Integrating multi-model ensemble and spatiotemporal analysis for enhanced climate modelling in Nigeria

This study integrates multi-model ensemble to enhance regional climate modelling in Nigeria. The variabilities of Global Climate Models (GCMs) outputs and biases often lead to inconsistencies in future climate projections. To address this, outputs from seven CMIP5-GCMs were bias corrected using the Climatic Research Unit (CRU TS v4.02) observation dataset. GCM datasets cover the years 1951-2005 (historical) and (2006-2100) under RCP4.5 and RCP8.5 scenarios. Three ensemble modeling methods, Arithmetic Mean (AM), Multiple Linear Regression (MLR), and Artificial Neural Network (ANN) were trained on 1951-1994 and evaluated on 1995-2005 using Mean Absolute Error, Root Mean Squared Error, and Nash–Sutcliffe Efficiency (NSE). Mann-Kendall (MK) trend test was then employed to assess climate vulnerability across the country’s agroecological zones. The results show that the ANN ensemble significantly improves accuracy of climate projections, particularly in the high-altitude regions. The average NSE values for AM, MLR, and ANN ensembles were -233%, -16%, and 81%, respectively. The annual MK trend test ranged from -2.47 to 0.35 for rainfall, and 5.93 to 7.06 for temperature. This integrated approach not only underscores the importance of multi-model ensembles in climate modelling but also provides valuable insights into the spatial and temporal patterns of future climate in Nigeria.