Geographic Information System: GIS based Neural Network for Appropriate Parameter Estimation of Geophysical Retrieval Equations with Satellite Remote Sensing Data
暂无分享,去创建一个
A method for appropriate geophysical parameter estimations in spatial and temporal domains with satellite remote sensing data based on neural network (NN) which is realized with GIS (it is referred to NN-GIS hereafter) is proposed. It is found that the proposed NN-GIS allow geophysical estimations with the most appropriate parameters for the areas and seasons in concern. One of the examples shows that SST estimation accuracy (0.245 K) is improved around 22.47% in comparison to the MCSST, typical regressive model based method (0.316 K) by applying the most appropriate parameters for the areas and seasons. It is also found that the proposed NN-GIS requires 2.63 times of computation resource in comparison to the existing regressive based methods.
[1] Adjacency Effects from Layered Box Shaped Clouds by Means of Monte Carlo Simulation , 2001 .
[2] Kohei Arai. Sea Surface Temperature (SST) Estimation Accuracy Assessment for ASTER/TIR - An Effectiveness of 8.3 μm of Water Vapour Absorption Band for SST Retrieval , 2000 .
[4] K. Arai,et al. Iterative estimation of the earth surface temperature and emissivity , 1995 .
[5] K. Arai,et al. Comparison of the sea surface temperature estimation methods , 1995 .