Data scarcity globally has impeded our understanding of hydrological processes. The situation is even worse in developing countries such as Malawi. In addition, there is a general agreement by scientists across the globe, that climate is changing and will continue to do so in the future. The extent and direction of change is very uncertain. This development is very worrisome to hydrologists and water managers, in their quest to design and construct hydraulic structures such as barrages, canals, bridges, dams, embankments, reservoirs and spillways. Consequently, there has been a plethora of climate change studies in an attempt to understand the change and come up with proper mitigation and adaptation measures. These studies are periodically done by the Intergovernmental Panel on Climate Change (IPCC), which conducts simulations for future climate caused by different emission scenarios. These studies have culminated into reports such as the Third, Fourth and Fifth Assessment Reports, termed AR3, AR4 and the recent AR5, respectively. In order to properly project future climate change, Global Climate Models (GCMs) are commonly used (Libanda & Nkolola, 2019). Knowledge of how climate will change in the future and the impacts this change might bring, is very vital for planning purposes.
In the recent past, the 21st century climate projections from GCMs participating in the World Climate Research Programme’s CMIP3 have been used to assess climate change impacts at regional and local scales. Currently, CMIP5 is into force and as such, and most simulations are being undertaken using these new generation GCMS. This is because these simulations provide the basis for many of the conclusions in the recent Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) (IPCC, 2013).
GCM simulations often give a myriad of model outputs, with varying capabilities; intended to accurately reproduce the past and predict the future. Owing to this huge number of outputs, selecting the best performing model has oftentimes posed a big challenge. To avert this challenge, evaluation of model outputs has often been used. Since the future climate change is very uncertain, it is often preferred to test the models in their capabilities to reproduce the past climate and this gives confidence in the models in predicting the future climate.
To understand the performance of CMIP5 GCM simulations, there have been so may evaluations that have been undertaken by different studies with differing objectives. Using 22 CMIP5 historical simulations of precipitation over East Africa (EA), Ongoma et al., (2018) noted that there is huge variation in the performance of individual models and that the overall performance of the models in that region is generally low. These findings are consistent with another study by Rupp et al. (2013), who also noted that there exist differences in the performance of models between regions of the United States. In some studies, it was discovered that model performance was dependent on model resolution (Akurut et al., 2014). This, therefore, requires some modification of the resolutions, if the model performance is to be fairly improved. In some studies, it has been observed that CMIP5 models perform better in some precipitation events including relatively intense events such as wet days (Koutroulis et al., 2016). It must be noted that when it comes to representing observed precipitation on short spatial and temporal scales, GCMs have limits. For instance, Mehran et al., (2014) reported poor reproduction of the observed precipitation over arid regions and certain subcontinental regions across the globe. When evaluating GCM model outputs, which are intended to be selected and applied to climate change impacts, there are always uncertainties. Accordingly, there is need to minimize these uncertainties as much as possible in order to have credible simulations. One way of achieving this is the use of ensemble mean of models. In some cases, the ensemble mean of models have demonstrated close approximation of the mean of the ensemble values with the observation (Nyeko et al., 2011). Woldemeskel et al., (2012) found that the model uncertainties were consistent using a number of groups of models, where ensemble run uncertainty was found to be more important in precipitation simulation than that in temperature. In addition, substantial improvement of the simulation of precipitation was achieved using multimodel ensembles of the CMIP5 and this provided a more realistic fine-scale features tied to local topography and land use (Nikiema et al., 2016). Despite the promise offered by the ensemble means for their reliability in offsetting the downside of individual model runs, there are also some discrepancies in their performance. In some cases, using fully-coupled models of the CMIP5 historical experiments, a study by Yang et al., (2014) observed the underestimation and overestimation of long and short rains, respectively, in East Africa.
Vincent et al., (2014) determined the nature of recent observed climate changes and projected future changes for Malawi, using a combination of GCM ensembles driven by RCP4.5 and RCP8.5 scenarios of IPCC’s AR5. In addition and using 11 GCMs and 21 RCMs, Erika & Reay, (2018) found that the current models are suitable for projecting temperature trends but not precipitation and that future plans will need to consider a range of future precipitation scenarios. Furthermore, using 18 CMIP5 historical simulations of precipitation over Malawi, to examine the performance against rain gauge data, Libanda & Nkolola, (2019) found that in general, most models poorly simulated the spatial standard deviation. In addtion, it was also found that while the selected models were well performing, they were also riddled with a myriad of deficiences. Owing to these findings, modelling improvement was therefore recommended for Malawi. .
Even though the above studies have demonstrated that the evaluation GCMs ability in replicating the observed climate in different regions globally has been conducted, very few have focused on Malawi in general and SRB in particular. The current study focuses on the SRB in Malawi. The SRB lies in the Sub-saharan Africa (SSA) and the SSA has been identified as being particularly vulnerable to future climate change due to its high exposure and low adaptive capacity (Niang et al., 2014). In addition, the SRB’s hydrological system represents Malawi’s most important water resource that provides essential water to key developmental sectors such as hydro-power generation, agriculture, public water supply, fisheries and navigation (International Water Association, 2011).
From the foregoing, the climate change impacts for Malawi and the SRB have been sparesly studied and mostly using the outdated greenhouse gas emission scenarios. In addition, most studies on climate change impacts for the basin, have focused on the averages; no much coverage has been done on the extremes, which are worrisome. Analysis of extreme rainfall events that were analysed at 43 stations across Malawi by Ngongondo et al., (2014) revealed a decrease in total annual rainfall, annual maximum 1-day and 5-day rainfall amount, number of heavy and extreme rainfall days. However, there was an increase in the consecutive number of wet and dry days. Studies on the current latest generation CMIP5 GCM simulations, the ones that motivate this paper, have not been extensively covered, thereby leaving a knowledge gap. It is believed that these latest GCM simulations provide fewer uncertainties while offering better and more reliable predictions. Owing to the identified gaps, this study therefore aims at assessing the ability of the current state-of-the-art 21st century GCM simulations, archived by CMIP5 in reproducing the past climate (precipitation) for the Shire River Basin (SRB) in Malawi. The assessment culminates into the selection of models that fairly reproduce that past precipitation and can therefore be used for future impact assessments.