Understanding the impact of climate change on Malawi’sagricultural sector and food supply chain requires robust projections ofprecipitation and temperature in order to design appropriate adaptation plans.
Ouranalysis indicates that the boundary conditions greatly influence theperformance of the RCMs. While the ERA-interim RCMs provide a reasonablecorrelation with the observed data for both precipitation and surfacetemperature (mean, maximum and minimum), the simulated RCMs and GCMs do not.Therefore, these models would be less useful for predicting specific extremeevents, such as droughts, and informing adaptation responses over thishistorical period. For example, the 2005 drought mentioned in the introductionis clearly visible in the observed data, with only 849.7mm of precipitationfalling over the country (22 percent less than the 1961-2005 average).
Themajority of the ERA-interim outputs and the average ensemble of these doindicate a downturn in precipitation in that year, however on average, neitherthe simulated ensemble RCMs or GCMs predict this. A lack of correlation between the models and the observeddata is not a barrier to use in itself; Climate models are not used to predictwhat will occur in any one year, but rather the trend in climatic change. Boththe RCMs and GCMs recreate the trending change in the temperature variableswith reasonable accuracy, however this is not true for precipitation where aclear signal is not seen; The projections for precipitation are highlydivergent across the models assessed in this paper.
Furthermore, the scale ofthe simulation outputs shows a bias for all variables, to a lesser or greaterdegree. While this analysis can only state that this uncertainty exists for thesimulations of the past, we suggest that this would also be true for futureprojections. As such, we find that these models, as they are, cannot easily beused to understand future changes in precipitation, but may have more utilityfor temperature projections, particularly if used for understanding the scaleof change in temperature rather than absolute values.Lake Malawi makes up over three-quarters of the easternborder and about a quarter of the country’s surface area. The Great Rift Valleypasses through the country from north to south causing elevations to rise from37 meters above sea level where the Shire River meets the border of Mozambique,to 3002 meters above seas level at the peak of Mulanje Massif in the ShireHighlands. This diversity in Malawi’s geographymakes climate modelling difficult. When this heterogeneity is coupled with therelatively low resolution of the GCMs (used as is, or as boundary conditions inthe RCMs), providing spatial analysis which would be useful on the scale of a climatechange impact assessment is not possible.
Additionally, the RCMs areatmospheric models and as such will not include the full complexity of theclimatic interaction with Lake Malawi. Expanding the RCMs into coupledatmospheric-lake models would likely improve the accuracy of the outputs. The authors suggest that future studies can use either theensemble RCMs or ensemble GCMs analysed in this paper to understand the trendsand degree of change seen in Malawi’s future temperature, however the RCMs doallow for better spatial understanding.
It should be noted that the absolutetemperature that these models predict is likely to be less accurate,particularly for mean and maximum temperatures. With respect to precipitation,the authors suggest that any impact and adaptation plans consider a range of potentialoutcomes, including the maximum, minimum and average projections from themodels, as well as a business as usual projection. This uncertainty highlightsthe need for further development in climate modelling for Malawi, and meansthat any adaptation planning will need to be designed and tested against arange of future scenarios.