An assessment of correlation on MODIS-NDVI and EVI with natural vegetation coverage in Northern Hebei Province, China

Abstract

Regional vegetation cover plays an important role in modeling ecosystem change and conservation. In this study, MODIS-NDVI (Normalized Difference Vegetation Index) and EVI (Enhanced Vegetation Index) image of 32 days composite were made in August 2008. 39 samples of grassland, 40 samples of shrub and 55 samples of forest were established in Northern Hebei Province in August 2008. The MODIS image we used was taken in the same period as the time of filed sampling. The MODIS-NDVI and EVI values of field samples were extracted from MODIS image. The optimal regression equation was developed between the NDVI values and the vegetation cover of field samples in grassland, shrub and forest respectively. The similar process was conducted for EVI values. The simulation precision of the equation was tested by independent field data. Lastly, the predictive validity of natural vegetation cover by using NDVI and EVI values was compared and assessed. The results show that all selected optimal regression equation pass significance testing of 0.05. As for forest, shrub and grassland, the coefficient Correlation (R2) of selected optimal regression equations based on NDVI values is higher than that based on EVI(from 0.488 to 0.644), and Mean Absolute Error (MAE) based on NDVI values is lower (from 0.0617 to 0.0916) than that based on EVI values. There is a conclusion that MODIS-NDVI are more correlation with field data of vegetation cover and have obvious advantages for predicting natural vegetation coverage than MODIS-EVI in our study area.