Modeling phosphorus trapping in wetlands using nonparametric Bayesian Regression

Abstract

Phosphorus‐enriched agricultural runoff from the Everglades Agriculture Area is believed to have caused ecological changes in the northern part of the Everglades wetlands. A number of efforts have been made to assess the effectiveness of using constructed wetlands as a means of phosphorus removal from the agricultural runoff. The objective of this study is to develop a predictive model for the total phosphorus effluent concentration of an Everglades wetland that has received this runoff for over 20 years. We used Bayesian nonparametric regression to develop a predictive model combining information from an Everglades wetland data set and a cross‐sectional data set. The prior model was based on the cross‐sectional data set and expert opinion; this prior model, when combined with data from the Everglades wetland, yielded the posterior model, which can be used to (1) estimate the probability of an outflow concentration standard violation and (2) provide the posterior distributions of effluent concentrations at different loading rates and water levels. The primary use of this model is to support decision making in sizing the proposed constructed wetlands in south Florida as well as keeping a practical management strategy.