Land Use Change Models in Idrisi
INTRODUCTION
Spatially explicit models of land use change focus on (1) the rate of change between two or more classes, (2) the location of change to one or more classes, or (3) the rate and location of change between two or more classes.
Models are usually calibrated using past time periods and/or hypothesizes of the driving factors of change.
Output may be the likelihood of each cell converting to a given class (at an unspecified time) or predicted land use at one or more (specified) dates.
MODELING APPROACHES
1. Markov Chain Models
Definition: Land use at time t+1 is based solely on land use at a
previous t
Commands in Idrisi:
Step 1: Markov (output =
suitability transition images and change matrix),
Step 2: CA-Markov (output =
predicted land use at specified date)
Comments: It is very simple, only needing land use at two time periods. Limitations: does
not consider drivers of change; can not be used over long time periods because
of stationary transition values; does not sufficiently differentiate between
suitable locations.
References:
Turner, M.G. 1987. Spatial simulation of landscape changes in
López, E., G. Bocco, M. Mendoza, and E. Duhau. 2001. Predicting land-cover
and land-use change in the urban fringe: a case in
2. Cellular Automata (CA)
Definition: Land use at time t+1 is based on a set of
rules that take into account some combination of starting land use, neighboring
land use, and other information.
Commands in Idrisi:
Step 1: Cellatom and/or CA-Markov (output = predicted land use at
specific date)
Comments: Location and rate of change are part of same
process (notable exception is work by White and Engelen). Able to represent (predict) complex patterns
associated with urban development. Capable of
representing emergent properties. Limitations:
difficult to calibrate; does not necessarily explicitly incorporate drivers of
change. In Idrisi may take a
LONG time to run.
References:
Clarke, K.C., S. Hoppen,
and L. Gaydos. 1997. A
self-modifying cellular automaton model of historical urbanization in the
Yeh, A. G.-O. and X. Li. 2003. Simulation of
development alternatives using neural networks, cellular automata, and GIS for
urban planning. Photogrammetric
Engineering and Remote Sensing 69(9): 1043-1052.
White, R. and G. Engelen. 2000. High-resolution
integrated modelling of the spatial dynamics of urban
and regional systems. Computers, Environment and Urban
Systems 24: 383-400.
White, R., G. Engelen and
3. Multi-Criteria Evaluation Model (MCE)
Definition: Also cartographic or overlay modeling. When suitability factors are added combined
to identify sites that have high suitability values for each land use
type. Land use at time t is located on
those cells that are most suitable.
Command in Idrisi:
Step 1: Decision Wizard (output
= transition suitability images for each class and predicted land use image)
OR
Step 1: MCE (output =
transition suitability images for each class)
Step 2: Rank (output =
suitability’s ranked in order)
Step 3: MOLA (output =
predicted land use based on number of conversions)
Comments: Incorporates hypothesized drivers of change.
In Idrisi can handle multiple objectives
quickly. Can combine factors based on different
distributions (ie does not assume all factors have
linear relationship to change). Can update input
factors. Limitations: unclear process of identifying
appropriate factors; unclear how to incorporate categorical factors; unclear
how to weight factors.
References:
Schneider, L.C. and R. G.
Pontius Jr. 2001. Modeling land-use change in the
4. Logistic Regression Based Model
Definition: Uses logistic regression to calculate relationship between
land use change and suitability factors between time t and t+1. Uses that relationship to
predict suitability at time t+2.
Land use at time t+2 is located on those cells most suitable.
Commands in Idrisi:
Step 1: Logi
(output = transition suitability image)
Step 2: CA-Markov (output = predicted land use at
specified time) OR
Fuzzy, Rank, Reclass
(output = predicted land use at specified time) OR
STchoice
(output = predicted land use at specified time).
Comments: Easy to incorporate categorical factors. Clear
how to test importance of each factor through more extensive statistical
diagnostics. Limitations: assumes logistic
relationship between factors and change; deterministic compared to CA
approaches.
References:
Schneider, L.C. and R. G.
Pontius Jr. 2001. Modeling land-use change in the
Wear, D.N. and P. Bolstad. 1998. Land-use changes in
southern Appalachian landscapes: spatial analysis and forecast evaluation. Ecosystems 1: 575-594.
5. GEOMOD
Definition: This module is like a dynamic (input variables update after
each time step) MCE that runs for multiple time steps
Commands: GeoMod
Comments: runs multi-time step simulations where amount of change is
specified by user and transition suitability values are defined based on change
“driver” images. Likelihood of change for a given
characteristic is determined based on the percent of cells converting in
calibration time period.
References:
Pontius, R.G. Jr, J.D. Cornell, and C.A.S. Hall. 2001.
Modeling the spatial pattern of land-use change with GEOMOD2: application and
validation. Agriculture, Ecosystems, and Environment
85: 191-203.
VALIDATION
1. Validate: calculates various kappa statistics
2. ROC (Relative Operating Characteristic): Compares transition
suitability image with reference land use image.
Reference: Pontius Jr, R G
and L Schneider. 2001. Land-cover change model
validation by a ROC method. Agriculture, Ecosystems &
Environment.
3. Crosstab, Histogram, Area,