Shallow Water Bathymetry Research Based On ICESat-2 And Sentinel-

Water depth are important hydrological features in shallow water areas and its measurements with high-precision has important application value.Traditional methods of water depth measurement have the limitations with complex operation,time-consuming,and high cost.The Ice,Cloud,and Land Elevation Satellite-2(ICESat-2)is equipped with an advanced topographic laser altimeter system(ATLAS),which uses photon-counting LIDAR and the auxiliary system to determine ground points elevation and latitude and longitude.It can be used for water level monitoring and shallow water bathymetry.At the same time,the joint bathymetry of ICESat-2 and Sentinel-2 also has broad application prospects.In this thesis,bathymetry based on ICESat-2 was studied,and inversion of water depth from remote sensing image was conducted by combination with ICESat-2 and Sentinel-2.Main research contents and results are summarized as follows:(1)For shallow water depth detection based on ICESat-2,the point cloud denoising method based on photon density and the point cloud denoising method based on moving median filtering was proposed.The sea surface and seafloor signal photons of four surface trajectories in the Oahu Island and six surface trajectories in the St.Thomas Island study area were extracted through rough and fine denoising.The elevation of seafloor signal photons was obtained.(2)The refraction correction and tidal correction were performed,and the ICESat-2bathymetry results were verified by using the measured bathymetry.The results show that the ICESat-2 bathymetry is very close to the measured data,and its average error is less than 0.52m,RMSE is 0.37-0.71m,R~2 is more than 0.99,and the maximum bathymetry measured by ICESat-2 can reach 35m,which verifies the feasibility of ICESat-2 shallow water bathymetry.(3)Taking Oahu as an example,the multi band logarithmic ratio model and the BP neural network model were constructed using ICESat-2 water depth and Sentinel-2 data.The trained model was used to retrieve water depth from Sentinel-2 multispectral images.The measured water depth is also used as training data to train the same model and invert the water depth to compare and verify the reliability of ICESat-2 and Sentinel-2 fusion bathymetry.Finally,the accuracy of the retrieved water depth is verified and analyzed using the measured water depth.The results show that the BP neural network model has better bathymetry results.The MAE is0.74-1.01m,the RMSE is 0.97-1.43 m,the R~2 is 0.89-0.95.This study verifies the reliability of active-passive fusion bathymetry using ICESat-2 altimetry combined with remote sensing images.