Abstract
Total phosphorus (TP) nutrient calculations for Coeur d Alene Lake tributaries in Idaho, USA were modeled using ordinary least squares (OLS). Results were revised from a March 2016 model using updated stream and loading data. The revised multivariate regression analyses examined total phosphorus as a function of land use/ land cover parameters. I used NHD24k stream data and included stream slope into the variable matrix. I tested prior variable selection using model selection. Since we have only 12 observations, only a maximum of nine variable can be included in the Multiple Regression Analysis. To select these nine variables, I used Random Forest Variance Importance for 10,000 simulations. Based on the model selection criteria, Model 4 (OLS model without wetland variables in the data set & with stream gradient in the model) followed by Model 1 (OLS model without wetland variables in the data set) seem to be the best supported by the data. Data set includes further updates for the June 2017 season.