Research On Sea Surface Temperature Estimation Algorithms For SLSTR Data

Sea surface temperature(SST)is an important parameter of material and energy exchange in the process of sea-atmosphere interaction,a basic parameter for studying changes in the atmosphere and marine environment,and a decisive factor for climate change on a regional and global scale.SST with high accuracy and high spatiotemporal resolution is an important data source in some research fields,such as marine monitoring,climate change prediction,atmospheric circulation anomaly analysis,marine primary productivity,and marine fisheries.The Sea and Land Surface Temperature Radiometer(SLSTR)mounted on Sentinel-3A provides high-quality global sea surface radiation observation data,but there are few researches on SST retrieval algorithms for SLSTR data.At the same time,due to the influence of cloud,the acquired satellite data are usually lack of useful information,which limits the application of satellite SST data.Researches regarding to remove the influence of cloud and generate all-weather temperature data with high quality and high accuracy is the focus of attention in the academic community.Considering that the all-weather SST is an inevitable demand for future SST applications,in response to the above problems,the research work of this thesis is as follows:Based on radiative transfer simulation,nine Split-Window Algorithms(SWAs)are studied for the SST retrieval accuracy of SLSTR data.In this thesis,nine widely used SWAs are selected as alternative algorithms for SLSTR SST retrieval,and a representative atmospheric profile dataset with global marine atmospheric conditions is selected from Seebor V5.1,and the corresponding radiative transfer simulation scheme is designed.This scheme is used to construct the training dataset and the validation dataset of the SWAs.Moreover,the coefficients of nine SWAs are obtained by fitting the training dataset.According to the calculated Standard Error of the Estimate(SEE),the nine SWAs have good fitting results.These SWAs are preliminarily validate using the simulation validation dataset.The results show that the MBE of the nine SWAs varied from-0.03 K to-0.16 K,and the RMSE is 0.33 K-0.58 K.The corresponding Sea Surface Emissivity(SSE)gray body model(Emis_con)and empirical model(Emis_emp)estimation formulas for SLSTR are given.And based on the theoretical model,a method for estimating SSE based on sea surface wind speed and satellite observed zenith angle is proposed-sea surface emissivity lookup table method(Emis_lut).Next,based on the three SSE estimation methods,SST was estimated from SLSTR data.Finally,the in-situ SST to validate the SST estimated by different SSE estimation methods and different SWA combinations.Three stations in different sea areas(deep sea station:station 51000 and station 42003;shallow sea station:station 41013)were selected for validation.The validation results in 2019 showed that:at the deep sea station,the SWAs based on Emis_luthave the best accuracy,the RMSE of the nine SWAs are 0.52K-0.66 K;the SWAs based on Emis_con have the worst accuracy,and the RMSE of the nine SWAs are 0.72 K-0.87 K.However,the results of the shallow station were just the opposite.The accuracy of the SWAs based on Emis_con is the best,with an RMSE of 0.57 K-0.65 K,and the accuracy based on Emis_lut is the worst,with an RMSE of 1.02 K-1.07 K.Through analysis of influence factors of SST,it is found that there is a positive correlation between the SST estimation error and the three influence factors.Carried out all-weather SST estimation research based on SLSTR data.The estimated SST of three algorithms with high accuracy:ULW1994,SR2000 and GA2008are used to reconstruct the all-weather SST of the SLSTR data based on the Random Forest(RF)method.First,the estimated SST of three algorithms(ULW1994,SR2000 and GA2008)were used as the label of the RF training model.Then,three different RF models(RF-ULW1994,RF-SR2000 and RF-GA2008)were constructed by combining the ERA5reanalysis data to estimate the all-weather SST.The generated all-weather SST image combines the characteristics of ERA5 reanalysis data and thermal infrared data,and the image quality has been greatly improved.The all-weather SST estimated by RF-ULW1994,RF-SR2000 and RF-GA2008 was validated with the in-situ SST of three stations(station 51000,station 51001 and station 51101).The results show that the RMSE of all-weather SST estimated based on RF-ULW1994 is 0.60 K-0.97 K,the RMSE of all-weather SST estimated based on RF-SR2000 is 0.54 K-0.82 K,the RMSE of all-weather SST estimated based on RF-GA2008 is 0.57 K-0.80 K.The accuracy of SST estimated by the three RF models in the clear conditions is higher than that of cloudy conditions.In summary,the RF-SR2000 model has the highest accuracy and applicability,and can be extended to all-weather SST estimation on a global scale.In this thesis,preliminary research have been made on the SLSTR SST retrieval algorithms,and the all-weather SST estimation method of SLSTR has been explored.The research results can provide high-precision,high-quality and high-resolution SST for marine monitoring,numerical forecasting and climate change assessment,and at the same time expand the application prospects of SLSTR data in the field of marine remote sensing.