Land Cover Classification Based on Multi-temporal MODIS NDVI and LST in Northeastern China

This paper investigated the regional land cover classification based on multi-temporal MODIS data. The study area lies in Northeastern China, where there are diverse and relative homogeneous land cover types. Through experiment, NDVI time-series data can be used to distinguish the woody (perennial) and herbaceous (annual), vegetation and non-vegetation categories depending on the seasonal differences. Grassland and cropland (one-crop-per-year), needle-leaf deciduous forest and broad leaf deciduous forest have similar phenological characteristics easy to be confused. We add the LST (land surface temperature) data to resolve this problem. But built-up area and bare land must depend on further information to be divided. Validated results with 363 ground truth filed samples; the result shows that the temperature-vegetation index (TVI) includes more information. The overall land cover classification accuracies with NDVI and TVI are 62.26% and 71.63% respectively. Based on this study, we concluded that TVI is more sensitive to land cover than NDVI, and MODIS data has its strength in the regional land cover mapping.