Preliminary results of temperature modelling in Nigeria using neural networks

Changing temperature is a major climatic concern, and its temporal variations affect activities such as agriculture, which is the major economic activity in north‐central Nigeria. We present preliminary results on the use of artificial neural networks to model temporal surface temperature variations recorded at Tropospheric Data Acquisition Network (TRODAN) stations that are located in north‐central Nigeria (7.29–9.93°N, 7.48–8.88°E). Training was undertaken using the Levenberg‐Marquardt backpropagation algorithm, and the networks were tested for interpolation and forecasting abilities. RMSEs in predictions were generally lower than 2 degC, and over 88% of the predictions had better accuracies than 2 degC.