Space versus Place in Complex Human-Natural Systems: Spatial and Multi-level Models of Tropical Land Use and Cover Change (LUCC) in Guatemala

David López-Carra Jason Davisa, Marta Jankowskab, Laura Grantc, Anna Carla López-Carrb, T Mitchell Aided, Matthew Clarkee

a Department of Geography, University of California, Santa Barbara, Human-Environment Dynamics Lab, 4836 Ellison Hall UC Santa Barbara (UCSB)Santa Barbara, CA 93106-4060 b Department of Geography, San Diego State University, 5500 Campanile Drive, San Diego, CA 92182c Department of Economics and Bren School of the Environment, UC Santa Barbara (UCSB)Santa Barbara, CA d Department of Biology, University of Puerto Rico, San Juan, PR 00931-3360e Department of Geography, Sonoma State University, Stevenson Hall 3066, 1801 E. Cotati Avenue, Rohnert Park, CA 94928

American Association for the Advancement of Science (AAAS) Annual Meeting: Mapping and Disentangling Human Decisions In Complex Human-Nature Systems Friday, February 18, 2011: 8:30 AM-11:30 AM140B (Washington Convention Center)

World Population Dynamics - 2 Big Trends: Urbanization & Aging

World Population: Average Annual Increase For Each Decade, 1750-2100 (projected) Source: "World

Population in Transition", Population Bulletin, by Thomas W. Merrick and PRB

0

20

40

60

80

100

1750 1800 1850 1900 1950 2000 2050 2100

Iin

cre

as

e (

mil

lio

ns

) Developing Regions

Developed Regions

1994 1996 1998 2000 2002 2004 2006 2008 2010Glo

ba

l p

ig a

nd

po

ultr

y p

rod

uctio

n

(to

nn

es x

106)

130

140

150

160

170

180

190

Year

1994 1996 1998 2000 2002 2004 2006 2008 2010

So

uth

Am

eri

ca

n s

oyb

ea

n

pro

du

ctio

n (

ha

x 1

06)

15

20

25

30

35

40

45

2001-2009163,326 km2

Aide, Clarke, Lopez-Carr, et al. (2010)..

Increasingly the driver is global urban consumption

NSF CHH Project: Latin America & Caribbean Demographic dynamics and LUCC

Total population change (1990 – 2000)

1990 – 426,465,0212000 – 503,388,073

diff 76,923,052 Municipality level change# of municipalities: 16,052# with negative growth : 4,200% with negative growth: 26.1%Year

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Pop

ula

tion

x 1

06

100

200

300

400

500

Rural Urban

Data from FAOSTATS Our analyses

Aide, Clarke, Lopez-Carr,, Levy, Grau, et al. (2010). Under review at Science. Global Land Project Open Science Meeting. Arizona State University. 17-19 October.

Tro

p an

d S

ubtr

op M

oist

Bro

ad F

ores

t

Tro

p an

d S

ubtr

op D

ry B

road

For

est

Tro

p an

d S

ubtr

op C

onife

r F

ores

t

Tem

p B

road

and

Mix

ed F

ores

t

Tro

p an

d S

ubtr

op G

rass

Sav

anna

Shr

ub

Tem

p G

rass

Sav

anna

and

Shr

ub

Flo

oded

Gra

ss a

nd S

avan

na

Mon

tane

Gra

ss a

nd S

hrub

Med

For

est W

oodl

and

Shr

ub

Des

ert a

nd X

eric

Shr

ub

Man

grov

e

Per

cent

ann

ual w

oody

cha

nge

(20

01 -

200

9)

-4

-3

-2

-1

0

1

2

3

4

Pantanal

Pampas

Mangrove

What municipalities are gaining and losing woody forest cover?

Aide, Clarke, Grau, Levy, Lopez-Carr, et al. (2010)..

What is the relationship between population change and woody vegetation change? Nada at the municipal level!

