sihong wu - diva portalkth.diva-portal.org/smash/get/diva2:372901/fulltext01.pdfseasonal courses of...

TRITA-LWR Lic Thesis 2054

ISSN 1650-8629

ISRN KTH/LWR/LIC 2054-SE

ISBN 978-91-7415-826-7

MODELING THE SEASONALITY OF CARBON, EVAPOTRANSPIRATION AND HEAT

PROCESSES FOR COLD CLIMATE

CONDITIONS

Sihong Wu

December 2010

Sihong Wu TRITA LWR Lic Thesis 2054

ii

© Sihong Wu 2010

Licentiate Thesis

KTH-Environmental Physics

Department of Land and Water Resources Engineering

Royal Institute of Technology (KTH)

SE-100 44 STOCKHOLM, SWEDEN

Modeling the Seasonality of Carbon, Evapotranspiration and Heat Processes for Cold Climate Conditions

iii

iii

ACKNOWLEDGEMENTS

First of all, I would like to acknowledge my main supervisor, Professor Per-Erik Jansson, for his continuous support and encouragement. Our traditional Monday meeting makes me learning lots of knowledge from you. I really appreciate that you spend a lot of time on me.

I also thank my co-supervisor, Professor Deliang Chen. He introduced the theory of statistical downscaling to me and guided me to make a clear planning for the future.

Moreover, I appreciate Prof. Xingyi Zhang and Dr. Pasi Kalori for providing materials for this study and sharing their valuable experiences with me. I also thank Prof. Qingke Zhu, Prof. Xinxiao Yu and Prof. Hongjiang Zhang for their warm care and helps.

Thanks to all people at LWR. It is an amazing experience to study at KTH.

I would like to thank China Scholarship Council (CSC) and School of Architecture and Built Environment (ABE) of Royal Institute of Technology (KTH) for providing the scholarship and the stipend.

Last but not least, I would like to thank my husband, Dr. Xiaoming Zhang and my parents for their endless support and love.


iv


v

v

ABSTRACT

The productivity of agricultural and forest ecosystems in regions at higher latitudes is to a large extent governed by low temperature and moisture conditions. Environmental conditions are acting both above- and below-ground and regulating carbon fluxes and evapotranspiration. How-ever, the understanding of various feedbacks between vegetation and environmental conditions is still unclear. In this thesis, two studies were conducted to understand the physical and biological processes. In the first study, the aim was to simulate soil temperature and moisture dynamics in the bare soil with seasonal frost conditions in China. In the second study, the aims were to model seasonal courses of carbon and evapotranspiration and to examine the responses of photosynthe-sis, transpiration and respiration on environmental conditions in a boreal Scots pine ecosystem in Finland. In both studies the CoupModel was applied to simulate the dynamic responses of the systems. Both sites represented investigations from which a high number of measurements were available. To understand to what extent the data could be used to increase the understanding of the systems, the Generalized Likelihood Uncertainty Estimation (GLUE) was applied. The GLUE method was useful to reduce basic uncertainties with respect to parameter ranges, model structures and measurements.

The strong interactions between soil temperature and moisture processes have indicated by a few behavioral models obtained when constrained by combined temperature and moisture criteria. Model performance on sensible and latent heat fluxes and net ecosystem exchange (NEE) also indicated the coupled processes within the system. Seasonal and diurnal courses were reproduced successfully with reduced parameter ranges. However, uncertainties on what is the most general regulation for transpiration and NEE are still unclear and need further systematic investigations.

Key words: net ecosystem exchange (NEE); sensible and latent heat fluxes; photosynthesis; respiration; nitrogen turnover; Scots pine forest; bare soil; cold climate; soil physical characteris-tics; CoupModel; GLUE


vi


vii

vii

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ......................................................................................................................... III

ABSTRACT .................................................................................................................................................... V

TABLE OF CONTENTS ........................................................................................................................... VII

LIST OF PAPERS ......................................................................................................................................... IX

INTRODUCTION ......................................................................................................................................... 1

Study objectives .................................................................................................................................................. 2

The scope of the study........................................................................................................................................ 2

METHODS ..................................................................................................................................................... 2

Study sites and measurements ............................................................................................................................. 2

CoupModel description and parameterization ..................................................................................................... 3

GLUE for model calibration and performance indices ........................................................................................ 3

RESULTS ........................................................................................................................................................ 4

Model performance on soil temperature and moisture (Paper I) .......................................................................... 4 Soil temperature performance ......................................................................................................................................................... 4

Soil moisture performance .............................................................................................................................................................. 4

Model performance on eddy covariance fluxes (Paper II) .................................................................................... 6 Seasonal courses............................................................................................................................................................................. 6

Diurnal courses ............................................................................................................................................................................. 7

Response functions for environmental factors (Paper II) .................................................................................... 8

DISCUSSION ................................................................................................................................................. 9

GLUE as a tool to increase the understanding of modeling processes ................................................................. 9

Process-oriented modeling to understand complex systems .............................................................................. 10

Response functions for environmental factors .................................................................................................. 11

CONCLUSIONS ........................................................................................................................................... 11

FUTURE WORK ........................................................................................................................................... 11

REFERENCES .............................................................................................................................................. 12


viii


ix

ix

LIST OF PAPERS

I. Wu SH, Jansson P-E and Zhang XY. 2010. Model for temperature, moisture and surface heat

balance in the bare soil with seasonal frost conditions in China. Submitted to European Journal

of Soil Science, June 2010.

