which includes the footprint of the tower (Hsieh et al. 2000).

The site is instrumented for continuous measurements of carbon, water and energy flux

using the eddy covariance method (Baldocchi 2003). Latent heat and carbon fluxes are

directly measured at the tower using an infrared gas analyzer (Licor 7500, Licor

Biosciences, Lincoln, NE) and a three-dimensional sonic anemometer (CSAT, Campbell

Scientific, Logan, UT). Soil moisture data are collected half-hourly using time-domain

reflectometers (CS616, Campbell Scientific) for both canopy and bare soil at five depths

(2.5, 12.5 22.5, 37.5, and 52.5 cm). These values are averaged across all depths to

produce one soil moisture profile for the site. Soil moisture data are later recombined into

shallow (0-20 cm) and deep (20- 60 cm) zones (see section 3.1). Tower flux data were

collected for three years (2002-2004) and include relatively wet (2004, 324 mm) and dry

(2003, 140 mm) years in terms of rainfall, as well as a year of typical rainfall amounts

(2002; see Kurc and Small 2007).

2.2. Remote Sensing Data and Processing

96

96

2.2.1. Vegetation indices

MODIS vegetation index data (MOD13Q1; (Huete et al. 2002)) were retrieved from the

Oak Ridge National Laboratory Data Archive and Acquisition Center (ORNL DAAC,

http://daac.ornl.gov/MODIS/modis.shtml) as 7 km x 7 km scenes. The vegetation index

scenes have a 250-m pixel grid with the central pixel containing the flux tower (see

section 3.1). Data files were downloaded in ASCII text format and processed in Matlab

(Mathworks, Inc.; Natick, MA).

We chose the vegetation index products over other MODIS products (e.g. leaf area index,

land surface temperature) because of its finer spatial resolution (250-m vs. 500-m for leaf

area index and 1-km for land surface temperature). EVI is less sensitive to soil and

atmospheric effects compared to other indices, namely the normalized difference

vegetation index (NDVI; (Huete et al. 2002). MODIS EVI products are available at 250

m resolution in 16-day composite records.

2.2.2. Temporal decomposition

MODIS data composites are designed to ensure that the final data products are of the

highest possible quality with respect to sensor view geometry (Huete et al. 2002).

Because the day of observation is retained in the MODIS Collection 5 composite data

products (Didan and Huete 2006), we “decompose” the images based on that information

to build a time series based on the day of observation and the availability of individual

pixel values.

97

97

By applying EVI on the date the data are recorded, we hoped to better represent the short-

term dynamics of semiarid systems. Decomposition by time stamping the observations

alters the overall variability in the EVI record. We better capture the finer variability in

carbon fluxes associated with short duration events using decomposition, by associating

the remote sensing data with fluxes measured on the same day.

2.2.3. Percent Cover Maps

High resolution (1-m) data from the Airborne Data Acquisition and Registration (ADAR)

5500 (Positive Systems, Inc.; see (Benkelman and Behrendt 1992) was used to generate

fine-scale vegetation cover information. Data were collected on August 23, 2001 in four

channels in the visible and near-infrared spectra (blue, green, red, near infrared).

Supervised classification of the area surrounding the tower was used to identify creosote

shrub cover from bare ground. After classification, the 1-m pixels were averaged up to

the 250-m resolution of MODIS pixels. A more in-depth discussion of the classification

process and results is given in Appendix A. Overall classification accuracy was 92.6%

and the kappa coefficient is 85.1%.

Because of the scale of this data and the generally low growth rates of large, mature

creosote shrubs (Allen et al. 2008; Medeiros and Pockman 2010) we assume the percent

cover mapping based on the ADAR image is representative of the percent cover for the

entire study period. These percent cover data are used in comparison with the MODIS

98

98

EVI data near the tower to determine the importance of increased shrub density on the

EVI-NEE relationship.

3. Approach

3.1. Flux Data Analysis

Tower flux data were collected at 10 Hz and aggregated into 30-minute records. Half-

hourly fluxes were then averaged to daily values for comparison to satellite-derived

vegetation indices. At the daily level, molar NEE was converted to mass flux [g m-2 d-1].

Soil moisture data were combined to produce values over two depths: 0-20cm and 20-60

cm. These two soil zones were then classified into “wet” and “dry” conditions. The

shallow soil zone was classified based on drydown curves. Drydown following all storm

events >8 mm were averaged and fit to an exponential model. The terminal soil moisture

value of the mean curve was used as the transition point between wet and dry conditions.

We defined the deep soil moisture threshold on transitions between positive NEE (> 0)

and negative NEE (< 0). Values of deep soil moisture on the first of five consecutive days

of positive NEE following at least five consecutive days of negative NEE were sampled.

Each instance of this transition was recorded, and a mean value of deep soil moisture

under the positive/negative NEE transition was used as the threshold. Five consecutive

days of positive NEE were used to ensure consistent net carbon flux behavior. The

transition indicates that photosynthetic activity is limited based on soil water stress. From

the wet/dry threshold values we can define four distinct soil moisture cases such that

99

99

Case 1 is a fully dry soil; Case 2 is wet in the shallow soil only; Case 3 is wet in the deep

soil only; and Case 4 is wet throughout the soil column.

3.2. MODIS-EVI and Tower-NEE Relationships

The link between vegetation greenness—identified by EVI—and NEE was evaluated

using the Pearson correlation coefficient. Correlations were calculated for each pixel

relative to the tower NEE flux over the three year record at the tower. We expect EVI and

NEE to be negatively correlated, i.e. increasing EVI is associated with decreasing

(negative) NEE, representing carbon uptake by vegetation. EVI-NEE relationships were

also considered for pixels within the tower footprint and from a larger 1.25-km (5 x 5

pixels) area around the tower.

Data also were analyzed in a temporal framework based on the soil moisture cases

described above. In addition, we examine the spatial patterns of correlation that emerge

based on the soil moisture cases. Linear regressions were performed on NEE as a

function of EVI to identify the best model to scale NEE based on the vegetation

characteristics in the tower footprint. Data from individual pixels were used in the

regression analysis, as well as the mean EVI value in the footprint and in the 1.25-km

tower area. Models were selected based on the coefficient of determination and

significance of the fit (only p < 0.05 were retained).

4. Results

4.1. EVI-NEE Relationships

100

100

EVI is not strongly correlated to NEE over much of the study area (Figure 1). Among the

significant correlations, only 15 of the 25 pixels within a 1.25-km grid around the tower

yield significant correlations (p < 0.05, r values range between -0.3 and -0.5). Strong

correlation to tower NEE mostly occurs far from the tower, including a region

approximately 1.5 km to the west, and a region to the southeast (Figure 1). Correlations

are low close to the tower and in the tower footprint, indicating potential effects of

landscape heterogeneity that is represented by the remotely sensed vegetation data.

The time series of NEE and EVI clarifies the temporal patterns associated with EVI-NEE

correlations (Figure 2). EVI roughly follows the trend for NEE during periods of net

carbon uptake (NEE < 0; e.g. October 2002, Figure 2). However, net respiration periods

exhibit poor correlation between NEE and EVI (e.g. 2003, Figure 2). Correlation between

increasing EVI and carbon uptake is clearest in the wet and normal years (2002, 2004)

but weak in the dry year (2003). Increasing EVI and net carbon uptake both follow the

timing of rainfall inputs. Moisture inputs from rainfall drive vegetation activity and net

uptake in the normal and wet years. That correlation is clearest in the wet/dry contrast

following precipitation events in the wet year (2004). In the three years of the study, all

three regions of interest—the tower footprint, 1.25-km tower area and highly correlated

pixels—express similar timing in EVI for the start and end of growing periods.

4.2. Deep Moisture Effects on EVI-NEE Relationships

The vertical distribution of soil moisture improves correlation between EVI and NEE

near the tower under certain conditions (Figure 3). In Cases 1 and 2, when the deep soil is

101

101

dry (Figures 3a, b), the correlation is weak across the study area (mean r = 0.02 and 0.16,

respectively; note that negative r values indicate stronger correlation, as noted above).

When there is moisture deep in the soil, in Cases 3 and 4 (Figures 3c, d), correlation

between EVI and NEE is improved in much of the study area (mean r = -0.40 and -0.38,

respectively). This is especially true in Case 4, when the entire soil column is wet, with

average correlation values near the tower of -0.28, compared to 0.10 in Case 1 (Figure 3,

inset). When the entire soil column is wet, the first MODIS pixel west of the tower,

which contains the largest contributing area to the tower, has a correlation coefficient of -

0.64. Higher correlations when deep soils are wet highlight the importance of soil

moisture in identifying the relationship between EVI and NEE.

The improved correlation (-0.64 in the tower footprint compared to -0.38 over the entire

data set), particularly in Case 4, is clear when it is used as the basis for generating direct

relationships between EVI and NEE (Figure 4). Cases 1 and 2, when deep soils are dry,

do not produce clear, significant relationships between EVI and NEE in the footprint or

in the 1.25-km tower region (Figure 4a, b). A slight trend toward increasing carbon

uptake (2 g m-2 d-1 at EVI = 0.15) at higher EVI exists but is not significant. Significant

correlations (p < 0.05) emerge in the footprint region during Case 4, when the entire soil

column is wet, (Figure 4d, Table 1), as well as in the 1.25-km tower region in Case 3

(Table 1).

Using a gross primary productivity (GPP) dataset for this tower (Kurc and Small 2007),

determined as the residual between NEE and modeled respiration, we can examine the

102

102

water limitation effect on vegetation (Figure 5). The tight link between EVI and

productivity when deep soil moisture is available is clear (Figure 5c, d, Table 2). The soil

moisture conditions that led to greater EVI-NEE correlation yield very strong predictive

relationships for GPP as well (r2 = 0.88 based on footprint EVI and 0.95 based on 1.25-

km tower region EVI). As in the EVI-NEE relationships, we also find a strong trend for

GPP when only the deep soils are wet (Figure 5c, r2 = 0.96 in the tower region).

Combining the two cases that represent deep soil moisture (Cases 3 and 4) we can build

models for NEE and GPP as a function of EVI that reflects the overall carbon response

(Figure 6). These trends provide predictive relationships for NEE and GPP (Table 3).

Deep soils are wet for approximately 26% of the data record in this study, thus this

relationship can be applied to scale tower fluxes for one-quarter of the three year span.

4.3. Spatial Analysis

The performance of EVI-NEE relationships is independent of the percent vegetation

cover in each pixel identified from the ADAR data (Figure 7). Percent cover in the tower

area ranged from 12% to 43%, with an average cover of 25%—similar to the 30%

reported by Kurc and Small (2007) for the study site. Over the range of percent cover

based on the spatially-aggregated ADAR data, EVI-NEE correlation coefficients do not

change significantly in the four soil moisture cases (Figure 7) or overall (not shown).

This insensitivity to vegetation cover is confirmed in the range of EVI values based on

percent cover, which are consistent over the entire data record as a function of percent

cover (not shown). Maximum and minimum EVI values are nearly unchanged as a

103

103

function of percent cover, with a mean minimum EVI of 0.095 and mean maximum EVI

of 0.170 across all percent woody cover values.

5. Discussion

This study assessed the spatial and temporal links between remotely sensed vegetation

activity and carbon flux measured at an eddy covariance tower. The analysis

demonstrates the limitations that exist when using vegetation indices to scale flux

measurements in semiarid shrublands. These limitations include sub-pixel vegetation

heterogeneity and the vertical soil moisture distribution of the region. Analyzing data

based on soil moisture improved correlation between vegetation indices and carbon flux

in the tower area, enabling spatial scaling of carbon flux based on MODIS EVI when the

deep soils are wet.

5.1. Soil Moisture Control on EVI-NEE Relationships

The soil moisture cases are a strong organizing factor in EVI-NEE and EVI-GPP

relationships in terms of their overall predictive power (Figures 4, 5; Tables 1, 2). When

deep soils are dry, the link between vegetation greenness and carbon flux is relatively

weak (Figures 4a, b). Only after moisture percolates to the deep soil is there a link

between EVI and carbon flux (Figures 4c, d, 5c, d). The need to represent moisture

dynamics in remote sensing-based models of carbon flux for semiarid systems has been

previously documented (Sjöström et al. 2011). However, other studies have linked carbon

flux to energy partitioning using the evaporative fraction (the ratio of latent heat flux to

available energy). In applying data regarding the vertical soil moisture profile, we can

104

104

quantify the response of vegetation to moisture directly available to roots (Fernandez

2007).

The link between deep soil moisture and carbon uptake is likely associated with

improved plant access to moisture below the shallow soil zone, where evaporative

demand quickly dries soil (Wythers et al. 1999). With moisture in the deeper soil, the

vegetation responds more strongly both in terms of leaf characteristics and photosynthetic

uptake. This pattern of greater carbon uptake when deep soils are wet is evident in

tower-derived GPP data (Figure 5). This strong correlation underscores the link between

moisture availability and vegetation activity in semiarid shrublands. When the moisture

limitation is relaxed, productivity is more directly related to vegetation greenness, and

productivity also dominates the net carbon flux (Figure 5).

In the cases where the entire soil is dry or only the shallow soil is wet, ecosystem carbon

flux is difficult to predict from EVI. This may be due to various physiological responses

to plant water stress (Cavanaugh et al. 2011; Ignace and Human 2009; Muldavin et al.

