Evaluation of SEAWAVE-Q Model for providing daily predictions from non-daily sampled atrazine surface-water concentration monitoring data

Abstract

For atrazine surface water monitoring data with sampling frequencies < daily (e.g., every 7 or 14 days), estimates of target quantities such as maximum m-day rolling averages and upper percentiles can be highly variable and biased. A method using bias factors (BFs) has been proposed by US EPA to minimize the bias. A BF is a multiplicative quantity designed to upwardly adjust estimates so that the adjusted estimates meet or exceed the true value at least 95% of the time (i.e., a BF-adjusted estimate is an upper 95% confidence limit of the true target quantity). BFs can only be determined from monitoring sites where the sampling frequency is daily (i.e., where the true value of a target quantity can be determined), but only a small proportion of monitoring sites possess this property. To increase the number of monitoring sites that can be used to determine BFs, the SEAWAVE-QEX modeling and conditional simulation approach has been proposed by EPA so that monitoring sites with < daily sampling frequencies can also be used to determine BFs. This approach generates conditional simulations (i.e., traces) of daily data, which are then used to determine “true” target quantity values from which BFs are determined. This raises the question of model performance in generating traces that are then used to determine “true” target quantities. In this poster we present an evaluation of the performance of the SEAWAVE-QEX model, and a comparison with other methods. A brief summary of issues related to bias factors determined from daily data also is included, as well as the additional issues that might arise when they are determined from simulated daily data.