II. Wu SH, Jansson P-E and Kolari P. 2010. Modeling seasonal courses of carbon fluxes and

evapotranspiration in response to low temperature and moisture in a boreal Scots Pine ecosys-

tem. Manuscript.


x


1

INTRODUCTION

Based on growing evidence, agricultural and forest ecosystems in the sensitive regions at higher latitudes quickly respond to climate variability and climate change, leading to dramatic changes in nitrogen and carbon cycles, water and heat dynamics, physical and chemical soil properties (IPCC, 2007, Hamp-ton et al., 2008; Challinor et al., 2009). There-fore, understanding the ecosystem mecha-nisms in these regions are of key importance in improving knowledge of the interactions between terrestrial ecosystems and climate change. A great amount of literatures have focused on the seasonal courses of carbon, water and heat fluxes for different ecosys-tems in these sensitive regions, e.g. the Nor-dic countries, North America, and northern China (Nyberg et al., 2001; Hu et al., 2006; Kolari et al., 2007; Högberg et al., 2009). Moreover, many studies have been done on responses of carbon, water and heat dynam-ics to environmental changes, especially un-der low temperature and moisture conditions (Schwarz et al., 1997; Mellander et al., 2006; Cheng et al., 2008).

In order to achieve a better understanding of biological and physical processes within the ecosystems, one way is to apply many high-resolution instruments to intensify the moni-toring frequencies. For instance, based on the eddy covariance technique, flux measure-ments are widely used to estimate momen-tum, heat, water, and carbon dioxide ex-change, as well as exchange of methane and other trace gases (Vesala et al., 2005). Other high-resolution measurements such as soil chamber measurements, meteorological measurements and soil temperature meas-urements also provide an access to more detailed observe processes in the ecosystems. The other way is to simulate these complex processes by using different types of models. Especially, it is quite common to assess the impacts of climate change on physiological and physical processes by using model simu-lations (Karmer et al., 2002).

Soil freeze – thaw cycles and snow events play important roles to affect soil physical properties, decomposition and respiration

processes in many regions at higher latitudes (Hansson et al., 2004; Gustafsson et al., 2004; Kettridge and Baird, 2008). Hence, dynamics of soil temperature and moisture during the shifting periods are very crucial to be under-stood, especially in response to climate varia-bility and climate change.

Boreal forests are exposed to low above- and below-ground temperatures during winter and early spring. Low temperatures in air and soil hinder the trees’ abilities to utilize water and nutrients (Bergh et al., 1998; Mellander et al., 2006). Effects of low temperature and moisture availability on reductions of physio-logical processes in boreal forests have been widely studied by means of various models (Tanja et al., 2003; Mäkelä et al., 2004; Bauer et al., 2008). However, most of the studies have only investigated one single process (photo-synthesis or respiration) or only assumed one possible environmental factor demonstrating a vital role to influence the processes. Only few studies have been conducted to examine the effects of low temperature and moisture availability on carbon fluxes and evapotran-spiration simultaneously in boreal forests (Svensson et al., 2008 a; Magnani et al., 2009).

In nature, transpiration, photosynthesis, carbon allocation, tree growth and water flow are all linked to each other. Furthermore, ambient environmental factors demonstrate combined influences on these processes. Therefore, there is an increasing interest to investigate these vital processes concurrently by using models. In this study, a one-dimensional physically-based model (Coup-Model) is used to simulate the seasonality of carbon, evapotranspiration and heat process-es simultaneously and evaluate the reductions of photosynthesis, transpiration and respira-tion due to low temperature and moisture availability in air and soil according to various response functions.

Since uncertainties in measurements, parame-ters and model structures are unavoidable, the onset of uncertainty analysis can give an aid to quantify the unknown uncertainties, which could improve the confidences of model predictions. A package of methodolo-gies of uncertainty analysis is available for model calibration, such as the Bayesian


2

method (Van Oijen et al., 2005), the Fuzzy method (Hall and Anderson, 2002), nonlinear regression (Vrugt et al., 2003) and the Gener-alized Likelihood Uncertainty Estimation (GLUE) (Beven, 2006). In this study, the GLUE framework was applied to estimate uncertainties in parameters, model assump-tions and measurements. When using GLUE, the performance of behavioral models is constrained by the subjective criteria, which are defined by users. Therefore, the respons-es of performances based on different per-formance indices were examined by this study as well.

Study objectives

The objectives of this study were to simulate and analyze the seasonal courses of carbon, evapotranspiration and heat processes, espe-cially under cold climate conditions – within the framework of uncertainty analysis (GLUE).The following specific objectives for this study were to:

a) estimate the impacts of criteria choices on parameter uncertainties and model behaviors, and investigate possible uncer-tainties in parameters and measurements (Paper I);

b) evaluate model performance with respect to soil temperature and moisture con-strained by 4 different criteria, using the GLUE uncertainty framework and identi-fy sensitive parameters which affect soil heat and water processes (Paper I);

c) model seasonal patterns of carbon fluxes and evapotranspiration by using eddy co-variance measurements and the GLUE method for a Scots pine forest ecosystem (Paper II);

d) describe the governing functions of pho-tosynthesis and water uptake in response to environmental conditions; and discuss to what extent we could distinguish the most important environmental factor be-tween air and soil to control the ecosys-tem responses in boreal regions (Paper II).

