Adaptive Irrigation System Based on Evapotranspiration and Soil Moisture Sensing for Precision Agriculture

Agriculture in Uganda faces mounting pressures from escalating food demand, climate variability, and declining freshwater availability, with irrigation alone accounting for over 80% of total freshwater withdrawals. To address the limitations of conventional evapotranspiration (ET) controllers—which often depend on expensive meteorological instrumentation—this study proposes a cost-effective, sensor-minimal approach that utilizes only ambient temperature and relative humidity measurements from a DHT11 sensor, combined with soil moisture sensing. ET estimation is performed using a Sugeno-type Adaptive Neuro-Fuzzy Inference System (ANFIS) trained via a hybrid learning algorithm that integrates least-squares estimation for linear parameters with gradient descent for nonlinear membership function tuning. The ANFIS model, achieving a mean absolute error of 0.035 and a coefficient of determination (R²) of 0.988, was implemented in MATLAB to generate ET values in real time. These values are evaluated alongside in-field soil moisture readings by an ESP32 microcontroller to determine optimal irrigation scheduling. The system incorporates both automatic and manual control modes, accessible through a MATLAB-based PC graphical user interface and an Android mobile application via Wi-Fi, enabling responsive and scalable deployment. This architecture delivers a precise, affordable, and adaptable irrigation management solution suitable for resource-constrained agricultural environments in Uganda. Keywords: Evapotranspiration; precision agriculture; adaptive irrigation; ANFIS