Turbidity-based sediment monitoring in northern Thailand: Hysteresis, variability, and uncertainty

Abstract

Annual total suspended solid (TSS) loads in the Mae Sa Catchment in northern Thailand, determined with an automated, turbidity-based monitoring approach, were approximately 62,000, 33,000, and 14,000 Mg during the three years of observation. These loads were equivalent to basin yields of 839 (603–1170), 445 (217–462), and 192 (108–222) Mg km−2 for the 74.16-km2 catchment during 2006, 2007, and 2008, respectively. The yearly uncertainty ranges indicate our loads may be underestimated by 38–43% or overestimated by 28–33%. In determining the annual loads, discharge (Q) and turbidity (T) values were compared against 333 hand-sampled total suspended solid concentrations (TSS) measured during 18 runoff events and other flow conditions across the three-year period. Annual rainfall varied from 1632 to 1934 mm; and catchment runoff coefficients (annual runoff/annual rainfall) ranged from 0.25 to 0.41. Measured TSS ranged from 8 to 15,900 mg l−1; the low value was associated with dry-season base flow; the latter, a wet-season storm. Storm size and location played an important role in producing clockwise, anticlockwise, and complex hysteresis in the Q-TSS relationship. Turbidity alone was a good estimator for turbidity ranges of roughly 10–2800 NTU (or concentrations approximately 25–4000 mg l−1). However, owing to hysteresis and high sediment concentrations that surpass the detection limits of the turbidity sensor during many annual storms, TSS was estimated best using a complex multiple regression equation based on high/low ranges of turbidity and Q as independent variables. Turbidity was not a good predictor of TSS fractions >2000 μm. Hysteresis in the monthly Q-TSS relationship was generally clockwise over the course of the monsoon season, but infrequent large dry-season storms disrupted the pattern in some years. The large decrease in annual loads during the study was believed to be related to depletion of fine sediment delivered to the stream by several landslides occurring the year prior to the study. The study indicated the importance of monitoring Q and turbidity at fine resolutions (e.g., sub-hourly) to capture the TSS dynamics and to make accurate load estimations in this flashy headwater stream where hysteresis in the Q-TSS signature varied at several time scales.