Jump to Main Content
Uncertainty in measured sediment and nutrient flux in runoff from small agricultural watersheds
- Harmel, R.D., King, K.W.
- Transactions of the ASAE 2005 v.48 no.5 pp. 1713
- agricultural watersheds, agricultural runoff, sediment yield, losses from soil, nitrate nitrogen, phosphates, nitrates, nonpoint source pollution, water pollution, measurement, water quality, sampling, storms, Texas
- Storm water quality sampling techniques vary considerably in the resources required for sample collection and analysis, and potentially in the resulting constituent flux estimates. However, quantitative information on sampling error is rarely available for use in selecting appropriate sampling techniques and for evaluating the effects of various techniques on measured results. In an effort to quantify uncertainty in constituent flux measurement for flow-interval sampling techniques, water quality data were collected from two small watersheds in central Texas. Each watershed was instrumented with two automated samplers. One sampler was programmed to take high-frequency composite samples to determine the actual load for each runoff event. The other sampler collected discrete samples, from which 15 strategies with 1.32 to 5.28 mm volumetric depth sampling intervals with discrete and composite sampling were produced. Absolute errors were consistently larger for suspended sediment than for NO3-N and PO4-P for both individual event and cumulative loads, which is attributed to differences in the variability of within-event constituent concentrations. The mean event-specific coefficient of variation (CV) ranged from 0.53 to 0.69 for sediment, from 0.38 to 0.39 for NO3-N, and from 0.18 to 0.21 for PO4-P. Event-specific CV values were correlated with the magnitude of absolute errors for individual event loads, with mean r values of 0.52 and 0.57 for the two sites. Cumulative errors were less than +/-10% for all sampling strategies evaluated. Significant differences in load estimate error resulted from changes in sampling interval, but increasing the number of composited samples had no effect; therefore, composite sampling is recommended if necessary to manage the number of samples collected.