A Combined Hierarchical/Consensus Reagonalization Approach to Aid Interpretation of Circulation-Induced Climate Change in Bolivia

Climate regionalization is an inseparable part of many climate change and environmental studies. Finding climatologically homogeneous regions enhance the precision in such studies and reduce the biases due to the uncertainties associated with climate model outputs in individual grid points. This research aims to divide Bolivia into smaller coherent climate subdivisions. Throughout the time, researchers and statisticians have developed different methods mainly including cluster analysis to perform regionalization but still the techniques are strongly dependent on the regional features like topography, latitude and so on. In this research, first, we apply a hierarchical climate regionalization method described in an open source R package, to cluster monthly precipitation and temperature separately based on a gridded observation in Bolivia spanning from 1979 to 2010. The clustering is performed separately to avoid arbitrary attribute scaling and information redundancy. A form of consensus clustering is then accomplished to take the categorical intersection of the two independent clusters and merge the two variables to create homogeneous regions. Result from this study shows that Bolivia can be divided into 10 climatically distinguishable subdivisions mainly based on topography and latitude, which are key climate control factors in the region. We apply the same selected regionalization to downscaled future climate projections to gain an understanding on how the regional atmospheric circulation might evolve and how those circulation changes may affect the future regional climate of Bolivia.