Evaluation of SDSM Performance in Simulating Rainfall and Temperature over Nigeria

Abstract

In this study, the performances of the Statistical Down-Scaling Model (SDSM) to simulate rainfall and temperature for the period 1995-2010 over Nigeria were evaluated. Twenty-year (1975-1994) station observations of rainfall and temperatures from 12 different locations scattered over the country were used as predictands in the SDSM model calibration. Large-scale atmospheric predictor variables were downscaled over each location and used in the SDSM to simulate daily rainfall and temperatures for the validation period. Model results were validated against four observation datasets (NMET, CRU, UDEL and GPCC) and biases in the simulations were derived. Results demonstrated that SDSM generally performed adequately well in reproducing the daily temperature and rainfall across the country. The biases in daily rainfall and temperature ranged between ± 1.0 mm/day and ± 0.79°C/day respectively. There were also good and significant agreements between the model simulations and the monthly observations of temperature (0.56≤ r ≤ 0.90) and rainfall (0.32 ≤ r ≤ 0.67) at p ≤ 0.01. The performances of the models varied across the stations and seasons with overall better simulations for temperature than rainfall.