Evaluation of Smartphone-Based Turbidity Monitoring in Two Philippine Rivers Using Normalized RGB Values

Abstract

River turbidity is an important water quality indicator, but frequent and spatially distributed monitoring remains difficult in resource-limited regions because conventional instruments are costly and require calibration and technical capacity. This study evaluated the applicability and limitations of smartphone-derived turbidity monitoring in two tropical coastal rivers, the Dalanas River and the Nalupa River, located in Barbaza, Antique Province, Philippines. A total of 50 paired observations were collected during dry-season low-flow conditions within a limited turbidity range of less than 30 NTU. Smartphone-derived turbidity information, field turbidimeter measurements, and laboratory-measured operational reference turbidity were compared, and supplementary event-level analyses were conducted using 10 river-day means. The field turbidimeter closely tracked laboratory-measured turbidity but showed systematic underestimation, particularly at higher turbidity levels. At the event level, the field turbidimeter had a mean bias of −3.09 NTU, MAE of 3.22 NTU, RMSE of 4.07 NTU, and Bland–Altman limits of agreement of −8.56 to 2.39 NTU. The smartphone-derived turbidity estimate showed a compressed response, with overestimation in the low-turbidity range and underestimation in the higher-turbidity range. Its event-level mean bias was close to zero, but the MAE and RMSE were 4.90 NTU and 5.23 NTU, respectively, with wider limits of agreement of −10.78 to 10.82 NTU. Normalized RGB values obtained from the HydroColor-derived output were also analyzed using log-ratio indices. The exploratory RGB log-ratio model explained approximately 77% of the observed variation in log-transformed laboratory turbidity (R² = 0.77, r = 0.88), but this represents an internal log-scale fit rather than validated NTU-scale prediction. These findings suggest that smartphone-derived RGB information can reflect relative changes in river turbidity under field conditions. Although the method cannot replace laboratory or field turbidimeter measurements, it may serve as a low-cost supplementary tool for rapid turbidity screening, citizen science, and STEM education in data-scarce regions.