Population change (1990-2000)

-20000 0 20000 40000 60000 80000 100000

Wo

od

y ve

ge

tatio

n c

ha

ng

e 2

00

1-2

00

9

(r -

co

rre

latio

n c

oe

ffici

en

t)

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

ns

p <0.1

p <0.1

n = 16,052, F = 0.21, P = 0.64, r2 = 0.000

Aide, Clarke, Lopez-Carr, et al. (2010).

So what is driving forest conversion?...

Two Latin Americas: 78% Urban but…• Argentina/Chile/Uruguay – 90% Urban• Guatemala, Ecuador, Bolivia – 50% Urban• And within these countries there is VAST variation

• These two Latin Americas are associated with two distinct deforestation Pathways:

• Pathway 1: High population growth, rural-rural migration, low technology, low yields, poverty, subsistence.

• Pathway 2: Falling population growth, urbanization, increased meat consumption, high yield, increasing affluence, high technology, export agriculture.

Macro-Scale demographic, political-economic, social, and ecological dynamics

Urban or Internation

al Destination

s

Rural Destinati

on

Agricultural Extensificati

on

Agricultural Intensificatio

n

Return to Top of Chart

Migration

Fertility regulation

Off-farm Labor

Household Responses

Local Variation

Land Management

Proximate and Underlying CausesWhy poverty-driven deforestation WITH rapid urbanization?

Disproportionate Scale Problem

Other response??

cropscropsforestforest

FARMFARM UNOCCUPIED UNOCCUPIED FORESTFOREST

SURROUNDING FARMSSURROUNDING FARMS

Poverty-driven forest conversion tends to target unoccupied forestland, the external frontier. Commercial agriculture often follows land consolidation and thus may or may not be converting old growth forest (internal frontier)

Internal (place) versus external (space) forest frontiers.

Methods: Data Sources

• 2000 Guatemalan Living Standards Measurement Survey– Independent variables: household fertilizer and tractor use

• 2003 Guatemala National Agriculture Census– Independent variables: 2003 percent land area in fallow and

crop yields for coffee, sugar, white corn, yellow corn

• 2001-2009 Forest Cover Change database for all of Latin America (funded by NSF)– Dependent Variables: percent woody cover in 2009 and percent

change in woody cover from 2001 to 2009– Independent variables: 1990 and 2000 population density

Methods: Variables

Dependent Variables = woody vegetation in 2009 in model 1, and the percentage change in municipal woody vegetation from year 2001 to 2009

Guatemala LUCC Multilevel Model

• i represent municipalities within jth departments • β1 is the intercept along with its independent error term

• β2 through βp are regression coefficients with corresponding explanatory variables x2ij through xpij

• εij represent an independent error term for x2ij through xpij

j

Guatemala LUCC GW Model

• Y i = woody vegetation in 2009 in model 1, and the percentage change in municipal woody vegetation from year 2000 to 2009 in model 2.

• β1 through βk are regression parameter estimates with

• corresponding explanatory variables xi1 through xik with independent error term

• The weighting function is based on distance, resulting in locations closer to the estimated point having more influence on the projected value than locations farther away.

i

iikk

ikii xy 1

0

Guatemala LUCC Multilevel Model Structure

(Municipalities within Departments)

Forest Cover Population

Results: OLS for Woody Vegetation Change 2001-2009: Intensive: Fertilizer vs

Extensive TractorsVariables Coefficient Standard Error

Population Density 2000 (persons/km^2)

-0.000 0.000

Percentage Population Density Change from 1990 to 2000

0.055 0.191

Percentage of Households Using Fertilizer

0.569* 0.182

Percentage of Households Owning a Tractor

-1.508+ 0.808

Café (kg/ha) 0.000 0.000

Sugar (kg/ha) -0.000 0.000

White Corn (kg/ha) -0.000 0.001

Yellow Corn (kg/ha) 0.000 0.001

Percentage of Land in Fallow 0.521 0.584

Model R2 = .043, Adjusted R2 = .016

Modeling Results: OLS for Woody Vegetation Cover 2009:

