Ensemble machine learning based model to estimate irrigation water requirement for wheat crop

Abstract

India faces a serious water shortage issue, as its population grows faster than the percentage of fresh water available, with only 4% of the world's fresh water available to 18% of the world's population. Agriculture sector is more water-consuming sector in India. India's irrigation system still faces two significant problems: low irrigation efficiency and a lack of optimization during irrigation. To address these problems, agriculturists ought to be aware of the water requirements for crops beforehand. Innovative fields like machine learning, a branch of artificial intelligence, have a big potential to improve irrigation. Verifying the suitability of the gradient boosting regressor machine learning algorithm-based model for estimating irrigation water requirements (IWR) is the aim of this research. The experiment is conducted in Ludhiana, a city in Central Punjab, India, with a hot, semi-arid climate that features scorching summers and chilly winters. The results demonstrate the remarkably high accuracy rate with coefficient of determination (R2) =0.98 for predicting IWR. The suggested model, which is based on a gradient boosting regression, allows the stakeholders to accurately estimate the amount of water needed for irrigation, the number of irrigation applications for the growing season of wheat crops, and the interval between irrigations.