Singularity maps applied to a vegetation index

Abstract

Agricultural drought quantification is one of the most important tasks in the characterisation process of this natural hazard. Recently, several vegetation indexes based on remote-sensing data have been applied to quantify it, being the Normalized Difference Vegetation Index (NDVI) the most widely used. Some index-based drought insurances define a drought event through the comparison of actual NDVI values in a given period with a NDVI threshold based on historical data of that period extrapolating this result spatially to the surrounded areas. Hence, the spatial statistical approach is very relevant and has not been deeply studied in this context. Drought can be highly localised, and several authors have recognised the critical role of the spatial variability. Therefore, it is important to delimit areas that will share NDVI statistical distributions and in which the same criteria can be applied to define the drought event. In order to do so, we have applied for the first time in this context the method of singularity maps commonly used in localisation of mineral deposits. The NDVI singularity maps calculated for each season and different years are shown and discussed in this context. For this study we have selected a region that includes the whole Autonomous Community of Madrid (Spain). The resulting singularity maps show that areas where the NDVI follows theoretically a spatial normal/log-normal distribution (α≅2) are widely scattered in the area of study and vary across seasons and years. Therefore, the extrapolation of normal/log-normal NDVI statistics should be applied only inside these areas.