2008; Potts et al. 2008) as well as non-plant activity, e.g. soil microbial respiration

(Barron-Gafford et al. 2011). In these periods, carbon flux can either be positive (release)

or negative (uptake) without a clear link between EVI and NEE (Figures 3a, b). As a

result, the correlation between EVI and NEE remains low across much of the study area

during this period (Figure 3a, b). Further study of these periods in the context of

respiration models (Barron-Gafford et al. 2011) and vegetation water use efficiency

strategies (Franco 1994, Ogle 2002), may clarify the drivers of NEE.

105

105

5.2. Spatial Patterns in NEE-EVI Relationships

The spatial pattern in correlation between EVI and NEE indicates that correlation is

generally weak in regions that are close to the tower (Figure 1). This precludes the direct

application of MODIS EVI to spatially scale carbon flux for this semi-arid creosote

dominated system. Nevertheless, the study area can be sub-divided into a region that is

similar in vegetation cover to the tower area. Further, because the EVI-NEE relationship

is not sensitive to the percent vegetation cover, the single model for NEE and GPP based

on deep soil moisture can be used for the entire creosote region irrespective of vegetation

density (Figure 7).

While the seasonal and interannual dynamics of EVI (Figure 2) can be linked to increases

in both leaf area and vegetation cover (Glenn et al. 2008), the relatively low growth rates

for creosote (Medeiros and Pockman 2010) suggest that any changes in percent cover are

more likely associated with small, seasonal herbaceous plants (Muldavin et al. 2008; Xia

et al. 2010). This trend is particularly relevant for the creosote-dominated areas near the

tower, where the high-resolution ADAR data were used to describe percent cover. The

low growth rate of creosote suggests that any potential changes in percent cover from

year to year are likely from the expansion of herbaceous plants. At the same time, the

creosote growth rate and its canopy dominance indicates that greenness detected at the

250-m scale in MODIS EVI on a seasonal basis is predominantly associated with leaf

area.

106

106

Developing a more robust assessment for vegetation cover by plant functional type or

species represents an avenue for further research based on this study that would improve

the ability to generalize tower fluxes. The consistency of EVI under different cover

percentages near the tower suggests that, for creosote-dominated parts of the landscape,

the EVI-NEE relationships may be more closely related to leaf area, which has been

indicated by other studies using EVI (Xiao et al. 2004; Zhang et al. 2005). Identifying

more clearly the change in percent cover and type over time would allow the long-term

application of the EVI-NEE relationship developed here.

5.3. Temporal Patterns in EVI-NEE Relationships

We find the decomposed time series, where data were assigned to the day of observation,

allows for better resolution of the temporal relationships between EVI values at a given

pixel and the daily flux tower record. Because these semiarid ecosystems are highly

sensitive to short-term moisture inputs, the decomposed time series allows us to resolve

the finer temporal variability for individual pixels that produce viable, high quality data

on certain days. Relationships built on these individual pixels enable carbon flux models

that fully capture brief events, such as the wet shallow/dry deep conditions described in

soil moisture Case 2.

The timing of EVI increase and decrease is consistent across all pixels, though the

magnitude of EVI may differ. The best correlated pixels reach maximum EVI values

similar to the footprint, but also have lower EVI during dry periods (Figure 2). This

increased dynamic range improves correlation, although these pixels may not be

107

107

representative of the vegetation that is measured at the tower. Spatial and temporal

differences in vegetation composition and density are not explicitly analyzed here, but

provide a promising avenue to extend this work.

The carbon flux models developed here are dependent on deep soil moisture to clarify the

relationship between MODIS EVI and NEE (and GPP). From a temporal standpoint, the

application of those models is dependent on the frequency and intensity of rainfall that

yields sufficient deep soil moisture. Interannual variation in rainfall patterns, such as

between the relatively dry year 2003 and the wet year 2004, may reduce or increase the

frequency of deep soil moisture, subsequently reducing or increasing the applicability of

the model.

5.4. Application for Carbon Flux Modeling

The soil moisture cases indicate that the vertical distribution of available moisture is

critical to link EVI to NEE and GPP in shrublands (Figures 4-6). This is consistent with

other findings on plant water use in semiarid ecosystems (Muldavin et al. 2008;

Porporato et al. 2002; Weltzin and McPherson 1997) and underscores the complex

connection between vegetation greenness and carbon flux among shrublands.

The improved performance of EVI in predicting NEE under wet conditions in deep soil

suggests the limitations of simple light use efficiency-based models of primary

productivity on these landscapes. While these models may incorporate a water-limitation

component, it is often based on root zone soil moisture (Knyazikhin et al. 1999;

108

108

Mahadevan et al. 2008; Xiao et al. 2011). However, the link between plant greenup and

carbon uptake is not as straightforward in semiarid shrublands. Only when moisture

limitation has been satisfied in the deep soil do we see a clear trend between greenness

and NEE in much of the landscape (Figure 4d, 5d). At other times, water use restrictions

may lead respiration to dominate the NEE signal, and other models will be needed to

appropriately estimate carbon flux.

6. Conclusions

In this study we demonstrate the importance of soil moisture as a limiting factor when

using satellite remote sensing to spatially scale carbon flux data. Using a simple

conceptualization of the vertical soil moisture profile, we isolate deep soil moisture as the

key variable linking vegetation indices to net carbon flux. This finding reinforces the

importance of water availability as the key limitation to carbon flux in semiarid systems.

Because carbon flux is more sensitive to EVI when deep soils are wet, scaling NEE and

GPP is straightforward under those conditions.

By combining the temporal characteristics of deep soil moisture with a spatial analysis of

the vegetation distribution, we demonstrate the ability to resolve some of the

complexities associated with spatially-distributed carbon flux modeling in heterogeneous

semiarid environments. Among the complicating effects associated with modeling carbon

flux from EVI, we demonstrate methods to isolate moisture conditions from greenness

and resolve sub-grid variability in vegetation conditions for a semiarid shrubland. While

the availability of deep soil moisture is limited in time—approximately 25% of the time

109

109

series includes wet conditions in the deep soil—it provides a direct, mechanistic approach

to resolve the effect of water limitation on the light-use efficiency construct that underlies

EVI-NEE and EVI-GPP relationships.

Having identified the period when water is not limiting, further studies applying species-

specific water use efficiency information could improve carbon flux estimation when

deep soils are dry. For example, applying stomatal conductance or plant water potential

data for creosote in conjunction with flux tower measurements would connect the

landscape processes described here to the plant-level response.

Acknowledgements

K. Hartfield in the Arizona Remote Sensing Center assisted with the processing of

ADAR data. S. Archer, D. Bunting, Z. Guido, K. Nelson, and Z. Sanchez-Mejia

improved previous drafts of the manuscript. Ameriflux data were used under the fair-use

policy. Funding for collection of data at the Sevilleta Long Term Ecological Research site

was provided by the National Science Foundation Long Term Ecological Research

program.

Appendix A. Vegetation Cover Accuracy Assessment

We classify shrub and bare soil areas on the landscape with a supervised classification

scheme based on 1-m ADAR data. While line-intercept vegetation transect data are

available near the tower, they do not include the creosote-dominated region immediately

upwind of the tower, which is the focus of this study. As a result, the accuracy

110

110

assessment for the classification was performed using a subset of the training data, based

on user-identified pixels. While this approach may lead to “optimistic” accuracy values,

i.e. an overly favorable estimate of classification, the classification performed was

intended to provide a simple estimate of vegetation cover. Differentiation between types

of cover is relatively restricted, since the classification was focused on a region that is

known to be predominantly shrub cover. Performance statistics for the classification are

given in Table A1.

This classification was then be resampled to the MODIS EVI 250-m scale as percent

cover. While complications may arise associated with identification of members for

classification groups and edge effects, this analysis gives us a first-cut at relating carbon

fluxes to percent cover on the landscape. The percent cover in the tower area ranged from

12 to 43%, with an average cover of 25%--similar to the 30% reported by Kurc and Small

(2007) for the study site. More research is necessary to provide detailed vegetation

information along with spectral data to build spatially extensive, fine scale land cover

maps for the entire site.

111

111

References

Allen, A.P., Pockman, W.T., Restrepo, C., & Milne, B.T. (2008). Allometry, growth and

population regulation of the desert shrub Larrea tridentata. Functional Ecology, 22, 197-

204

Asrar, G., Myneni, R.B., & Choudhury, B.J. (1992). Spatial heterogeneity in vegetation

canopies and remote sensing of absorbed photosynthetically active radiation: A modeling

study. Remote Sensing of Environment, 41, 85-103

Badeck, F.-W., Bondeau, A., Böttcher, K., Doktor, D., Lucht, W., Schaber, J., & Sitch, S.

(2004). Responses of spring phenology to climate change. New Phytologist, 162, 295-309

Baldocchi, D.D. (2003). Assessing the eddy covariance technique for evaluating carbon

dioxide exchange rates of ecosystems: past, present and future. Global Change Biology,

9, 479-492

Barron-Gafford, G.A., Scott, R.L., Jenerette, G.D., & Huxman, T.E. (2011). The relative

controls of temperature, soil moisture, and plant functional group on soil CO2 efflux at

diel, seasonal, and annual scales. J. Geophys. Res., 116, G01023

Benkelman, C.A., & Behrendt, R.H. (1992). The Airborne Data Acquisition and

Registration (ADAR) System 5000: A CCD-Based System for High Re solution

Multispectral Remote sensing In I.S.f.P.a.R. Sensing (Ed.), Invited and Presented Papers

of the XVIIth Congress (pp. 172-176). Washington.

Cavanaugh, M.L., Kurc, S.A., & Scott, R.L. (2011). Evapotranspiration partitioning in

semiarid shrubland ecosystems: a two-site evaluation of soil moisture control on

transpiration. Ecohydrology, 4, 671-681

112

112

Didan, K., & Huete, A. (2006). MODIS vegetation index product series collection 5

change summary. In, MODIS VI C5 changes

Fensholt, R., & Sandholt, I. (2003). Derivation of a shortwave infrared water stress index

from MODIS near- and shortwave infrared data in a semiarid environment. Remote

Sensing of Environment, 87, 111-121

Fensholt, R., Sandholt, I., & Rasmussen, M.S. (2004). Evaluation of MODIS LAI,

fAPAR and the relation between fAPAR and NDVI in a semi-arid environment using in

situ measurements. Remote Sensing of Environment, 91, 490-507

Fernandez, R.J. (2007). On the frequent lack of response of plants to rainfall events in

arid areas. Journal of Arid Environments, 68, 688-691

Franco, A.C., Soyza, A.G., Virginia, R.A., Reynolds, J.F., & Whitford, W.G. (1994).

Effects of plant size and water relations on gas exchange and growth of the desert shrub

&lt;i&gt;Larrea tridentata&lt;/i&gt. Oecologia, 97, 171-178

Friedlingstein, P., Cox, P., Betts, R., Bopp, L., von Bloh, W., Brovkin, V., Cadule, P.,

Doney, S., Eby, M., Fung, I., Bala, G., John, J., Jones, C., Joos, F., Kato, T., Kawamiya,

M., Knorr, W., Lindsay, K., Matthews, H.D., Raddatz, T., Rayner, P., Reick, C.,

Roeckner, E., Schnitzler, K.G., Schnur, R., Strassmann, K., Weaver, A.J., Yoshikawa, C.,

& Zeng, N. (2006). Climate–Carbon Cycle Feedback Analysis: Results from the C4MIP

Model Intercomparison. Journal of Climate, 19, 3337-3353

Gao, X., Huete, A.R., Ni, W., & Miura, T. (2000). Optical–Biophysical Relationships of

Vegetation Spectra without Background Contamination. Remote Sensing of Environment,

74, 609-620

113

113

Gilmanov, T.G., Tieszen, L.L., Wylie, B.K., Flanagan, L.B., Frank, A.B., Haferkamp,

M.R., Meyers, T.P., & Morgan, J.A. (2005). Integration of CO2 flux and remotely-sensed

data for primary production and ecosystem respiration analyses in the Northern Great

Plains: potential for quantitative spatial extrapolation. Global Ecology and Biogeography,

14, 271-292

Glenn, E.P., Huete, A.R., Nagler, P.L., & Nelson, S.G. (2008). Relationship between

remotely-sensed vegetation indices, canopy attributes and plant physiological processes:

What vegetation indices can and cannot tell us about the landscape. Sensors, 8, 2136-

2160

Goerner, A., Reichstein, M., Tomelleri, E., Hanan, N., Rambal, S., Papale, D., Dragoni,

D., & Schmullius, C. (2011). Remote sensing of ecosystem light use efficiency with

MODIS-based PRI. Biogeosciences, 8, 189-202

Hamerlynck, E.P., & Huxman, T.E. (2009). Ecophysiology of two Sonoran Desert

evergreen shrubs during extreme drought. Journal of Arid Environments, 73, 582-585

Hsieh, C.-I., Katul, G., & Chi, T.-w. (2000). An approximate analytical model for

footprint estimation of scalar fluxes in thermally stratified atmospheric flows. Advances

in Water Resources, 23, 765-772

Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., & Ferreira, L.G. (2002).

