Do quadratic and Poisson regression models help to predict monthly rainfall?

Agricultural water scarcity in the primarily rainfed agricultural system of Jigawa State in Nigeria is more related to the variability of rainfall. Rainfed subsistence farming systems in this state generally obtain low crop yields and production as a result of highly erratic rainfall seasons. Thus, predicting rainfall in the region is of great significance as it could help the government to improve sustainable rainfed agriculture in the region. To enable the design of a model capable of accurate predictions, this paper summarizes recent scientific studies aimed at predicting rainfall in Nigeria and around the world utilizing artificial and mathematical models. According to this review, it is evident that quadratic and Poisson regression models have not yet been considered in other studies about monthly rainfall prediction. Additionally, few recent studies have used solar radiation and sunshine duration as input parameters for their models. Consequently, quadratic (QM) and Poisson regression (PRM) models are proposed for predicting the monthly rainfall in Jigawa State in the north–west of Nigeria. Monthly meteorological parameters including rainfall, average temperature, minimum temperature, maximum temperature, relative humidity, sunshine duration, solar radiation, and wind speed data spanning 10 y (2008–2017) obtained from the Nigerian Meteorological Agency were used in this study. Furthermore, temporal correlation and spatial correlation were applied to measure the relationship between monthly rainfall data and other meteorological parameters for the selected region. Moreover, the proposed models (QM and PRM) were compared with the most prominent rainfall artificial models (multilayer feed-forward neural network, cascade feed-forward neural network, and radial basis neural networks) to show the predictive accuracy of the proposed model. The results demonstrate that the developed PRM model is superior in predicting the value of monthly rainfall with reported values of 0.887 and 0.0542 for the parameters of R 2 and root mean squared error, respectively.