Abstract: Landsat ETM + image in 2002 were selected to interpret and draw the vegetation landscape map in the middle reach of Lancang River. The spatial extent gradients in relative area and spatial grain gradients in different pixel sizes were set, and the changes of landscape pattern index in the landscape scale with spatial extent and grain were analyzed, then the curve regression was used to fit the change trend of landscape pattern index with spatial extent and grain. The results show that the landscape pattern indices in the river valley are scale dependent on the spatial scale, and with the change of spatial extent and grain, most landscape pattern indices have obvious response, and can be simulated or predicted by curve estimation model; the simulation curves of landscape pattern indices such as the average patch shape index in spatial extent analysis and the standard deviation of patch area in spatial grain analysis have obvious scale inflection points, and the relative area(spatial extent) is 0.1–0.2, the spatial pixel sizes are 240–300 m. So, it can provide theoretical support for the scale determination of landscape pattern analysis in river valley area.