A Regression Tree based Assessment of Wastewater Treatment Facility Compliance with 1 Current and More Stringent Ammonia Discharge Limits

8 A regression tree-based diagnostic approach is developed to evaluate factors affecting US 9 wastewater treatment plant compliance with ammonia discharge permit limits. Discharge Monthly 10 Report (DMR) data from a random sample of 106 municipal treatment plants: 5% of the EPA 11 ICIS/ECHO database, for the period of 2004-2008 is used to fit this model. Predictor variable 12 importance is determined using random forests, and selected variables, including previous month’s 13 ammonia discharge, influent flow rates and plant capacity utilization, are used to fit a regression 14 tree. The tree models are used to evaluate compliance with existing ammonia discharge standards 15 and with anticipated more stringent discharge limits under consideration in many states. The model 16 predicts that the ability to meet both current and future limits primarily depends on the previous 17 month’s treatment performance. The likelihood of violating existing ammonia limits in any month 18 depends on the previous month’s performance and the plant inflow rate; while non-compliance 19 with more stringent future limits is associated only with the previous month’s treatment 20 performance. More stringent discharge limits result in up to 18% increases in predicted ammonia 21 concentration relative to discharge limits. In-sample validation shows that the regression trees can 22