1. G. P. Patil December 2004 – January 2005 G. P. Patil December 2004 – January 2005 Regional Vulnerability Assessment Issues and Approaches

2. ReVA, Risk Assessment, and Other Programs Monitoring to establish status and trends Association analyses to suggest probable cause where degradation is observed Prioritization of the role of individual stressors as they affect cumulative impacts and risk of future degradation Analysis of trade-offs associated with future policy decisions and development plans Development of strategies to restore areas and reduce risk Important stressors identified Monitoring data used to develop exposure models Suggest new indicators Suggest improvements to monitoring design and indicators Opportunities for regional risk reduction associated with local restoration. Evaluate success in improving decision-making EMAP ReVA Restoration/ Risk Mgmt. Condition, biophysical, and stressor data from other programs Models From Other Programs Integrated Assessment Methods Data for model development Optimization Strategies

3. ECOLOGICAL RESEARCH MULTI-YEAR PLAN Diagnosis Condition of Streams, Estuaries and Landscapes Restoration Forecasting Assessment DESIRED ENVIRONMENTAL CONDITION Assessment

4. Examples of Key Elements from the EPA 2003 Strategic Plan that ReVA will Enhance Improving accountability: Assessing the state of the environment Strengthening partnerships Enhancing risk assessment approaches Enhancing science and research to:  Determine implications and consequences of global change

5.  Improve TMDL and impaired water body assessments and targeting  Sustain, clean-up, and restore communities and the ecological systems that support them  Protect, sustain, and restore the health of natural habitats and ecosystems  Identify and help promote sustainable practices at multiple scales Enhance environmental results through the improved use of quality environmental information by Decision Makers and the Public … environmental data, indicators, and tools

6. From the 2001 ORD Research Strategy Important issues and needs identified by clients and stakeholders :  “The complexity of emerging environmental issues will place a premium on integrated, multimedia, multidisciplinary research”  “Expand our understanding of environmental research so that the findings of economics, sociology, psychology and other social sciences can be incorporated into decision-making”  “New arrangements among EPA and individual citizens, states, tribes, industry, and other organizations must be developed”

7. From the 2001 ORD Research Strategy Important goals that ReVA contributes to:  Integration across environmental science and technology  Anticipating future environmental trends and issues

8. Summary of Important Cross-ORD Contributions of ReVA Multi-scaled environmental targeting approach … because we can’t sample everywhere Methods to extend and enhance results of existing field-based, monitoring programs (e.g., EMAP) Integration of environmental condition estimates in space provide a means of assessing multiple stressors and environmental endpoints at multiple scales

9. Summary of Important Cross-ORD Contributions of ReVA Approach and framework for evaluating alternative future policies and management strategies relative to:  Optimizing for multiple ecological resources and processes (key to sustainability)  Multiple stressors that span EPA Program Offices  Assessing economic and social consequences and feedbacks of different options  Overall quality of life and sustainability

10. Charge Questions Question 1 … the overall approach What are the strengths and limitations of ReVA’s approach to assess current and future regional conditions? How could this approach be improved?

11. Charge Questions Question 2 … Web-based Tool How effective is ReVA’s web-based, Environmental Decision Toolkit (EDT) in communicating ecological condition and vulnerability to decision-makers at regional to local scales? How could the EDT be improved?

12. Charge Questions Question 3 … Consequences of alternative futures How useful is the ReVA approach in allowing decision-makers to determine the overall consequences of future development, mitigation, conservation, and restoration activities? How could this approach be improved?

13. Charge Questions Question 4 … Challenges of Down- scaling and Up-scaling What are the challenges and issues of applying the ReVA approach (including data, models, and integration) at finer scales (e.g., community and watershed scales)? How well has ReVA identified and prioritized research activities and alternative applications related to methods for decision-making at multiple scales?

