One approach to regional drought classification
Book part (Published version)
MetadataShow full item record
Over the past several decades, drought events in Serbia have become increasingly frequent. There is no temporal pattern and relatively homogeneous regions tend to exhibit varying degrees severity and non-uniform spatial distributions of drought. The impact of scarce or excessive precipitation during a specific time period largely depends on the local geography, climate, vegetation and other factors. Agriculture in Serbia is traditionally rainfed and the question is often raised, who has suffered more from drought and who should receive more subsidies. The task of the government, as the highest instance in the decision-making process, is to make such decisions based on a set of local parameters reflecting a large number of regions. The approach leading to a general decision using a series of local indicators needs to be objective and founded upon both physical observations and statistical decision-making methods. The main objective of this paper is to propose an approach to the characte...rization of individual areas within a region (the lowland region of Vojvodina) from a drought severity perspective during the growing season, and to then validate the procedure in areas of another region (the hilly region of Sumadija), whose characteristics differ to a large extent but which is under the administrative jurisdiction of the same legislative bodies in Serbia. The proposed approach is based on the Standardized Precipitation Index (SPI) and on statistical pattern recognition. Using SPIs derived for individual local areas, a random vector is generated to represent drought severity in the extended region, and then scattering matrices are applied to reduce dimensions and design a parametric classifier for the reduced space, leading to the final, general decision. The proposed algorithm is not subject to any geographical constraints and is not affected by the surface areas of the regions or their inter-relationships. The end result is a clear picture of the areas within the analyzed region which are influenced by local factors and, as such, representative of the analyzed region, and those that stand out as good representatives for final decision-making purposes. The procedure was implemented using SPIs from 20 meteorological stations, covering a period of 56 years.
Source:Droughts: New Research, 2013, 243-266
- Nova Science Publishers, Inc.