Simulation of land use dynamics in the upper reaches of the Dadu river
大渡河上游地区土地利用动态模拟分析

Based on remote sensing imageries of 1967,1987 and 2000, and a digital elevation model with a scale of 1:250000, the key forces driving land use change and controlling land use pattern in the upper reaches of the Dadu river are found out from such biophysical and socioeconomic factors as terrain, elevation, roads, water system, urban and rural residential areas, and then the probability maps for each land use type are created by using Logistic stepwise regression, of which the goodness of fit is evaluated for all equations with the ROC (Relative Operating Characteristics) method. In this study, CLUE-S model which has the capability of modeling changes in quantity and location simultaneously, is applied to simulate temporal and spatial changes in land use from 1967 to 1987 and from 1987 to 2000 for an area of 18665 km2 which covers the counties of Rangtang, Jinchuan and Barkam. Comparisons for validation between simulated land use maps and actual land use maps of 1987 and 2000 find that Kappa index reaches to 0. 86 and 0. 89 respectively, indicating a successful simulation. For a better understanding of the future land use changes in the region, the same model is further put into application to predict spatial distribution of land use changes in 2010 for three scenarios associating with current governmental policy of "grain to green". The results of scenario analysis demonstrate that CLUE-S model can play key roles in land use planning and ecological construction, and is also a key part of decision-support system. In the scenario analysis, the changes in quantity are specified on purpose, and thus the simulation is focused on land use changes in location. This is because land use and cover changes in the upper reaches of the Dadu river are mainly driven by policies, especially for forest land and cultivated land, and changes in area are usually determined by government. Therefore, such an application of CLUE-S model is more suitable to regions characterized by policy-driven land use change, in which once the changes in quantity, such as areas of forest cutting, afforestation, grazing-forbidden, or planned reserves, are specified, then their changes in locations can be predicted with the same model. Furthermore, the spatial resolution of modeling can reach to a level of single grid cell.