Application of self-organizing maps technique in downscaling GCMs climate change projections for Same, Tanzania

High resolution surface climate variables are required for end-users in climate change impact studies; however, information provided by Global Climate Models (GCMs) has a coarser resolution. Downscaling techniques such as that developed at the University of Cape Town, which is based on self-organizing maps (SOMs) technique, can be used to downscale the coarse-scale GCM climate change projections into finer spatial resolution projections. The SOM downscaling technique was employed to project rainfall and temperature changes for 2046–2065 and 2080–2100 periods for Same district, Tanzania. This model was initially verified using downscaled NCEP reanalysis and observed climate data set, and between NCEP reanalysis and GCM controls. After verification the model was used to downscale climate change projections of five GCMs for 2046–2065 ( future-A) and 2080–2100 ( future-B) periods. These projections were then used to compute changes in the climate variables by comparing future-A and B to the control period (1961–2000). Verification results indicated that the NCEP downscaled climate compared well with the observed data. Also, comparison between NCEP downscaled to GCM downscaled showed that all the four GCM models (CGCM, CNRM, IPSL, and ECHAM) compared well with the NCEP downscaled temperature and rainfall data. Future projections (2046–2065) indicated 56 mm and 42 mm increase in seasonal total rainfall amounts for March–April–May (MAM) and October–November–December (OND) seasons (23% and 26% increase), respectively, and an increase of about 2 °C in temperature for both seasons. Furthermore, future projects show that during MAM there will be 2 days decrease in dry spells, and 8 days increase in seasonal length while for OND, there will also be 2 days decrease in dry spells, and 40 days increase in the seasonal length. Future-B projects 4 °C rise in temperature, and 46.5% and 35.8% increase in rainfall for MAM and OND, respectively. It is recommended to investigate the effect of increased rainfall and temperature on agricultural production.