The scope of the study

The results in both papers are all based on a one-dimensional physically-based model

(CoupModel) by using the GLUE calibration method. In Paper I, dynamics of soil temper-ature and moisture in the bare black soil during 2005 - 2007 were simulated. Different model performances were obtained, depend-ing on the different criteria applied. Based on previous experiences in different responses of performance indices and parameter per-formances from Paper I, seasonal courses of net ecosystem exchange (NEE) and sensible and latent heat fluxes in a boreal Scots pine ecosystem were simulated successfully (Paper II). Responses of photosynthesis, transpira-tion and respiration processes on low tem-perature and moisture availability in air and soil were evaluated as well.

METHODS

This section includes the descriptions of the two study sites, measurements, the Coup-Model, the GLUE method and correspond-ing performance indices.

Study sites and measurements

The two sites in this study are both located at higher latitudes, where air temperature is quite low and snow is usually present during the whole winter period.

In Paper I, the study site was located at Hai-lun Agricultural Ecology Experimental Sta-tion, Heilongjiang province in the northeast of China (47°26’ N and 126°38’ E). After approximately 100 years of cultivation, a 400 m2 bare black soil plot was established in 1985. Hourly meteorological data including precipitation, global radiation, net radiation, air temperature, wind speed, relative air hu-midity and vapor pressure during 2005 – 2007 were used as driving variables. The mean annual temperature was 2.0, 1.9, and

3.9 °C and the annual precipitation was 544, 548, and 405 mm in 2005, 2006, and 2007, respectively. Hourly soil temperature at the depths of 20, 40 and 60 cm determined by thermometer probes during the whole 3 years and daily mean values of soil water content at the depths of 20, 40, 90, 150 and 250 cm determined by neutron probes during the growing seasons of the three years were used as calibrated data. In order to remove the


3

impact of systematic errors by neutron probes, soil water storages in the five corre-sponding depths were calculated.

In Paper II, the study site was located at the SMEAR II field station (Station for Measur-ing Forest Ecosystem-Atmosphere Relations), in Hyytiälä, southern Finland (61°51’ N and 24°17’ E). The area is part of the southern boreal zone with an average air temperature of 3 °C and average annual precipitation of 700 mm (Markkanen et al., 2001). The site is dominated by Scots pine planted in 1962.

High resolution data were measured with the eddy covariance technique, the soil chamber method, the soil heat flux, soil temperature and meteorological conditions with an hourly resolution. Low resolution data were forest inventory data, soil moisture and snow data. Eddy covariance fluxes during 2007 – 2008 were applied to constrain model performanc-es of carbon flux and sensible and latent heat fluxes.

The detailed descriptions of the measure-ments are found in the Methods sections of Paper I and Paper II.

CoupModel description and parameteri-zation

CoupModel is a one-dimensional physically-based model for simulating thermal and hydrological processes and the corresponding biological processes that regulate carbon and nitrogen transfer in a soil-plant-atmosphere environment (Jansson and Moon, 2001; Jansson and Karlberg, 2004), and is a cou-pling of the former SOIL and SOIL-N mod-els (Jansson and Halldin, 1979; Eckersten et al., 1998).

In Paper I, modules of soil heat, soil water, soil evaporation, snow and radiation process-es were used to build the model structure for simulating soil temperature and moisture dynamics.

In Paper II, the same modules in Paper I have been used, but also, modules of plant biotic, plant abiotic and soil carbon and ni-trogen processes have been used to model seasonal patterns of sensible and latent heat fluxes and NEE. Moreover, specific response

functions for temperature and moisture in air and soil were chosen to assess reductions of photosynthesis, transpiration, respiration due to temperature and moisture stresses.

A detailed technical description of the model was given by Jansson and Karlberg (2004). The specific equations are listed in the Ap-pendix section in Paper I and Paper II.

A number of 26 parameters were chosen for the GLUE calibration procedure (Paper I, Table 2). These parameters were highly relat-ed to soil heat and water processes. The prior ranges of the 26 parameters were set based on experiences and literature studies in this region or adopted from previous applications of the model (Alvenäs and Jansson, 1997; Zhang et al., 2007; Svensson et al., 2008 a).

In total, 46 parameters were selected for GLUE calibration (Paper II, Table 1). The prior uncertainty ranges of parameters were defined mainly based on previous studies (Mellander et al., 2006; Svensson et al., 2008 a; Norman et al., 2008; Svensson et al., 2008 b).

GLUE for model calibration and perfor-mance indices

A full documentation of the Generalized Likelihood Uncertainty Estimation (GLUE) could be found in Beven and Binley (1992) and Beven (2006). The number of behavioral models varied depended upon user-defined criteria applied. The behavioral models were considered as an ensemble to demonstrate an equally good ability to produce similar results, the so-called equifinality.

Because GLUE is flexible in the choice of performance index, the choice of perfor-mance index is important and it should re-flect the best understanding of possible er-rors in the observed data (Beven, 2006). Performance indices and acceptance criteria are different in the two papers and they are described below.

In Paper I, in order to investigate the effects of performance indices, three different per-formance indices were used, including coeffi-cient of determination (R2), mean error (ME) and root mean square error (RMSE).


4

A total of 25 000 simulations were performed using Monte Carlo random uniform sampling.

