Homogenisation of a soil properties map by principal component analysis to define index agricultural insurance policies

Extreme climatic events are a serious concern for agriculture and its related activities in the entire world. Some recent studies have shown that there is a change in seasonal patterns and an increasing frequency of extreme climatic events, adding higher risk on farming activities. Therefore, government institutions should implement agricultural risk management policies or enable the private sector to develop appropriate tools.One of the main strategies for transferring agricultural risk is crop insurance. An alternative for traditional agricultural insurance is insurance based on easily measurable indicators, which use a predetermined index value as the basis for defining the indemnities. However, index insurance cannot be based only on isolated measurements. It should also be integrated into a complete monitoring system that uses many sources of information and tools (e.g., index influence areas, crop production risk maps, crop yields, and insurance claims statistics).To establish index influence areas, it is necessary to have secondary information indicating the type of climate and soil homogeneity of the study area. Over a homogeneous area, index measurements on crops of interest will be similar, but differentiating index values for areas with different soil and climatic characteristics will reduce basis risk.This study assesses two conventional agricultural and geographic methods (control and climatic maps) based on expert criteria. They were compared with one statistical method of multi-factorial analysis (factorial map). All these methods claim to homogenise soil and climatic characteristics.The three resulting maps were evaluated by agricultural and spatial analysis. The factorial map showed more homogeneous classes than the climatic map but the later lost all the soil characteristics information that will influence the index value variations. On the other hand, the factorial map obtained fewer classes than the control map, although retaining the main information on soil variability in the study area. These results obtained from the statistical method (factorial map) demonstrate that this method has generalised efficiently climatic, topographic and soil characteristics of the complete dataset.