The crop/grassland of savannah is a complex ecosystem in which the relation between vegetation productivity and precipitation is uncertain due to high interannual climate variability and anthropogenic activities. This posed a serious threat to biodiversity, food security and socioeconomic development. In view of this, previous studies have emphasized on identifying the effective modelling approaches for quantifying these relationships, mostly by comparing the ordinary least square (OLS) and geographically weighted regression (GWR) models. Although, the conventional regression failed to successfully model these relationships, most previous research who compare the two techniques for studying the influence of precipitation on vegetation only used normalized vegetation difference index (NDVI) metrics for various locations without referring to specific vegetation type. In this study, we investigated the relationships between the NDVI metrics acquired from a 15-year Moderate resolution imaging spectroradiometer (MODIS) time series data and mean total annual precipitation generated from the inverse distance weighted interpolation technique using the ground observations data (15 years) of 42 weather stations in Nigeria. The study compared OLS and GWR modelling approaches in a crop/grassland dominated savannah. OLS did not find any significant relationship between the NDVI metrics and mean total annual precipitation. In contrast, the GWR modelling shows that the relationship exists. The rainfed crops (R2 = 0.66), mosaic croplands/vegetation (R2 = 0.65) and mosaic vegetation/croplands (R2 =58) were found to respond more strongly to mean total annual precipitation using GWR. The GWR found the highest R2 values of 0.66 and 0.97 for the individual observations and global estimate respectively. The rainfed crops, mosaic croplands/vegetation and mosaic vegetation/croplands showed the largest variability, and were much more sensitive to variability in the precipitation than other vegetation types.