Frontiervs. Settled and UrbanVariables Coefficient Standard Error

Population Density 2000 (persons/km^2)

-0.000* 0.000

Percentage Population Density Change from 1990 to 2000

0.388*** 0.115

Percentage Population Density Change from 1990 to 2000^2

-0.089** 0.032

Percentage of Households Using Fertilizer

-0.064* 0.032

Percentage of Households Owning a Tractor

-0.449** 0.143

Café (kg/ha) 0.000*** 0.000Sugar (kg/ha) -0.000* 0.000White Corn (kg/ha) -0.000** 0.000Yellow Corn (kg/ha) 0.000 0.000Percentage of Land in Fallow -0.538*** 0.104

Model R2 = .274, Adjusted R2 = .251

Results: LUCC Multilevel Model. Woody Cover in 2009 is near PAs and remote areas. Less forest in settled rural areas

Spatial Modeling Results: Moran’s I for Woody Vegetation Change

2001-2009Variable Moran’s I Z Score P Value Autocorrelation

Change in Woody Vegetation 2001 - 2009

.024 3.772 < .0001 Strongly Clustered

Woody Vegetation 2009 .094 12.586 < .0001 Strongly Clustered

OLS Change 2000 – 2009 Standardized Residuals

.010 1.780 .072 Clustered

OLS 2009 Standardized Residuals

.070 9.421 < .0001 Strongly Clustered

GWR Change 2000 – 2009 Standardized Residuals

.005 1.159 .246 Random

GWR 2009 Standardized Residuals

.003 .715 .474 Random

Figure 1. Getis-Ord Gi* maps for woody vegetation in 2009 and change in woody vegetation 2001-2009 with hot spots in red and cold spots in blue.

Figure 2. Coefficient estimates for percent households using fertilizer for the percent woody cover in 2009 model (left) and percent change in woody cover from 2001 to 2009 model (right).

Figure 4. Coefficient estimates for percent population density change from 1990 to 2000 (left), and percent of land in fallow (right) for the percent woody cover in 2009 model.

Figure 3. Coefficient estimates for percent households owning tractors for the percent woody cover in 2009 model (left) and percent change in woody cover from 20001to 2009 model (right).

Conclusionss

• More variation was found internally within model types than between multi-level and spatial models.

• However, the interpretation and utility of the results may be notably distinct in a Geographically Weighted Regression (GWR) versus a multi-level model.

Conclusions• Ordinary Least Squares (OLS) regression

suggests that population increase and density, agricultural intensification in the form of fertilizers and tractors, and higher crop production for sugar cane and white corn are negatively associated with forest cover in 2009 while coffee production is associated with higher forest cover.

Conclusion• Examining forest change during the first decade

of the 2000s for Guatemala, we observe that areas that increasingly relied upon mechanized equipment and/or fertilizers have more thoroughly captured and put into production available agricultural land.

Conclusions• What did GWR and Mlevel regressions tell us

beyond these findings?

• The multi-level model suggests significant differences exist at the municipal and departmental levels and indicates maintains a positive relationship between coffee production and forest cover at both levels of analysis.

• The GWR indicates where these association of changes are most salient. A clear trend emerges: The southwest to northeast gradient of decreasing population density, higher but decreasing forest cover, and lower but increasing technological inputs is particularly illuminated by the GWR.

Conclusions• Why do we care?

• Space is important and WHERE things happen is crucial for policy and management.

• Coupled human-natural systems take home message: The debate in geography and cognate sciences over the importance of space vs. place is often framed by qualitative vs. quantitative research. It need not be so. Place in coupled human-natural systems can be quantified and measured.

El fin. Gracias!

Sunset or Sunrise over Guatemala’s Forests?

Which way is LUCC heading?Which way is LUCC heading?