Comparison Of Temperature And Emissivity Separation Methods

Land surface temperature is a mainly parameter of thermal infrared remote sensing retrieval. It is very important to obtain land surface temperature for the studies of energy balance of earth and atmosphere, and ecological environmentIt is only relied on thermal infrared remote sensing to get the temporal and spatial distribution of global or regional scale land surface temperature. The material of the universe, as long as its temperature over absolute zero, will continue eradiate infrared radiation. The infrared energy eradiated from land surface objects at normal temperature is mostly in the infrared area, it is not only related with the surface characters of the material, but also is the function of composition and temperature. In the atmosphere transmission, it can permeate two atmospheric windows which is 3 - 5μm and 8 - 14μm .Thermal infrared remote sensing is using on-board sensor to collect and record the features of land surface objects' thermal infrared information and use it to identify objects and retrieve land surface parameters, such as temperature, humidity and thermal inertia and so on.Land surface temperature extracting using thermal infrared remote sensing has aroused the attention of many scientists, for its characteristics which does not destroy the thermodynamic state of surface and has covered a wide area, plentiful information and high resolution. So it is very attractive for land surface temperature or emissivity retrieval with thermal infrared remote sensing. But the sensor can not get the temperature and emissivity directly because of the impact of land surface parameters and its heat characteristic and atmospheric composition. Radiance measured by thermal infrared sensor is the function of surface temperature and emissivity. So, the temperature and emissivity separation is a key issue for the land surface temperature retrieve. Temperature and emissivity are two leading factors of land surface objects thermal infrared radiation feature, once temperature was acquired, the corresponding emissivity can be obtained, and vice versa.In the study area, in Tongliao, inner Mongolia of China, one Landsat7 ETM+ image scene(path and row is 123/030,acquirtion date is May 16,2000) and one ASTER image scene(path and row is 124/082,acquire data is April 27,2002) were chosen. For the characteristic of imagery data, the author use different methods for the land surface temperature retrieval, and choosing temperature data, multi annual month average temperature getting from institute of geographic sciences and natural resources research, CAS(Chinese Academy of Sciences), to validate the results of this paper.For Landsat7 ETM+ remote sensing image, because of its relatively high spatial resolution, it is suitable for studying of land surface temperature (LST) and climate, and analyzing the spatial distribution of heat. For there is only one thermal infrared channel on Landsat ETM+ sensor, it could not retrieve land surface temperature and land surface emissivity (LSE) at the same time, so the resulting of surface temperature retrieval accuracy is not high. This paper estimated the land surface emissivity by NDVI (normalized different vegetation index) method, then bring it into different land surface temperature retrieval formulation to acquire land surface temperature. This study chooses: (1)Jimenez-Munoz et al.'s generalized single-channel method, (2) the Z. H. Qin et al.'s mono-window algorithm, (3) Malaret et al.'s absolutely land surface temperature method. These three algorithms need less parameters. Through the comparison of the three methods, the results show that Jimenez-Munoz et al.'s method is better than the others.For ASTER data, the author chooses reference channel emissivity(RCE) algorithm and temperature and emissivity separation(TES) algorithm. Band 13 was chosen as a reference channel to isolate the temperature and emissivity information when retrieve the land surface temperature using RCE algorithm for it can more truly reflect land surface environment. TES algorithm utilize Normalization Emissivity Method(NEM), Ratio method(RAT) and Minimum and Maximum Difference(MMD), it synthetizes the advantages of the above three methods and improve on their deficiency. It is also the official algorithm for extracting temperature and emissivity from ASTER data. It use normalized emissivity method (NEM) to estimate initial emissivity value for each band, and ratio algorithm(RAT) to compute the ratio of the channel eissivity to the average of the whole emissivity, and maximum minimum difference(MMD) to correct the emissivity, and Plank law to compute land surface temperature, and using this temperature to recomputed the emissivity for each band, and return the NEM till the error for the neighbor temperature satisfy the require. The result indicates that the both algorithm can quickly obtain LST from ASTER data. The TES algorithm based on NEM,RAT and MMD and mended single TES, so the precision is higher than RCE.Although this paper has done some works, but there are still some pending issues for further study.Owing to condition constraints, the study did not make the field measurement of the land surface emissivity estimating. Detailed land cover information can be used if available to determine the land surface emissivity with the better precision in the future study. After all, the method based on vegetation index is just a used method when there is no detail information of land cover types. The later study will use these retrieval algorithm in several other types of surface area to verify if these methods could get high precision result, and use other methods(mainly aim at temperature retrieval algorithms for only one thermal infrared channel remote sensing data). The error of land surface temperature retrieval with ASTER data is come from that assuming atmospheric correction. The accurate atmospheric correction or profile data of atmosphere in real time could improve the precision of the result. In addition, the lack of meteorological data of the study area make more difficult for evaluating the accuracy. In the later study, the author will increase the fixed point real time data to solve this problem.