The proper management of water resources is one of the key elements of sustainable development in cities and countries. Governments and water agencies worldwide have to consider all the processes involved in water demands – being agriculture, industries or human use – to properly address supply problems. Climate change is an added challenge to water resources management given its impacts on the hydrological cycle. In a warming planet, it is expected to be observed increasing evaporation which would yield to more precipitation. However, observational and modeling studies normally find that these changes have non-linear interactions and regional impacts can be presented in different ways. Changes in variability of precipitation and evaporation can produce more frequent droughts or floods in different regions, depending on these interactions.
Global Climate Models (GCMs) provide projections of precipitation, evapotranspiration and other variables to support the necessary water management policies. While the typical horizontal resolution of GCM projections has been increasing in recent years, the sub-grid processes are not simulated with the adequate complexity in the models. One of the methods for addressing this issue is dynamical downscaling using a Regional Climate Model (RCM). This study presents the application of different model configurations/scenarios from Coupled Model Intercomparison Project Phase 5 (CMIP5) to force run-off models in the Cantareira Reservoir (a system that provided water for the Sao Paulo metropolitan area in southeast Brazil) and downscaled versions produced using the Weather Research and Forecasting (WRF) model. We compare the performance of CMIP5 ensembles and the RCM downscaling for the simulation of recent year, with focus on the severe 2013-14 drought observed in Brazil. This also allows for better understanding of the value added by the dynamical downscaling. GCM and RCM precipitations predictions are used as input to a run off model based on the Australian Water Balance Model. This model was calibrated based on 10 years of historical records to provide effective inflow into the Cantareira reservoir. The management of the system was then simulated with minimal outflows and legal limits according with an acceptable risk under dry historical records. This simulation then provides an estimate of the impact of the climate forecast into the system according with the policies used in the management.