Scientific Investigations Report 2012–5030
AbstractUrban development alters important physical, chemical, and biological processes that define urban stream ecosystems. An approach was developed for quantifying the effects of these processes on aquatic biota, and then linking those effects to endpoints that can be used for environmental management. These complex, interacting systems are challenging to model from a scientific standpoint. A desirable model clearly shows the system, simulates the interactions, and ultimately predicts results of management actions. Traditional regression techniques that calculate empirical relations between pairs of environmental factors do not capture the interconnected web of multiple stressors, but urban development effects are not yet understood at the detailed scales required to make mechanistic modeling approaches feasible. Therefore, in contrast to a fully deterministic or fully statistical modeling approach, a Bayesian network model provides a hybrid approach that can be used to represent known general associations between variables while acknowledging uncertainty in predicted outcomes. It does so by quantifying an expert-elicited network of probabilistic relations between variables. Advantages of this modeling approach include (1) flexibility in accommodating many model specifications and information types; (2) efficiency in storing and manipulating complex information, and to parameterize; and (3) transparency in describing the relations using nodes and arrows and in describing uncertainties with discrete probability distributions for each variable. In realization of the aforementioned advantages, a Bayesian network model was constructed to characterize the effect of urban development on aquatic macroinvertebrate stream communities through three simultaneous, interacting ecological pathways affecting stream hydrology, habitat, and water quality across watersheds in the Northeastern United States. This model incorporates both empirical data and expert knowledge to calculate the probabilities of attaining desired aquatic ecosystem conditions under different urban stress levels, environmental conditions, and management options. Ecosystem conditions are characterized in terms of standardized Biological Condition Gradient (BCG) management endpoints. This approach to evaluating urban development-induced perturbations in watersheds integrates statistical and mechanistic perspectives, different information sources, and several ecological processes into a comprehensive description of the system that can be used to support decision making. The completed model can be used to infer which management actions would lead to the highest likelihood of desired BCG tier achievement. For example, if best management practices (BMP) were implemented in a highly urbanized watershed to reduce flashiness to medium levels and specific conductance to low levels, the stream would have a 70-percent chance of achieving BCG Tier 3 or better, relative to a 24-percent achievement likelihood for unmanaged high urban land cover. Results are reported probabilistically to account for modeling uncertainty that is inherent in sources such as natural variability and model simplification error. |
First posted April 2, 2012 For additional information contact: Part or all of this report is presented in Portable Document Format (PDF); the latest version of Adobe Reader or similar software is required to view it. Download the latest version of Adobe Reader, free of charge. |
Kashuba, Roxolana, McMahon, Gerard, Cuffney, T.F., Qian, Song, Reckhow, Kenneth, Gerritsen, Jeroen, and Davies, Susan, 2012, Linking urbanization to the Biological Condition Gradient (BCG) for stream ecosystems in the Northeastern United States using a Bayesian network approach: U.S. Geological Survey Scientific Investigations Report 2012–5030, 48 p. (Available online at http://pubs.usgs.gov/sir/2012/5030/.)
Abstract
Introduction
Purpose and Scope
Methods
USGS Urban Stream Data
Biological Condition Gradient
Assigning BCG Tier Membership in the Northeastern United States
Bayesian Network Model
Creating a Bayesian Network Model
Developing the Northeastern U.S. Bayesian Network Prior Model Using Expert Elicitation
Model Structure
Variable Selection and Discretization
Conditional Probability Tables
Prior Weights
Updating the Prior Model with Data
Predicting Effects of Urbanization on Biota
Prior Bayesian Network Model
Data-Only Bayesian Network Model
Posterior Bayesian Network Model
Assessing the Value of a Bayesian Network Approach
Benefits of Using a Bayesian Network for Urban Development Modeling
Unresolved Issues
Conclusions
References Cited
Appendix 1. Distribution Forms for Bayesian Updating
Appendix 2. Supplemental Prior and Posterior Conditional Probability Tables, Data Tables, and Bayesian Network Diagrams