Using long term NOAA-AVHRR NDVI time-series data for desertification monitoring in South Tunisia

In this work, we demonstrate the possibility of use of only long term NOAA-AVHRR NDVI time-series data, without need to any precipitation data, for assessing desertification conditions in South Tunisia. Our idea is based on the study of the minimum NDVI trend in stead of the maximum one. Motivation behind this idea is that the minimum NDVI is much less correlated to the precipitation compared to the maximum NDVI. We propose to use linear regression to determine trends in minimum NDVI. Negative trends indicates a consistently environmental degradation and that the studied pixel is under desertification. More significative results have been obtained throw the Application of the t-student test to compute the 90% confidence interval. Results suggest that the trend must be less than a threshold value of (-10^-4) in stead of zero. Obtained results are validated by comparison with a list of sites that are classified as desertification threatened sites by the a qualified institute. Indeed, the proposed method succeeds to extract all the indicated desertification sites.