Priority Substances List Assessment Report for Ammonia in the Aquatic Environment

Appendix D Development of Exposure CDFS for The North Saskatchewan River

The following steps were taken to enable development of exposure CDFs for the North Saskatchewan River.

• A rough sensitivity analysis was carried out to determine which input variables had the most influence on predicted ammonia levels in the river (see "Sensitivity analysis of river parameters on the size of the potentially toxic zone" in technical supporting document [  Environment Canada, 2000]).
  
  
• Based on the results of the sensitivity analysis, five important input variables were identified; these were river flow, temperature, pH, effluent flow and total ammonia concentration. There were concerns with uncertainties about the monthly geometric mean for ammonia levels in the North Saskatchewan River. Thus, monthly arithmetic means were estimated for the normally distributed input variables (river temperature and pH) and monthly geometric means for the lognormally distributed variables (river and effluent flow, effluent ammonia concentration). Using the calculated means, normal distributions were fit to temperature and pH (the latter is already on a log scale), and lognormal distributions were fit to the remaining
  
  
• For each of these five parameters, the available data were analysed to develop distributions of each. Each parameter distribution was then fed into a statistical package (i.e., Crystal Ball) to expand the distribution using Latin Hypercube sampling so as to develop a larger data set with which to make predictions. The measured correlations between the input variables (e.g., effluent flow and ammonia concentration) were specified prior to the sampling exercise.
  
  
• Five hundred sets of parameters were then developed and fed into CORMIX to develop predictions for a typical August at six distances downstream of the effluent outfall (1, 2, 5, 10, 15 and 20 km). The outputs from this exercise were 500 estimates of un-ionized ammonia levels (mid-plume) at each of several distances between 1 and 20 km downstream of the plant. Lognormal distributions were then fit to the data at each distance downstream.
  
  
• This distribution of data points was divided into ranges equal in size (each range was 0.000 58 units). The number of data points in each range was used to develop an exposure CDF for each of the distances (1, 2, 5, 10, 15 and 20 km) downstream of the Gold Bar WWTP.