Identifying factors contributing to nitrogen and phosphorus levels using a GIS-based prediction model in the Lake Erie watershed, Ontario
Lake Erie is the most susceptible of the Great Lakes to eutrophication, a process during which a waterbody becomes excessively enriched in nutrients. Many streams and rivers from southwestern Ontario drain into Lake Erie, all of which contribute to the eutrophication of Lake Erie and final drainage into Lake Ontario. Nitrogen (N) and phosphorus (P) are key nutrients for plant growth and are typically associated with non-point source pollution from agricultural fields. Increased eutrophication can lead to harmful algal blooms with significant consequences to drinking water and aquatic wildlife habitat. The levels of N and P in these streams are affected by different factors. This project studies the impacts of five different categories of factors (landcover, soil type, soil hydraulic conductivity, slope, and weather) on the N and P levels in streams within the Ontario portion of the Lake Erie watershed. A Random Forest machine learning model was trained on 80% of the data and was used to assess the importance of each of the factors in predicting N and P levels and associated levels in unmonitored streams. The remaining 20% was used as an additional test of the performance of the model. This process produced a layer for each of the identified factors, a model that can predict N and P levels at a point based on the factors affecting the upstream drainage area, and maps showing the predicted N and P levels generated by the predictive model. The most important factors for predicting N and P levels identified by the predictive model were temperature, precipitation, agricultural land cover, natural land cover, and slope with an R2 of 0.438 and an MAE of 1.373.