Four different criteria for model acceptance were proposed (Paper I, Table 4). Explicitly, C1 was composed by only soil temperature constraints. In C1, both R2 > 0.97 and -1 < ME < 1 for soil temperature were used to constrain the model performance. C2 was constituted by only soil moisture constraints. In C2, both absolute ME < 1/3 of total posterior absolute ME and RMSE < 1/3 of total posterior RMSE for soil water storage were applied to constrain the model perfor-mance. A combination of C1 and C2 as joint criteria was called C3. Because of the low number of behavioral models constrained by C3, the behavioral models probably caused over-parameterization problems and were not robust enough to represent the natural conditions in reality. Hence, the thresholds were changed until a reasonably high number of behavioral models were obtained. The corresponding combined criteria were named C4.

Among the 25 000 simulations, 266, 261, 6, and 112 simulations were identified as behav-ioral models constrained by the 4 criteria respectively.

In Paper II, a multi-criteria set was proposed with respect to Nash-Sutcliffe R2 and mean error (ME). Explicitly, Nash-Sutcliffe R2 should be above 0.4 for sensible heat flux, 0.2 for latent heat flux and 0.4 for NEE. Because no perfect enclosure on the energy balance from the eddy covariance measurements was avail-able, a criterion for the mean value was only constrained on the NEE corresponding to a threshold of ME, ± 0.15 g C m-2 day-1. Based on this multi-criteria, 333 out of 35 000 be-havioral models were identified.

RESULTS

Model performance on soil temperature and moisture (Paper I)

Soil temperature performance

Generally, for all 4 groups of simulations, the mean of simulated soil temperature at 20, 40 and 60 cm showed reasonably good perfor-mance with observations (R2 ≥ 0.94).

Fig. 1 demonstrates the upper and lower uncertainty boundaries and the mean of total accepted soil temperature at 40 cm, which were constrained by the 4 different criteria respectively, together with observed data.

When constrained by C3, behavioral models showed the narrowest uncertainty ranges during the 3 years (Fig. 1c) and a similar mean of simulated soil temperature as in Fig. 1a, where soil temperature was constrained by C1. The reason for this could be explained by the fact that the means and standard devi-ations (SDs) of parameters constrained by C3 were narrower than those constrained by other criteria. The largest uncertainty ranges for soil temperature were found in Fig. 1b when constrained by C2. However, in Fig. 1d, where soil temperature and moisture were constrained by C4, the results were more or less intermediate to the results presented in Fig. 1a and Fig. 1b. This was caused by more intermediate means and SDs of parameters constrained by C4, as compared with those by C1 and C2 respectively.

Soil moisture performance

Model performance with respect to moisture constrained by C2 showed a similar pattern as constrained by C3 and C4. Therefore, Fig. 2 only shows a comparison of soil water storage in 2 different soil depths, resulting from C1 and C2. Constrained by C1, the width of the uncertainty range of soil water storage from 0 - 45 cm during the 3 years (Fig. 2a) was one-and-a-half times the width of that constrained by C2 (Fig. 2b). Moreover, with the increase of soil depth to 260 cm, the width of uncertainty range of soil water stor-age in the final year constrained by C1 (Fig. 2c) was increased by more than 100 % com-paring with the width constrained by C2 (Fig. 2d). Constrained by C1 and C2 respectively, the lower uncertainty boundaries (the 5th percentile) presented more uncertainties than the upper uncertainty boundaries (the 95th percentile) with increasing soil depth and with time since start of measurements.


5

Figure 1 Model performance of soil temperature at 40 cm during 2005 – 2007, when con-strained by the 4 proposed criteria respectively. (a, b, c, and d) represent the model con-strained by C1, C2, C3 and C4 respectively. The upper and lower uncertainty boundaries (grey solid line), the mean of total accepted soil temperature constrained by corresponding criteria (black solid line), and measured soil temperature at 40cm (red dotted line) were plot-ted.

Figure 2 Model performance of soil water storage from 0-45 cm and from 0-260 cm during 2005 – 2007 respectively. (a and c) represent the model constrained by C1. (b and d) repre-sent the model constrained by C2. Black bar represents averaged 5-day precipitation. The upper and lower uncertainty boundaries (grey solid line), the mean of total accepted soil water storage constrained by corresponding criteria (black solid line), and calculated soil water storage (black triangle) were plotted.

So

il T

em

pe

ratu

re a

t 4

0 c

m /

oC

-30

-20

-10

0

10

20

30

(a)

-30

-20

-10

0

10

20

30

(b)

05-1-1 05-7-1 06-1-1 06-7-1 07-1-1 07-7-1 08-1-1

So

il T

em

pe

ratu

re a

t 4

0 c

m /

oC

-30

-20

-10

0

10

20

30

(c)

05-1-1 05-7-1 06-1-1 06-7-1 07-1-1 07-7-1 08-1-1

-30

-20

-10

0

10

20

30

(d)

Soi

l Wat

er S

tora

ge

(0-4

5 c

m)

/ m

m

90

140

190

240 0

50

100

150 90

140

190

240

Pre

pci

pita

tion / m

m 5

day

s-1

0

50

100

150

05-1-1 05-7-1 06-1-1 06-7-1 07-1-1 07-7-1 08-1-1

Soi

l Wat

er S

tora

ge

(0-2

60 c

m)

/ m

m

600

700

800

900

1000

1100

1200

05-1-1 05-7-1 06-1-1 06-7-1 07-1-1 07-7-1 08-1-1

600

700

800

900

1000

1100

1200

(a) (b)

(c) (d)


6

In 2007, climate conditions were warmer and drier than the conditions in 2005 and 2006. The mean of simulated soil water storage in the whole profile (0 - 260 cm) in 2007 was lower than that in 2005 and 2006, in response to the climate conditions. However, the cal-culated soil water storage in the same depth showed little response to climate conditions. This phenomenon could be partly explained by measurement uncertainties of the soil water content in the different layers. Fur-thermore, it could also be due to parameter uncertainties that led to underestimation of soil storage in the deeper layers.

