Spatial Group Mapping and Modeling

1
Spatial Group Mapping and Modeling

• Progress Summary
• NSF Meeting, 13-15 Jan 2006
• Kunming, China
• John Vogler

2

• Overview
• Large-scale Mapping
• Large-scale Modeling
• Fuzzy Cognitive Mapping
• MTCLIM
• Small-scale Modeling
• What next?

3
(No Transcript)
4
(No Transcript)
5
Large-scale Mapping
New Datasets - Thailand

• Landsat ETM image acquired 29 Feb 2004
• Aerial Photographs (via Alan Zieglar)
• PKEW jan1954, dec1995 _at_ 150,000
• jan2002 _at_ 125,000
• Mae Sa jan1954, jan1968-70, dec1995 _at_
  150,000
• jan2002 _at_ 125,000
• 150,000 (and larger-scale) thematic layers from
• USER (Louis Lebel)
• FFORCCT (Chatchai Royal Thai Forestry Dept.)
• MCC (Methi Ekasingh Chalermpol)
• 20m DEM and Topographic Moisture Index (Mae Sa)

6
Large-scale Mapping
New Datasets - Laos

• Landsat ETM image acquired 25 March 2004
• Detailed thematic layers for N. Laos districts
• from Khamla
• 150,000 (and larger scale) thematic datasets
• from Yokoyama (EWC visiting researcher)
• for all of Laos including
• - Admin, village, hydro, landuse,
• road, builtup areas, elev points
• - Contours and derived 30m DEMs
• - b/w orthorectified IKONOS(?) images

7
Large-scale Mapping
New Datasets - Xishuangbanna

• Landsat ETM image acquired 25 March 2004
• Township-level 2000 Census data for Yunnan
  Province
• Township boundaries (CBIK)
• Daily observations for climate stations (CBIK)
• Jinghong (1954 2001)
• Menghai (1958 2001)
• Mengla (1957 2001)
• Damenglong (1958 1996)
• 20m DEM and Topographic Moisture Index (Nam Ken)
  

8
Large-scale Mapping
9
Large-scale Modeling
Cellular Automata model

• Develop annual dynamic simulations of land cover
• to the years 2025 and 2050
• for detailed simulation regions along road
  corridor
• based on 3 interrelated LCLUC scenarios
• 1) agricultural intensification
• 2) road development
• 3) growth of markets

10
Large-scale Modeling
Cellular Automata

• Mathematical object defined as
• n-dimensional cellular space, consisting of cells
  of equal size
• Cells in one of a discrete number of states
• Cells change state as the result of a transition
  rule
• Transition rule is defined in terms of the states
  of cells that are part of a neighbourhood
• Time progresses in discrete steps. All cells
  change state simultaneously.

11
Large-scale Modeling
Cellular Automata example
Conways Life (Gardner, 1970)
12
Large-scale Modeling
Cellular Automata model developments

• Xishuangbanna model characteristics
• coded and run using IDL
• grows rubber and rice annually (active classes)
• from 1988 1999
• using 3 x 3 neighborhood, 30m res. cells,
  90x90km domain
• random seeding to start
• restricted areas include parks and protected
  areas
• calculates suitability scores for both active
  classes
• reconciles rice vs. rubber
• outputs annual maps
• landscape and class-level pattern metrics
• passive classes include forest, swidden, barren,
  urban, water
• factor level and within-factor weights from AHP
  

13
Large-scale Modeling
Cellular Automata model developments

• Analytic Hierarchy Process Questionnaires
• Glean expert knowledge on conversion to rubber
• Synthesized to determine relative weights of
  conversion factors
• Factors (inputs) Weights (normalized 0-1)
• d2procsuit 0.274
• elevsuit 0.816
• market price time-varying blanket weight
• lcluwgt 0.296
• - forest .55
• - swidden 1
• - rubber 1
• - rice .24
• - urban, water, barren 0
• rubber score
• d2procsuit (wgt) elevsuit (wgt) mpwgt
  lcluwgt (wgt)

rice score d2streamsuit riceslpsuit
14
Fuzzy Cognitive Mapping
Consensus Social Cognitive Map of Rubber
Production Damenglong and Meungpong Combined
Most Central Variables Rubber Inputs Income Pests
Price
Connections gt ABS(0.1, -0.1)
Feedbacks
Technology
0.13
- 0.41
0.15
0.15
0.34
0.31
- 0.46
0.1
0.63
0.19
0.19
- 0.3
0.1
0.15
- 0.2
0.23
0.13
Least Central Variables (Centrality lt
0.5) Labor State Farms Policy Physical
Environment Government Ext. Credit
0.23
0.25
0.1
0.15
0.2
Net Causal Relationships after Additively
Superimposing 16 FCMs and Normalizing results
15
MT-CLIM
Mountain Climate Simulator for Excel
(Numerical Terradynamic Simulation Group, U. of
Montana)