Overview of the radiometric and biophysical performance of the MODIS vegetation

indices. Remote Sensing of Environment, 83, 195-213

Huxman, T.E., Snyder, K.A., Tissue, D., Leffler, A.J., Ogle, K., Pockman, W.T.,

Sandquist, D.R., Potts, D.L., Schwinning, S., & Yu (2004). Precipitation pulses and

carbon fluxes in semiarid and arid ecosystems. Oecologia, 141, 254-268

114

114

Ignace, D.D., & Human, T.E. (2009). Limitations to photosynthetic function across

season in Larrea tridentata (creosotebush) growing on contrasting soil surfaces in the

Sonoran Desert. Journal of Arid Environments, 73, 626-633

IPCC (2007). Climate Change 2007: The Physical Science Basis. Cambridge: Cambridge

University Press

Jolly, W.M., Nemani, R., & Running, S.W. (2005). A generalized, bioclimatic index to

predict foliar phenology in response to climate. Global Change Biology, 11, 619-632

King, D.A., Turner, D.P., & Ritts, W.D. (2011). Parameterization of a diagnostic carbon

cycle model for continental scale application. Remote Sensing of Environment, 115,

1653-1664

Knyazikhin, Y., Glassy, J., Privette, J.L., Tian, Y., Lotsch, A., Zhang, Y., Wang, Y.,

Morisette, J.T., Votava, P., Myneni, R.B., Nemani, R.R., & Running, S.W. (1999).

MODIS Leaf Area Index (LAI) and Fraction of Photosynthetically Active

RadiationAbsorbed by Vegetation (FPAR) Product (MOD15) Algorithm Theoretical

Basis Document.

Kurc, S.A., & Benton, L.M. (2010). Digital image-derived greenness links deep soil

moisture to carbon uptake in a creosotebush-dominated shrubland. Journal of Arid

Environments

Kurc, S.A., & Small, E.E. (2007). Soil moisture variations and ecosystem-scale fluxes of

water and carbon in semiarid grassland and shrubland. Water Resources Research, 43

Lajtha, K., & Whitford, W.G. (1989). The effect of water and nitrogen amendments on

photosynthesis, leaf demography, and resource-use efficiency in Larrea tridentata, a

desert evergreen shrub. Oecologia, 80, 341-348

115

115

Mahadevan, P., Wofsy, S.C., Matross, D.M., Xiao, X., Dunn, A.L., Lin, J.C., Gerbig, C.,

Munger, J.W., Chow, V.Y., & Gottlieb, E.W. (2008). A satellite-based biosphere

parameterization for net ecosystem CO2 exchange: Vegetation Photosynthesis and

Respiration Model (VPRM). Global Biogeochem. Cycles, 22, GB2005

Medeiros, J.S., & Pockman, W.T. (2010). Carbon gain and hydraulic limits on water use

differ between size classes of Larrea tridentata. Journal of Arid Environments, 74, 1121-

1129

Muldavin, E.H., Moore, D.I., Collins, S.L., Wetherill, K.R., & Lightfoot, D.C. (2008).

Aboveground net primary production dynamics in a northern Chihuahuan Desert

ecosystem. Oecologia, 155, 123-132

Neal, A.L. (2012). Soil moisture controls on spatial and temporal patterns of carbon

dioxide fluxes in drylands. In, Hydrology and Water Resources: University of Arizona

Notaro, M., Liu, Z., Gallimore, R.G., Williams, J.W., Gutzler, D.S., & Collins, S. (2010).

Complex seasonal cycle of ecohydrology in the Southwest United States. J. Geophys.

Res., 115, G04034

Noy-Meir, I. (1973). Desert Ecosystems: Environment and Producers. Annual Review of

Ecology and Systematics, 4, 25-51

Ogle, K., & Reynolds, J.F. (2002). Desert dogma revisited: coupling of stomatal

conductance and photosynthesis in the desert shrub, Larrea tridentata. Plant, Cell &

Environment, 25, 909-921

Porporato, A., D'Odorico, P., Laio, F., Ridolfi, L., & Rodriguez-Iturbe, I. (2002).

Ecohydrology of water-controlled ecosystems. Advances in Water Resources, 25, 1335-

1348

116

116

Potts, D.L., Scott, R.L., Cable, J.M., Huxman, T.E., & Williams, D.G. (2008). Sensitivity

of Mesquite Shrubland CO2 Exchange to Precipitation in Contrasting Landscape

Settings. Ecology, 89, 2900-2910

Sala, O.E., & Lauenroth, W.K. (1982). Small Rainfall Events: An Ecological Role in

Semi-Arid Regions. Oecologia, 53, 301-304

Scott, R.L., Huxman, T.E., Cable, W.L., & Emmerich, W.E. (2006). Partitioning of

evapotranspiration and its relation to carbon dioxide exchange in a Chihuahuan Desert

shrubland. Hydrological Processes, 20, 3227-3243

Sims, D.A., Luo, H., Hastings, S., Oechel, W.C., Rahman, A.F., & Gamon, J.A. (2006a).

Parallel adjustments in vegetation greenness and ecosystem CO2 exchange in response to

drought in a Southern California chaparral ecosystem. Remote Sensing of Environment,

103, 289-303

Sims, D.A., Rahman, A.F., Cordova, V.D., El-Masri, B.Z., Baldocchi, D.D., Bolstad,

P.V., Flanagan, L.B., Goldstein, A.H., Hollinger, D.Y., Misson, L., Monson, R.K.,

Oechel, W.C., Schmid, H.P., Wofsy, S.C., & Xu, L. (2008). A new model of gross

primary productivity for North American ecosystems based solely on the enhanced

vegetation index and land surface temperature from MODIS. Remote Sensing of

Environment, 112, 1633-1646

Sims, D.A., Rahman, A.F., Cordova, V.D., El-Masri, B.Z., Baldocchi, D.D., Flanagan,

L.B., Goldstein, A.H., Hollinger, D.Y., Misson, L., Monson, R.K., Oechel, W.C.,

Schmid, H.P., Wofsy, S.C., & Xu, L. (2006b). On the use of MODIS EVI to assess gross

primary productivity of North American ecosystems. J. Geophys. Res., 111, G04015

117

117

Sjöström, M., Ardö, J., Arneth, A., Boulain, N., Cappelaere, B., Eklundh, L., de

Grandcourt, A., Kutsch, W.L., Merbold, L., Nouvellon, Y., Scholes, R.J., Schubert, P.,

Seaquist, J., & Veenendaal, E.M. (2011). Exploring the potential of MODIS EVI for

modeling gross primary production across African ecosystems. Remote Sensing of

Environment, 115, 1081-1089

Sjöström, M., Ardö, J., Eklundh, L., El-Tahir, B.A., El-Khidir, H.A.M., Hellström, M.,

Pilesjö, P., & Seaquist, J. (2009). Evaluation of satellite based indices for gross primary

production estimates in a sparse savanna in the Sudan. Biogeosciences, 6, 129-138

Smith, M.O., Ustin, S.L., Adams, J.B., & Gillespie, A.R. (1990). Vegetation in deserts: I.

A regional measure of abundance from multispectral images. Remote Sensing of

Environment, 31, 1-26

Tang, X., Liu, D., Song, K., Munger, J.W., Zhang, B., & Wang, Z. (2011). A new model

of net ecosystem carbon exchange for the deciduous-dominated forest by integrating

MODIS and flux data. Ecological Engineering, In Press, Corrected Proof

Turner, D.P., Ritts, W.D., Cohen, W.B., Maeirsperger, T.K., Gower, S.T., Kirschbaum,

A.A., Running, S.W., Zhao, M.S., Wofsy, S.C., Dunn, A.L., Law, B.E., Campbell, J.L.,

Oechel, W.C., Kwon, H.J., Meyers, T.P., Small, E.E., Kurc, S.A., & Gamon, J.A. (2005).

Site-level evaluation of satellite-based global terrestrial gross primary production and net

primary production monitoring. Global Change Biology, 11, 666-684

Weltzin, J.F., & McPherson, G.R. (1997). Spatial and temporal soil moisture resource

partitioning by trees and grasses in a temperate savanna, Arizona, USA. Oecologia, 112,

156-164

118

118

Wu, C., Munger, J.W., Niu, Z., & Kuang, D. (2010). Comparison of multiple models for

estimating gross primary production using MODIS and eddy covariance data in Harvard

Forest. Remote Sensing of Environment, 114, 2925-2939

Wylie, B.K., Johnson, D.A., Laca, E., Saliendra, N.Z., Gilmanov, T.G., Reed, B.C.,

Tieszen, L.L., & Worstell, B.B. (2003). Calibration of remotely sensed, coarse resolution

NDVI to CO2 fluxes in a sagebrush-steppe ecosystem. Remote Sensing of Environment,

85, 243-255

Wythers, K.R., Lauenroth, W.K., & Paruelo, J.M. (1999). Bare-soil evaporation under

semiarid field conditions. Soil Science Society of America Journal, 63, 1341-1349

Xia, Y., Moore, D.I., Collins, S.L., & Muldavin, E.H. (2010). Aboveground production

and species richness of annuals in Chihuahuan Desert grassland and shrubland plant

communities. Journal of Arid Environments, 74, 378-385

Xiao, J., Davis, K.J., Urban, N.M., Keller, K., & Saliendra, N.Z. (2011). Upscaling

carbon fluxes from towers to the regional scale: Influence of parameter variability and

land cover representation on regional flux estimates. J. Geophys. Res., 116, G03027

Xiao, J., & Moody, A. (2005). A comparison of methods for estimating fractional green

vegetation cover within a desert-to-upland transition zone in central New Mexico, USA.

Remote Sensing of Environment, 98, 237-250

Xiao, J.F., Zhuang, Q.L., Law, B.E., Chen, J.Q., Baldocchi, D.D., Cook, D.R., Oren, R.,

Richardson, A.D., Wharton, S., Ma, S.Y., Martini, T.A., Verma, S.B., Suyker, A.E.,

Scott, R.L., Monson, R.K., Litvak, M., Hollinger, D.Y., Sun, G., Davis, K.J., Bolstad,

P.V., Burns, S.P., Curtis, P.S., Drake, B.G., Falk, M., Fischer, M.L., Foster, D.R., Gu,

L.H., Hadley, J.L., Katul, G.G., Roser, Y., McNulty, S., Meyers, T.P., Munger, J.W.,

119

119

Noormets, A., Oechel, W.C., Paw, K.T., Schmid, H.P., Starr, G., Torn, M.S., & Wofsy,

S.C. (2010). A continuous measure of gross primary production for the conterminous

United States derived from MODIS and AmeriFlux data. Remote Sensing of

Environment, 114, 576-591

Xiao, X., Hollinger, D., Aber, J., Goltz, M., Davidson, E.A., Zhang, Q., & Moore Iii, B.

(2004). Satellite-based modeling of gross primary production in an evergreen needleleaf

forest. Remote Sensing of Environment, 89, 519-534

Zhang, L., Wylie, B.K., Ji, L., Gilmanov, T.G., Tieszen, L.L., & Howard, D.M. (2011).

Upscaling carbon fluxes over the Great Plains grasslands: Sinks and sources. J. Geophys.

Res., 116, G00J03

Zhang, Q., Xiao, X., Braswell, B., Linder, E., Baret, F., & Moore Iii, B. (2005).

Estimating light absorption by chlorophyll, leaf and canopy in a deciduous broadleaf

forest using MODIS data and a radiative transfer model. Remote Sensing of Environment,

99, 357-371

120

120

Footprint Tower Area β0 β 1 r2 β0 β 1 r2 Case 1 best pixel 0.39 14.02 0.00 -11.24 128.75 0.37 mean 2.70 -8.10 0.00 -0.23 17.61 0.01 Case 2 best pixel -3.29 29.27 0.06 -9.32 97.86 0.54 mean -4.29 36.87 0.05 -6.16 62.28 0.13 Case 3 best pixel -36.93 273.37 0.23 67.73 -503.33 0.78 mean -2.21 20.22 0.01 5.57 -42.12 0.13 Case 4 best pixel 5.85 -53.97 0.41 3.23 -33.17 0.49 mean 1.62 -22.49 0.11 0.47 -11.46 0.01

Table 1. Linear regression coefficients and coefficient of variation for regressions of

NEE on EVI. Regressions were based on pixels from the two pixels in the tower

footprint and the 5 x 5 pixel area around the tower. Best pixel coefficients represent

the pixel that has the best fit for each soil moisture case. Mean values are based on the

average EVI for the footprint or tower area. Bold indicates regressions are significant

(p < 0.05).

121

121

Footprint Tower Area β0 β1 r2 β0 β1 r2 Case1 best pixel -1.15 12.10 0.15 -5.74 56.73 0.35 Mean -1.30 13.58 0.14 -0.53 7.37 0.02 Case2 best pixel -1.73 17.68 0.18 -3.76 32.93 0.50 Mean -0.18 5.97 0.01 0.23 1.67 0.00 Case3 best pixel -4.63 39.92 0.64 -5.31 45.35 0.96 Mean -2.97 26.76 0.25 -3.44 30.64 0.39 Case4 best pixel -4.30 43.00 0.88 -3.16 34.86 0.95 Mean -4.18 42.82 0.85 -3.03 32.72 0.54

Table 2. Linear regression coefficients and coefficient of variation for regressions of

GPP on EVI. Best pixel coefficients represent the pixel that has the best fit for each

soil moisture case. Mean values are based on the average EVI for the footprint or

tower area. Bold indicates regressions that are significant (p < 0.05).

