Modeling Deforestation Risks for the Maya Biosphere Reserve, Guatemala

1
Modeling Deforestation Risks for the Maya
Biosphere Reserve, Guatemala

• Wolfgang Grunberg

School of Renewable Natural Resource Sciences The
University of Arizona Tucson, Arizona, 85721,
USA July 14, 2000
2
Acknowledgement

• The author would like to thank the following
  organizations and individuals for their
  indispensable help
• ART Group - The University of Arizona
• CARE Guatemala
• CONAP - CEMEC
• CI - ProPeten
• WCS - Gainesville
• Perfecto Carillo, Teresita Chinchilla, Gary
  Christopherson, Reno Fiedler, Georg Grünberg, D.
  Phillip Guertin, Vinicio Montero, Howard R.
  Gimblett, Gustavo Rodriguez Ortiz, Marco Antonio
  Palacios, Victor Hugo Ramos, Steven Sader,
  Claudio Saito, Norman Schwartz, William W. Shaw,
  Carlos Soza, Laura Stewart, and Craig Wissler

3
Overview

• The Maya Biosphere Reserve (MBR)
• Landscape, People, Deforestation
• Methods Results
• Data - Types, Sources, and Accuracy
• Spatial Analysis
• Roads, Settlements, Soil Results
• Deforestation Probability Surface
• 1986-99 Deforestation Probability Results
• Forecasting Deforestation
• 1999 Deforestation Forecast Results
• 2001 Deforestation Scenario Results
• Discussion
• Deforestation Model
• Future Improvements
• Conclusions

4
Guatemala, Central America

• Area 108,890 km2
• Climate Tropical hot, humid in lowlands cooler
  in highlands
• Terrain Mostly mountains with narrow coastal
  plains and rolling limestone plateau (Peten)
• Population 12,300,000 (2.68 growth rate)
• Ethnic Groups
• 56 Ladino (Mestizo)
• 44 Mayas and other indigenous Peoples
• Literacy 55.6
• Labor Force
• Agriculture 58
• Services 14
• Manufacturing 14
• Commerce 7
• Construction 4
• Other 3

According to the CIA World Factbook 1999
5
The MBR and its Buffer Zones (ZAM and ZUM)

• Founded 1990
• 21,130 km2 Reserve and Buffer Zone
• Hilly Limestone Carst Landscape
• 100-300 m Elevation
• 25 C Mean Annual Temperature
• 1600 mm Yearly Precipitation Average
• Predominantly Tropical Lowland Forest

6
The Agricultural Frontier
7
Slash and Burn
8
Road Construction
9
Oil Pipeline and Ranching
10
The Peoples and their Primary Occupation

• Itza Maya - Majority in 1 Settlement
• Native Mayan population
• Swidden Agriculture (Corn), Agroforestry, Forest
  Products
• Ladino Petenero - Majority in 6 Settlements
• Non-Immigrant Population of Hispanic Descent
• Wage Labor, Swidden Agriculture, Agroforestry
• Highland Mayas - Majority in 25 Settlements
• Recent Immigrants from Guatemalas Central
  Highlands
• Swidden Agriculture
• Ladino Sureño - Majority in 134 Settlements
• Recent Immigrants of Hispanic and Mayan Descent
• Swidden Agriculture and Ranching

11
Maya House with Corn Field
12
Ladino House along a Road
13
Methods - Data Used and Their Sources

• 1986, 90, 93, 95, 97, and 99 Forest Change
  Detection Images based on NDVI analysis of 30 m
  resolution TM Landsat Images
• Maine Image Analysis Laboratory, University of
  Maine
• 1200,000 Soil Map, reclassified for agricultural
  suitability
• CONAP and FAO
• 194 Settlement locations and associated
  socio-economic data from 1820 to 1999
• CARE Guatemala and CEMEC-CONAP
• Roads and associated attributes
• CEMEC-CONAP, WCS-Gainesville, and SEGEPLAN
• Administrative boundaries
• CEMEC-CONAP and WCS-Gainesville
• The Vector and Raster Themes have a Root Mean
  Square Error of 400 Meter to Each Other

14
Methods - Spatial Analysis

• Settlement Points Analysis
• 20 concentric 1 km buffers per settlement and
  analysis year
• Averaged deforestation distance decay curves
  according to socio-economic categories
• Soil Polygons Analysis
• Reclassification according to agricultural
  suitability
• deforestation per soil category and analysis
  year
• Road Lines Analysis
• Only perennial roads were included in the models
• The entire area is assumed to be easily
  penetrated on foot, with mules, or with
  4-wheel-drive vehicles
• Perennial roads, however, are significant for
  market access and public transportation

15
Buffering the El Naranjo Settlement

• Founded 1981 Ladino Sureño Majority in
  Transition from Agriculture to Ranching 3500
  Inhabitants in 1996

