Abstract:

High-precision remote sensing monitoring of dynamic changes in wetlands is of great practical significance for wetland conservation and restoration. Taking the wetland of Bosten Lake in Xinjiang as the research object, we use the multiple endmember spectral mixture analysis (MESMA) method to extract the vegetation, water body, and bare land area from Landsat images, verify the accuracy by UAV images, and then combine with the trend analysis method to explore the spatial and temporal change characteristics and trends of Bosten Lake wetland from 2000 to 2022. The results show that: (1) The MESMA classification results verified by the resampling accuracy of UAV images showed that the goodness of fit (R²) of vegetation image element is 0.75 and the R² of water body image element R² is 0.84, indicating that the classification results were consistent with the actual feature conditions. (2) From 2000 to 2022, the vegetation area of Bosten Lake increased by 536.65 km², an increase of 183.14%; the water area decreased by 595.76 km², a decrease of 37.07%; the bare land area increased by 99.12 km², an increase of 25.42%. (3) The area of vegetation in the wetlands of Bosten Lake with an increasing trend accounts for 30.6% of the total area, which is located in the northwestern part of the Great Lake and the northern part of the Small Lake; on the contrary, the area of water with a decreasing trend accounts for 34.6% of the total area, which is located in the northern and eastern shores of the Great Lake and the wetlands of the Small Lake. To accurately grasp the spatial and temporal changes of Bosten Lake wetlands and their trends, it provides a reference basis for monitoring and protecting inland wetlands in the arid zone.

Key words: MESMA, mixed pixel, wetland, spatiotemporal variation, Bosten Lake