Modelling Mean Surface Temperature of Nigeria Using Geostatistical Approach.

Abstract

Abstract Understanding spatial variability of mean surface temperature (MST) of Nigeria is necessary for ecological restoration and national planning toward effects of unstable climate conditions. This study aimed to develop MST model derived from two geostatistical procedures and multiple linear regression (MLR) model using measurements of monthly MST in Nigeria. The geostatistical models includes ordinary kriging (OK) developed in two dimensions with isotropy and in three dimensional plane with anisotropy and regression kriging (RK) that employs both correlation with explanatory variables and spatial autocorrelation simultaneously. Six statistics were considered to evaluate the performance of the approaches used. The results revealed that in the fitted MLR model all the predictors are significant and the model explains 62% variability in the MST values. The one-leave-out cross-validation indicates that RK produced less errors compared to OK model with R2 value of 78%. The OK with zonal anisotropic shows that the spatial continuity in the directions of north and north east are stronger than in the directions of east and south east at a distance of 930 kilometres and 460 kilometres respectively. The kriging weights for OK and RK were similar as shown in the maps.