A modified cosine correction algorithm is developed for correcting atmospheric and topographic effects on Landsat TM imagery over rugged terrain. First, a relationship between offset component (such as path radiance) and terrain height is evaluated using 6S code as well as actual satellite data in shadowed area. It is shown that the offset component decreases significantly in visible bands as the height increases, while it can be assumed constant in infrared bands. Next, we examine the relationship between local topography and sensor radiance, which is expected to follow the cosine law of solar incident angle. Their correlation is high in all bands, but is higher in infrared bands which contains more direct solar irradiance due to high reflectance of vegetation. We apply the modified cosine correction (offset correction and the cosine law correction) to actual Landsat TM images with high solar elevation angle. The results can be judged good not only by visual interpretation but also from a statistical viewpoint. It is emphasized that the accurate DEM is necessary both for ortho rectification and cosine law correction.
By using NOAA/AVHRR multi-temporal data, the primary production was estimated in the semi-arid grassland of Inner Mongolia, China. Regression analysis was applied between ground truth biomass and vegetation index variables derived from NDVI, PVI, MSAVI and GEMI. The results were evaluated for vegetation index variable types, function types and atmospheric effects of AVHRR data. The results showed that the primary production of the region can be effectively estimated by using the quadratic-type function of MSAVI at the peak growth season, and MSAVI showed less sensitivity to the atmospheric effect.
The soil moisture mapping was attempted by analysing observation data from several artificial satellites. Until now, only a few studies attemted a wide area mapping using satellite data. Clearly, the estimation of soil moisture in a wide area is important from a view point of water resources management. Therefore, we used the volumetric soil moisture as a criterion variable of the statistical estimation. The measurements of the volumetric soil moisture were carried out at the time synchronized with the satellite visit time. We used the backscattering coefficient of SAR data, the inclination of ground, the roughness of ground surface and the vegitation index as the predictor variables. These variables were caluculated on the basis of the geographical information of Digital National Land Information and multiple satellite data. Soil moisture mapping was conducted by using the results of multiple regression analysis between the mesurement of soil moisture and satellite data.
Thermal infrared (TIR) multi-spectral remote sensing has been successfully applied and become a useful tool in the fields of earth sciences. The objective of this study is to evaluate the spatial distribution of thickness of desert varnish (DV) coating on the arid land surface using TIR remote sensing data. TIR emissivity spectra were measured on surface covered with DV coatings and freshly broken surface of rock samples collected in Cuprite, Nevada. The measured spectra of rock surface with DV coatings were low in contrast comparing with ones of broken surface, and they have the absorption features from 9.0 to 10.0μm due to the DV coatings. The silicified rocks have two emissivity minima in TIR region ; 8.25 μpm and 9.2μm. The depth of each minimum decreases as the relative thickness and amount of DV coating on the sample surface increase. Two models for thermal emission of rock surface were examined using the measured emissivity spectra. The first is the linear abundance mixing model and the second treats the DV coating as a layer which absorbs and transmits the electromagnetic wave in the TIR region. We found that the second model can explain the measured spectra better. Thickness of DV coating was investigated using airborne TIR remote sensing data acquired over Cuprite area, Nevada in 1990. The proposed thickness index based on the model described above, was calculated using emissivity spectra derived from the remote sensing data. The spatial distribution of the DV thickness index in Cuprite area derived from the airborne data showed good agreement with the relative extent of DV coating.
Quality of satellite imagery is affected by cirrus clouds, which are often overlooked in the prescreening process. Some image processing techniques have been used to remove the thin cloud effect, but their applicability is limmited because of lack of physical background. In this paper, we note the fact that the cirrus clouds are located in upper cold atmosphere, and propose a correction method based on the assumption that the digital numbers in visible bands increase in propotion to the difference from surface temprerature. It is shown that the method works well for an actual Landsat TM image after some parameter fitting.
This is a study to examine the observation capability of radiometric intensity by JERS-1 OPS. Several ground targets with different spectral characteristics as large as to be identifiable by OPS such as bare soil areas, grass areas and ponds were selected. The spectral characteristics of the ground targets and the irradiances were measured on the ground at the same time with JERS-1 overpass. The measurements on the ground were carried out twice, one in winter and the other in summer. The radiometric intensity of the ground targets at an altitude of the satellite were calculated using the atmospheric transmittance and the path radiances of the optical path from the ground surface to the satellite, which were estimated by LOWTRAN 7. On the other hand, Digital Numbers of the ground targets were obtained from the OPS images acquired at the same dates with the measurements on the ground and they were transformed to radiances using the conversion equation provided by NASDA. The observation capability of radiometric intensity by OPS was examined by comparing the calculated radiances with the observed radiances by OPS of the ground targets for each band in each test case. Both values were plotted in graphs. Most of the points were generally dispersed and not included in the coincidence area deffined as an range within±10% from the line of 45' in the graphs. But there may be errors of about 12% for the calculated radiances and 2% for the observed radiances. Considering that the coincidence area is an area within ±24% from the line of 45°, about 80% of the points were included in the coincidence area. From the above, it may be said that the differences between the observed radiances by OPS and the calculated radiances of the ground targets may be included in the range within ±10% with an accuracy of about 80%. The winter case showed a better agreement than the summer case, why the estimation of the atmospheric transmittance and the path radiance are considered to be closer to the reality in winter than in summer. It is better to carry out the evaluation test in winter for the examination of radiometric observation capability by optical sensors.