Model performance on eddy covariance fluxes (Paper II)

Seasonal courses

In general, the performances of calibrated eddy covariance fluxes demonstrated good agreements for the behavioral simulations (Fig. 3). The obtained range of Nash-Sutcliffe

R2 was 0.41 - 0.71 for sensible heat flux (H), 0.20 - 0.54 for latent heat flux (LE), and 0.40 - 0.62 for NEE.

Seasonal courses of H and LE in 2007 and 2008 (Fig. 3) demonstrated quite high similar-ities. However, the uncertainty band of H was much narrower than that of LE, reflect-ing to the criteria used to constrain their performances. Simulated H showed a ten-dency to overestimate H in spring and early summer period in 2007, which was not ap-pearing in 2008. Additionally, a big drop of H was observed by eddy covariance meas-urements at the end of 2007 that was not in the simulated values. A small uncertainty in the simulated range was observed during winter period. Hence, it may suggest a small uncertainty in the aggregated mean values of eddy covariance measurements for that peri-od. Simulated LE reproduced the seasonality of observed LE successfully in both years. Nevertheless, a systematic overestimation of

Figure 3 Model performance on H, LE and day-time and night-time NEE based on 333 be-havioral models. The grey band is the 5% - 95% probability band; black solid line represents measured data; red dashed line represents the mean of accepeted simulated data.


7

LE was noticed during summer in both years, which has not shown in simulated H.

Simulated night-time NEE showed a good agreement with observed night-time NEE in the two years without any systematic devia-tion, although night time fluxes are lower than the corresponding day-time fluxes. Simulated day-time NEE captured the major dynamics of the two years, except for May and June in 2007. The uncertainty band of day-time NEE was larger than that of night-time NEE. Obviously this indicated the larger uncertainty in the model performance to describe photosynthesis than the corre-sponding respiration processes.

The timing of simulated carbon uptake in

spring was more precise in 2008 than that in 2007. Simulated performances of H, LE and NEE in 2008 has demonstrated a consistent agreement to measured data. The high agreements implied the semi-coupled stomata control model for photosynthesis and tran-spiration had an ability to predict the seasonal courses of H, LE and NEE. However, the spring of the first year did not show equal convincing indications that the model is of general applicability.

Diurnal courses

In order to better illustrate the seasonal pat-terns of H, LE and NEE, the diurnal courses of the three fluxes for February, April, May and September were plotted in Fig. 4. The

H (W m-2)

LE (W m-2) NEE (g C m-2 day-1)

0 6 12 18 24

0

100

200

0 6 12 18 24

0

100

200

February

0 6 12 18 24

-15

-10

-5

0

5

0 6 12 18 24

0

100

200

0 6 12 18 24

0

100

200

April

0 6 12 18 24

-15

-10

-5

0

5

0 6 12 18 24

0

100

200

0 6 12 18 24

0

100

200

May

0 6 12 18 24

-15

-10

-5

0

5

0 6 12 18 24

0

100

200

0 6 12 18 24

0

100

200

September

0 6 12 18 24

-15

-10

-5

0

5

Measured

Simulated

H Figure 4 Diurnal measured (solid line) and simulated (dashed line) courses of H, LE and NEE for February, April, May and September based on one behavioral model.


8

diurnal courses were the results from a be-havioral model being picked up randomly from the 333 behavioral models.

Simulated H showed systematic overestima-tions especially in the morning hours. This may be explained by the model assumption of no heat storage in the air beneath the reference height. Therefore, the model re-sponded more rapidly to changes in the gen-eration of sensible heat from the land surface compared with the reality.

A surprisingly good agreement was obtained for LE in all the four selected months, imply-ing a realistic description of transpiration regulations in the model or minor effects of vapor storage beneath the reference height. In April and May, day time values of H was two times as large as corresponding LE val-ues. In September H and LE were of the same magnitudes. Based on simulated net radiation, H and LE were the major compo-nents of the energy balance.

Diurnal courses of simulated NEE were systematically delayed with about 2 hours for the starting of carbon uptake compared with observed NEE in the four months. Especial-ly during May observed NEE responded faster than observed H and LE. However, the magnitudes of simulated NEE, except for May were similar as the observed data. There was no easy explanation to the dis-crepancies obtained between diurnal patterns for simulated and measured NEE fluxes. Diurnal discrepancies of NEE may be a problem both in the model and in the meas-urements. The monthly mean day approach was useful to reduce uncertainties from both measurements and simulations.

Response functions for environmental factors (Paper II)

Based on 333 behavioral models, the season-ality of response functions of photosynthesis and transpiration for temperature and mois-ture in air and soil is obtained (Fig. 5 and Fig. 6).

For photosynthesis regulations, air tempera-ture regulated the starting and ending of the growing season, whereas water stress origi-nating from soil governed the late spring and

summer period. In addition, nitrogen re-sponse showed a similar seasonal pattern with increased photosynthesis efficiency from spring to summer. Nitrogen response showed a relatively stable limitation during the whole year demonstrating an important major limiting factor for photosynthesis in this region. The water response on the pho-tosynthesis originated mostly from soil tem-perature effect on the water uptake.

