Assessment of small watershed stormwater sampling methods

ABSTRACT Measurements of contaminants in runoff are essential for stormwater management. Sampling accuracy is difficult to assess due to the cost and labor of field studies and the large variation in runoff behavior between sites, rain events, and pollutants. Stormwater sampling requires autosamplers to be programmed based on weather forecasts to accurately sample events, and forecasting errors will affect results in ways that have not been thoroughly examined. We developed an easily interpretable idealized model for runoff pollution and applied the model to sampling methods for a variety of typical conditions observed in small watersheds with moderate to high runoff potential. We examined forecast errors and found a large impact on sampler accuracy, in some situations leading to biases in concentration estimates of greater than 20% along with substantial uncertainty. We found time-paced samplers to generally underestimate concentrations. We analyzed sampler robustness to pollutant and watershed parameters and found that runoff volume had a large effect on sampler accuracy. We developed a novel sampler timing approach that outperformed conventional strategies for overforecast events, and a novel cost-free change to sample compositing that can improve accuracy for underforecast events.