• Extrapolates precipitation, max and min
  temperatures
• at one location (site)
• using daily climate data from known location
  (base)
• and DEM (elevation, slope, aspect)
• site latitude and lapse rate also required
• Daily observations for climate stations
  (Jianchu)
• Jinghong (1954 2001)
• Menghai (1958 2001)
• Mengla (1957 2001)
• Damenglong (1958 1996)

16
Small-scale Modeling
MMSEA Climate simulations thus far

• Present climate (1998-2002 NCEP/NCAR) w/
  present LCLU
• (Control)

17
Small-scale Modeling
Climate simulations thus far

• Present climate (1998-2002 NCEP/NCAR) w/
  extreme deforestation

18
Small-scale Modeling
Climate simulation results to date

• Deforestation of MMSEA increases precipitation
  in Indochina
• peninsula while decreasing precipitation in
  southeastern China

• Using RegCM3 for E
• SE Asia domain
• Daily (April 15 Dec 1)
• Extreme Control
• lt Difference in ensemble
• June-July-August
• precipitation

19
Small-scale Modeling
Climate simulations ahead

• Present climate (1998-2002 NCEP/NCAR) with 2025
  LCLU
• Present climate (1998-2002 NCEP/NCAR) with 2050
  LCLU
• Control climate (PCM 2045-55 Present CO2) with
  present LCLU
• Control climate (PCM 2045-55 Present CO2) with
  2050 LCLU
• Projected 2050 climate (PCM 2045-55 SRES A2
  CO2)
• with present LCLU
• Projected 2050 climate (PCM 2045-55 SRES A2
  CO2)
• with 2050 LCLU

20
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Changing Land Use and its Effects (CLUE)
  modeling framework

• Spatial policies
• restrictions
• Parks protected areas
• Restricted areas
• Agricultural
• development zones

• LCLU type-specific
• conversion settings
• Transition sequences
• (From-to matrix)
• Conversion elasticity
• (min and max t)

CLUE
LCLU change allocation
LCLU requirements (demand)
Location characteristics
Location factors soil, access., topography, biocl
imate, demography, socio-economic, etc.
scenarios
Lclu specific location suitability
aggregate lclu demand
Logistic regression
trends
advanced models
Source The CLUE Group, Wageningen University,
Netherlands, website http//www.dow.wageningen-ur
.nl/clue/
21
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• CLUE Allocation Procedure

• Some allocations reversible
• Some allocations dependent
• on earlier time steps

LCLU type specific settings
Conversion Elasticity ( ELASu )
Competitive Strength ( ITERu )
Allowed conversions
If No, then update competitive strength for
those types not meeting demand
Is total lclu area for each type equal to the
demand?
Calculation of change
Land cover/use ( t )
LCLU ( t 1)
Yes
For each grid cell i, calc total probability for
each lclu type TPROPi,u Pi,u ELASu ITERu
Grid cell specific settings
Location suitability ( Pi,u )
Spatial policies
Neighborhood weights
Regional demand
Source The CLUE Group, Wageningen University,
Netherlands, website http//www.dow.wageningen-ur
.nl/clue/
22
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Data Requirements
• - LULC
• - Masks and Protected Areas (WDPA)
• - Socio-economic (Income, GDP, Malnutrition
  rate, Illiteracy, etc.)
• - Demographic (Population Density)
• - Bioclimatic (Mean Temp and Precip, etc.)
• - Geographic (Distance to )
• - Topographic (Elevation, slope, aspect)
• - Soils/Geomorphology (Soil type, Soil
  Degradation, Landform)
• LCLU Requirements (Demand) Exercise
• - by country (6 countries intersect MMSEA)
• - by modeled LCLU type ( of total country,
  of pixels, area)
• - for years 2025 and 2050

23
Future Steps

• What Next? (among other things)
• Successfully map rubber over time
• Refine and expand large-scale CA model
• Continue to explore Agent-based modeling
• Complete small-scale CLUE modeling MMSEA
• Incorporate narratives/livelihoods into model
• What are we doing in northern Thailand?