An Improved Cross-Platform Atmospheric Correction Approach for Landsat-5 Sensor in Turbid Waters using MODIS Sensor

Abstract

All pixels in satellite data are essentially mixture pixels that composed of many sub-pixels with different bio-optical properties due to the continuum of variation of water dynamic conditions and the intrinsic mixed nature of water body. Generally, it is more possible to successfully find the “clear water” from data with high spatial resolution than these with low spatial resolution, because the mixed pixel in low spatial resolution data is more complicated and mixing than these in high spatial resolution data. To account for this, an improved cross-platform atmospheric correction model (ICAC) has been developed for removing the atmospheric effects from the Landsat-5 image. The accuracy and stable of ICAC model is evaluated through comparison between the satellite-derived and synchronized field-measured remote sensing reflectance. The results indicate that use of ICAC model can produce 7.19, 8.76, 5.28, and 14.42 % uncertainty in deriving remote sensing reflectance at three visible and one near-infrared bans, respectively, from Landsat-5 data. By comparison, using the ICAC algorithm in removing atmospheric influence on Landat-5 data in Taihu Lake could decrease by 6.32, 12.80, 25.45, and 32.96 %, respectively, at three visible bands and one NIR band to traditional cross-platform atmospheric correction algorithm. The improvements are very significant.