Multimodeling Framework for Predicting Water Quality in Fragmented Agriculture-Forest Ecosystems

Both livestock and wildlife are major contributors of nonpoint pollution of surface water bodies. The interactions among them can substantially increase the chance of contamination especially in fragmented agriculture-forest landscapes, where wildlife (e.g. white tailed deer) can transmit diseases between remote farms. Unfortunately, models currently available for predicting fate and transport of microorganisms in these ecosystems do not account for such interactions. The objectives of this study are to develop and test a multimodeling framework that assesses the risk of microbial contamination of surface water caused by wildlife-livestock interactions in fragmented agriculture-forest ecosystems. The framework consists of a modified Soil Water Assessment Tool (SWAT), KINematic Runoff and EROSion model (KINEROS2) with the add-on module STWIR (Microorganism Transport with Infiltration and Runoff), RAMAS GIS, SIR compartmental model and Quantitative Microbial Risk Assessment model (QMRA). The watershed-scale model SWAT simulates plant biomass growth, wash-off of microorganisms from foliage and soil, overland and in-stream microbial transport, microbial growth, and die-off in foliage and soil. RAMAS GIS model predicts the most probable habitat and subsequent population of white-tailed deer based on land use and crop biomass. KINEROS-STWIR simulates overland transport of microorganisms released from soil, surface applied manure, and fecal deposits during runoff events at high temporal and special resolutions. KINEROS-STWIR and RAMAS GIS provide input for an SIR compartmental model which simulates disease transmission within and between deer groups. This information is used in SWAT model to account for transmission and deposition of pathogens by white tailed deer in stream water, foliage and soil. The QMRA approach extends to microorganisms inactivated in forage and water consumed by deer. Probabilities of deer infections and numbers of infected animals are computed based on a dose-response approach, including Beta Poisson and Maximum Risk models, which take into account pathogen variation in infectivity. An example of the Multimodeling framework performance for a fragmented agriculture-forest ecosystem will be shown in the presentation.