Research on the construction method of reservoir DEM based on multi-source data fusion
To address the challenges of difficulty and high cost in obtaining underwater terrain of reservoirs,this paper proposes a method for constructing a digital elevation model(DEM)of reservoirs using long-term time series images and water level data.Taking the Miyun Reservoir as the study area,a total of 370 scenes of Sentinel-1 and 297 scenes of Sentinel-2 images from 2014 to 2023 were collected based on the Google Earth Engine(GEE)cloud platform.The object-oriented water body extraction was conducted using the SNIC segmentation and OSTU threshold algorithm.According to the water level on the day of image acquisition,the images were clustered into 45 groups using the K-Means method,and the extraction results of water bodies in each group were merged to obtain 45 storage capacity curves corresponding to different water levels.A total of 467 601 elevation points were extracted from the storage capacity curves,and a DEM with a resolution of 5 m was constructed using spatial interpolation methods based on TIN creation.The storage capacity calculated based on the DEM was in basic agreement with the observed storage capacity of the Miyun Reservoir Management Office(R2=0.999 8).Based on this,the relationship model of"water level-area-storage capacity"and the distribution characteristics of water depth in the Miyun Reservoir were explored.The results show that the Miyun Reservoir exhibits characteristics of"deeper in the southwest,shallower in the northeast,"and"deeper at the entrance of Baihe River,shallower at the entrance of Chaohu River."The water level is strongly linearly correlated with the area(R2=0.994 7),and the water level is significantly correlated with the storage capacity in a quadratic polynomial relationship(R2=1).According to the established relationship,the relative error of area calculation is approximately 2％,and the relative error of storage capacity calculation is less than 1％,which can provide effective data for reservoir monitoring and management.
万方数据
维普