For transpiration regulations, the atmospher-ic stress was demonstrated by a ratio of tran-spiration rate, which is defined by the poten-tial transpiration as calculated with the Lohammar equation divided by the transpira-tion assuming a maximal conductance. The ratio in both years was quiet low (Fig. 6). The wide range of the parameter, Conduct Max(1) (Paper II, Table 3), may have caused the low values for the atmospheric response as calcu-lated by the Lohammar equation. During winter or during conditions with low transpi-rations, the ratio was quite high but it did not have any important impact because of the low atmospheric demand. However during the growing season systematic patterns was identified. The second year showed more stable conditions with similar atmospheric stresses from May to September, whereas the first year showed a systematic increase of the transpiration efficiency from spring to sum-mer.

In a similar way the results showed a tenden-cy to reduce the water uptake due to the soil water stress in the beginning of growing seasons in both years (Fig. 6). Moreover, the reduction of potential transpiration to actual transpiration was mainly governed by the soil temperature reduction function (dashed line (Fig. 6) as the combined effect of both soil temperature and moisture). In both years a systematic increase of water uptake efficiency occurred from spring to summer.


9

DISCUSSION

GLUE as a tool to increase the under-standing of modeling processes

It is important to notice that ensembles of behavioral models have been obtained based on different criteria in both papers. These

Figure 5 Simulated daily response functions for photosysthesis processes. The different responses originate from air temperature (upper), transpiration (middel), and nitrogen (lower). The grey band is the min-max response band based on 333 behavioral models; black soild line represents the mean of 333 behavioral individual responses; dashed line represents the multiplicative mean by multiplying 3 mean values of behavioral individual responses.

Figure 6 Simulated daily reduction ratio by atmospheric stresses using the Lohammar equation (upper). Simulated daily response functions for transpiration processes. The different responses originate from soil temperature (middel), soil moisture (lower). The meanings of symbols are analogous as that in Figure 5.


10

ensembles demonstrated good agreements on measured data. Moreover, the posterior pa-rameter distribution functions were changing into either normal or log-normal distributions in most cases, as compared to the prior uni-form distribution. This indicates that the parameter uncertainties could be reduced successfully by applying different criteria. Further, according to reliable behavioral simulations, uncertainties in measurements were identified. For instance, in Fig. 3, a big drop of H was observed by eddy covariance at the end of 2007, which was not simulated by the Coupmodel. This suggests that it might be an error in eddy covariance meas-urements. Hence, using the GLUE method, not only the parametric uncertainties were reduced, but the measurement uncertainties were distinguished also, which helps model-ers to communicate with experimental scien-tists to increase the understanding of model-ing processes.

The ensembles of behavioral models also provide equally-good parameter sets (equifi-nalities), which give us more reliable explana-tions as compared with a single optimal simu-lation, e.g., using a conventional stepwise manual calibration.

However, the flexibility of choosing perfor-mance index and defining criteria causes different model performances in response to different criteria. For instance, in Paper I, the four criteria were proposed to assess the possible performances. For a single con-straint, the coefficient of determination (R2) was more efficient in discarding models with poor performance regarding soil temperature, in contrast to the mean error (ME) that pro-duced a relatively low efficiency in the rejec-tion of poor performance on soil tempera-ture. Similarly, the root mean square error (RMSE) was more efficient in rejecting mod-els with poor performance regarding soil water storage as compared with ME. The combined constraints normally increased the number of rejected simulations substantially. Therefore, the multi-objective criteria could considerably reduce the parametric uncertain-ty, which was also concluded by Franks et al. (1997) when they applied a distributed hydro-logical model.

Process-oriented modeling to understand complex systems

CoupModel is a process-oriented model coupled with water, heat, carbon and nitro-gen processes. Since the processes are cou-pled together, it is worth understanding the responses and feedbacks between the pro-cesses. In Paper I, a fairly simply model structure was defined, including soil heat and water processes as major components. How-ever, soil moisture performance was im-proved when only constraining the model by using the soil temperature constraints. In this case the physically correct models enabled us to make use of one type of measurements to describe also other related phenomena. In Paper II, a more complex model structure was used compared with that in Paper I. Behavioral models were retained only by H, LE and NEE constraints. At the same time, the performance of other measurements such as soil chamber CO2 measurements, soil temperature, soil moisture and snow depth was improved. By analyzing posterior param-eter performance, the impacts of parameters to other unconstrained processes were ob-tained even if the coupling is very complex for the biological processes.

High resolution monitoring data combined with low resolution data provide an oppor-tunity to understanding the crucial processes deeply, e.g., photosynthesis, evapotranspira-tion, respiration, etc. By applying these meas-urements the complexity of the model struc-ture will indeed increase and it will be more difficult to identify the correct and most appropriate holistic model. There is an open issue to find the balance between the com-plexity of the model and the quality of data.

Generally, simple model structure has a high-er chance to reproduce better performance than the complex coupled model structure does. For example, ICBM is a carbon balance model without connection to water, which was used by Juston et al. (2010). They achieved impressive simulations by using the GLUE method, but they have no chance to understand the relation between carbon and water from such a simple model. The pro-cess-oriented models give perspectives to interpret the performance by different ways,


11

even though the compromise performance for different calibrated variables was usually obtained.

Response functions for environmental factors

In Paper II the regulations of temperature and moisture in air and soil were emphasized. As displayed in Fig. 5 and Fig. 6, the study demonstrated how the separate response for photosynthesis and decomposition could be estimated based on available data and as-sumed regulating functions. The problem was to a large extent the correlation between soil and plant processes. For example, the nitrogen response on photosynthesis was strongly correlated with the corresponding water uptake response on photosynthesis in the model. Hence, it was not possible to precisely distinguish the separate importance of these factors with any significance.

