ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Data-Driven Irrigation Scheduling Framework Integrating IoT Sensors and Deep Learning-Based Soil Water Tension Prediction

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2601228.(doi:10.13031/aim.202601228)
Authors:   Deus Francis Kandamali, Denis Olgen Kiobia, Peter Cosmas Ngimbwa, Hasan Mirzakhaninafchi, Ali P. Tavandashti, Shekhar Thapa, Canicius Mwitta, Glen Rains
Keywords:   Crop-specific modeling; decision support system; IoT-based irrigation; precision irrigation; root-zone soil moisture; soil moisture forecasting; soil water tension; temporal convolutional networks.

Abstract. Precise irrigation scheduling is essential for improving crop water management under spatially and temporally variable soil, crop, and weather conditions. Conventional irrigation practices often rely on fixed schedules or reactive interpretation of current soil moisture measurements, which may limit their ability to anticipate near-term crop water stress. This study presents a data-driven irrigation scheduling framework that integrates IoT-based soil water tension sensing, crop- and season-specific metadata, and short-term soil moisture forecasting for row-crop production fields. Soil water tension data were collected using IoT-enabled Watermark sensors installed at depths of 8, 16, and 24 inches across four production fields at the D.A.T.A. farm in Tifton, Georgia. Seasonal metadata, including crop type and planting date, were used to compute days after planting (DAP) and construct crop-dependent root-zone weighting schemes for estimating field-level root-zone soil water tension. Machine learning and deep learning models were evaluated for 24-h prediction using historical soil moisture and meteorological data. Model performance was assessed using MAE, RMSE, and R2. Among the evaluated models, the temporal convolutional network (TCN) achieved the strongest performance, with an RMSE of 3.7 kPa and an R2 of 0.92, outperforming XGBoost, RNN, GRU, and LSTM models. Predicted soil water tension values were linked to crop-specific irrigation threshold logic to generate field-level irrigation recommendations. The framework demonstrates the potential of combining IoT sensing, phenology-aware root-zone soil water representation, and 24-h forecasting to support practical irrigation decision-making.

(Download PDF)    (Export to EndNotes)