MODIS Data Land Use Classification

Land use status quo survey is an important survey for national situation and power. It is also a basic work which remote sensing technical was applied to earliest and researched most on. In the past, macro land use classification used NOAA/AVHRR data mainly. However, MODIS has better spectral and spatial resolution, and it has seven bands designed for land cover specially. MODIS has many advantages and great application values for land use classification. In this study, MODIS 32-day composites of 2003 are used for land use classification research in three Northeast Provinces. The study has two steps. In the first, continuous tree cover, herbaceous cover and bare cover quantitative estimating were made, using the training data derived from land cover classifications of TM images, and basing on multi-variable stepwise regression algorithm and regression tree algorithm; Then, based on the continuous land surface map of tree, herbaceous and bare, the land use classification was made using expert classifier. The overall accuracy is 70.79%. The result from this study shows that the MODIS data is performing well in land use classification and it has highly practical value and application potential to macro-monitoring of land use situation. Moreover, the new approach for large area land-use survey provided by this study has good application foreground.