Evaluating Standards Using Data Collected From Regional Probabilistic Monitoring Programs

Eric P. Smith 1, Keying Ye 2

1 Department of Statistics Virginia Tech, Blacksburg, VA 24061 Email: epsmith@vt.edu
Phone: (540) 231-7929 Fax: (540) 231-3863
2 Department of Statistics, Virginia Tech, Blacksburg, VA 24061

Under section 303 (d) of the Clean Water Act, states must identify water segments, where loads of pollutants are violating numeric water quality standards. A common approach is to use data collected at a single site to make a decision to list or not list the site. This might be based on a raw score approach in which a stream segment is listed as impaired when greater than 10% of the measurements of water quality conditions exceed a numeric criteria or a Binomial test that the probability of violation is 0.10. While this approach is useful for single site evaluation, it gives little information about regional violation rates. In addition, for the approach to have reasonably good error rates, moderately large sample sizes are requires (20 samples). Recent changes in monitoring programs focus on a more regional assessments of environmental conditions. To evaluate violations of standards we propose a regional model that is more closely related to the sampling designs based on rotating panel surveys. The approach uses mixed models and small area estimation techniques. We show how to use the small area estimation techniques along with the accepting sampling by variables method to make the decision to list or not list the segment. In addition, the method results in an evaluation at the regional level. Some of the benefits and limitations of the approach are discussed.

Keywords: TMDL, Monitoring, Standards, Acceptance Sampling by Variables, Binomial Distribution, Small Area Estimation, BLUP