LC Mapping and Modeling Group

1
LC Mapping and Modeling Group

• Progress Summary
• NASA Meeting, 1 June 2006
• Honolulu, HI
• John Vogler Jeff Fox

2

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

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 SPOT(?) 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 Mapping

• Above derived primarily from photo interp of
  125k, 2002
• Detailed participatory mapping this summer?
  (Pornwilai)

10
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

11
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.

12
Large-scale Modeling
Cellular Automata example
Conways Life (Gardner, 1970)
13
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 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
  

14
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
15
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
16
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)

17
MT-CLIM
18
Small-scale Modeling
Climate simulations

• 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

19
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/
20
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/
21
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Data Requirements (Raw data cleaned, vector
  raster, 1km GRIDs cut to 6 different regions,
  GRIDs coverted to ASCII)
• - Initial LC (year 2000 same LC used by Omer)
• - Masks and Protected Areas (WDPA)
• - Socio-economic (income, GDP, malnutrition
  rate, illiteracy, etc.)
• - Demographic (population density (dynamic
  variable))
• - Bioclimatic (subset of bioclimate variables
  from WorldClim)
• - Geographic (distance to road, river, market
  (to road is dynamic variable))
• - Topographic (elevation, slope, aspect)
• - Soils/Geomorphology (soil type, soil
  degradation, landform)

22
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LCLU Requirements (Demand Scenarios)
• - by region (6 countries intersect MMSEA)
• - by modeled LCLU type (vary by region)
• - for years 2025 and 2050 specify of total
  pixels (spreadsheet)
• - linear step increases from 2000 - 2025 and
  2025 2050
• - Converted to ASCII text demand file for each
  region
• demand.in1
• 50
• 19106 4184 1884 7 25318 120116
• 19480 4301 1922 7 25216 119637
• 19854 4418 1960 7 25114 119158
• ... To 2025 and 2050

23
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LCLU Requirements (Allowed Conversions Matrices)
• - by region (6 countries intersect MMSEA)
• - by modeled LCLU type (vary by region)

24
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LCLU Requirements (Allowed Conversions Matrix)
• - by region (6 countries intersect MMSEA)
• - by modeled LCLU type (vary by region)
• - converted to ASCII matrix

allow.txt
1 1 0 0 1 0 0 1 1 0 0 0 0 1 0 0 1 1 1 1 0 1 1 1 1
0 0 0 0 0 1 1 1 1 1 0 1 0 1 0 0 1 0 0 0 0 0 1 0 1
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 1 1 0
0 0 0 0 0 0 1 0 1 1 1 1 0 0 1 1 0 0 0 0 0 0 0 1 0
1 1 1 0 1 0 1 1 1 1 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 1 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 0
1 1 0 1 0 0 0 0 0 1 0 1 1 1 0 0 0 1 0 0 1 0 0 0 0
0 0 1 1 0 1 1 0 0 1 1 0 1 0 0 0 0 0 1 1 1 1 1 1 1
0 1 1 0 1 0 0 0 0 0 1 1 1 1
25
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LCLU Requirements (Conversion Elasticities)
• - Codes for allowed changes and behaviors of LC
  types
• 3 Parameters
• 1) No consideration of present land cover to
  high preference for current land cover (0 1)
  
• 2) Minimum number of years a cell must remain
  in specific LC type
• (e.g. regrowth of forest from grassland to
  forest)
• 3) Maximum number of years a cell can remain in
  specific LC type
• (e.g. crop rotations)

26
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Spatial Policy Requirements (Restricted Areas)
• - by region (6 countries intersect MMSEA)
• - conversion restricted in parks and protected
  areas
• - WDPA restricted grid cells recoded to 9998
  (no change)
• - Active cells recoded to 0 NoData cells to
  9999
• - With exception of Laos, all region models use
  restricted areas
• - But scenarios can be run with/without
  restricted areas

27
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LC Location Characteristics (Allocation
  Regressions)
• - by region (6 intersect MMSEA) and by model LC
  types (14 17)
• - Binary logistic regression using SPSS
• - All GRIDs stripped of NoData values and fed
  into SPSS
• - Modeled LC types become dependent variables
  with value 0 or 1
• - Hypothesized drivers are independent
  variables
• - Separate equations for every LC type modeled
  in each region
• - Constant and variable coefficients retained
  for use in CLUE
• - Goodness of fit measured using ROC
  characteristic, not R2

28
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• LC Location Characteristics (Drivers)
• - driving variables numbered sequentially
• - most driving variables are stable
• - dynamic variables change annually
• - new ASCII grid accessed each year
• - Dynamic population density (US Census IDB)
• using projected annual growth rates
• - Dynamic distance to road

6 1 1 0 1 7 -4.266 5 -0.027
22 -0.013 20 -0.175 26 0.003 1 0.041
27 8 1.178 6 -0.012 22 -0.023 20 0.028
26 -0.001 1 0.066 27 -0.007 21
29
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Main Parameter File
• - by regional model (6 intersect MMSEA)

16 1 8 33 560 1039 1 (not 0.0083333333333) 97.5374
9999999 21.137499999959 0 1 2 3 4 5 6 7 8 9 10 11
12 13 14 15 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0.5
5 2000 2050 1 1 1 0 1 25 0 0

• 1 Number of land use types
• Number of regions
• Max. number of independent variables in a
  regression equation
• Total number of driving factors
• Number of rows
• Number of columns
• Cell area
• 8 xll coordinate
• 9 yll coordinate
• 10 Number coding of the land use types
• 11 Codes for conversion elasticities
• 12 Iteration variables
• 13 Start and end year of simulation
• 14 Number and coding of explanatory factors that
  change every year
• 15 Output file choice 1, 0, -2 or 2
• 16 Region specific regression choice 0, 1 or 2
• 17 Initialization of land use history 0, 1 or 2
• 18 Neighborhood calculation choice 0, 1 or 2
• 19 Location specific preference addition

30
Small-scale Modeling
MMSEA Land Cover / Land Use Simulations

• Overall MMSEA Results

Increase
Little/No change
Decrease
31
LCLUC Simulations 2000 2050
2000
32
LCLUC Simulations 2000 2050
2005
33
LCLUC Simulations 2000 2050
2010
34
LCLUC Simulations 2000 2050
2015
35
LCLUC Simulations 2000 2050
2020
36
LCLUC Simulations 2000 2050
2025
37
LCLUC Simulations 2000 2050
2030
38
LCLUC Simulations 2000 2050
2035
39
LCLUC Simulations 2000 2050
2040
40
LCLUC Simulations 2000 2050
2045
41
LCLUC Simulations 2000 2050
2050
42
Future Steps

• What Next?
• Map rubber over time at larger scale
• Refine and expand large-scale CA modeling
• Incorporate narratives/livelihoods into model
• Explore Agent-based modeling
• Refine small-scale CLUE modeling MMSEA
• - Neighborhood characteristics
• - Dynamic distance to roads
• - Model without protected areas
• - Addition of more exploratory drivers
• - More LC elasticity testing
• Publications