LakeTrophicModelling

LakeTrophicModelling predicts and classifies lake trophic state using Random Forests trained on in situ water quality (including the National Lakes Assessment) and GIS-derived land use/land cover and lake morphometry data to model chlorophyll-a concentrations.

Key Features:

• Data Integration: Integrates in situ water quality from the National Lakes Assessment with GIS-derived land use/land cover and lake morphometry data for modeling.
• Random Forests: Uses Random Forests to model chlorophyll-a concentrations and capture complex interactions among predictors.
• Dual-Model Framework: Implements a Full Model (in situ + GIS) with MSE 0.09, adjusted R² 0.8, and 69% classification accuracy, and a GIS-Only Model with MSE 0.22, adjusted R² 0.48, and 49% classification accuracy.
• Prediction Probabilities: Provides prediction probabilities to quantify uncertainty, ranging 0.42–1 for the Full Model and 0.33–0.96 for the GIS-Only Model.

Scientific Applications:

• Ecosystem Monitoring: Predicts trophic state to monitor ecosystem condition and risks such as harmful algal blooms.
• Resource Management: Informs management of nutrient inputs and other environmental drivers affecting lake health.
• Research and Education: Supplies modeled chlorophyll-a and uncertainty estimates for limnology research and environmental science education.

Methodology:

Collects in situ water quality, GIS, and other datasets; develops Random Forest models to predict chlorophyll-a concentrations; classifies lakes into trophic states based on modeled chlorophyll-a; and estimates uncertainty via prediction probabilities.