Based on the GLUE method, ensembles of seasonal responses were produced. Due to different parameter uncertainties within dif-ferent processes, different widths of uncer-tainties bands for responses were calculated as well. Comparing to the ranges of uncer-tainties bands of water and nitrogen respons-es, the uncertainty range of air temperature response was much smaller, suggesting that the uncertainties of calibrated parameters for the air temperature response function were probably reduced substantially.

Previous studies of Scots pine forest in Hyytiälä have mainly been focusing on air temperature as an adaption factor but in our study we have indicated that the same adap-tion can be related to water uptake and soil temperatures.

CONCLUSIONS

The main objective of this thesis was to im-prove the understanding of the seasonal courses of carbon, evapotranspiration and heat processes, especially under cold climate conditions by merging of mathematical mod-els and data from experimental studies. The modeling approach provides a basic under-

standing on different biological and physical processes. The GLUE method was a useful and applicable tool to quantify the uncertain-ties in parameters, model assumptions and measurements based on user-defined criteria.

The seasonal patterns of carbon, evapotran-spiration and heat processes were successful-ly simulated but with a number of equifinali-ties. This demonstrated that it was not possible to identify the unique explanations to the observed phenomena. However a number of clear tendencies of responses for environmental factors were demonstrated showing the importance of considering the coupling between various components of the ecosystem. Both the evapotranspiration and the NEE flux showed a response to soil temperature conditions via different direct and indirect ecosystem mechanisms.

High resolution data such as the eddy covari-ance technique was very useful to reduce the parameter uncertainty but the study also indicated that we do need more data to pre-cisely demonstrate the general validity of the behavioral simulations.

FUTURE WORK

The study of seasonality of carbon, evapo-transpiration and heat processes especially under cold climate conditions is good starting point for a further understanding of the impacts of climate variability and climate change on carbon, water and heat dynamics in regions at higher latitudes. The role of nitrogen, soil temperature or air temperature as acclimation factors on photosynthesis in a boreal Scots Pine ecosystem is one of the interesting topics for continued research. Further research will test a number of more optional models on regulation of NEE, also based on data from new sites. It will for example be of high interesting to test new acclimation function for photosynthesis based on long-term response, e.g., the re-sponse as a function of growing temperature sum.


12

REFERENCES

Alvenäs G and Jansson P-E, 1997. Model for evaporation, moisture and temperature of bare soil: calibration and sensitivity analysis. Agricultural and Forest Meteorology, 88: 47-56.

Bauer J, Herbst M, Huisman JA, Weihermüller L and Vereecken H, 2008. Sensitivity of simulated soil heterotrophic respiration to temperature and moisture functions. Geoderma, 145:17-27.

Bergh J, McMurtrie RE and Linder S, 1998. Climate factors controlling the productivity of Nor-way spruce: A model-based analysis. Forest Ecology and Management, 110: 127-139.

Beven KJ and Binley A. 1992. The future of distributed models-model calibration and uncertainty prediction. Hydrological process, 6: 279-298.

Beven KJ, 2006. A manifesto for the equifinality thesis. Journal of hydrology, 320: 18-36.

Challinor AJ, Ewert F, Arnold S, Simelton E and Fraser E, 2009. Crops and climate change: progress, trends, and challenges in simulation impacts and informing adaptation. Journal of Experimental Botany, 60: 2775-2789.

Cheng HY, Wang GX, Hu HC and Wang YB, 2008. The variation of soil temperature and water content of seasonal frozen soil with different vegetation coverage in the headwater region of the Yellow River, China. Environmental Geology, 54: 1755-1762.

Eckersten H, Jansson P-E and Johnsson H, 1998. SOILN model, ver.9.2, User’s manual. Swedish Agricultural University, Uppsala. 113 pp.

Franks SW, Beven KJ, Chappell NA and Gineste P, 1997. The utility of multi-objective condi-tioning of a distributed hydrological model using uncertain estimates of saturated areas. Proceedings of the International Congress on Modelling and Simulation (MODSIM '97), Hobart, 8-11 December 1997, 1, 335-340.

Gustafsson D, Lewan E and Jansson P-E, 2004. Modeling water and heat balance of the boreal landscape - comparison of forest and arable land in Scandinavia. Journal of applied meteorology, 43: 1750-1767.

Hall J and Anderson M, 2002. Handling uncertainty in extreme or unrepeatable hydrological processes - the need for an alternative paradigm. Hydrological Processes, 16(9): 1867-1870.

Hampton SE, Izmesteva LR, Moore MV, Katz SL, Dennis B and Silow EA, 2008. Sixty years of environmental change in the world’s largest freshwater lake - Lake Baikal, Siberia. Global Change Biology, 14, 1947-1958.

Hansson K, Simunek J, Mizoguchi M, Lundin L-C and Genuchten MT, 2004. Water flow and heat transport in frozen soil: numerical solution and freeze-thaw applications. Vadose zone journal, 3: 693-704.

Hu HP, Ye BS, Zhou YH and Tian FQ, 2006. A land surface model incorporated with soil freeze-thaw and its application in GAME/Tibet. Science in China Series D: Earth Science, 49: 1311-1322.

