AI-Based Decision Support System for Precision Irrigation - third paper

Precision farming has a timing problem—water at the wrong moment and you either stress the crop or waste a resource you can't get back, and more and more, the job of getting that timing right is falling to automated decision systems rather than guesswork. IoT sensors are part of the answer, streaming soil and climate readings around the clock, but a stream of numbers doesn't irrigate anything on its own; it needs a layer on top that reads the conditions and decides. That layer is what this paper builds: an AI-based decision support system for precision irrigation that runs on live IoT sensor data. The model is deliberately lightweight—trained on past readings of soil moisture, soil temperature, and relative humidity, it learns to forecast when a crop needs water, and we pose irrigation as a classification problem, looking at the current conditions and then deciding to water or wait. On data from a simulated agricultural environment, the model spotted crop stress and supported timely watering, something a fixed threshold can't do: a static cutoff answers to a single number and ignores the rest, where a trained model weighs everything at once and adjusts as the field changes. Pair continuous IoT data with machine learning, and irrigation stops running on rigid rules—it runs on prediction, and it scales.