Abstract: Abstract: Landform, land cover and landscape pattern are important underlying surface conditions for basin erosion and the dominating factors for regional water and soil erosion. It is essential to study the law of water and soil loss in Loess Plateau area to find out how to quantize the features of basin underlying surface in a scientific and rational manner, establish a underlying surface indicator system suitable for describing the erosion environment of the Loess Plateau area, and study the interaction and coupling mechanism between various quantitative parameters and sediment generation arising from watershed erosion. By applying multi-disciplinary cross theory knowledge, combining GIS (geographic information system) and RS (remote sensing) technology, and taking Dalihe Watershed of Loess Plateau in north Shaanxi as the study area, this paper studied the relationship between the fractal dimension of vegetation distribution pattern and water and soil loss. The paper established a Brownian motion fractal dimension calculation model for the vegetation pattern of Dalihe Watershed based on NDVI (normalized difference vegetation index); the model illuminated the space distribution characteristics of NDVI of Dalihe Watershed. An erosion sediment yield model was also established for Dalihe Watershed, and the comparative study was performed on the advantages and disadvantages of the fractional Brownian motion (FBM) fractal dimension of vegetation pattern and the vegetation index NDVI in respect of comprehensive quantization of watershed vegetation parameters. The phase relation between fractal dimension of vegetation pattern and soil erosion intensity was analyzed, and the opinion that water and soil conservation for Dalihe Watershed should focus on small watersheds and the control measures thereof was proposed. In order to further study the general effects of FBM fractal dimension of vegetation coverage on vegetation coverage characteristics in different watersheds, and the application of fractal dimension in hydrology, soil erosion and other prediction models as a fixed parameter, substituting the existing quantitative indicator of vegetation, the paper established a multivariate linear regression model of erosion sediment yield for Dalihe Watershed through comprehensive analysis of the precipitation, runoff, sediment yield and other data of the watersheds monitored by such 3 hydrologic stations as Qingyangcha, Lijiahe and Caogeping, which are located at the upstream, middle reach and downstream of Dalihe Watershed respectively. The secondary rainfall runoff erosion power was set as the aggressive external force input, the FBM fractal dimension of geomorphologic characteristics in watershed controlled by station, the FBM fractal dimension of vegetation landscape pattern, and the vegetation index NDVI were used to describe the underlying surface features; the GIS was adopted as the platform, and the predicting outcomes of 2 groups of different parameters were compared and analyzed. The results show that: With fractal dimension of vegetation pattern as vegetation quantization parameter, both the relative error and absolute error between the simulation value of sediment transport module of the model and the measured value are smaller than that with NDVI as vegetation quantization parameter. Among 38 storm floods, 13 of them present that the relative error between the simulated value and measured value is lower than 10% with fractal dimension of vegetation pattern as vegetation quantization parameter in the rainstorm-sediment transport module of the model, accounting for 34.21% of the total; 21 of them present a relative error lower than 20%, accounting for 55.26% of the total; 32 of them present a relative error lower than 50%, accounting for 86.85% of the total; in the above 38 storm floods, the average relative error between the simulated value and measured value is 25.19%, including 31 storm floods showing an absolute error of lower than 300 t/km2 between the simulated value and measured value. FBM fractal dimension of vegetation pattern can better reflect the relationship between vegetation coverage and water and soil loss. In addition, the analysis shows that fractal dimension of vegetation pattern presents a negative correlation with soil erosion intensity and the correlation coefficient is 0.506 6. It indicates that soil erosion intensity shows downtrend along with the increasing of FBM fractal dimension of vegetation pattern, which means that FBM fractal dimension of vegetation pattern significantly affects soil erosion intensity.