14. ReVA Regional Vulnerability Assessment Moving from monitoring status and trends to targeting risk management activities

15. ReVA Regional Vulnerability Assessment Informing Decisions Through Synthesis and Forecasting

16. What is the Question? How do we assess cumulative impacts to prioritize risk management? Who is the Client? EPA Program Offices, EPA Regional Offices, State and Local Administrators – National- to local-scale decision makers What have we done so far? Developed regional-scale statistical models that predict environmental conditions Evaluated existing and newly developed methods for the synthesis of existing spatial data Developed a web-based tool that allows decision-makers to ask multiple assessment questions Developed a future scenario that incorporates major drivers of ecological change so that potential changes can be assessed

17. Integrated Science for Ecosystem Challenges (ISEC) – Strategic Priorities Synthesize existing information Improve understanding of effects of multiple stresses Improve assessments and forecasts under alternative policy and management options

18. ReVA’s Working Hypotheses Spatial connections (upstream-downstream, transportation network, shortest air distance) are important in determining the total ramifications of local human activities. Therefore, cumulative ecological condition over large regions is related to large scale patterns as well as small scale decisions. Spatial variability reduces the efficiency of bottom up approaches (and increases the efficiency of top down approaches) when assessing ecological condition over large regions. Sustainability can only be achieved by maintaining regional variability. Some areas must be reserved to maintain regional biodiversity. Some areas are vulnerable to human disturbance.

19. ReVA’s Research Partnerships All of ORD’s Labs and Centers Other Federal Agencies – USGS, USFS, TVA Academia – Florida Atlantic University, Penn State, University of Maryland, Duke University, North Carolina State University, University of NC - Charlotte

20. ReVA’s Client Partnerships EPA Program Offices – Office of Air, Office of Water, Office of Environmental Information EPA Regional Offices: the Mid-Atlantic, the Southeast, and the Mid-West State government agencies – Maryland, and Pennsylvania Cross-jurisdictional alliances – North and South Carolina Councils of Government Feedback Demonstrations Tech transfer

21. ReVA is… Risk assessment research – comparative, cumulative, multi- scale Visualizing spatial patterns of condition and impact Identifying current and future environmental vulnerabilities Projecting potential impacts from major drivers of ecological change Enabling trade-off analyses through “what if” scenarios Informing diagnosis of current conditions Linking environmental health with economic and human health Working directly with clients

22. ReVA Supports Research on: New Indicators New Spatial Models Integration Methods Socio-Economics Decision Tools Quantifying Error and Uncertainty Issues of Scale Information Technology

23. Making Information Accessible: ReVA’s Decision-Support Tools Multiple clients, multiple approaches, multiple tools Both web-based and PC-based New tools and improved existing tools ReVA’s Web-based Decision-Support System ATtILLA

24. Steps in the Process: Acquire/prepare relevant spatial data Develop models, project current conditions Synthesize into information Forecast future conditions Assess implications Engage Decision-makers: look at real-life decisions, alternative futures Make information accessible Improve, refine approaches Look ahead to future needs, opportunities We are here 1998 1999 2001 2002 2004

25. ReVA is estimating condition across the map using existing data

26. ReVA synthesizes environmental data and model results to inform decision-making High Vulnerability Low Vulnerability Providing indices of relative condition and vulnerability

27. Looking Ahead: A Need for New Approaches Despite compliance with environmental regulations, biological populations are declining. Major drivers of change include: Land use change Resource extractions Pollution and pollutants Exotic invasive species Climate change

28. Future Scenarios – Drivers of Change (Futures, Part 1) Land use change – Bird migration scenarios Groundwater vulnerability Landscape indicators N and P loadings Mining – permitted areas Risk of timber harvest (not yet incorporated) “Clear Skies” scenarios for ozone, PM, N and S Human population demographics/ quality of life indicators Risk of NIS (with/without climate change) [ Projected change in forest species (climate change) – future work ] Enabling insights into cumulative/aggregate impacts

29. Value of Approach: Futures Analysis - Results Under projected future conditions, 3 watersheds drop out of the best categories – one on the coast, the others in W VA – for the same reasons: loss of forest habitat and invasive exotic species.

30. Value of Approach: Futures Analysis - Results Most vulnerable area is the foothills region. This result despite using an urban growth model for land use change

31. Applying ReVA Approach and Information to Decision-making (Futures, part 2) Evaluating alternative “Smart Growth” strategies Identifying where to set aside lands for conservation Assessing impacts of alternative incentives for pollution prevention Investigating solutions for “cross boundary” issues associated with air and water quality (e.g. cross-media trading) Estimating impacts of new road development (water quality, air quality, quality of life) Tracking progress/performance

32. Jones, et al., 2000. Landscape Ecology Evaluating Alternative Risk Management Options: Linking Nutrient Loadings with Restoration Potential Current Loadings 10% increase in riparian forest 10% decrease in riparian forest Scenario with restoration Scenario with continued development

33. Turning Spatial Data into Information for Decision Makers Web-based, interactive integration and visualization Data diagnostics and data preparation Integration of data in selectable subgroups Weighting in support of multi-criteria decision making Data access (summarized by reporting unit)

