Downscaling global climate models statistically and generating projections of changes in precipitation and temperature at meteorological stations in Zambia

Abstract

Zambia has been experiencing adverse impacts of climate change. Generation of climate information about changes in future precipitation and temperature is useful in designing adaptive measures. Currently, global climate models (GCMs) are primary tools utilized to simulate the present and future climate under different greenhouse emission scenarios. Owing to their coarse resolution of approximately 100 – 300 km per grid box, GCMs outputs are not suitable for direct use in assessing climate change impacts and designing adaptation strategies at local-scale. In this study, a statistical downscaling approach has been used to downscale GCMs at meteorological station level in order to improve GCMs coarse resolution to match with local needs for impact assessment. Further, the study used the downscaled time series to generate projected changes in precipitation and temperature at meteorological stations of Zambia for the period 2020 – 2049 relative to 1971 – 2000. A non-parametric analogue method based on nearest neighbour was used to downscale daily precipitation, minimum temperature and maximum temperature over 19, 13 and 11 meteorological stations, respectively, across Zambia from three GCMs: CanESM2, CNRMCM5 and MPI-ESM-MR under RCP4.5 and RCP8.5 emission scenarios. ERA-Interim reanalysis and station datasets for a common period 1981 – 2010 were used to train the downscaling model. Findings presented are based on the ensemble of models. The ensemble mean at each station and local variable was computed from at least two GCMs with the same sign of change. Minimum and maximum temperatures are projected to increase at each meteorological station under both emission scenarios. The increase is higher towards the south of Zambia and for emission scenario RCP8.5 as compared to RCP4.5 scenario. Results also show decrease in precipitation over most stations in the Northern and Eastern parts of the country, increase in the western and Southern parts and exhibit a mixed signal for stations in the central part of the country. The downscaled precipitation and temperature scenarios can be used as inputs in climate impact models such as crop and hydrological models. Key words: CMIP5 models, Statistical downscaling, Temperature, Precipitation, Climate models, Representative concentration pathways