Objective Long-term coal mining has caused significant surface subsidence and water resources forma-tion in the Huainan and Huaibei mining areas,which makes accurate measurement of water depth crucial for ra-tional utilization of water resources.Methods The method of remote sensing inversion of water depth can efficient-ly and conveniently obtain large-scale water depth data.In this study,a multi-source data fusion inversion model was proposed.By integrating Sentinel-2 satellite remote sensing data and water depth data collected by unmanned ships,model analysis methods based on single band,multi-band and logarithmic ratio were established respec-tively.The CatBoost machine learning algorithm was used to optimize the model with the best fitting results among the three conventional models.Results Through experimental verification,it was found that the goodness of fit of the green single-band model,the multi-band model and the green and red ratio logarithmic model were 0.675,0.692 and 0.260 respectively.The root mean square error were 2.40,2.34 and 3.62m respectively.After theop-timizationby using the CatBoost machine learning algorithm,the inversion accuracy of the traditional model was improved.The goodness of fit was 0.755,0.762 and 0.386 respectively.The root mean square error were 2.03,1.96 and 3.17m respectively.The CatBoost single-band model had the highest inversion accuracy in the 4～12m interval,and the CatBoost multi-band model had the highest inversion accuracy in the 0～4m and 12～16m.Conclusion The research proves that the multi-source data fusion technology can effectively improve the accuracy of water depth inversion,provide a scientific basis for water governance and water resources management in min-ing areas and show its potential application in environmental monitoring and resource management.
维普
万方数据
