The dynamics of shifting cultivation captured in an extended Constraint Cellular Automata land use model

Abstract

The Constraint Cellular Automata (CCA) land use model has found many applications in research and planning practices. A CCA model (the RUHUNUPURA model) was set up for the Ruhunupura area of Sri Lanka using the METRONAMICA® software. Preliminary model simulations revealed that the model is not capable of capturing the proper dynamics of the ‘chena’ land use class, which is a particular practice of shifting cultivation. Essentially, the ‘short memory’ of the model can neither account for the gradual decrease in soil fertility after an area of forest has been cleared for cultivation, nor can it capture the process of regen eration once the plot is fallowed. Chena is a common land use class (15% of the model area in 2001) and by its nature highly dynamic. Therefore, we extended the CCA model with a component that modifies land suitability as a function of the time since the last land use transition. The RUHUNUPURA model was set up with eleven land use classes out of which seven are dynamic (chena, build-up area, homestead, forest, paddy, other crops, and shrub and other uncultivated area), and four are static (inland fresh water, brackish water, un-modelled land, and sea) over the simulation period. In this model, the transition potential of a cell from one state to another is determined by five factors; neighbourhood interaction, zoning, accessibility, suitability and age, the latter being the newly introduced component. The direct effect of age is set to apply only for the land use classes chena and shrub and other uncultivated area. On one hand ‘age’ acts to linearly decrease the potential of chena lands to remain chena over time and keeps the effect constant once a chena is 6 years old. On the other hand ‘age’ linearly increases the potential of shrub and other uncultivated areas to change into chena over time, and becomes constant when the shrub and other uncultivated area reaches 5 years. The extended RUHUNUPURA model was calibrated for the period 1985 to 2001 and the results were assessed in terms of location-to-location overlap as well as structural similarity at multiple scales. These results give confidence in the representation of land use dynamics for all the main land use classes (shrub and other uncultivated area, homesteads, chena, forest, and paddy) except for the class other crops. The class build-up area, which is expected to become more important in future developments, did not expand much over the calibration period and therefore remains untested. A scenario was formulated to examine the model’s ability to capture chena dynamics, and transformed in to a model simulation from 2001 to 2030. The simulation results show that the model adapts to produce shortening fallow periods and long cultivation periods of chena, as a result of increasing pressure on land. We conclude that the model extension is crucial for a model area which includes shifting cultivation. The extension affects not only the land use class chena, but through spatio-temporal interactions that are already present in the original CCA model the whole land use system is better represented. Category/Keywords: Natural Resource Management / Land use model, Cellular Automata, Shifting cultivation, Sri Lanka .