Wesley R. Brooks | USGS Science

USGS2016

Predicting recreational water quality advisories: A comparison of statistical methods

Epidemiological studies indicate that fecal indicator bacteria (FIB) in beach water are associated with illnesses among people having contact with the water. In order to mitigate public health impacts, many beaches are posted with an advisory when the concentration of FIB exceeds a beach action value. The most commonly used method of measuring FIB concentration takes 18–24 h before returning a result. In order to avoid the 24 h lag, it has become common to ”nowcast” the FIB concentration using statistical regressions on environmental surrogate variables. Most commonly, nowcast models are estimated using ordinary least squares regression, but other regression methods from the statistical and machine learning literature are sometimes used. This study compares 14 regression methods across 7 Wisconsin beaches to identify which consistently produces the most accurate predictions. A random forest model is identified as the most accurate, followed by multiple regression fit using the adaptive LASSO.

Environmental Modelling and Software↗

USGS2013

Partial least squares for efficient models of fecal indicator bacteria on Great Lakes beaches

At public beaches, it is now common to mitigate the impact of water-borne pathogens by posting a swimmer's advisory when the concentration of fecal indicator bacteria (FIB) exceeds an action threshold. Since culturing the bacteria delays public notification when dangerous conditions exist, regression models are sometimes used to predict the FIB concentration based on readily-available environmental measurements. It is hard to know which environmental parameters are relevant to predicting FIB concentration, and the parameters are usually correlated, which can hurt the predictive power of a regression model. Here the method of partial least squares (PLS) is introduced to automate the regression modeling process. Model selection is reduced to the process of setting a tuning parameter to control the decision threshold that separates predicted exceedances of the standard from predicted non-exceedances. The method is validated by application to four Great Lakes beaches during the summer of 2010. Performance of the PLS models compares favorably to that of the existing state-of-the-art regression models at these four sites.

Journal of Environmental Management↗

USGS2013

Characterizing response of total suspended solids and total phosphorus loading to weather and watershed characteristics for rainfall and snowmelt events in agricultural watersheds

Understanding the response of total suspended solids (TSS) and total phosphorus (TP) to influential weather and watershed variables is critical in the development of sediment and nutrient reduction plans. In this study, rainfall and snowmelt event loadings of TSS and TP were analyzed for eight agricultural watersheds in Wisconsin, with areas ranging from 14 to 110 km2 and having four to twelve years of data available. The data showed that a small number of rainfall and snowmelt runoff events accounted for the majority of total event loading. The largest 10% of the loading events for each watershed accounted for 73–97% of the total TSS load and 64–88% of the total TP load. More than half of the total annual TSS load was transported during a single event for each watershed at least one of the monitored years. Rainfall and snowmelt events were both influential contributors of TSS and TP loading. TSS loading contributions were greater from rainfall events at five watersheds, from snowmelt events at two watersheds, and nearly equal at one watershed. The TP loading contributions were greater from rainfall events at three watersheds, from snowmelt events at two watersheds and nearly equal at three watersheds. Stepwise multivariate regression models for TSS and TP event loadings were developed separately for rainfall and snowmelt runoff events for each individual watershed and for all watersheds combined by using a suite of precipitation, melt, temperature, seasonality, and watershed characteristics as predictors. All individual models and the combined model for rainfall events resulted in two common predictors as most influential for TSS and TP. These included rainfall depth and the antecedent baseflow. Using these two predictors alone resulted in an R 2 greater than 0.7 in all but three individual models and 0.61 or greater for all individual models. The combined model yielded an R 2 of 0.66 for TSS and 0.59 for TP. Neither the individual nor the combined models were substantially improved by using additional predictors. Snowmelt event models were statistically significant for individual and combined watershed models, but the model fits were not all as good as those for rainfall events (R 2 between 0.19 and 0.87). Predictor selection varied from watershed to watershed, and the common variables that were selected were not always selected in the same order. Influential variables were commonly direct measures of moisture in the watershed such as snowmelt, rainfall + snowmelt, and antecedent baseflow, or measures of potential snowmelt volume in the watershed such as air temperature.

Wisconsin↗