Statistical Analysis of Climate Variability and Prediction of Rainfall in Morocco

The worldwide variations such as climate crisis and global warming are influencing considerably living species and earth due to their implication in the severity and rate of extreme weather. Thus, the seasonal precipitation prediction is of great importance due to the geographical, geomorphological, and seasonal characteristics of Atlantic and Mediterranean positions showing significant events impacting rainfall. In this research, we present a precipitation forecasting model based on multivariate time series where several factors affect the accuracy performance to mitigate extreme events induced by climate change. The purpose is to study the relationship between several climate variables, including surface pressure, wind speed, relative humidity, corrected precipitation, and temperature range, influencing precipitation and the design of an effective multivariate forecasting model showing the temporal evolution of different climate factors. Additionally, the prediction model was designed using a vector autoregression model and tested on monthly data collected from the NASA portal between 1981 and 2021. The introduced model extracts the variations of the time series characteristics and the interactive relationship of several related variables. We performed the Augmented Dickey-Fuller (ADFuller) test for stationarity and the cointegration causality test. The accuracy of the rainfall forecast model was evaluated using regression metrics, including mean absolute error, mean squared error, and root mean squared error, which showed effectiveness in predicting the monthly rainfall of Morocco. Our results showed a decrease in rainfall and a variation in the aridity index in Morocco between 1981 and 2021, reaching a peak in 1996 with a drought index value of 24.25 (semi-humid climate) and a minimum value in 2017 with an aridity index of 5.57 (arid climate). We conclude that during the twentieth and the beginning of the twenty-first centuries (1981–2021), Morocco experienced a change in climate levels from semi-humid to semi-arid and currently knows arid climate. Our model can capture real-world behavior, providing better forecasting performance, while the choice of the best lag value and the split interval of the time series is a challenging task.