Application of Linear Mixture Model to Time Series Avhrr Ndvi Data

This research explores the possibility of using linear mixture model to generate fraction images from 1km time series NOAA AVHRR monthly composite NDVI data. The percentages of forest, grassland and farmland in the pixels are determined by applying the Constrained Least Squares (CLS) method over the study area in the northeast region of China. The validation of the model for AVHRR monthly composite NDVI data is performed by comparing the resulting fraction images with the classification results derived from coincident multi-temporal Landsat TM data and NOAA AVHRR monthly composite NDVI data using conventional methods. The results show that linear unmixing techniques, with improved estimates of endmember values and more sophisticated methods to fully utilize the included information, can have great potential when applied to coarse spatial time series AVHRR NDVI data for global studies.