| Near-surface air temperature(NSAT),is an integrated parameter of radiation exchange and energy balance at the interface between the surface and the atmosphere,controlling most of the land surface processes,such as photosynthesis,respiration and evapotranspiration(ET).Simultaneously,NS AT,as the key parameter for energy cycle,water cycle and carbon cycle of earth system,and as one of most important input parameters of land data assimilation system,plays a large role in driving regional weather and climate.Additionally,the NSAT anomaly,not only leads to drought,freezing damage,heat injury,and forest fires and other agroforestry meteorological disasters,but also affects the derivation and propagation of crops,pest and pathogen prevalence,and then threats to food safety and public health.Consequently,for many studies,it is crucial to have access to reliable estimates of NSAT over large spatial and temporal scales,e.g.,climate,hydrology,ecology,agriculture and terrestrial biological activity,as well as the land surface processes and global climate change.Compared to conventional meteorological measurements of the NSAT,the remote sensing technology provided a new important means to acquire the NSAT.The current methods for NSAT retrieval based on thermal infrared bands can be classified in three ways:(1)directly estimating NSAT from the statistical relationship between the land surface temperature(LST)and NSAT,such as the empirical-statistical methods,the artificial neural network(ANN)methods,atmospheric profile methods,etc.;(2)estimating NSAT based on Normalized Difference Vegetation Index,normalized Difference vegetation Index(NDVI)and LST feature space relationship and assumed the dense vegetation LST is equal to the NSAT,such as TVX(Temperature-vegetation Index)methods;(3)estimating NSAT based on the surface energy balance equation.However,these methods can not to estimate the NS AT directly from the thermal radiation transfer equation.These methds indirectly used the thermal infrared remote sensing data as LST or brightness temperature to estimate the NSAT,lacking of the radiation transfer process mechanism.Therefore,it is very valuable and significant to study the NSAT retrieval directly from the thermal infrared remote sensing data based on the thermal radiation transfer equation.In this paper,the three-channel algorithm,the two-channel algorithm and their extention algorithms were developed on the basis of the thermal radiation transfer equation.For the MODIS(MODerate Resolution Imaging Spectroradiometer)sensor,the best channels used for these retrieving algorithms were analyzed.And the methods for estimating the related parameters used in these retrieving algorithms were also given,including the land surface emissivities ?i,the atmospheric transmittance ?i,temperature parameter Li,and brightness temperature Ti(i is for MODIS band 29,31 and 32).Then,sensitivity analysis of the algorithm has been performed for evaluation of probable NSAT estimation error due to the possible errors in transmittance and emissivity.Finally,we applied our developed retrieving algorithms to the Southern Great Plains(SGP)of USA as the test site for validation.Two MODIS/Terra images in January and July in 2006 representing two seasons were used for the validation.The ground measurements of NSAT in SGP were used to evaluating the retrieving algorithms accuracy.Through this study,the main conclusions are as follows:(1)Based on the thermal radiation transfer equation,the three-channel algorithm and the two-channel algorithm were developed to retrieve NSAT with the linearization of Planck's radiance function by though Taylor expansion at T0(NSAT)and the relationship between the effective mean atmospheric temperature and NSAT.Compared to the pioneers' studies,these retrieving algorithms provided a new method for NSAT retrieval.In essence,the two-channel algorithm was consistent with the three-channel algorithm.The biggest difference between these two algorithms was that the two-channel algorithm used the upwelling atmosphere radiation at 53° zenith insteading the downwelling atmosphere radiation in the three-channel algorithm.Though this replace,the equation unknown dimensions in the three-channel algorithm were reduced from 3 to 2.Thus,the three-channel algorithm was changed to the two-channel algorithm.These two algorithms involve only two essential parameters(transmittance and emissivity).(2)When the LST known,the three-channel algorithm and the two-channel algorithm can be made further optimization,namely,the three-channel extention algorithm and the two-channel extention algorithm were developed.In other words,these two algorithms can be called two-channel algorithm and single-channel algorithm.Compared to the three-channel algorithm and the two-channel algorithm,these two algorithms provided an insight to study the relationship between the LST and NSAT.Similar to the three-channel algorithm and the two-channel algorithm,these two algorithms involve only two essential parameters(transmittance and emissivity),because the LST parameter can be got from the split-window and the generalized single-channel algorithm.(3)The estimation of MODIS thermal infrared band atmospheric transmittance must need to consider the synergies between the water vapor content of atmosphere and the viewing zenith.The use of binary quartic polynomial function can be accurately portrayed the relationship among the atmospheric transmittance,the water vapor and the viewing zenith.The fitting was nearly perfect,namely,the R2 are greater than 0.9996 and the RMSE is less than 0.005.(4)The sensitivity analysis showed that the three-channel algorithm,the two-channel algorithm,the three-channel extention algorithm and the two-channel extention algorithm were sensitivity to transmittance but not to the emissivity.Assuming an error of 0.005 in atmospheric transmittance estimate for the MODIS thermal channels,the average NSAT error with these four algorithms were above 2?,2.5?,1? and 3?,respectively.However,assuming an error of 0.005 in surface emissivity estimate for the MODIS thermal channels,the average NSAT error with these four algorithms were below 1.5?,0.6?,0.3? and 0.2?,respectively.(5)Among these four retrieving algorithms,the two-channel extention algorithm had the highest accuracy with R,BIAS,MAE and RMSE were 0.9594,-1.06?,2.61? and 3.51?,respectively.This showed that the two-channel extention algorithm can satisfy the needs of the other practical application and simultaneously indicated that the MODIS/Terra band 31 was the best band for NSAT retrieval,with slope near to 1 and intercept near to 0.However,the accuracy of the three-channel algorithm,the three-channel extention algorithm and the two-channel algorithm were not good,with R,BIAS,MAE and RMSE were 0.5521?-1.88??7.33? and 7.94?,0.4811?0.22??7.90? and 7.32?,0.2251?0.91??9.13?and 8.86?,respectively.Simultaneously,these three retrieving algorithms were sensitive to the seasonal characteristics of the atmosphere,which causing underestimated in July(summer)and overestimated in January(winter).(6)The NSAT retrieval using MOD07 data with the ambient lapse rate(-6.5K/km)showed systemic underestimation,with R,BIAS,MAE and RMSE were 0.9786,-4.15?,4.49? and 2.01?,respectively.In addition,this retrieving algorithm was sensitive to the seasonal characteristics of the atmosphere,which causing the NSAT retrieval underestimated larger in July(summer)than in January(winter),namely,BIAS was-5.93? in July(summer)and-1.38? in January(winter).However,the deviation of the algorithm to 1:1 straight line in seasons were reverse,namely,the slop was 0.5558 in July(summer)and 0.6228 in January(winter).The main innovation of this paper was developed four retrieving algorithms of NSAT directly from the thermal radiation transfer equation,including the three-channel algorithm,the three-channel extention algorithm,the two-channel algorithm and the two-channel extention algorithm.All of these four retrieving algorithms can be drived only by land surface emissivity and the atmospheric transmittance of thermal bands.Compared to the three-channel retrieving algorithm,the other three algorithms can make more satellite thermal infrared data(namely,the thermal infrared bands of the sensor is smaller than 3,such as AVHRR/NOAA,AATSR/ENVISAT,TIRS/Landsat8,VIRR/FY-3 and IRS/HJ-1B,etc.Appendix 3 demonstrates in full detail about current satellite thermal infrared sensors.)used effectively with higher accuracy,especially for the two-channel extention algorithm. |
PDF Full Text Request