A systematic method for generating land use patterns using stochastic rules and basic landscape characteristics: results for a Colombian hillside watershed

For some applications existing land use change models may be cumbersome to use because of substantial data requirements or tight integration with other simulation modules. This paper describes a spatially explicit stochastic methodology for simulating land use changes at a watershed level without the need to describe the complex relationships between biophysical, economic and human factors that affect actual land use decisions at the farm and field levels. Land use changes were made on a grid cell basis and for each year, and were governed by a set transition rules that accounted for the impact of various landscape characteristics on the direction of land use change. The model was applied to the 3246 ha Cabuyal River watershed in Colombia. Three different analyses were carried out. First, a simple model was created and its transition probabilities were based on observed frequencies of actual conversions between forest, pasture and scrub in the period 1946-1970, accounting for the impact of three landscape characteristics (distance to roads; distance to streams; land use in a 3×3 grid cell neighborhood). The model was then run for the same time period. The fit between the actual and simulated land use pattern in 1970 was good in that the areas covered by each land use class were similar, though the simulated landscape was more fragmented. Secondly, land use changes were simulated for three explorative scenarios for the year 2025: Business as Usual (BU), Ecological Watershed (EW), and Corporate Farming (CF). For this analysis, the model was expanded with three additional landscape characteristics (slope, aspect and elevation), and three different sets of land use transition rules and transition probabilities were created based on hypothetical assumptions about the possible affect of the six landscape characteristics on land use changes. The model generated three different, but plausible land use patterns for 2025. The portion of bare soil, for example, ranged from 4.0% under the EW scenario to 20.3% under the BU scenario, and cropland varied from 13.5% under the EW scenario to 26.9% under the CF scenario. Thirdly, land use changes were simulated for a period of 100 years for all scenarios. This demonstrated that, using the present set of transition rules and transition probabilities, the percentage occupied by each land use class would reach a near-equilibrium after 100 years under the EW and BU scenarios, but not under the CF scenario.