Algorithm Development For Retrieving Surface Temperature From Thermal Infrared Data Onboard The Chinese Gaofen-5 Satellite

Sea surface temperature?SST?and land surface temperature?LST?play a significant role in the physical processes between the surface and the atmosphere.High-quality SST data is needed for many applications,such as understanding the marine physico-chemical characteristics and monitoring the dynamic of the thermal condition on ocean surface.Knowledge of the LST is also of fundamental importance for a variety of fields,such as computing soil moisture and monitoring surface thermal anomaly.Thus,the research on the SST and LST has great realistic meaning.The satellite data offers the only possibility for measuring SST and LST over extended regions.With the rapid development of remote sensing from space,the researchers have great interest to the use of high-resolution images.The project of China High-resolution Earth Observation System?CHEOS?has been carried out since May,2010.As the fifth satellite of a series of satellites for CHEOS project,the Gaofen-5?GF-5?is scheduled to be launched in the late 2017.The multiple spectral-imager?MSI?,which has a 40-meter spatial resolution for the thermal infrared?TIR?bands,is one of the sensors onboard the GF-5 satellite.The high-resolution surface temperature data from GF-5 are especially needed in some fields,including explaining the formation mechanism of urban heat island effect,monitoring and evaluating accurately the water temperature distribution range caused by the nuclear power station near the coast,while these tasks can't be finished well by the TIR data at the coarse resolution.Therefore,it is imperative to develop the algorithms for SST and LST retrieval from GF-5satellite data.The aim of this study is to develop the SST and LST retrieval algorithms for the GF-5 satellite.The current research status of the satellite-derived SST and LST is reviewed.The basic conceptions and fundamental laws in the TIR remote sensing field are given,and the relevant data and the generation process of simulated data are introduced.Based on the physical mechanization of the development of the split-window?SW?algorithm,this study focuses on the algorithm development for SST and LST retrieval from GF-5 data.The main contents are as follows:?1?Since the emissivities of sea water in the SW channels are very close to unity,the emissivity value of 1 for both SW channels was used to develop the SST retrieval algorithm.The SST errors caused by the linearization of the Planck function and the hypothesis of the equal atmospheric equivalent temperatures in two SW channels(T_ai=T_aj)were analyzed.Based on the radiative transfer equation,the second-order Taylor series of the Planck function was considered to revise the error resulted from the linearization of the Planck function.For the correction of the error resulted from the T_ai=T_aj hypothesis,T_ai and T_aj were calculated using the thermal-path atmospheric upwelling radiance and the atmospheric transmittance.It was found that there is a linear relationship between T_ai and T_aj.If this linear relationship was used for the derivation of the SW algorithm,the RMSE of 0.62 K can be obtained,while that is 1.74 K for the hypothesis of T_ai=T_aj.?2?Writing the revision of the error from the linearization of the Planck function for two channels and combining the linear relationship between T_ai and T_aj,the SST retrieval algorithm was obtained that could be simplified to the quadratic SW equation.This means that the second-order Taylor series of the Planck function and the linear relationship between T_ai and T_aj were implicitly considered in the derivation of the quadratic SW algorithm.The error of the quadratic SW algorithm was RMSE=0.30 K.Validation using Matchup Data Set from EUMETSAT Ocean and Sea-Ice Satellite Application Facility?OSI-SAF?gives the RMSE of 0.41 K,showing that the quadratic split-window algorithm can be used to estimate SST.?3?The influence of the emissivity on the quadratic relationship between T_s–T_i and T_i-T_j?T_s is the surface temperature,T_i and T_j are the at-sensor brightness temperatures of channels i and j?was studied.The results showed that the quadratic relationship between T_s–T_i and T_i-T_j was changed as the channel mean emissivity???became small.Furthermore,the quadratic relationship was also affected by the difference????of the two channel emissivities.When?=0.90,??=-0.02,the RMSE of 1.71 K can be produced by the quadratic SW algorithm.To solve this problem,the constant in the quadratic SW algorithm was reparameterized by maintaining the other coefficients the same as those obtained for the black body condition.The variation trend of the constant was analyzed.It was shown that the constant was influenced by the atmospheric water vapor content?W?and near-surface air temperature?T₀?.To reparameterize the constant,an exponential approximation between W and T₀ was used.A LST retrieval algorithm was then proposed.The error?RMSE?generated by the proposed algorithm was 0.70 K.For a given emissivity,such as?=0.90,??=-0.02,the RMSE of the improved algorithm is 0.85 K.?4?The developed LST retrieval algorithm was validated by using ASTER L1B satellite data as a proxy.Compared with AST₀8 temperature product,the developed algorithm underestimates the LST about 0.8 K in the two study areas,in which the geographic and climatic conditions are different.The relatively small difference between AST₀8 and the retrieved LST indicates that the developed algorithm can be used for estimating LST from GF-5 TIR data with the satisfactory accuracy.