16
Perennial Road
17
Results - Deforestation Distance Decay
Curves According to the Settlements Primary
Occupation

• Exclusion of Wage Labor Settlements from the
  Model due to Minimal Deforestation Impact

18
Results - Deforestation and Agricultural
Soil Suitability

• More Deforestation on Well Draining Soils than on
  Poorly Draining Soils

19
Methods - Deforestation Probability Surface

• Cell by Cell Logistic Regression for Each
  Analysis Year (1986 to 1999) using 5 Stratified
  Random Samples (gt 1,100,000 cells)
• Dependent Variable Deforested (1) / Forested
  (0)
• Independent Variables LN distance to Roads,
  LN Distance to Settlements, Well (1) / Poorly
  (0) Draining Soils

20
1986 - Deforestation Probability Surface
Results 1986
Observed Deforestation
21
1990 - Deforestation Probability Surface
Results 1990
Observed Deforestation
22
1993 - Deforestation Probability Surface
Results 1993
Observed Deforestation
23
1995 - Deforestation Probability Surface
Results 1995
Observed Deforestation
24
1997 - Deforestation Probability Surface
Results 1997
Observed Deforestation
25
1999 - Deforestation Probability Surface
Results 1999
Observed Deforestation Man Caused Wild Fires
(Summer 1998)
26
Methods - Forecasting Deforestation

• Forecasting Deforestation for 1999
• Forecasted Deforestation Probability Surface
  based on
• 1997s probability surface regression
  coefficients
• Roads and settlements observed in 1999
• Deforestation Forecast based on
• Percent deforestation in 1997s deforestation
  probability zones
• Comparing 1999 Observed and Forecasted
  Deforestation
• The 2001 Deforestation Scenario
• Forecasted Deforestation Probability Surface
  based on
• 1999s probability surface regression
  coefficients
• 2001 roads scenario
• Deforestation Forecast based on
• Percent deforestation in 1999s deforestation
  probability zones

27
Results - The Forecasted 1999 Deforestation Proba
bility Surface

• The 1999 Forecast is based on the 1997 Regression
  Coefficients and in 1999 Observed Roads and
  Settlements

28
Results - Forecasting Percent Area Deforested

• The 1997 and 1999 Observed Probability Zone
  Deforestation Percentages were used respectively
  for the 1999 Deforestation Projection and 2001
  Scenario

29
Results - 1999 Deforestation Forecast

• 1999 Forecasted Deforestation for Each
  Probability Zone Area of Forecasted 1999
  Deforestation Probability Zone x of Zone
  Deforested in 1997
• 1999 Observed vs Predicted Deforestation

30
Results - Testing the 1999 Deforestation Forecast

• Difference between 1999 Predicted and Observed
  Deforestation

31
Results - The Forecasted 2001 Deforestation
Probability Surface

• The 2001 Forecast is based on the 1999 Regression
  Coefficients and a 2001 Roads Scenario

32
Results - The 2001 Scenario

• 2001 Predicted Deforestation vs Observed
  Deforestation

33
Results - The 2001 Scenario Continued

• The 2001 Scenario forecasts an increase in
  deforestation since of 14.5 (295 km2) since
  1999

34
Discussion - The Models

• Pros
• Can be used to estimate impacts of new roads and
  settlements in scenarios
• Simple model with relatively good results
• Uses common spatial features such as roads,
  settlement points, and simple soil maps
• Cons
• Does not account for spatial and temporal
  autocorrelation
• Does not account for road and settlement age
• Does not predict deforestation location
• Forecasting beyond 2 years is questionable due to
  changing deforestation trends

35
Discussion - Room for Improvements

• Age of Road and Settlement Factor needs to be
  included
• Spatial and temporal autocorrelation need to be
  addressed
• Differentiate settlement deforestation impacts
  according to their socio-economic qualities
• River traffic and oil-pipelines need to be
  considered as access ways
• Water availability for ranching and agriculture
  could be included
• Slope and aspect data of adequate resolution in
  combination with better soil maps may turn this
  regional model into a more localized version

36
Discussion - Mostly Obvious Conclusions and
Suggestions

• Clear relationship between the presence of roads
  and settlements deforestation
• Simplicity of model is advantageous for
  forecasting deforestation in agricultural
  frontiers on a regional scale
• Suggestions for reducing deforestation risks
• Control access to roads
• Avoid building perennial roads or upgrading
  existing intermittent roads to a perennial status
• Pipelines and rivers need to be considered as
  possible access routes
• Avoid any new settlements in low deforestation
  risk areas
• Consider supporting a forestry or wage-labor
  based economy
• In an agricultural frontier, regional
  deforestation trends are not only controlled by
  access but also by soil quality

37
Thank You for Your Participation Contact
InformationWolfgang Grunberg School of
Renewable Natural Resources, The University of
Arizona, Tucson, AZ 85721, USA Phone 1 (520)
621 3045 e-mail wolfgang_at_srnr.arizona.edu