Abstract:

Cloud contamination has a significant effect on the accuracy in estimating vegetation fraction based on remote sensing.MVC method can apparently eliminate the influence of cloud elimination on the timeseries NDVI images.But in the regions with longterm cloudy or rainy weather,it is difficult to completely eliminate the cloud coverage with this method. In this paper,BISE algorithm was adopted to identify and analyze noise from the timeseries MODIS/NDVI data in the hilly southeastern Hubei province.Based on it,vegetation fraction was calculated according to the developed dimidiate pixel model.Compared with the vegetation fraction data derived from forest resource inventory,Mean Absolute Error (MAE) was 0149 and Mean Relative Error (MRE) 277%.Moreover,the obtained timeseries vegetation fraction data was calculated based on division into five degrees (i.e.<20%,20%～40%,40%～60%,60%～80% and >80%).The results show that about 75% to 84% of the study region was covered by vegetation fraction more than 40% and the area ratio of vegetation fraction more than 60% was almost 80% during July to September.Except the low degree of vegetation fraction (less than 20%),other types had a remarkable seasonal change,and especially the high degree (more than 80%) and the intermediatetohigh degree (60%～80%) interchanged.