Optimizing Water Consumption and Improving Productivity in Afghanistan‘s Greenhouses Through IoT and Machine Learning

Agriculture is the main pillar of Afghanistan’s economy. In recent years due to insufficient rainfall, precipitation and melting of glaciers and lack of proper management of the greenhouse’s environment, farmers were not able to have their harvest in the best possible way and they suffer from drought on a significant scale. This research aims to consider and evaluate smart irrigation methods and enhancement of greenhouses productivity using IoT and Machine Learning ML in Afghanistan. The authors examined the sensor data from a special tomato plant to increase productivity under limited water sources to meet future production needs. The real data has been collected through IoT sensors inside a simulated greenhouse between June-August 2022. The authors deployed six different machine ML models, in which the XGBoost model outperformed the others with a stable and balanced accuracy metrics, which makes it possible to supply the required water and avoid excessive water usage. In addition, we provided a webpage (visualization screen) for the farmers using the Ambient IoT platform to allow them better supervise the greenhouse environment. The result of this study revealed that XGBoost is the best fit model with high and balanced accuracy metrics for such classification problems. Furthermore, the outcome of this study will help the farmers and involved authorities to predict and manage their agricultural water needs, overcome expected drought, and inform them to take necessary actions on time.