122

122

Footprint Tower Area β0 β 1 r2 β0 β1 r2 NEE best pixel 6.30 -53.73 0.25 6.73 -56.64 0.37 Mean 2.02 -21.00 0.04 1.77 -19.63 0.07 GPP best pixel -6.21 55.91 0.60 -5.95 53.05 0.78 Mean -6.25 54.94 0.63 -4.42 40.47 0.54

Table 3. Regression coefficients and coefficients of variation for NEE and GPP

models based on the presence of deep soil moisture. The “best pixel” model for NEE

and GPP is identified as the best model based on the highest value of the coefficient

of variation.

123

123

Number in Class

Training Reference Accuracy Number of Samples Vegetation Bare

Vegetation 95.8 2038 1952 86

Bare 93.8 2142 133 2009

Total 4180 2085 2095

Reliability 93.6 95.9

Number in Class

Test Reference Accuracy Number of Samples Vegetation Bare

Vegetation 93.3 705 658 47

Bare 91.8 705 58 647

Total 1410 716 694

Reliability 91.9 93.2

Overall training performance = 94.8%. Training κ = 89.5%. Overall test performance = 92.6%. Test κ = 85.1%. Table A1. Accuracy assessment statistics for the training and test periods of the supervised classification used to define vegetation cover.

124

124

Figure 1. Spatial plot of correlation coefficient between tower NEE and MODIS EVI.

Negative values indicate increasing correlation. Pixels that had correlations that were

not significant at p < 0.05 are set to zero.

125

125

Figure 2. Time series of NEE (gray bars, left axis), precipitation (black bars, right inset axis) and EVI (right axis). Average values

of EVI are shown for the tower footprint (open circles) and the 1.25-km region around the tower (points). Both regions show a

general trend of increasing EVI slightly before the start of carbon uptake (NEE < 0). Missing EVI data in 2003 are due to quality

control issues on the MODIS sensor.

126

Figure 3. Correlation coefficients between tower NEE and MODIS EVI based on soil

moisture cases 1 (a), 2 (b), 3 (c) and 4 (d). Inset in Figure 4d shows the values for the

1.25-km region surrounding the tower, which are greater in Case 4 than in other cases.

(a) (b)

(c) (d)

127

Figure 4. NEE as a function of EVI for soil moisture cases 1 (a), 2 (b), 3 (c), and 4 (d).

Points represent the EVI from the pixel with the best fit (see Table 1) plotted with NEE

measured at the tower. Gray bars indicate the standard deviation of EVI for all pixels in

the 5 x 5 pixel area around the tower. Dashed lines show the linear regression of NEE as

a function of EVI, with reported coefficient of determination and p-value. The positive

trend in NEE in the upper figures and negative trend in the lower figures is clear on the

individual pixel level.

128

Figure 5. Gross primary productivity as a function of EVI for soil moisture cases 1 (a), 2

(b), 3 (c), and 4 (d). Trends in GPP as a function of EVI are clearer for an individual pixel

rather than based on the mean trend (see Table 2), with a slight decrease in GPP with EVI

when deep soils are dry and increasing GPP with EVI when deep soils are wet.

129

Figure 6. NEE (top) and GPP (bottom) as a function of tower area EVI. Circles are when deep soils are dry, triangles when deep soils are wet. Linear fits are based on the wet deep soil conditions. Data points are for the individual pixel that has the highest coefficient of determination (see Table 3).

130

Figure 7. Correlation between EVI and NEE as a function of percent cover in each 250-m

pixel, under each of the four moisture conditions (arranged as in Figure 3). The link

between EVI and NEE shows little variation based on percent cover irrespective of soil

moisture conditions.

131

APPENDIX C: LATERAL HYDROLOGIC REDISTRIBUTION SUBSIDIZES

CARBON FLUX IN SEMIARID LANDSCAPES

Andrew L. Neala, Paul D. Brooksa, Shirley A. Kurca,b

aDepartment of Hydrology and Water Resources, University of Arizona

1133 James E. Rogers Way, Tucson, AZ, USA.

2School of Natural Resources and Environment, University of Arizona,

Biological Sciences East, Tucson, AZ, USA

for submission to Advances in Water Resources

Abstract

Lateral hydrologic redistribution controls the spatial availability of resources for

ecosystem productivity, but generally is not considered in models of carbon flux. By

linking simple models of lateral hydrologic redistribution and ecosystem carbon

dynamics, we explicitly quantified the effect of spatially variable moisture on carbon

flux. Simulations were performed on nested subwatersheds at Walnut Gulch

Experimental Watershed in southern Arizona, USA. Lateral hydrologic redistribution

approximately doubled plant available water near the watershed outlet relative to

precipitation alone. Net carbon uptake was nearly eight times higher at high topographic

index (TI), i.e. near the outlet, compared to low TI. Integrated across catchments,

hydrologic redistribution resulted in 25-120% higher carbon uptake relative to a model

that did not consider redistribution. Precipitation characteristics also influenced carbon

uptake via runoff and rainfall magnitude was more important than interstorm interval on

132

carbon. These results indicate that spatial heterogeneity in moisture availability controls

carbon dynamics based on TI.

1. Introduction

Terrestrial carbon dioxide fluxes are an important component of global biogeochemical

cycles [1, 2] but responses to climate, particularly changes in the amount, timing, and

duration of moisture available for vegetation, are unclear [3-5]. The spatial and temporal

availability of plant available water is controlled by the timing, duration, and intensity of

rainfall and infiltration [6, 7] which have received significant attention over the last

decade [7, 8]. Topographically driven redistribution of runoff also affects plant available

water, but largely has been studied from the standpoint of understanding flowpaths and

residence times of streamflow [9, 10], with less attention paid to how this affects

moisture availability to vegetation [11, 12]. Observations of available moisture at the

watershed level indicate that vegetation optimizes water use across climate regimes [13-

16], suggesting that catchments offer a natural study domain for quantifying

ecohydrological interactions. Refining these relationships between rainfall, available

moisture, and carbon dioxide flux requires coupling hydrologic and ecological processes

within a watershed, where moisture availability is dynamic in space and time [17].

Semiarid ecosystems are particularly sensitive to climate [18], with carbon flux generally

limited to brief periods following rainfall, often described as “pulses” [19-21]. This

sensitivity has led to insights into how semiarid vegetation is distributed on the landscape

133

[22, 23] and how the ecosystem responds to moisture dynamically across scales [24, 25].

The explicit representation of lateral moisture redistribution following intense rainfall

may indicate a different carbon flux response at the catchment scale than predicted

without incorporating redistribution [26].

Lateral redistribution of moisture is largely controlled by the topography and connectivity

of a catchment and supplements precipitation-derived moisture at points lower along a

flowpath [27, 28], particularly in semiarid systems [29, 30]. As a result of this moisture

redistribution, the topography of a catchment is linked to hydrologic partitioning [31],

respiration [32] and reduced plant water stress [33]. Therefore, moisture and net carbon

flux are associated with geomorphic characteristics, such as topography. Moisture

redistribution also promotes feedback loops for vegetation [28, 34-37], where vegetation

structure enhances nutrient and moisture availability below the canopy [35, 38-40].

Concentration of resources can, in turn, support enhanced uptake and release of carbon

[41-43]. Landscape connectivity joins patches to hillslope and catchment scales [44, 45],

and provides the mechanism to concentrate resources along flowpaths [28, 46, 47]. Thus,

spatial redistribution of moisture alters resource availability within a watershed.

While previous research has shown greater rates of carbon dioxide uptake in areas that

concentrate soil moisture from runoff [23, 26] and reduced rates of carbon dioxide uptake

in areas that generate runoff [27], most models of carbon flux in semiarid systems do not

consider moisture redistribution explicitly at the catchment scale [48, 49]. This is based

134

on the assumption that runoff infiltrates along relatively short distances [50, 51].

However, intense storms characteristic of monsoon precipitation dynamics in semiarid

systems often produce runoff due to induced soil anisotropy, increasing overland flow

over long distances [28, 52]. As a result, runoff-based moisture redistribution is likely to

be more significant during the wet season. In order to better understand carbon dynamics

in semiarid ecosystems, lateral moisture redistribution and carbon flux should be linked

in a simple model framework that captures runoff behavior at appropriate spatial scales.

The objective of this paper is to quantify spatially explicit carbon fluxes resulting from

moisture redistribution. Here, we combine a topographically-driven model of moisture

routing with a simple model of vegetation activity to quantify the spatial distribution of

carbon flux. We use a distributed runoff approach that routes moisture based on a 30-m

grid [53] with a simple runoff process. We hypothesize that explicit accounting of

moisture redistribution will indicate spatial heterogeneity in vegetation-driven carbon

dynamics currently underrepresented in most carbon flux models. We determine the

influence of moisture redistribution on ecosystem carbon fluxes using a simple model of

carbon flux response to moisture inputs [54]. We discuss these results in the context of

topographic structure and interannual precipitation variability.

2. Methods

2.1. Site Description and Data

135

In this study, we used digital elevation data and daily rainfall and runoff data from the

Walnut Gulch Experimental Watershed (WGEW) in southern Arizona. Data span six

years (1999-2004) representing a range of wet and dry conditions (data archived at

www.tucson.ars.ag.gov/dap). WGEW is located at the ecotone between the Sonoran and

Chihuahuan deserts and is characterized by a mix of desert shrub and grass species [55].

Mean annual temperature is 17°C [56] while annual precipitation is approximately of 350

mm [57] and falls primarily during the North American Monsoon from July through

September [58]. Annual discharge from the outlet of WGEW typically is <1% of

precipitation, indicating that most of the precipitation is lost to evapotranspiration [59].

Runoff at WGEW is generally infiltration-excess as opposed to saturation-excess [60].

We focused this study on subwatersheds 102, 104 and 106 in the Lucky Hills shrubland

at WGEW (Figure 1, inset), because the subwatersheds are instrumented with runoff

flumes, are nested, and their spatial scale ensures minimal effect of spatial heterogeneity

of rainfall and runoff generation [29]. Lateral redistribution of moisture was determined

using runoff data measured at flumes in the WGEW subwatersheds [59].

2.2. Hydrologic Redistribution

We quantified catchment-scale hydrologic redistribution and available water using a

distributed modeling approach. Data from the flumes at the outlet of each subwatershed

were used to quantify the runoff ratio (runoff over precipitation) from individual rainfall

events. Because of the relatively small spatial extent of the three subwatersheds, we were

able to use precipitation data from a rain gauge in subwatershed 106 to represent the

136

entire watershed. Additionally, because runoff is primarily infiltration-excess in our study

area, runoff was calculated as a fraction of current rainfall above a minimum threshold

and was routed through the watershed rapidly as a function of topography [25],

Topographic characteristics of the watershed were developed based on the topographic

index (TI). TI characterizes the relative balance between runoff lost from any point on the

landscape relative to the flow into that point from upslope [61]. TI was derived for our

subwatersheds from a 30-m digital elevation model (DEM) as:

⎟⎟⎠

⎞⎜⎜⎝

⎛=

βtanln ATI (1)

where A is the area upslope of a pixel and β is the slope at that pixel using the Spatial

Analysis tools in ArcGIS (v. 9.3, ESRI, Inc., Redlands, CA).

A general runoff scheme [e.g. 62] was applied to simulate runoff for each 30-m grid cell.

The water balance for each cell is based on the following equation:

( )UQPkQ += (2)

where Q is the runoff from a given grid cell (mm), P is daily precipitation (mm), QU is

the runoff from all contributing cells (mm) and k is the runoff ratio as a function of TI.

Values of k were identified empirically based on runoff ratios (Q/P, determined from

daily Q data for each subwatershed and P from the raingauge in the watershed) for all

subwatersheds at Lucky Hills. The scale-dependence of runoff volumes was satisfied by

linking k to TI [63-65], similar to a scaled runoff coefficient approach used in other

137

studies [e.g. 66]. Runoff was routed along flowpaths determined from the DEM using the

flow direction tool in ArcGIS (v. 9.3, ESRI, Inc., Redlands, CA). Runoff model

performance was measured against flume data at the outlet of the three subwatersheds

using the root mean squared error (RMSE). Low values of RMSE indicate good estimates

of runoff over the whole watershed.

In our model, the effective precipitation (Peff) is defined as the depth of moisture

remaining on a given 30-m cell after runoff is removed, i.e.:

QQPP Ueff −+= (3).

Peff is used as the input for carbon flux modeling with runoff, i.e. Peff replaces P in

equations 4 and 5. To compare results of the runoff model to measured water balance, the

available moisture (P-Q) and fractional available moisture, (P-Q)/P, were determined for

the three subwatersheds and compared to Peff at the outlet of each subwatershed.

Variability in Peff over the entire watershed was confirmed to be conservative in each year

of simulation, with area-weighted ΣPeff/P ≈ 1.

2.3. Carbon Flux

To quantify the importance of moisture redistribution in the dynamics of carbon dioxide

flux, we developed a simple model of ecosystem response to rainfall inputs. This model

is based on a conceptual relationship between the magnitude of a rainfall event and the

duration of ecosystem (vegetation and microbial) activity that follows [67]. The duration

of microbial and other heterotrophic respiration (Rh) is calculated as:

138

Dh = min 0.2P , 2( ) (4)

where Dh is the duration of Rh in days. Similarly, the duration of carbon uptake by

photosynthesis (Da, in days), termed gross primary productivity (GPP), and autotrophic

respiration (Ra) are determined as:

Da =min(0.17P + 0.14, 7)

0

⎧⎨⎪

⎩⎪ P ≥ 5

P < 5. (5)

Equations 4 and 5 were used to simulate ecosystem activity on the 30-m grid under no

runoff conditions (as above) as well as with runoff included (i.e. Peff instead of P).

