Skill of CMIP5 models in simulating rainfall over Malawi

Abstract

Unravelling future projections of precipitation at the local scale is vitally important for building a climate-resilient economy and for the formulations of National Policies on Climate Change. Central to the entire discipline of climate projections is the use of models. However, model performance varies from one region to another. Therefore, the goal of this study is to examine the performance of 18 Coupled Model Intercomparison Project Phase 5 (CMIP5) historical simulations of precipitation over Malawi against rain gauge data. The period of study runs from 1982 to 2005. Statistical analyses were extensively carried out at both spatial and temporal scales using the following metrics: correlation coefficient (R), bias, percentage bias (PBias), standard deviation (STDEV), root mean square error (RMSE), and trend. At spatial scales, Hovmoller diagrams (HD) were used to analyse model simulations. Results indicate that the models adequately reproduce the expected annual cycle of rainfall although ~ 77% of them overestimate rain gauge data. Further, only nine of the models analysed correlate positively with rain gauge data. The correlation ranges from − 0.2 to 0.43. Seasonal root mean square errors (RMSEs) are largest during the core of the rainy season (December–February), the beginning (September–October), and the end (March–May), respectively. Rainy gauge data showed that the highest standard deviation was in the north-eastern parts of the country and around the Lake Malawi region. In general, most models poorly simulated the spatial standard deviation. Although there are large variations in model performance, models that generally perform better than others are: CNRM-CM5, EC-EARTH, GISS-E2-H, and MPI-ESM-LR. While these models are identified as well performing, their deficiencies have also been extensively discussed in this work, and therefore, caution needs to be exercised by end users when using these models to make decisions pertaining to climate change adaptation and mitigation strategies.