Quantitative Evaluation of an Air-monitoring Network Using Atmospheric Transport Modeling and Frequency of Detection Methods

Abstract

A methodology has been developed to quantify the performance of an air-monitoring network in terms of frequency of detection. Frequency of detection is defined as the fraction of "events" that result in a detection at either a single sampler or network of samplers. An "event" is defined as a release to the atmosphere of a specified amount of activity over a finite duration that begins on a given day and hour of the year. The methodology uses an atmospheric transport model to predict air concentrations of radionuclides at the samplers for a given release time and duration. Another metric of interest determined by the methodology is called the network intensity, which is defined as the fraction of samplers in the network that have a positive detection for a given event. The frequency of detection methodology allows for evaluation of short-term releases that include effects of short-term variability in meteorological conditions. The methodology was tested using the U.S. Department of Energy Idaho National Laboratory Site ambient air-monitoring network consisting of 37 low-volume air samplers in 31 different locations covering a 17,630 km region. Releases from six major facilities distributed over an area of 1,435 km were modeled and included three stack sources and eight ground-level sources. A Lagrangian Puff air dispersion model (CALPUFF) was used to model atmospheric transport. The model was validated using historical Sb releases and measurements. Relevant 1-wk release quantities from each emission source were calculated based on a dose of 1.9×10 mSv at a public receptor (0.01 mSv assuming release persists over a year). Important radionuclides were Am, Cs, Pu, Pu, Sr, and tritium. Results show the detection frequency was over 97.5% for the entire network considering all sources and radionuclides. Network intensity results ranged from 3.75% to 62.7%. Evaluation of individual samplers indicated some samplers were poorly located and added little to the overall effectiveness of the network. Using the frequency of detection methods, alternative sampler placements were simulated that could substantially improve the performance and efficiency of the network.