A Bayesian Hierarchical Model for Continental-Scale Prediction of Water Quality in US Lakes

• Activity Number: 678 - New Methods in Spatial and Spatiotemporal Modeling and Assessment
• Type: Contributed
• Date/Time: Thursday, August 2, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Meridith Bartley* and Ephraim Hanks
• Companies: Penn State University and The Pennsylvania State University
• Keywords: spatially varying coefficients; cross-scale interaction; limnology; water quality; missing data; bayesian hierarchical modeling

Abstract

In addition to providing diverse ecosystems for a variety of aquatic organisms, replenishment for fresh groundwater, and providing ample opportunities for recreation, the vast system of lakes in the continental US store and transfer vital nutrients such as Carbon (C), Nitrogen (N), and Phosphorus (P). Understanding the water quality in these lakes allows for informed ecosystem management and better predictions of the environmental impacts of climate change. With data for 50,000 lakes in in the northeast United States, we develop a model for predicting nutrients for all lakes in the continental US. We propose a bayesian hierarchical model that models regional differentiation through spatially-varying coefficients, correlation in vital nutrients through a joint multivariate model, and cross-scale interaction effects of covariates. Missing data are imputed jointly with model fitting. Increasing our understanding of these water properties on a continental-scale may lead to a more comprehensive understanding of relationships between ecological drivers and linked nutrient cycles.