151 Environmental Toxicology and Chemistry, Vol. 21, No. 1, pp. 151–162, 2002  2002 SETAC Printed in the USA 0730-7268/02 $9.00 + .00 SPATIAL PATTERNS IN BENTHIC BIODIVERSITY OF CHESAPEAKE BAY, USA (1984–1999): ASSOCIATION WITH WATER QUALITY AND SEDIMENT TOXICITY BENJAMIN L. PRESTON* Carolina Environmental Program, University of North Carolina–Chapel Hill, Chapel Hill, North Carolina 27599-1105, USA ( Received 15 February 2001; Accepted 2 July 2001) Abstract—Non–point-source pollution is an increasing source of stress to aquatic, estuarine, and marine ecosystems. Such pollution may be of unknown etiology, distributed over extensive spatial scales, and comprised of multiple stressors. Current stressor-based paradigms for ecological risk assessment (ERA) may be insufficient to characterize risk from multiple stressors at regional spatial scales, necessitating the use of effects-based approaches. Historical data (1984–1999) for benthic macroinvertebrate biodiversity in Chesapeake Bay, USA, were incorporated into a geographic information system (GIS) and spatial analysis tools were used to model zones within the bay predicted to be of low or high anthropogenic impact. Data for benthic water quality and sediment toxicant concentrations from each of these zones were subsequently analyzed and compared to identify associations between benthic biodiversity and potential stressors. A number of stressors were significantly associated with high-impact zones, including increased nitrogen and phosphorus concentrations, low dissolved oxygen, heavy metals, pesticides, polycyclic aromatic hydrocarbons, and polychlorinated biphenyls. The spatial autocorrelation among multiple stressors suggests that traditional stressor-based approaches to ERA may result in the a priori exclusion of ecologically relevant stressors. Considering the effects of individual stressors rather than net effects of multiple stressors may result in underestimation of risk. The GISs are a useful tool for integrating multiple data sets in support of comprehensive regional ERA. Keywords—Ecological risk assessment Geographic information systems Water quality Toxicity Benthos INTRODUCTION During the past decade, the use of risk-based decision- making in environmental management has increased signifi- cantly, as evidenced by the proliferation of the ecological risk assessment (ERA) paradigm as the standard method for char- acterizing ecological impacts associated with anthropogenic activities [1]. This paradigm largely centers on identifying potential ecological stressors, characterizing their toxicity through laboratory methods, and subsequently inferring effect concentrations and ecological consequences. As such, ERA currently focuses on characterizing the stressor while effects are estimated through various forms of data manipulation. This approach to risk assessment is valid when information re- garding the potential ecological consequences of contaminant exposure is required before their release, such as in the de- velopment of novel chemical compounds or the siting of an industrial facility that will generate potentially toxic effluent. However, future challenges to environmental quality will likely be quite different from those that have historically been of concern. As water quality criteria become more rigorous and pollution-prevention controls more effective, acute point- source pollution will have a decreasing influence on environ- mental quality. The future challenge to risk assessors and man- agers will be the cumulative effects of chronic exposure to multiple stressors of unknown etiology distributed over vari- able temporal and spatial scales [2]. To meet this challenge, development of effects-based approaches to ecological risk assessment will be necessary [3], whereby observed changes in ecosystem structure and function are used to monitor eco- logical stress and identify stressors. A potential barrier to effects-based risk assessment is the * prestonb@pewclimate.org. availability of sufficient data to characterize ecological effects. Natural populations are frequently associated with significant temporal and spatial variability [4–7], necessitating robust data sets to differentiate natural variability from anthropogenic im- pacts. If such data are to be used in hazard identification, then extensive data collection regarding the distribution and mag- nitude of a broad range of potential stressors must be per- formed as well. Although such data dependence frequently may be perceived as a prohibitive obstacle to effects-based ERA, such data sets are readily available for many regions of the United States. For example, state environmental agencies conduct routine biological monitoring, water quality assess- ment and toxicity screening in pursuance of the requirements of the Clean Water Act. Temporally or spatially extensive en- vironmental monitoring projects also have been sponsored by federal agencies, such as the U.S. Environmental Protection Agency’s Environmental Monitoring and Assessment Program and the U.S. Geological Survey’s National Water Quality As- sessment Program [8,9]. However, historically few concerted attempts have been made to coordinate monitoring programs, even within the same region, and significant variation may exist with respect to spatial resolution and monitoring fre- quency among data sets [9]. Thus, a more difficult challenge is developing methods for integrating multiple large data sets into a construct amenable to ERA so that such valuable data resources can be used effectively in environmental manage- ment [9]. Geographic information systems (GISs) are robust tools for managing data associated with natural landscapes, and their use in the analysis of environmental monitoring data may en- hance the ERA process. The automated functions of com- mercial GISs allow rapid quantification of distance, area, and gradient, and more complex operations can be executed to