Model Performance Evaluations: How Many Receptors Should be Used for Short Term Standard Evaluations for Gaussian Models

摘要

In USEPA Guidance for conducting Model Performance Evaluations, there is no clear guidance for how many receptors should be placed around monitor sites used in developing the performance metrics used for evaluating gaussian models. Historical performance evaluations involving longer averaging times than 1-hour have typically taken only a receptor located at the monitor site into account. With longer averaging times, the hourly variability in the wind direction and speed in the model inputs is smoothed reducing the need to consider more than one receptor site for each monitor. More recently, one-hour standards have become more prevalent for pollutants with impacts that are typically evaluated using use Gaussian models. However, when 1-hour standards are being evaluated, the wind direction and wind speed become more critical in the evaluation as the hourly values cannot benefit from the averaging over time of the wind direction and speed. Since the advent of the 1-Hour SO_2 and NO_2 Standards, studies of gaussian model performance have been presented that have used between a single receptor located at the monitor site and matrices of various sizes centered on the monitor site. The conclusions of these studies have varied seemingly without regard to the number of receptors considered in the analysis. After being part of the team that performed a one year on site performance evaluation in 2018 using a 9x9 receptor grid centered on each monitor site (Long and Beekman, 2018) and observing the existence of gradients within the receptor grids around the individual monitors, the question as to the optimal receptor configuration for this or any model performance study focused on hourly model performance was brought to the fore. This study reprocessed the modeling results for the national seivice data base case used by Long and Beekman to examine the impacts on the reported results by using various receptor arrangements ranging from a single receptor at the monitor location to the 9x9 receptor configuration in the original paper. This allows an evaluation of the impacts of differing numbers of receptors on calculated model performance, allowing insight into the effect of receptor numbers on calculated model performance.