Bolivia has a high percentage of mountainous terrain, making the country vulnerable to anthropogenic climate change. In order to inform about the possible risks, the Bolivian Ministry for the Environment and Water (MMAyA) and the World Bank have asked the Department of Atmospheric Sciences at the University of Utah to perform a corresponding study. The outcomes of the study serve to strengthen Bolivia's capacity in regional climate change scenario planning for resource management and to adapt to future changes in climate.

The study uses dynamic regional downscaling to generate locally relevant climate data from relatively coarse-scale predictions of global climate models (GCMs). To this end we nest the WRF V3.4 regional climate model into the output of four different GCM predictions. The results are local climate predictions that are informed by both local specifics (e.g., the terrain) and the global models.

The particular WRF configuration for the study is determined from the performance of about ~100 short climate test simulations. Each simulation uses different parameterizations and is compared against available satellite and in-situ observations. The best performing version is selected for this study. The climate model data are taken from the archives of the 5th phase of the international Climate Model Intercomparison Project (CMIP5) under the RCP8.5 scenario and are selected according to criteria measuring global and regional simulation performance (Reichler and Kim 2007).

The regional simulations are carried out at spatial resolutions of up to 1 km to adequately resolve the complex Bolivian topography with its high Andean mountains. Another novel aspect of the study is the relatively long simulation length of 21 years, which allows effective filtering of interannual noise and robust statistical testing. As shown below, we use a basic triple nested domain layout, with two extra domains over river basins that are of particular interest for this project.

The outer domain d01 (South America) has a spatial resolution of ~38 km and covers a large part of South America and its adjacent oceans. The outer domain is nested into the grid of atmospheric reanalysis and selected GCM predictions to provide the necessary lateral boundary and initial conditions. The intermediate domain d02 (Andes) has a resolution of ~9.5 km and covers entire Bolivia and its surrounding countries. Domain d03 has a grid point resolution of ~3 km and covers most of Bolivia. Of particular interest to this study are two river basins, the so-called "Cuenca del Rio Mizque" (d04) and "Cuenca del Rio Pirai" (d05), which are both run at a spatial resolution of 1 km. The table compares the design of the five domains.

In order to visualize the simulation results we developed the Climate Explorer web tool. The tool allows interactively exploring the simulation results. Both results from the global and regional models can be directly compared with each other.