LAUSR

Drought monitoring is a complex phenomenon, as several climatic variables are required to accurately monitor and forecast drought. Furthermore, inappropriate existence of gauge stations scattered over the region without any comprehensive drought monitoring framework might end up to misleading conclusions. In this study, we aimed to develop a novel regionalized drought monitoring framework, which requires minimal drought monitoring stations. For this purpose, we considered K-means clustering algorithm based on the transient behavior to identify homogenous drought characterization regions. We applied our proposed framework on 52 meteorological stations across Pakistan in such way that each cluster consists of those meteorological stations that have a similar pattern with respect to the natural behavior of drought severity. We found nine meaningful clusters and the sum of square of the deviations in each cluster is very low. Further, the correlation within these clusters confirms the results of transition-based clustering method. The scatter plots and the Pearson correlation were used to assess the performance of the developed structure of homogenous drought characterization regions. Results show that instead of using individual observatory, any station located within the cluster can be considered for monitoring and forecasting drought of whole region. In summary, minimal and appropriate selection of optimal drought monitoring stations may incorporate to study overall regionalized behavior of drought. However, the choice of weather station depends on the existence of climatic parameters, its reliability, and historical availability of data on environmental variables.