Predicting rainfall in Nineveh Governorate in northern Iraq using machine learning time-series forecasting algorithm

Rainfall forecasting addresses several challenges in water resource management and socioeconomic development, despite its complexity and non-linearity. To overcome this issue, machine learning and advanced time series data statistical approaches have been used. Water scarcity in the Nineveh Governorate in northern Iraq leads to significant ecological and economic issues, and the lack of a comprehensive rainfall forecasting system could exacerbate this problem. This study utilizes the Prophet model to present long-term rainfall forecasting up to 2030 on a daily scale for Nineveh, based on time series data from five rain stations from 1981 to 2021. The results demonstrate that the Prophet model yields high accuracy for daily rainfall forecasting and effectively captures seasonal variations, as evidenced by the evaluation metrics and trend analysis. We found that rain station locations, elevations, and local climatic conditions influenced long-term rainfall forecasting. Our findings also revealed that annual rainfall is predicted to decline by approximately 314 to 191 mm (39%), 160 to 120 mm (25%), 200 to 160 mm (20%), and 297 to 253 mm (14.8%) at Rabbia, Haddar, Sinjar, and Mosul stations, respectively, by 2030. However, Shekhan station is projected to receive a 20% increase in rainfall from 370 to 447 mm by 2030. Moreover, lower rainfall zones in the southeast and southwest will expand towards the central and southwestern regions, with a concentration of increased rainfall in the northeast areas of the governorate. These findings highlight the need for proper management and consideration of the consequences of climate change, which are likely to have debilitating impacts on local hydrological systems.