| Recently,with the development of remote sensing technology and the increasing attention to the urban environment,scholars have paid more and more attention to the study of urban thermal environment.Land surface temperature(LST)has proved it's irreplaceable in monitoring the urban thermal environment,and,thus,accurate LST retrieved from satellite remote sensing data is crucial for the development of urban thermal environment research.At present,there are many difficulties in LST retrieval in urban areas.For example,the estimation of land surface emissivity(LSE)and atmospheric correction need further study.Meanwhile,the continuous launch of the China High Resolution Earth Observing System(CHEOS)satellites of China National Space Administration(CNSA)increases the application demand of high-resolution data in urban areas.In this paper,the research on the LST retrieval methods in urban areas was based on the thermal infrared data of Gaofen-5(GF-5)to promote the application of CHEOS in urban areas.First,the LSE observation experiment in Chinese typical city(i.e.,Chongqing)was carried out,and the LSEs of typical underlying surfaces of this city were observed.Based on the emissivity data obtained in the observation experiment,the LSE database of typical underlying surfaces in urban areas was established.In addition,an improved NDVI threshold method was used to accurately estimate LSE of urban areas.Second,the temperature and emissivity separation(TES)algorithm for GF-5 was trained based on the four thermal infrared(TIR)channels in GF-5 Visual and Infrared Mutispectral Imager(VIMI),and the related parameters in the TES algorithm are adjusted to make it suitable for GF-5 data.Considering that the GF-5 satellite cannot provide atmospheric correction products(i.e.,ground radiance and atmospheric downward radiation),random forest(RF)was used for the atmospheric parameters estimation with some atmospheric related parameters(i.e.,the near-surface temperature-Ta and column water vapor-CWV)as the main inputs.The RF-based atmospheric correction method and TES algorithm were combined to accurately estimate LST and named as RF-TES.For the four TIR bands,the accuracy of the upward radiation of the atmosphere is 0.032 to 0.046 W/(m2ˇsrˇ?m);the accuracy of the downward radiation of the atmosphere is 0.070 to 0.091 W/(m2ˇsrˇ?m);and the accuracy of the transmittance is 0.002 to 0.004.The accuracy of RF-TES is validated against the radiation transmission simulation data and 202 sets of ground observation LST data.For the radiation transmission simulation data,the RF-TES method yields the Root Mean Squared Error(RMSE)of 1.16 K.Further validation against the 202 sets of ground observation LST data shows that the RMSE of the RF-TES is 4.12 K.Therefore,when the atmospheric correction products are not available,RF can be used to estimate atmospheric parameters to achieve an approximate atmospheric correction,and the RF-TES method can be used for GF-5 LST retrieval.Third,based on the characteristics of VIMI TIR bands,LST integrated method for GF-5 LST retrieval based on the split window algorithm(SWA)was developed.For VIMI TIR bands,band 9 and band 10 are in the range of 8-10 ?m,and the band 11 and band 12 are in the range of 11-13 ?m.The band combination method used in this paper is combining the adjacent bands,that is,band 11 and band 12 are selected as combination 1,and band 9 and band 10 are as combination 2.In this paper,17 SWAs are selected to train,filter,and analyze the sensitivity of the algorithm,and 8 SWAs were finally obtained.Based on the 8 SWAs,this paper proposes an integrated method for GF-5 to further improve the accurate of LST estimate.The validation results show that the integrated method has the best retrieving performance in each CWV intervals.The validation results based on the radiation transmission simulation data set are: RMSE = 1.09 K(combination 1),RMSE = 2.95 K(combination 2);the verification result based on the measured data of 202 groups of stations is: RMSE = 1.94 K(combination 1),RMSE = 5.44 K(combination 2),thus,it is clear that combination 1 performs better than combination 2.Based on the combination 1,the accuracy in the urban district is 1.96 K,and the accuracy in sub-urban is better than 1 K,indicating that the LST integrated method based on combination 1 can estimate a relatively accurate LST in the urban areas.In this paper,the preliminary research and measurement of LSE in urban areas were conducted,and the LST retrieval method based on GF-5 VIMI TIR data for urban areas was trained,filtered,analyzed and integrated.The research results provide a reference to the algorithm for generating GF-5 urban LST products which is helpful to promote the application of CHEOS in urban areas. |
PDF Full Text Request