Machine Learning Regression to Predict Soil Moisture in Domestic Garden Environments

Abstract

Due to the rapid growth of the global population and limited water resources, water shortages have become an urgent problem for our society. Over 71% of freshwater withdrawals in the world are for irrigation purposes. Thus, more accurate and robust soil moisture modeling is needed to create more efficient irrigation systems, which, in turn, may lead to substantial water savings. However, existing soil moisture modeling methodologies have limited accuracy and low temporal resolution. In this study, the accuracy of using a machine learning model for high temporal resolution soil moisture modeling is demonstrated. A multimodal sensing system is designed and implemented to create a high temporal resolution dataset in the water-scarce region of South Africa. This data is then used to evaluate the accuracy of different algorithms for soil moisture modeling, where the Random Forest regressor shows promising results.