Carbon flux values for shrub functional types [54] were used to calculate GPP, Ra and Rh

under “relaxed” and “active” states. Active states occur when Da and Dh are greater than

zero; otherwise GPP, Ra and Rh are at their relaxed values. Our model focuses on the

photosynthesis and respiration responses to growing-season rainfall inputs only when the

canopy is assumed to be fully developed (July-September) and does not incorporate the

growth of new leaf area at the start of the season [54]. Using this assumption, we expect

to capture the majority of carbon flux for these semiarid systems which receive most of

their rainfall during the growing season (July-September) [67]. Estimates of GPP, Ra and

Rh were combined into the net ecosystem exchange of carbon dioxide (NEE = -GPP + Ra

+ Rh) based on the sum of each of the three fluxes for the growing season.

The model assumes that vegetation is uniformly distributed and static. Therefore all

points in the watershed have the same carbon flux magnitudes (i.e. the same values of

139

GPP, Ra and Rh) under relaxed and active states and the flux values do not vary between

years. Any two points can differ in relaxed or active states based on the variability in Peff

(equation 3). Each point under a given relaxed or active state will have the same flux rate,

but the number of days that any point is active (Da > 0 and/or Dh > 0) can vary. The

model treats each year independently rather than serially based on vegetation dynamics.

It simulates the expected gains and losses in carbon on the basis of moisture

redistribution but does not apply those changes as vegetation growth or soil carbon loss.

Whether carbon flux results in changes to leaf area, patch density and/or soil carbon is

outside the scope of this model.

3. Results

3.1. Moisture Redistribution

Growing season precipitation ranged between 159 mm and 274 mm from 1999-2003

(Table 1), the period during which runoff data were consistently available for all three

subwatersheds. Catchment water balance for the three subwatersheds indicates that much

of the water that enters the subwatersheds during the growing season via precipitation is

retained, i.e. not lost to Q. Runoff ranged from 1 mm to 48 mm, generally with higher

runoff depths in the upper subwatersheds 102 and 106, compared to the lower

subwatershed 104. Fractional available moisture, (P-Q)/P, ranged between 0.81 and 0.99

in subwatershed 104. The distributed runoff model operating on a per-pixel basis

generated flow estimates at the outlet of the three subwatersheds comparable to measured

Q (Table 2). Root mean squared error (RMSE) for runoff in subwatershed 104 was 0.40

140

mm, but smaller subwatersheds had greater error (RMSE = 0.46 mm in subwatershed 102

and RMSE = 0.71 mm in subwatershed 106, Table 2). Mean errors relative to measured

flow were 14%, 7% and 6% for subwatersheds 104, 102 and 106, respectively, indicating

this simple model captures topographically driven moisture redistribution.

Areas of the watershed with TI values less than 7.5 experienced a deficit of available

water relative to P (Peff < 1.0), while areas with TI values larger than 7.5 exhibited next

subsidy of water with upslope (Figure 2). Areas with the highest TI values had Peff

approximately double the seasonal P demonstrating the role of moisture redistribution

from low TI to high TI along flowpaths concentrating moisture in downslope locations.

Overall, the majority of the catchments (~43%) were net sources of runoff with Peff <1.0,

approximately 41% exhibited Peff near 1.0 and relatively small parts of the watershed

(15%) at high TI received a greater amount of moisture than they would directly from

rainfall.

3.2. Carbon Flux Modeling

The duration of autotrophic activity increased with increasing TI (Figure 3) from a low of

approximately 25 days to a maximum of 75 days. A logistic fit best represents this trend

(Figure 3), because Da has a lower limit based on the minimum Peff in the catchment and

a maximum based on the accumulation of flow near the watershed outlet and the duration

of the summer growing season (100 days).

141

Similarly, NEE was more negative, i.e. carbon uptake was greater, with increasing TI.

The range of NEE values span positive and negative values of NEE up to TI = 8.5, above

which NEE was always negative within the range of annual precipitation values (83-

274mm). NEE appears to follow a logistic shape, increasing toward a limit with

increasing TI, which is expected given its relationship to Da Individual logistic curves

for each year (Figure 5) vary in their upper and lower bounds and the inflection points

shift from TI = 9.3 to TI = 11.2 between wet (year 1999) and dry (2004) conditions.

Interestingly, interannual variability in NEE is lowest at TI = 6.3 (72 g m-2 season-1) but

was highest at TI = 10.3 (242 g m-2 season-1, Figure 5).

NEE in the driest year varied between 20 and 29 g m-2 season-1 among the subwatersheds.

However, NEE under dry ecosystem conditions was 10 g m-2 season-1, suggesting that

small rainfall amounts stimulate net carbon loss. Interannual range of NEE varied

considerably as a function of TI, increasing from 72 g m-2 season-1 at TI = 6.3 to 222 g m-

2 season-1 at TI = 10.3 (Figure 7). Although TI is linked to the runoff/runon

characteristics at each pixel, flow connectivity can also be used to describe the effect of

moisture redistribution on carbon flux. As more pixels contribute runoff to areas of flow

convergence, Peff increases. This is reflected in the increase in carbon uptake as a function

of relative connectivity (Figure 7). Runoff connectivity also increases the variability in

NEE on an interannual basis (Figure 7) reaching a peak of 230 g m-2 season -1 at 45% of

maximum connectivity.

142

Comparing results of the carbon flux model with and without runoff, lateral redistribution

results in higher rates of carbon uptake in the three subwatersheds compared to the model

without runoff (Figure 8). As growing season precipitation decreased, the model with

hydrologic redistribution exhibits greater change in NEE relative to the model without

runoff, i.e. the slope of the linear trend is steeper when runoff is included than when it is

not included. The component fluxes of NEE, GPP, Ra and Rh, increase with increasing

TI. GPP and Ra grow at a faster rate compared to Rh, and thus dominate the net carbon

flux (Figure 9). For example, Rh varied between 220 and 225 g m-2 season-1, while GPP

varied between 456 and 827 g m-2 season-1 based on TI. As a function of seasonal P, the

insensitivity of Rh leads to greater carbon loss (Ra + Rh) in the driest year. Over the six

study years, mean watershed Rh varied between 187 and 263 g m-2 season-1 while GPP

varies from 288 to 541 g m-2 season-1.

4. Discussion

4.1. Runoff and Carbon Flux Modeling

Available moisture, P-Q, reflects the water retained in a catchment that is available to

infiltrate and thus promote ecosystem function [14]. Our model indicates areas within a

watershed where runoff contributions yield substantially more available moisture than

precipitation (Figure 2). By simulating the water balance on 30-m pixels, we identify the

influence of spatially explicit flowpaths on Peff within each of the three nested

subwatersheds. This relaxes the assumptions of uniform catchment-scale moisture [14,

143

16, 68, 69] and simple bifurcating flowpath models [31] to simulate carbon dioxide

exchange along physically-based flowpaths.

Our analysis shows that available moisture from the runoff model is comparable to

measured available moisture in the three subwatersheds (Tables 1, 2). Substantially

greater volumes of moisture (Peff > 2) become available via upslope runoff because of

flow convergence [63, 70] particularly close to the watershed outlet (Figure 2). This

runoff infiltrates downslope because of enhanced infiltration rates [11, 71]. Runoff-

vegetation feedbacks have been observed at patch scales [22, 72] and linked to increased

infiltration from runoff [39, 73-75]. The role of vegetation in controlling infiltration is not

directly considered in this study, but areas where moisture is retained (i.e. at high TI

where Pfrac> 1, Figure 2) support longer periods of vegetation activity (Figure 3).

Values of Da indicate the effective length of the growing season based on the availability

of moisture (as Peff) such that the lowest regions in the catchment are active for

approximately 50 days longer than those at the top of the watershed. Longer growing

periods result in an increase in the rate of carbon uptake relative to upslope areas (Figure

3). The interannual variability in Da (and NEE) establishes the upper and lower bounds of

vegetation carbon flux activity that result from different rainfall amounts. Values of mean

Da (ranging from 25 to 75 days) indicate vegetation activity is dependent on topography.

The model shows that areas lower in the watershed support longer periods of vegetation

activity because the accumulated runoff from upslope increases Peff and therefore

144

increases soil moisture via infiltration. The shape of the logistic curve of Da describes

how the landscape structure drives ecosystem response through runoff. For example, the

steep increase in Da between TI = 8 and TI = 10 reflects the consistent accumulation of

moisture in these areas. The relative rate of moisture accumulation diminishes above TI =

10 as less runoff is generated and more moisture is stored. Thus, the rate of increase in Da

slows per unit value of TI.

The runoff model used here yields reasonable estimates of runoff (i.e. low RMSE and

relative error) based on subwatershed outflow estimates and indicates that redistribution

of moisture to downslope areas increases the amount of water available for ecosystem

use. However, we note that accumulation of moisture in this study is accounted for as

precipitation, not as infiltrated moisture, which is treated implicitly. Spatial differences in

soil moisture have been identified across a range of soil-vegetation combinations [50, 76,

77], due to soil characteristics (e.g. hydraulic conductivity) and vegetation effects (e.g.

macropores and preferential flowpaths) influencing infiltration. These moisture patterns

directly affect plant water and nutrient availability [78-81]. Site-level differences in

vegetation and soil properties may yield different sensitivities in the response of carbon

flux to moisture, e.g. reduced sensitivity of vegetation to moisture inputs on fine-textured

soils [80]. To validate the results from our study in other environments, detailed

examination of the physical and vegetation effects on spatial moisture heterogeneity is

needed.

145

4.2. Topographic Distribution of Carbon Fluxes

Prior efforts to model carbon dioxide flux response to “pulse”-type moisture inputs have

not dealt with moisture redistribution directly [48, 49, 82]. Using our model, net carbon

uptake is lower at low TI (Figure 4), and greater at high TI (maximum mean flux 200 mg

m-2 over the growing season). Low TI regions may be net sources of carbon dioxide to

the atmosphere during dry conditions, primarily due to Rh consuming carbon previously

stored in soil [83]. Based on the structure of our runoff model, TI predicts spatially

distributed carbon fluxes (Figure 4) resulting from moisture redistribution. Topographic

controls on net carbon flux shown here are similar to observed and simulated vegetation

patterns from other studies, which indicate increased vegetation activity where moisture

(and other resources) are concentrated along hillslopes [22, 27, 31, 32, 35].

We demonstrate interactions between topography, flow routing, and NEE, on the basis of

moisture redistribution that are conceptually similar to continental scale work on

hydrologic partitioning that relates topographic structure to a dimensionless metric of

catchment hydrologic partitioning, the Horton index [16, 31]. The Horton Index, which is

the ratio of catchment-scale ET to plant available water, varies spatially due to moisture

redistribution along a flowpath [31] and by vegetation type [13]. The increase in carbon

uptake lower in the watershed is conceptually consistent with increased partitioning in

favor of evapotranspiration [31], linking the increase in vertical flux of moisture to

carbon uptake along flowpaths.

146

The spatial distribution of NEE rates also varies based on the degree of hydrologic

connectivity, the relative length of all flowpaths contributing to a point (Figures 6, 7).

These results suggest that quantifying hydrologic connectivity is necessary to simulate

ecosystem function because total watershed net carbon flux rates are between 25% and

120% lower (i.e. indicating more carbon uptake) than when runoff is not considered.. As

contributing area increases, mean carbon uptake increases, and the range of NEE

increases (Figures 6, 7). Hydrologic connectivity is particularly important to represent

hydrologic features in semiarid landscapes [63, 84] by quantifying the potential runoff

contribution to each point following a storm.

4.3. Carbon Flux Response to Interannual Precipitation Variability

The frequency and timing of large storms in 2000 (P = 237mm) led to more runoff and

thus more carbon uptake near the watershed outlet, where the effect of runoff generation

is strongest, (-274 gCO2 m-2 season-1) than during the wettest year (1999; NEE = -260

gCO2 m-2 season-1, P = 274mm). Greater C uptake in the year 2000 primarily occurred

due to moisture subsidy in areas of high TI values following larger storms. For example,

the year 2000 had a larger mean storm compared to 1999 (6.6 mm and 6.2 mm,

respectively) but the average interstorm period was 3.9 days (compared to 2.7 in 1999).

These observations emphasize the importance of the stochastic nature of rainfall on

ecohydrologic function in drylands [7, 85], and further how rainfall interacts with

landscape and lateral hydrologic redistribution

147

Prior studies of ecosystem sensitivity to rainfall indicate larger carbon fluxes (GPP, Ra

and Rh) when total P increases [4, 86], but that increase is also dependent on the

frequency and magnitude of storms [8, 87]. The change in carbon fluxes is further evident

when runoff is incorporated into the model framework (Figures 4-8). By including runoff

when simulating carbon dynamics, our results suggest that NEE becomes more sensitive

to growing season P (Figure 8), evident from steeper slopes of P-NEE relationships

compared to NEE estimated without runoff. Future climate states are expected to have

altered P regimes, including reduced seasonal P or reduced storm frequency [88, 89].

