Quantifying calibration strategies for capacitive soil-moisture sensors (V1.2 SKU: FA3003-1) in automated drip irrigation for upland, rainfed farms

Abstract

Upland, rainfed farms in the Philippines require precise irrigation to manage scarce water resources efficiently. To address this, we developed an Arduino-based automated drip irrigation system using six low-cost analog capacitive soil-moisture sensors and a normally closed solenoid valve, triggered by a simple threshold control mechanism. The sensors were calibrated in the laboratory using loamy sand and evaluated through two strategies: a pooled general model and sensor-specific models. Exponential regression yielded the best monotonic fit (R² = 0.94–0.97). Compared to the general model (RMSE = 2.42–3.37%), sensor-specific calibrations improved both precision (CV < 10%) and accuracy (RMSE = 1.81–2.67%). Although mean differences and R² values were not statistically significant (paired T-test, α = 0.01), the general equation remains practical for field deployment. However, sensor-specific equations are preferable when higher accuracy is required. At higher moisture levels (36–46% VWC), the sensors tended to overestimate, suggesting increased error near the wet end of the scale. Overall, the low-cost system functioned reliably, demonstrating a viable, moisture-responsive irrigation control solution for small- to medium-sized upland plots.