Abstract: The lake is the most sensitive geographical unit to environmental change, and the study on the response mechanism of lake fluctuation is of great significance for ecosystem stability. In this paper, based on Landsat data, meteorological data, hydrological monitoring data and cropland area from 1999 to 2018, the improved normalized differential water index (MNDWI) and visual correction methods were used to extract lake distribution information in the Wulanbuhe Desert by ArcGIS technique, and the relationship between the driving force factors and lake fluctuation was analyzed by statistical method. The results showed that large lakes area (area size more than 100 hm²) and small lakes number (area size less than 100 hm²) were dominant in the southeastern margin of the Wulanbuhe Desert in study period, respectively. The trend analysis showed that the area and number of large lakes had a significant decrease trend (R=0.624 > R_18,0.01=0.561 and R=0.648 > R_18,0.01=0.561), and the small lakes was not significant. In terms of lake spatial distribution pattern, the central part of the study area was a sustainable distribution area of large lakes. Typical year analysis indicated that the number of large lakes varied in 11-23, and the area changed between 2 208 and 4 581 hm². The lake change was mainly affected by the yearly actual amount of yellow water diversion, cropland area and under groundwater table. Among them, the yearly water diversion have affected all lakes (P≤0.01), while the cropland area and under groundwater table had significant effects on large lakes (P≤0.01) and small lakes (P≤0.05), respectively. The multiple regression model constructed with the above mentioned key factors could predict the changes in the area size and number both of large lakes with an accuracy of over 75.7% and small lakes with an accuracy of over 60.5%.

Key words: lakes, diversion of Yellow River, groundwater depth, planting scale, Wulanbuhe Desert