Crop Area and Yield Monitoring Using Low Resolution Satellite Images

Estimation of inter-annual crop area and crop production variation is central to the evaluation of agricultural performance and food production at the global scale. Remote sensing is only one of the tools involved in this process, and in particular low resolution satellite images show a high potential for monitoring large agricultural surfaces at a high temporal frequency. Several methods have been developed in the last decades to use these data for deriving quantitative area and production estimates and at the same time to overcome the problems due to the mixed nature of their pixels. This thesis gives a contribution to ameliorate the exploitation of low resolution satellite images and to develop preoperational systems for agricultural monitoring, especially in countries where the early knowledge of agricultural production can help to improve the planning of food aid interventions. The research first uses a long time series of NOAA AVHRR satellite images to update winter wheat acreages in a test area in central Italy over an 18 years period. At this aim, a SAM (Spectral Angular Mapping) method is applied to temporal NDVI profiles, isolating their information content linked to acreage variation. The results allow the identification of winter wheat area changes for each of the 18 test years compared with the average. Next, the work concentrates on the use of SPOT VEGETATION images for crop monitoring for food security. First, low resolution land cover classifications are used to extract NDVI profiles for agricultural vegetation. Then, an index (CNDVI) developed for Europe is applied to a test area in Eastern Africa with high resolution land cover data (AFRICOVER), allowing the extraction of crop specific temporal NDVI profiles for each administrative unit. The index proves to be significantly more sensitive in revealing crop stress conditions than average NDVI values, as they are frequently used for early warning in similar conditions. Operational applications of the index are presented and further research is discussed.

REMBOLD Felix; 

2005-11-25

JRC30901