Our results suggest changes in precipitation could reduce carbon uptake more than

expected without considering runoff (Figure 8). For example, total rainfall in 2002 was

40 mm less than in 2000, but the average storm was approximately 0.5 mm larger and

average interstorm periods were similar (3.7 and 3.8 days respectively). However, the

distribution of interstorm periods is more strongly skewed toward shorter interstorm

periods (0-2 days) in 2000 compared to 2002 (not shown). Carbon uptake was

approximately 10% greater in 2000 at the watershed outlet (Figure 4). Alternatively,

comparing years with similar interstorm distributions (2002 and 2003) but different mean

event size (7.1 mm and 5.7 mm, respectively), carbon uptake rates at the watershed outlet

are 85% greater in 2002. While details of P characteristics and their effect on NEE are

not the focus of this study, results from this limited dataset indicate the reduction in mean

storm size may have a greater effect on carbon dynamics than the duration between

storms. Further study of the complexities of P dynamics on runoff generation and

148

ecosystem response is needed to fully describe the connections between storm size,

interstorm duration and carbon flux.

Enhancement of Rh occurs in the driest year of the study (2004) leading to net carbon loss

from the ecosystem in both the runoff and no runoff models (Figure 7). But contrary to

the minimal effect of runoff anticipated from patch studies [54], our model shows that

lateral moisture redistribution supports carbon uptake lower in the watershed (Figure 4)

due to small amounts of runoff accumulating downslope. Because GPP is more sensitive

to moisture than Rh (Figure 9), the accumulation of small runoff volumes may be

sufficient to reverse an area from a source to a sink of carbon based on high connectivity

near the bottom of the watershed (Figure 7; [90]).

The link between Peff (Figure 2) and longer growing periods, based on Da (Figure 3) may

also indicate regions where vegetation may better resist climate stress because more

moisture is available. Only locations with TI > 10 were net carbon sinks in the driest year

(2004, Figure 4); the remainder of the watershed was a net source of carbon to the

atmosphere. If climate shifts toward less frequent storms and lower seasonal P, i.e.

toward rainfall regimes similar to year 2004, then locations at TI > 10 are more likely to

sustain carbon uptake (Figures 4, 8) compared to locations that are consistently water

stressed.

4.4. Understanding Semiarid Landscapes Across Scales

149

This study estimates the spatial heterogeneity in carbon dioxide that may not be captured

at larger scales, e.g. in studies of regional models or catchment water balance, or that is

lost using integrated measurements, e.g. eddy covariance [13, 91, 93-95]. These carbon

flux estimates bridge the spatial resolution gap from plot and patch scale studies, such as

investigations of landscape evolution and vegetation pattern [22, 96], to larger scales. By

applying this approach at an intermediate spatial scale, several new avenues for study

become available. These include comparisons among patch vegetation dynamics [22, 23],

plant hydraulic behavior [97-99], landscape models of water and nutrient distribution [91,

100] and large-scale coupling of hydrologic partitioning and vegetation [13, 16].

Our results indicate a scale-dependence in carbon dioxide flux that is similar to the scale-

dependence of catchment hydrologic partitioning and vegetation [31], and both trends

depend on the accumulation of moisture along flowpaths. Trends of NEE as a function of

TI (Figure 5) suggest that NEE stabilizes between -250 and -300 g m-2 season-1 at TI > 12

in the wettest years. In the driest year, NEE stabilizes at -100 g m-2 season-1 around TI =

14. Both of these TI thresholds suggest that the scale-dependence of NEE is minimal at

scales slightly larger than the Lucky Hills subwatersheds (4.5ha). Larger scales are

required in dry years in order to concentrate enough moisture to support persistent carbon

uptake.

The coupled hydrologic and carbon model used here is consistent with the concept of

islands of fertility—that areas of the landscape concentrate resources and subsequently

150

experience greater biological activity [35]. The hydrologic routing component of the

model predicts moisture availability which in turn supports biological function. While our

results assume static, uniform vegetation distribution, vegetation dynamics have minimal

effect on annual carbon flux compared to precipitation [101]. Vegetation dynamics may

alter the spatial distribution of productivity and respiration [102, 103] but the overall

effect on net carbon flux is not immediately clear and warrants further study. Changes in

precipitation, and resulting changes in hydrologic redistribution, have a greater effect on

areas where biological function is enhanced due to runoff (Figures 4, 6, 7). This study

identifies locations on the landscape where carbon flux is sensitive to change—

specifically areas of high TI—and estimates the magnitude of that variability (~250 mg

m-2 season-1).

5. Summary

This study quantifies the high degree of spatial heterogeneity of carbon dioxide flux

(~200 g m-2 season-1 difference within the watershed) that results from lateral moisture

redistribution in semiarid landscapes. By applying topographic information in a simple

runoff model, we determine that substantial amounts of water are redistributed to points

lower in the watershed. This redistribution of moisture supports longer periods of

ecosystem carbon flux, resulting in higher rates of carbon dioxide uptake during the

growing season. The response of carbon flux to interannual variations in seasonal

precipitation is modulated by topography, such that larger than expected carbon fluxes

occur when runoff is considered. As a result, topography and interannual precipitation

151

control the pattern of carbon flux in a watershed. These results reinforce connections

among topography, hydrologic and ecological functions in semiarid ecosystems. Explicit

simulation of redistribution as performed here is important to understand the long-term

response of these ecosystems to changing climate, as including runoff increased the

sensitivity to total seasonal rainfall.

Acknowledgements

Funding for this research was provided by SAHRA STC 9876800, NSF EAR 0910831, EAR-0724958, and DE-SC0006968. Datasets were provided by the USDA-ARS Southwest Watershed Research Center. Funding for these datasets was provided by the United States Department of Agriculture, Agricultural Research Service. The authors would like to thank S. Archer for helpful comments on an early draft of this manuscript.

References

[1] Friedlingstein P, P Cox, R Betts, L Bopp, W von Bloh, V Brovkin, et al. Climate–Carbon Cycle Feedback Analysis: Results from the C4MIP Model Intercomparison. Journal of Climate. 19 (2006) 3337-53. [2] IPCC. Climate Change 2007: The Physical Science Basis. Cambridge University Press, Cambridge, 2007. [3] Jones C, J Lowe, S Liddicoat, R Betts. Committed terrestrial ecosystem changes due to climate change. Nature Geosci. 2 (2009) 484-7. [4] Thomey ML, SL Collins, R Vargas, JE Johnson, RF Brown, DO Natvig, et al. Effect of precipitation variability on net primary production and soil respiration in a Chihuahuan Desert grassland. Global Change Biology. 17 (2011) 1505-15. [5] Sitch S, C Huntingford, N Gedney, PE Levy, M Lomas, SL Piao, et al. Evaluation of the terrestrial carbon cycle, future plant geography and climate-carbon cycle feedbacks using five Dynamic Global Vegetation Models (DGVMs). Global Change Biology. 14 (2008) 2015-39. [6] Borgogno F, P D'Odorico, F Laio, L Ridolfi. A stochastic model for vegetation water stress. Ecohydrology. (2010) [7] Laio F, A Porporato, L Ridolfi, I Rodriguez-Iturbe. Plants in water-controlled ecosystems: active role in hydrologic processes and response to water stress: II. Probabilistic soil moisture dynamics. Advances in Water Resources. 24 (2001) 707-23.

152

[8] Porporato A, F Laio, L Ridolfi, I Rodriguez-Iturbe. Plants in water-controlled ecosystems: active role in hydrologic processes and response to water stress - III. Vegetation water stress. Advances in Water Resources. 24 (2001) 725-44. [9] Jencso KG, BL McGlynn, MN Gooseff, SM Wondzell, KE Bencala, LA Marshall. Hydrologic connectivity between landscapes and streams: Transferring reach- and plot-scale understanding to the catchment scale. Water Resour Res. 45 (2009) W04428. [10] McGuire KJ, JJ McDonnell, M Weiler, C Kendall, BL McGlynn, JM Welker, et al. The role of topography on catchment-scale water residence time. Water Resour Res. 41 (2005) W05002. [11] Guntner A, A Bronstert. Representation of landscape variability and lateral redistribution processes for large-scale hydrological modelling in semi-arid areas. Journal of Hydrology. 297 (2004) 136-61. [12] Tague CL, LE Band. RHESSys: Regional Hydro-Ecologic Simulation System—An Object-Oriented Approach to Spatially Distributed Modeling of Carbon, Water, and Nutrient Cycling. Earth Interactions. 8 (2004) 1-42. [13] Brooks PD, PA Troch, M Durcik, E Gallo, M Schlegel. Quantifying regional scale ecosystem response to changes in precipitation: Not all rain is created equal. Water Resour Res. 47 (2011) W00J8. [14] Troch PA, GF Martinez, VRN Pauwels, M Durcik, M Sivapalan, C Harman, et al. Climate and vegetation water use efficiency at catchment scales. Hydrological Processes. 23 (2009) 2409-14. [15] Horton RE. The role of infiltration in the hydrologic cycle. Trans-Amer Geophys Union. 14 (1933) 446-60. [16] Voepel H, B Ruddell, R Schumer, PA Troch, PD Brooks, A Neal, et al. Quantifying the role of climate and landscape characteristics on hydrologic partitioning and vegetation response. Water Resour Res. 47 (2011) W00J9. [17] Hwang T, L Band, TC Hales. Ecosystem processes at the watershed scale: Extending optimality theory from plot to catchment. Water Resour Res. 45 (2009) W11425. [18] Diffenbaugh NS, F Giorgi, JS Pal. Climate change hotspots in the United States. Geophys Res Lett. 35 (2008) L16709. [19] Noy-Meir I. Desert Ecosystems: Environment and Producers. Annual Review of Ecology and Systematics. 4 (1973) 25-51. [20] Jenerette GD, RL Scott, TE Huxman. Whole ecosystem metabolic pulses following precipitation events. Funct Ecol. 22 (2008) 924-30. [21] Collins SL, RL Sinsabaugh, C Crenshaw, L Green, A Porras-Alfaro, M Stursova, et al. Pulse dynamics and microbial processes in aridland ecosystems. J Ecol. 96 (2008) 413-20. [22] Franz TE, KK Caylor, EG King, JM Nordbotten, MA Celia, I Rodríguez-Iturbe. An ecohydrological approach to predicting hillslope-scale vegetation patterns in dryland ecosystems. Water Resour Res. 48 (2012) W01515. [23] Caylor KK, TM Scanlon, I Rodriguez-Iturbe. Ecohydrological optimization of pattern and processes in water-limited ecosystems: A trade-off-based hypothesis. Water Resour Res. 45 (2009) W08407.

153

[24] Jenerette GD, GA Barron-Gafford, AJ Guswa, JJ McDonnell, JC Villegas. Organization of complexity in water limited ecohydrology. Ecohydrology. (2011) n/a-n/a. [25] Maneta M, S Schnabel, V Jetten. Continuous spatially distributed simulation of surface and subsurface hydrological processes in a small semiarid catchment. Hydrological Processes. 22 (2008) 2196-214. [26] Porporato A, P D'Odorico, F Laio, L Ridolfi, I Rodriguez-Iturbe. Ecohydrology of water-controlled ecosystems. Advances in Water Resources. 25 (2002) 1335-48. [27] Popp A, M Vogel, N Blaum, F Jeltsch. Scaling up ecohydrological processes: Role of surface water flow in water-limited landscapes. J Geophys Res. 114 (2009) G04013. [28] Ivanov VY, RL Bras, ER Vivoni. Vegetation-hydrology dynamics in complex terrain of semiarid areas: 2. Energy-water controls of vegetation spatiotemporal dynamics and topographic niches of favorability. Water Resour Res. 44 (2008). [29] Smith MW, LJ Bracken, NJ Cox. Toward a dynamic representation of hydrological connectivity at the hillslope scale in semiarid areas. Water Resour Res. 46 (2010) 18. [30] Reaney SM, LJ Bracken, MJ Kirkby. Use of the Connectivity of Runoff Model (CRUM) to investigate the influence of storm characteristics on runoff generation and connectivity in semi-arid areas. Hydrological Processes. 21 (2007) 894-906. [31] Thompson SE, CJ Harman, PA Troch, PD Brooks, M Sivapalan. Spatial scale dependence of ecohydrologically mediated water balance partitioning: A synthesis framework for catchment ecohydrology. Water Resour Res. 47 (2011) W00J3. [32] Riveros-Iregui DA, BL McGlynn. Landscape structure control on soil CO2 efflux variability in complex terrain: Scaling from point observations to watershed scale fluxes. J Geophys Res-Biogeosci. 114 (2009) 14. [33] Emanuel RE, HE Epstein, BL McGlynn, DL Welsch, DJ Muth, P D'Odorico. Spatial and temporal controls on watershed ecohydrology in the northern Rocky Mountains. Water Resour Res. 46 (2010) W11553. [34] Istanbulluoglu E, O Yetemen, ER Vivoni, HA Gutiérrez-Jurado, RL Bras. Eco-geomorphic implications of hillslope aspect: Inferences from analysis of landscape morphology in central New Mexico. Geophys Res Lett. 35 (2008) L14403. [35] Rango A, SL Tartowski, A Laliberte, J Wainwright, A Parsons. Islands of hydrologically enhanced biotic productivity in natural and managed arid ecosystems. Journal of Arid Environments. 65 (2006) 235-52. [36] Wu X, S Archer. Scale-Dependent Influence of Topography-Based Hydrologic Features on Patterns of Woody Plant Encroachment in Savanna Landscapes. Landsc Ecol. 20 (2005) 733-42. [37] Gutierrez-Jurado HA, ER Vivoni, JBJ Harrison, H Guan, Pp. Ecohydrology of root zone water fluxes and soil development in complex semiarid rangelands. Hydrological Processes. 20 (2006) 3289-316. [38] Schlesinger WH, AM Pilmanis. Plant-soil interactions in deserts. Biogeochemistry. 42 (1998) 169-87. [39] Bhark EW, EE Small. Association between plant canopies and the spatial patterns of infiltration in shrubland and grassland of the Chihuahuan Desert, New Mexico. Ecosystems. 6 (2003) 185-96.

