Calibration and validation of low-cost AWLR sensors for hydrological applications

This research evaluates the field performance of the Automatic Water Level Recorder (AWLR) string-gauge system for supporting flood rating-curve development in a tropical river environment. The study was conducted at Namorambe Weir, Indonesia, where the AWLR system was installed within a stilling-well configuration to reduce the direct influence of turbulence, wave action, sediment disturbance, and floating debris. Sensor performance was evaluated through controlled calibration against reference water levels, QA/QC screening of the field time series, and empirical rating-curve development using ADCP-measured discharge. The controlled calibration showed a highly linear response, with measurement deviations not exceeding 2 mm and a coefficient of determination of R²=0.99997. The continuous AWLR dataset consisted of 4,761 valid records collected at 10-minute intervals, corresponding to approximately 33.1 days of monitoring. Empirical rating curves were developed using 12 independent ADCP discharge measurements paired with AWLR-derived stage observations. The segmented power-law model provided a better complexity-penalized fit than the single power-law model, as indicated by higher Adjusted R² and lower RMSE, MAE, AIC, and BIC values. Bootstrap resampling with 1,000 iterations was used to quantify parameter uncertainty and generate 95% confidence intervals for the fitted parameters and 95% confidence bands for the rating curves. The results indicate that the protected AWLR string-gauge system can provide reliable high-resolution water-level data for hydrological monitoring and rating-curve development in tropical river conditions. The limited hysteresis observed in the stage–discharge relationship was interpreted as a hydrodynamic characteristic of the monitored cross-section, not as direct evidence of sensor accuracy. Overall, the system shows potential as a relatively simple and low-resource option for continuous flood monitoring, with possible integration into IoT-based telemetry and flood early-warning applications. However, formal cost and energy-consumption assessments were beyond the scope of this study.