Högberg P, Bhupinderpal-Singh, Löfvenius MO and Nordgren A, 2009. Partitioning of soil res-piration into its autotrophic and heterotrophic components by means of tree-gridling in old boreal spruce forest. Forest Ecology and Management, 257: 1764-1767.

IPCC, 2007. Climate change 2007: Synthesis Report. Cambridge University, United Kingdom. Jansson P-E and Halldin S, 1979. Model for the annual water and energy flow in a layered soil.

In: Halldin, S. (Ed.), Comparison of Forest and Energy Exchange Models. Society for Ecological Modelling, Copenhagen. 145-163.

Jansson P-E and Karlberg L, 2004. Coupled Heat and Mass Transfer Model for Soil-Plant - At-mosphere Systems. Royal Institute of Technology. Stockholm. 321 pp.

Jansson P-E and Moon DS, 2001. A coupled model of water, heat and mass transfer using ob-ject-orientation to improve flexibility and functionality. Environmental Modelling and Soft-ware, 16: 37-46.


13

Juston J, Andrén O, Kätterer T and Jansson P-E, 2010. Uncertainty analyses for calibrating a soil carbon balance model to agricultural field trial data in Sweden and Kenya. Ecological Model-ling, 221(16): 1880-1888.

Karmer K, Leinonen I, Bartelink HH, Berbigier P, Borghetti M, Bernhofer Ch, Cienciala E, Dolman AJ, Froer O, Gracia CA, Granier A, Grünwald T, Hari P, Jans W, Kellomäki S, Loustau D, Magnani F, Markkanen T, Matteucci G, Mohren GMJ, Moors E, Nissinen A, Peltola H, Sabaté S, Sanchez A, Sontag M, Valentini R and Vesala T, 2002. Evaluation of six process-based forest growth models using eddy-covariance measurements of CO2 and H2O fluxes at six forest sites in Europe. Global Change Biology, 8: 213-230.

Kettridge N and Baird A, 2008. Modelling soil temperatures in northern peatlands. European Journal of Soil Science, 59, 327-338.

Kolari P, Lappalainen HK, Hänninen H and Hari P, 2007. Relationship between temperature and the seasonal course of photosynthesis in Scots pine at northern timberline and southern boreal zone. Tellus, 59B: 542-552.

Magnani F, Nolè A, Ripullone F and Grace J, 2009. Growth patterns of Pinussylvestris across Europe: a functional analysis using the HYDRALL model. Journal of Biogeosciences and For-estry, 2: 162-171.

Markkanen T, Rannik Ü, Keronen P, Suni T, and Vesala T, 2001. Eddy covariance fluxes over a boreal Scots pine forest. Boreal Environment Research, 6: 65-78.

Mellander P-E, Stähli M, Gustafsson D and Bishop K, 2006. Modelling the effect of low soil temperatures on transpiration by Scots pine. Hydrological Processes, 20: 1929-1944.

Mäkelä A, Hari P, Berninger F, Hänninen H and Nikinmaa E, 2004. Acclimation of photosynthe-sis capacity in Scots pine to the annual cycle temperature. Tree Physiology, 24: 369-378.

Norman J, Jansson P-E, Farahbakhshazad N, Butterbach-Bahl K, Li C and Klemedtsson L, 2008. Simulation of NO and N2O emissions from a spruce forest during a freeze/thaw event using an N-flux submodel from the PnET-N-DNDC model integrated to CoupModel. Ecological Modelling, 216: 18-30.

Nyberg L, Stahli M, Mellander P-E and Bishop K H, 2001. Soil frost effects on soil water and runoff dynamics along a boreal forest transect: 1. Field investigations. Hydrological Process, 15: 909-926.

Schwarz PA, Fahey TJ and Dawson TE, 1997. Seasonal air and soul temperature effects on pho-tosynthesis in red spruce (Picearubens) saplings. Tree Physiology, 17: 187-194.

Svensson M, Jansson P-E, Gustafsson D, Kleja DB, Langvall O and Lindroth A, 2008 a. Bayesi-an calibration of a model describing carbon, water and heat fluxes for a Swedish boreal forest stand. Ecological Modelling, 213: 331-344.

Svensson M, Jansson P-E and Kleja DB, 2008 b. Modelling soil C sequestration in spruce forest ecosystems along a Swedish transect based on current conditions. Biogeochemistry, 89 (1): 95-119.

Tanja S, Berninger F, Vesala T, Markkanen T, Hari P, Mäkelä A, Ilvesniemi H, Hänninen H, Nikinmaa E, Huttula T, Laurila T, Aurela M, Grelle A, Lindroth A, Arneth A, Shibistova O and Lloyd J, 2003. Air temperature triggers the recovery of evergreen boreal forest photosynthesis in spring. Global change biology, 9: 1410-1426.

Van Oijen M, Rougier J and Simth R, 2005. Bayesian calibration of process-based forest models: bridging the gap between models and data. Tree Physiology, 25: 915-927.

Vrugt JA, Gupta HV, Bouten W and Sorooshian S, 2003. A Shuffled Complex Evolution Me-tropolis algorithm for optimization and uncertainty assessment of hydrologic model pa-rameters. Water Resources Research, 39(8): 1201.

Zhang KL, Peng WY, Wang L, Fu PA and Xu XL. 2007. Variation of soil temperature and soil moisture on black soil profile in seasonal frozen area of Northeast China (in Chinese). Geographical research, 26: 314-320.

sihong wu - diva portalkth.diva-portal.org/smash/get/diva2:372901/fulltext01.pdfseasonal courses of...

Documents