154

[40] Schlesinger WH, JA Raikes, AE Hartley, AF Cross. On the Spatial Pattern of Soil Nutrients in Desert Ecosystems. Ecology. 77 (1996) 364-74. [41] Peters DPC, BT Bestelmeyer, JE Herrick, EL Fredrickson, HC Monger, KM Havstad. Disentangling Complex Landscapes: New Insights into Arid and Semiarid System Dynamics. BioScience. 56 (2006) 491-501. [42] Cable JM, GA Barron-Gafford, K Ogle, M Pavao-Zuckerman, RL Scott, DG Williams, et al. Shrub encroachment alters sensitivity of soil respiration to temperature and moisture. J Geophys Res. 117 (2012) G01001. [43] Barron-Gafford GA, RL Scott, GD Jenerette, TE Huxman. The relative controls of temperature, soil moisture, and plant functional group on soil CO2 efflux at diel, seasonal, and annual scales. J Geophys Res. 116 (2011) G01023. [44] Miller GR, JM Cable, AK McDonald, B Bond, TE Franz, L Wang, et al. Understanding ecohydrological connectivity in savannas: a system dynamics modelling approach. Ecohydrology. (2011) n/a-n/a. [45] Okin GS, AJ Parsons, J Wainwright, JE Herrick, BT Bestelmeyer, DC Peters, et al. Do Changes in Connectivity Explain Desertification? Bioscience. 59 (2009) 237-44. [46] Tromp-van Meerveld HJ, JJ McDonnell. On the interrelations between topography, soil depth, soil moisture, transpiration rates and species distribution at the hillslope scale. Advances in Water Resources. 29 (2006) 293-310. [47] Caylor KK, S Manfreda, I Rodriguez-Iturbe. On the coupled geomorphological and ecohydrological organization of river basins. Advances in Water Resources. 28 (2005) 69-86. [48] Shen WJ, GD Jenerette, DF Hui, RP Phillips, H Ren. Effects of changing precipitation regimes on dryland soil respiration and C pool dynamics at rainfall event, seasonal and interannual scales. J Geophys Res-Biogeosci. 113 (2008) 15. [49] Reynolds JF, PR Kemp, K Ogle, RJ Fernandez. Modifying the 'pulse-reserve' paradigm for deserts of North America: precipitation pulses, soil water, and plant responses. Oecologia. 141 (2004) 194-210. [50] Puigdefábregas J. The role of vegetation patterns in structuring runoff and sediment fluxes in drylands. Earth Surface Processes and Landforms. 30 (2005) 133-47. [51] Mayor ÁG, S Bautista, J Bellot. Scale-dependent variation in runoff and sediment yield in a semiarid Mediterranean catchment. Journal of Hydrology. 397 (2011) 128-35. [52] Sela S, T Svoray, S Assouline. Soil water content variability at the hillslope scale: Impact of surface sealing. Water Resour Res. 48 (2012) W03522. [53] Tague C, C McMichael, A Hope, J Choate, R Clark. APPLICATION OF THE RHESSys MODEL TO A CALIFORNIA SEMIARID SHRUBLAND WATERSHED. JAWRA Journal of the American Water Resources Association. 40 (2004) 575-89. [54] Huxman TE, KA Snyder, D Tissue, AJ Leffler, K Ogle, WT Pockman, et al. Precipitation pulses and carbon fluxes in semiarid and arid ecosystems. Oecologia. 141 (2004) 254-68. [55] Skirvin S, M Kidwell, S Biedenbender, JP Henley, D King, CH Collins, et al. Vegetation data, Walnut Gulch Experimental Watershed, Arizona, United States. Water Resour Res. 44 (2008) W05S8.

155

[56] Keefer TO, MS Moran, GB Paige. Long-term meteorological and soil hydrology database, Walnut Gulch Experimental Watershed, Arizona, United States. Water Resour Res. 44 (2008) W05S7. [57] Adams DK, AC Comrie. The North American Monsoon. Bulletin of the American Meteorological Society. 78 (1997) 2197-213. [58] Goodrich DC, TO Keefer, CL Unkrich, MH Nichols, HB Osborn, JJ Stone, et al. Long-term precipitation database, Walnut Gulch Experimental Watershed, Arizona, United States. Water Resour Res. 44 (2008) 5. [59] Stone JJ, MH Nichols, DC Goodrich, J Buono. Long-term runoff database, Walnut Gulch Experimental Watershed, Arizona, United States. Water Resour Res. 44 (2008) W05S. [60] Goodrich DC, TJ Schmugge, TJ Jackson, CL Unkrich, TO Keefer, R Parry, et al. Runoff simulation sensitivity to remotely sensed initial soil water content. Water Resour Res. 30 (1994) 1393-405. [61] Beven KJ, MJ Kirkby. A physically based, variable contributing area model of basin hydrology / Un modèle à base physique de zone d'appel variable de l'hydrologie du bassin versant. Hydrological Sciences Bulletin. 24 (1979) 43-69. [62] Lesschen JP, JM Schoorl, LH Cammeraat. Modelling runoff and erosion for a semi-arid catchment using a multi-scale approach based on hydrological connectivity. Geomorphology. 109 (2009) 174-83. [63] Mueller EN, J Wainwright, AJ Parsons. Impact of connectivity on the modeling of overland flow within semiarid shrubland environments. Water Resour Res. 43 (2007). [64] Vivoni ER, D Entekhabi, RL Bras, VY Ivanov. Controls on runoff generation and scale-dependence in a distributed hydrologic model. Hydrol Earth Syst Sci. 11 (2007) 1683-701. [65] Parsons AJ, RE Brazier, J Wainwright, DM Powell. Scale relationships in hillslope runoff and erosion. Earth Surface Processes and Landforms. 31 (2006) 1384-93. [66] Fentie B, B Yu, MD Silburn, CAA Ciesiolka. Evaluation of eight different methods to predict hillslope runoff rates for a grazing catchment in Australia. Journal of Hydrology. 261 (2002) 102-14. [67] Kurc SA, EE Small. Soil moisture variations and ecosystem-scale fluxes of water and carbon in semiarid grassland and shrubland. Water Resour Res. 43 (2007). [68] Sivapalan M, MA Yaeger, CJ Harman, X Xu, PA Troch. Functional model of water balance variability at the catchment scale: 1. Evidence of hydrologic similarity and space-time symmetry. Water Resour Res. 47 (2011) W02522. [69] Thompson SE, CJ Harman, AG Konings, M Sivapalan, A Neal, PA Troch. Comparative hydrology across AmeriFlux sites: The variable roles of climate, vegetation, and groundwater. Water Resour Res. 47 (2011) W00J7. [70] Bracken LJ, J Croke. The concept of hydrological connectivity and its contribution to understanding runoff-dominated geomorphic systems. Hydrological Processes. 21 (2007) 1749-63. [71] Mayor ÁG, S Bautista, EE Small, M Dixon, J Bellot. Measurement of the connectivity of runoff source areas as determined by vegetation pattern and topography:

156

A tool for assessing potential water and soil losses in drylands. Water Resour Res. 44 (2008) W10423. [72] Ludwig JA, BP Wilcox, DD Breshears, DJ Tongway, CI Anton. Vegetation Patches and Runoff-Erosion as Interacting Ecohydrological Processes in Semiarid Landscapes. Ecology. 86 (2005) 288-97. [73] Duniway MC, KA Snyder, JE Herrick. Spatial and temporal patterns of water availability in a grass-shrub ecotone and implications for grassland recovery in arid environments. Ecohydrology. 3 (2010) 55-67. [74] Moran MS, EP Hamerlynck, RL Scott, JJ Stone, CD Holifield Collins, TO Keefer, et al. Hydrologic response to precipitation pulses under and between shrubs in the Chihuahuan Desert, Arizona. Water Resour Res. 46 (2010) W10509. [75] Snyder KA, SL Tartowski. Multi-scale temporal variation in water availability: Implications for vegetation dynamics in arid and semi-arid ecosystems. Journal of Arid Environments. 65 (2006) 219-34. [76] Moran MS, RL Scott, EP Hamerlynck, KN Green, WE Emmerich, CDH Collins. Soil evaporation response to Lehmann lovegrass (Eragrostis lehmanniana) invasion in a semiarid watershed. Agricultural and Forest Meteorology. 149 (2009) 2133-42. [77] Wainwright J, AJ Parsons, AD Abrahams. Plot-scale studies of vegetation, overland flow and erosion interactions: case studies from Arizona and New Mexico. Hydrological Processes. 14 (2000) 2921-43. [78] Sandvig RM, FM Phillips. Ecohydrological controls on soil moisture fluxes in arid to semiarid vadose zones. Water Resour Res. 42 (2006) W08422. [79] Wheeler CW, SR Archer, GP Asner, CR McMurtry. Climatic/edaphic controls on soil carbon/nitrogen response to shrub encroachment in desert grassland. Ecological Applications. 17 (2007) 1911-28. [80] Fravolini A, KR Hultine, E Brugnoli, R Gazal, NB English, DG Williams. Precipitation Pulse Use by an Invasive Woody Legume: The Role of Soil Texture and Pulse Size. Oecologia. 144 (2005) 618-27. [81] Schwinning S, JR Ehleringer. Water use trade-offs and optimal adaptations to pulse-driven arid ecosystems. J Ecol. 89 (2001) 464-80. [82] Ogle K, JF Reynolds. Plant responses to precipitation in desert ecosystems: integrating functional types, pulses, thresholds, and delays. Oecologia. 141 (2004) 282-94. [83] Carbone M, C Still, A Ambrose, T Dawson, A Williams, C Boot, et al. Seasonal and episodic moisture controls on plant and microbial contributions to soil respiration. Oecologia. 167 (2011) 265-78. [84] Lexartza-Artza I, J Wainwright. Hydrological connectivity: Linking concepts with practical implications. CATENA. 79 (2009) 146-52. [85] Rodríguez-Iturbe I, V Isham, DR Cox, S Manfreda, A Porporato. Space-time modeling of soil moisture: Stochastic rainfall forcing with heterogeneous vegetation. Water Resour Res. 42 (2006) W06D5. [86] Archibald SA, A Kirton, MR van der Merwe, RJ Scholes, CA Williams, N Hanan. Drivers of inter-annual variability in Net Ecosystem Exchange in a semi-arid savanna ecosystem, South Africa. Biogeosciences. 6 (2009) 251-66.

157

[87] Chou WW, WL Silver, RD Jackson, AW Thompson, B Allen-Diaz. The sensitivity of annual grassland carbon cycling to the quantity and timing of rainfall. Global Change Biology. 14 (2008) 1382-94. [88] Fatichi S, VY Ivanov, E Caporali. Simulation of future climate scenarios with a weather generator. Advances in Water Resources. 34 (2011) 448-67. [89] Diffenbaugh NS, JS Pal, RJ Trapp, F Giorgi. Fine-scale processes regulate the response of extreme events to global climate change. Proc Natl Acad Sci U S A. 102 (2005) 15774-8. [90] Belnap J, JR Welter, NB Grimm, N Barger, JA Ludwig. Linkages between Microbial and Hydrologic Processes in Arid and Semiarid Watersheds. Ecology. 86 (2005) 298-307. [91] Jenerette GD, RL Scott, GA Barron-Gafford, TE Huxman. Gross primary production variability associated with meteorology, physiology, leaf area, and water supply in contrasting woodland and grassland semiarid riparian ecosystems. J Geophys Res-Biogeosci. 114 (2009). [92] Baudena M, G Boni, L Ferraris, J von Hardenberg, A Provenzale. Vegetation response to rainfall intermittency in drylands: Results from a simple ecohydrological box model. Advances in Water Resources. 30 (2007) 1320-8. [93] Sjöström M, J Ardö, A Arneth, N Boulain, B Cappelaere, L Eklundh, et al. Exploring the potential of MODIS EVI for modeling gross primary production across African ecosystems. Remote Sensing of Environment. 115 (2011) 1081-9. [94] Zhang L, BK Wylie, L Ji, TG Gilmanov, LL Tieszen, DM Howard. Upscaling carbon fluxes over the Great Plains grasslands: Sinks and sources. J Geophys Res. 116 (2011) G00J3. [95] Lee D, J Kim, KS Lee, S Kim. Partitioning of catchment water budget and its implications for ecosystem carbon exchange. Biogeosciences. 7 (2010) 1903-14. [96] Caylor KK, P D'Odorico, I Rodriguez-Iturbe. On the ecohydrology of structurally heterogeneous semiarid landscapes. Water Resour Res. 42 (2006) W07424. [97] Tietjen B, E Zehe, F Jeltsch. Simulating plant water availability in dry lands under climate change: A generic model of two soil layers. Water Resour Res. 45 (2009) 14. [98] Scott RL, WL Cable, KR Hultine. The ecohydrologic significance of hydraulic redistribution in a semiarid savanna. Water Resources Research. 44 (2008). [99] Hultine KR, DG Williams, SSO Burgess, TO Keefer. Contrasting patterns of hydraulic redistribution in three desert phreatophytes. Oecologia. 135 (2003) 167-75. [100] Le Maitre DC, SJ Milton, C Jarmain, CA Colvin, I Saayman, JHJ Vlok. Linking ecosystem services and water resources: landscape-scale hydrology of the Little Karoo. Front Ecol Environ. 5 (2007) 261-70. [101] Williams CA, JD Albertson. Contrasting short- and long-timescale effects of vegetation dynamics on water and carbon fluxes in water-limited ecosystems. Water Resources Research. 41 (2005) 13. [102] Emanuel, RE, DA Riveros-Iregui, BL McGlynn, HE Epstein. On the spatial heterogeneity of net ecosystem productivity in complex landscapes. Ecosphere. 2 (2011) 86.

