Study on Area Estimation of Small Water Using MODIS Data

In the low-resolution MODIS satellite remote sensing images, the edge of a small water body is often blurred due to the resolution limitation. The existence of mixed pixels is not conducive to water information extraction and observation. Two Dimension Scattered Points (2DSP)-Classification(C) NDVI method is introduced to area estimation of the small water body. This method first analyzes band 1 and band 2 of MODIS data through 2-dimensional visualization, second, selects sample point sets of surface categories distinguished by eyes to train BP neural network and then, discriminates and classifies pixels in the study region to get mixed pixels, finally estimates water area by using linear spectral mixing model. Above-mentioned method is applied to estimate area of the West Lake in Hangzhou. Estimate error is discussed in detail. The results shows that 2DSP-C NDVI method which has a better adaptability and a higher precision in quantitative calculation for small water area than other water information extraction methods, is applicable to real-time monitoring of small size water such as small lakes and reservoirs.