A Statistical Modeling Methodology for the Detection, Quantification, and Prediction of Ecological ThresholdsEPA Grant Number: R832447
Title: A Statistical Modeling Methodology for the Detection, Quantification, and Prediction of Ecological Thresholds
Investigators: Richardson, Curtis J. , Qian, Song S.
Institution: Duke University
EPA Project Officer: Hiscock, Michael
Project Period: February 1, 2005 through June 30, 2007 (Extended to August 31, 2008)
Project Amount: $278,876
RFA: Exploratory Research: Understanding Ecological Thresholds In Aquatic Systems Through Retrospective Analysis (2004) RFA Text | Recipients Lists
Research Category: Ecosystems , Water , Aquatic Ecosystems
The primary objective is to develop a statistical modeling approach for the detection and quantification of ecological thresholds that can be used in aquatic ecosystems worldwide. We will use a Bayesian hierarchical modeling approach to integrate the single-species change point analysis method developed by Qian et al. (2003) to study the interactions among fast and slow responding species or ecological metrics indicating alternative stable states. Under the structural approach, the resulting species metric thresholds will be integrated to form an ecosystem-level threshold distribution. A risk-based criteria approach will be developed to assess ecological resilience and predict the threshold for alternative state changes.
The project will use two long-term (13 -year) datasets collected following EPA QA/QC guidelines by the Duke University Wetland Center in the Everglades. These studies provided both: (1) long-term observations along a man-made nutrient and hydrologic gradient, and (2) experimental data from a replicated mesocosm P-threshold experiment. The database includes ecological metric data from fast (periphyton), intermediate (macroinvertebrates and macrophytes), and slow processes (peat accretion, nutrient accretion in the peat) from both the natural state and alternative states. Data is also available from a series of explicit spatial analysis sampling studies of the responses of the phosphorus-limited Everglades wetland ecosystems to anthropogenic and hydrologic shifts.
Our new approach includes: (1) Selecting from a collection of species ecological metrics the best indicators of ecological change in the Everglades, based on a screening analysis of the relationship between metric responses over time and space. These ecological attributes will cover five trophic levels, allowing for the study of the relationships slow changing soil variables with fast and intermediate biotic responses. Importantly, the data characterize both internal (soil P concentrations and soil efflux data) and external driving variables (water column P loadings and nutrient) concentrations). (2) Modifying and applying our new integrated Bayesian change point methodology to the selected metrics. (3) Developing and refining the Bayesian hierarchical model to integrate the single species metric threshold distribution across time and spatial scales.
The statistical change point analyses are uniquely suitable for this study, because they include information on both mean and variance statistics in the population response, address nonlinear responses, and can provide a risk assessment probability analysis, leading to an “early warning” of ecosystem threshold approach. The study’s goal is the development of a new, integrated easily tractable threshold approach applicable to other aquatic ecosystems worldwide, not revalidating the P threshold for the Everglades. This will be accomplished in part by reducing the number of required metrics needed in the final model to the minimum number required to allow accurate predictions of shifts in alternative ecological states.
This study will provide a general methodology for integrating threshold information from multiple species ecological metrics, allow for prediction of changes of alternative stable states, and provide a risk assessment tool that can be applied to adaptive management. The integrated threshold distributions will provide a quantification of ecosystem resilience response to environmental disturbance.