BIAS CORRECTION OF PRECIPITATIONAND TEMPERATURE IN CORDEX SIMULATION OVER SOME SELECTED STATIONSIN NIGERIA

Abstract

Considering the importance of simulation of climate system in this present day, it is very important to correct some biases, since it has been established that biases exist from simulated data gotten from climate models. Due to reasons such as parameterization, initial value problem, etc, Regional Climate Models driven by Global Climate Models will always contain some biases compared to observational data. Therefore, this study attempt to bias correct precipitation and temperature of simulated data to be close to observational data of some stations. This study made use of five stations to carryout temporal analysis; namely: Calabar, Gusau, Ibi, Ikeja, and Maiduguri while five additional stations were added to carryout spatial analysis for more accuracy; namely: Bauchi, Minna, Ilorin, Lokoja and Asabar. Two packages under quantile mapping were chosen to execute this bias correction and the two packages representing the methods are Quantile-Quantile (QQ) method and Parametric Transformation (PTF) Method. The study used daily data of the stations mentioned earlier for observed and RegCM3 data driven by two GCMs namely; Met Office Hadley Centre (MOHC) and Max Plank Instsitute (MPI) was used for simulated data, The data were splitted into historical and future, the historical data runs from 1975 – 2000, while the future data runs from 2010 – 2030 Representative Concentration Pathway (RCP 4.5 Scenario). R software was used to run the both methods of bias correction while excel and ArcGIS was used to analyse temporally and spatially respectively. Coefficient of Effieciency helped to identify which of the two methods best bias correct the simulated data using taylor’s diagram, which evidently was seen to be Parametric Transformation (PTF) method and also it was observed that the amongst the two GCMs, MOHC allow for better performance of bias correction. Therefore, Parametric Transformation (PTF) method was then used for future impact scenario on RegCM3 driven by MOHC. Fits generated when historical bias correction was carried out was then used fit the future impact scenario, where after the analysis it was compared to raw simulated data used for future impact scenario and series of overestimation and underestimation was observed which would have affected decision making if not corrected.