Thermal Infrared Emissivitive Characteristics Of Soil On The Ground Experimental Condiction

The temperature and moisture content of soil have played important roles in many fields such as agriculture, hydrology and meteorology, and then have significant impacts on the existence and developmen of mankind t. It would be greatly meaningful that if the quantitative and spatial distribution of soil temperature and moisture can be detected quickly. With the advantages of fast and up-to-date data acquisition , large coverage and repeat observation, remote sensing technique has now become one of the most important tools for land monitorring. The thermal infrared remote sensing has great advantage in soil temperature and moisture monitor-ring because of its sensitivity to the changes of temperature and water status.The main methods used for soil moisture inversion by TIR remote sensing were Thermal Inertia method and Temperature -Vegetation Index method, both of them need the soil temperature as an known input parameter, therefore, the inversion of soil moisture can be translated into soil temperature inversion. The radiancy received by TIR sensor was the function of land surface temperature(LST) and emissivity, as a result, the land emissivity became a key factor for LST inversion accuracy . However, the emissivity was very unstable because of its suscep-tivity to soil particle size, soil moisture and soil sand content and so on. Therefore, it was needed to study the relation between emissivity and those factors mentioned above.This thesis focused on revelation of relation between soil characteristics and its emissivity. The impacts of environment factors on emissivity acquired process were analyzed based on simulative data. Moreover, The features of soil emissivitive spectra was analysed using different methods, and several prediction models were established according to soil characteristics based on ground measured soil emissivity data, and some of the soil characteristics were in-versed successfully. By simulating ASTER and MODIS sensor's TIR bands using measured soil emissivity, the soil sand content and moisture content were predicted, furthermore, these two sensor's potential of retrieving soil characteristics with band's emissivity was assessed.This study used Iterative Spectrally Smooth Temperature/Emissivity Separation Algorithm (ISSTES) to separate the soil temperature and emissivity. After relevant analysis, it was found that the separation effect of ISSTES Algorithm was best when using the smooth index defined by Bower and 7.9983-8.4139μm range as the inversion band, the accuracy of emissivity separated was higher than 0.001. The further calculation with simulative emissivity data revealed that the emissivity calculative error which resulted from atmospheric downward radiance (DWR) was much larger than that from instrument noice and sample temperature. The smoothing effect of five-points Moving average (MA), Savitzky Golay (S-Golay) and Wavelet denoising (WD) were also compared on the purpose of determining which method was most suitable for soil emissivity smoothing, The result shows that the smoothing effect of Wavelet denoising (WD) was best.By comparing the emissivity spectra of sand, paddy soil, red soil and black soil, it was found that different kinds of soil possessed of particular spectral feature in the TIR spec-trarange. Both of paddy soil and red soil had similar impacts on their emissivity spectral, their emissivity increased with soil particle size at first, then it began decreasing under certain particle size level. There have directive ratio and inversive ratio relationship between emissivity and moisture content and sand content respectively. However, there still be somewhat difference in changing extent for different spectral range. For soil sand content, the 8-9.5μm range changed significantly, whereas the 11-13μm range was almost changeless; for soil moisture content, 9.5-12μm range had obvious variation , but the emissivity spectral curves intercross nearby two ends of TIR window, which indicated that these ranges had little correlation with moisture content. Observed from the spectral curves of black soil with different organic material (OM) content, it was found that the emissivity of black soil was relatively high because of high OM content , the average emissivity was up to 0.93-0.98, however, the spectral curves were tangly, and represented little regularity for soil OM content. Results of single correlation analysis shows that the correlation coefficient between emissivity and OM content was lower than 0.14, which mean that there was no sinificant correlation betweeen them, but this conclusion still needed for further research.After filtered by the WD, soil emissivity data preprocessed with reciprocal of logarithm of the emissivity(l/logE), first differential (FD), baseline correct (BL), normalization (NOR), and multiple scattering correct (MSC), with the original data (NO), were modeled to predict soil sand content and moisture content using two linear models, Partial least squares regression method (PLSR) and Principal component regression (PCR). On the whole, the prediction precision of PLSR model was better than PCR's for both sand content and moisture content predictions. Concretely, the FD method gave the best precision under PLSR model for soil sand content prediction, and BL method was the best with PLSR model for soil moisture content. The ASTER and MODIS sensor's TIR bands emissivity and bands ratio data were simulated with measured soil emissivity, and the soil sand content and moisture content were predicted using these two sensor's simulative data. The results suggested that good linear correlation between sensors' data and sand/moisture content existed, and ASTER's prediction precision was better than MODIS's. As to different spectral variables, the band emissivity of ASTER had better inversion effect for both sand content and moisture content, conversely, the band ratio data of MODIS gave the best prediction precision.