Abstract: Bailongjiang watershed, a typical soil erosion area located in the ecotone among the Loess Plateau, Tibet Plateau and Qinba Mountains, was selected as a case to study the landscape ecological risk of soil and water erosion at the watershed scale. We integrated the revised universal soil loss equation (RUSLE) and landscape ecological risk index (LERI) by raster layers in a GIS workstation to identify and assess the spatiotemporal characteristics of landscape ecological risk (LER). The factors used in RUSLE and LERI were obtained by using different data from meteorological stations, soil surveys and maps, digitized topographic maps, NDVI, DEM and Landsat TM images. And then, the LER of water and soil erosion was classified into five levels: low, slight, moderate, intense and severe, and dynamically monitored from 1990 to 2010 with the spatial analyst tool of ArcGIS 9.3. The results showed that: (1) the value of LER was high and the proportion of above the moderate erosion risk level was up to 48%. As to the spatial distribution pattern, the high risk area mostly located along the Bailongjiang and its tributaries, Minjiang River, northern and northwestern parts of Diebu County. (2) The change of LER pattern was small and tended to be weakened during the period of 1990-2010; especially after 2002, the low erosion risk area increased obviously with an increase of 280.89 km², showing that the performed water and soil conservation measures were very effective. (3) The higher risk of the soil and water erosion mainly occurred on the slope cultivated land and unused land, while the lower risk type was the forestland. At the same time, the ecological risk of water and soil erosion in Wudu district was the highest (55.38%). Consequently, more attention should be paid on these towns, such as Majie, Luotang, Anhua, Hanwang, Yulong, Putian, Xiongchi, Jiegan, and northern Liangshui of Wudu District, to control the soil and water loss in the future. Our study showed that the RUSLE-GIS model was a useful tool for evaluating and mapping LER of water and soil erosion quantitatively.

Key words: geographically weighted Poisson model (GWPR), recruitment trees, Moran’s I., spatial distribution, spatial scale