158

[103] O'Donnell, FC, KK Caylor. A model-based evaluation of woody plant encroachment effects on coupled carbon and water cycles. J Geophys Res. 117 (2012) G02012. [104] Scott RL. Using watershed water balance to evaluate the accuracy of eddy covariance evaporation measurements for three semiarid ecosystems. Agricultural and Forest Meteorology. 150 (2010) 219-25.

159

106 102 104 Year P Q P-Q P-Q/P Q P-Q P-Q/P Q P-Q P-Q/P 1999 274 48 226 0.82 29 245 0.90 24 250 0.91 2000 237 45 192 0.81 40 197 0.83 32 205 0.86 2001 188 1 187 0.99 4 184 0.98 2 186 0.99 2002 200 34 166 0.83 17 182 0.91 12 188 0.94 2003 159 25 134 0.84 15 144 0.91 11 148 0.93 Total 1059 154 905 105 953 81 978

Table 1. Precipitation, runoff, available moisture (P-Q), and the ratio of available moisture to P for the three subwatersheds for each growing season. Moisture that does not leave the subwatersheds via runoff remains available on the landscape to support ecosystem activity. Values of (P-Q)/P are comparable to [59, 104]

160

106 102 104 Year Qobs Qmod P-Qmod/P Qobs Qmod P-Qmod/P Qobs Qmod P-Qmod/P 1999 48 18 0.93 29 29 0.89 24 35 0.87 2000 45 15 0.94 40 24 0.90 32 29 0.88 2001 1 12 0.94 4 20 0.90 2 24 0.87 2002 34 13 0.94 17 21 0.89 12 25 0.87 2003 25 9 0.94 15 15 0.90 11 18 0.89 Total 154 67 0.96 105 109 0.94 81 131 0.92

RMSE 0.71 0.46 0.40 Table 2. Table of measured growing season runoff (Qobs, mm) and simulated runoff (Qmod, mm) for the study period and simulated wetness No runoff data were available for comparison in 2004. The underestimation of runoff in subwatershed 106 is likely due to inaccurate routing resulting from its small size (~5400 m2 or 6 30-m pixels). Despite differences between Qmod and Qobs, the fraction of available moisture is similar to measured values (see Table 1).

161

Figure 1. Site map of WGEW showing the subwatersheds in the Lucky Hills region.

104

102 106

101

103 105

162

Figure 2. Mean fraction of rainfall available (Pfrac) on the landscape as a function of topographic index. Pfrac values are determined as the ratio of Peff to P on an event basis and averaged over the study period. Error bars indicate one standard deviation of Pfrac. The line indicates a Pfrac of 1, which represents the distribution of P if runoff is not used to model carbon flux.

163

Figure 3. Mean duration of vegetation activity (Da) for subwatershed 104 based on the

distributed runoff method. Error bars indicate one standard deviation of Da. The linear fit

(dashed line) has a coefficient of determination of 0.36. The logistic fit shown has a

coefficient of determination of 0.96, and indicates an inflection point in the Da trend at TI

= 9.8.

164

Figure 4. NEE as a function of TI for each year of the study period. NEE becomes more

negative (i.e. carbon uptake increases) with increasing TI. The range of variation is

lowest at TI = 6.3 (72 g m-2 season-1) but is high when TI is high (TI = 10.3, range of

NEE = 242 g m-2 season-1).

165

Figure 5. NEE as a function of TI (as in Figure 4) with logistic curves fitted for each year.

All curves fit at r2 > 0.95, except for 2003 at r2 = 0.88.

166

Figure 6. Interannual range of NEE (ΔNEE) as a function of TI. The lowest ranges, i.e. least variable NEE between years, are found around TI = 6 and the most variable at TI > 10.

167

Figure 7. Mean NEE (left panel) and interannual variability in NEE (ΔNEE) as a function of the relative connectivity of each pixel (λ). Connectivity is the relative number of pixels that contribute flow to a point in the watershed, e.g. λ = 0.5 means half as many pixels contribute to that point as contribute to the outlet.

168

Figure 8. NEE response to precipitation in the three subwatersheds, as indicated in each panel. Black points represent NEE estimated with runoff and gray points indicate NEE with no runoff. The slope for the no runoff model is shown in the left panel. All fits are significant at p<0.01.

169

Figure 9. Mean component fluxes of NEE as a function of TI (left panel) and seasonal P (right panel). Note that GPP and (Ra + Rh) have similar magnitudes around TI = 6. Also, (Ra+Rh) exceeds GPP in the driest year.

170

APPENDIX D: USING LANDSAT ETM+ NDVI TO CLASSIFY VEGETATION

COVER

1. INTRODUCTION

The analysis in this appendix was developed for the study in Appendix B as a method to

develop more precise landscape classification. This classification was intended to identify

regions that included vegetation characteristics similar to the source area immediately

surrounding the eddy covariance flux tower. The analysis examined the change in

vegetation greenness detected via the Landsat Enhanced Thematic Mapper Plus (ETM+).

Vegetation greenness near the study site is described as bimodal, with peaks in the spring

and summer (Pennington and Collins 2007). Because of the evergreen nature of creosote

shrubs, the change in greenness associated with creosote-dominated parts of the

landscape should be minimal compared to those dominated by seasonal vegetation, e.g.

grasses and forbs (Peters et al. 1997)

A revised version of this approach, using data from the Landsat Thematic Mapper sensor

and using data throughout the three-year study period, is in development as a complement

to the current analysis in Appendix B. The new analysis, in conjunction with high-

resolution spatial data, should better depict changes in fractional vegetation cover over

the study period. With a method to identify changes in fractional cover, the effect of leaf

area on vegetation indices over the same time period can be more accurately described.

171

2. METHODS

Data from the Landsat 7 ETM+ were analyzed to resolve the sub-grid variability in

MODIS scenes from Appendix B. Landsat scenes were downloaded for January and

August 2002. Landsat data were corrected for atmospheric effects using the LEDAPS

tool (Masek et al. 2006). Individual Landsat ETM+ bands were used to compute the

normalized difference vegetation index (NDVI), calculated as:

redNIR

redNIRNDVIρρρρ

+−

= (2)

Landsat ETM+ NDVI was used to determine relative land cover dominance between

evergreen creosotebush and seasonal vegetation (grasses and forbs). Seasonal change in

NDVI (ΔNDVI) was used to characterize the distribution of shrubs based on the

difference between scenes taken during the growing (August) and non-growing (January)

season. The areas dominated by creosote (an evergreen shrub) are expected to have

relatively stable NDVI values compared to other vegetation types (i.e. grasses, forbs and

deciduous shrubs). Landsat ΔNDVI data were resampled to the 250-m scale to

correspond to MODIS pixels in Appendix B.

3. RESULTS

The seasonal vegetation dynamics, described by ΔNDVI in the Landsat image, are

minimal in the region close to the tower, i.e. the value of NDVI does not change (Figure

1). High ΔNDVI (~0.3) at the 30-m Landsat scale is evident to the northeast of the tower,

172

along the edge of the shrub/grass ecotone (Kurc and Small 2007), as well as along the

drainages . The diminished EVI (annual footprint ΔEVI = 0.06) and NDVI (ΔNDVI in

the footprint ~ 0.10) variability near the tower is associated with the evergreen character

of creosotebush, which reduced correlation to tower NEE. From the ΔNDVI values, we

delineate regions that have vegetation characteristics similar to the tower footprint, and

apply scaling rules in those locations, where similar flux patterns can be expected.

To produce an estimate of carbon flux when deep soils are wet, the ΔNDVI values from

the Landsat ETM+ image were resampled to MODIS EVI resolution, i.e. from 30-m to

250-m. Resampled pixels with ΔNDVI values less than or equal to the tower area were

considered similar to the tower. Applying the mean tower area equations for NEE and

GPP identified in Table 3, we can simulate carbon fluxes for the creosote-dominated

region when deep soils are wet. The average annual NEE for these pixels for the entire

three year record is -6.19 g m-2 d-1 and mean GPP is 8.95 g m-2 d-1. The spatial pattern of

NEE in 2002 (Figure 9), indicates consistent carbon uptake when deep soils are wet, and

similar net flux rates across the study area. Carbon uptake is slightly greater at the head

of drainages in the south and north of the study area.

4. DISCUSSION MODIS EVI pixels that were highly correlated to tower flux exhibit a greater dynamic

range of EVI, suggesting differences in vegetation between correlated pixels and the

tower area, potentially due to increased bare space or annual vegetation. To the northeast

of the tower the landscape transitions from shrub-dominated to grass-dominated. A

173

stream crosses the scene toward the western edge, and wide swales drain the

northwestern edge of the creosote region. The ΔNDVI values in the Landsat image

corroborate the differences in vegetation characteristics along the drainages and in the

northeast region of the study area (Figure 8). The source area of the tower is

characterized by relatively uniform cover by creosote shrubs and flat terrain (Kurc and

Small 2007). Linking spatial patterns in the Landsat and MODIS images, we identify a

region that is similar to the tower area, based on the maximum ΔNDVI in the tower

source area.

The seasonality of vegetation, described by ΔNDVI, indicates two main patterns (Figure

8). The central portion of the study area, which is dominated by creosote, has a low range

of NDVI values, indicating minimal change in vegetation characteristics (leaf area or

fractional cover) between the growing and non-growing seasons. This result is expected

in light of the perennial nature of creosote and the low density of vegetation cover (Qi et

al. 1994). Alternatively, locations toward the margins of the study area and in the

drainages have a greater range of NDVI from growing to non-growing periods,

suggesting these areas are more densely vegetated and include seasonal vegetation rather

than the perennial creosote near the tower.

The goal of the ΔNDVI analysis was to provide a simple approach to determine regions

in the MODIS scene (7 km x 7 km) that have vegetation similar to the tower source area,

so the NEE and GPP models could be applied over similar vegetation conditions.

174

While the example shown (Figure 9) is applied only for one year (2002, the same year as

the Landsat NDVI figures), concerns were raised about the long-term applicability of this

method under interannual vegetation change. For example, assuming that the distribution

of creosote shrubs based on an analysis in 2002 was static over the three year study

period. Also, the ability to discern cover type reliably using NDVI on the basis of

phenology was a concern if this approach were applied over the three year period.

5. CONCLUSIONS AND FUTURE WORK

The ΔNDVI analysis to identify different vegetation communities in a scene may be

suitable for simple demonstration purposes to highlight the NEE and GPP models

developed in Appendix B. However, there remain potential complications in the

interpretation of the results, namely, the application of the method over longer time spans

in which vegetation characteristics may change. Further, because the model developed in

Appendix B used MODIS EVI (and not NDVI), methodological consistency would be

best served by reanalyzing using Landsat EVI products. Improvements to this approach

would also include repeat analysis to monitor changes in vegetation distribution over

time. This could be done by linking Landsat data to high-resolution remote sensing data,

such as the U.S. Department of Agriculture’s National Agricultural Imagery Program

(NAIP) data sets, which provide more detailed information about vegetation cover.

6. REFERENCES

175

Kurc, S.A., & Small, E.E. (2007). Soil moisture variations and ecosystem-scale fluxes of

water and carbon in semiarid grassland and shrubland. Water Resources

Research, 43

Masek, J.G., Vermote, E.F., Saleous, N., Wolfe, R., Hall, F.G., Huemmrich, F., Gao, F.,

Kutler, J., & Lim, T.K. (2006). A Landsat surface reflectance data set for North

America, 1990-2000. Geoscience and Remote Sensing Letters, 3, 68-72

Pennington, D., & Collins, S. (2007). Response of an aridland ecosystem to interannual

climate variability and prolonged drought. Landscape Ecology, 22, 897-910

Peters, A.J., Eve, M.D., Holt, E.H., & Whitford, W.G. (1997). Analysis of Desert Plant

Community Growth Patterns with High Temporal Resolution Satellite Spectra.

Journal of Applied Ecology, 34, 418-432

Qi, J., Chehbouni, A., Huete, A.R., Kerr, Y.H., & Sorooshian, S. (1994). A modified soil

adjusted vegetation index. Remote Sensing of Environment, 48, 119-126

176

Figure 1. ΔNDVI between the annual maximum (August) and minimum (January) NDVI in 2002. Upper image shows ΔNDVI at the 30-m resolution of Landsat and the lower image is at the resampled 250-m resolution. ΔNDVI values near the tower (marked with a cross) are all below 0.1, which was used as the threshold to identify regions that are creosote-dominated.

177

Figure 2. Map of total NEE (g m-2 d-1) in 2002 during periods when deep soils are wet. Regions that have NEE of 0 g m-2 d-1 are outside the creosote-dominated area determined from the resampled Landsat image (Figure 8).