Water Need Estimation for Smart Irrigation: From Traditional to Machine Learning Approaches
In recent years, the development and popularization of the Internet of Things have allowed spatial-temporal fine-grained data for agriculture, which favored the advancement of data-driven approaches and the deepening of precision agriculture. In this context, this work investigates and contributes to state-of-the-art specifically in water need estimation for IoT-based precision irrigation. For that, we first survey the area and each of their key subjects, build up a graph-shaped map to guide solution designers, and investigate and extrapolate research trends. After, we recognize that data quality is an issue to machine learning approaches applied to IoT data, addressing possible solutions. Finally, we test the feasibility of machine learning approaches to water need estimation based on soil moisture forecast in a real case with a 2–4-year history of 12 fields from 4 farms of diverse geographic and climatic characteristics, with 55 crops from 8 different crop types. We show that machine learning is a promising alternative to traditional approaches, and we also provide a guide for machine learning-based soil moisture forecast modeling.