34. How are Decisions Made to Reduce Risk for Vulnerable Ecosystems? Multiple Criteria Stakeholder Input, Politics, Economics, Feasibility, Scientific Understanding Evaluation of Trade-offs Costs/ Benefits of Alternatives 6 8

35. What Makes an Ecosystem “Vulnerable”? Condition Pristine, Good, Stressed, Degraded Sustainability f (ecosystem sensitivity; stressors affecting) Value to Society Aesthetics, Economic Opportunities, Goods and Services What Drives Risk Management Decisions? Feasibility, Clear Options, Economics What works where?, Range of method applicability Need to Address Multiple Assessment Questions

36. Current Uses of Approaches and Tools R3 – Strategic Planning – Vulnerable populations, Watershed health, Responsible development - outreach and partnerships MD DNR, PA DEP, Baltimore County – outreach, identification of priority areas for protection, alternative scenarios of development SEQL project – alternative scenarios of development, opportunities for cross-media trading, focus on quality of life

37. Future Work in ReVA Water supply modeling with USGS – Mid Atlantic (Region 3) and SEQL Initiate work in SE (Region 4) – pilot on vulnerability of human and wildlife populations to air toxics Pilot work in Midwest (Region 5) – 1) decision support for hazardous wastes mitigation – Net Environmental Benefits Assessment; and 2) IT research – webservices in support of compliance reporting and analysis for state water data

38. Research Issues/Needs Indicator/ Model Domains of Scale Changing Reporting Units (reaggregation of data, model results) Quantifying error and uncertainty Feedbacks and interactions Incorporating thresholds Minimizing degradation, optimizing opportunities “Translators” where data don’t address questions specifically

39. Research Opportunities Nested broad – to fine –scale applications (e.g. Baltimore Co., MD, Mid-Atlantic; SEQL, Southeast) New methods to interpolate (e.g. Bayesian methods) Model output as surface maps (finer-scale) Cross- media trading Estimating error

40. Getting to Outcomes Decisions incorporate ReVA approach and information Client partnerships at regional, state, and local levels Demonstrations of application of approach and information at different scales Multiple clients – Multiple decision support tools Easy access to data and tools

41. ReVA as a Screening Tool  Provides an early warning  Identifies where we need to look closer  Puts issues in perspective  Identifies opportunities  Identifies future issues  Targets use of limited resources

42. Remaining Agenda Data and models Acquiring and preparing spatial data; Developing and applying statistical landscape models Integration Synthesis into Information Data Issues, Assessment Questions Forecasting Drivers of Change Applications Engaging Decision-makers, addressing real decisions Making Information Accessible Multiple Decision-makers, multiple needs

43. Data Issues The ReVA data analysis effort includes exploring a number of integration and modeling techniques in order to address key ReVA objectives. Integrating data from multiple sources can be difficult for the following reasons:

44. Data Issues Different scales/units of metric data. Data need to be reaggregated to one scale or spatial unit prior to integration and comparison. Applicability of metric throughout the region. Do data apply throughout the region or only in particular areas? Accuracy of the data. Are the metric errors small enough for the data to be meaningful?

45. Data Issues Metric sensitivity to small changes. How large a change in data is required to cause a change in the relevant endpoint? Organizational level. What does the metric relate to, e.g. population, community, ecosystem, watershed, forestry, etc.? Ease of interpretation. Can the metric be easily interpreted by the public and decision-makers? Is the metric relevant to societal values and risk management objectives?

46. Data Issues Range of data for the metric. Can the total range of the metric be specified? Is there a reference condition that can be used to specify an "ideal" value? Is literature available to determine the threshold of the metric, e.g. undesirable values based on toxicology or serious ecological impacts? Losing information when integrating. What integration method will ensure that the least amount of information from the data is lost?

47. Data Issues Uniqueness of metrics - correlation between variables. When integrating data, need to make sure that metrics are actually different measures of the same thing and/or are not calculated from other variables or develop methods to adjust for this bias. Correlated variables are variables that are dependent on each other and actual measure the same effect. One wants variables to be independent so that the variables are actually measuring different things. Social values or stakeholder interests may necessitate weighting of variables especially when the variables also represent decision criteria.

48. Data Issues Different data will have different significance based on what is important to the interested parties. For example, if stakeholders were looking at social value data, rankings would depend on what is important to the stakeholders. Relative rankings of the area of interest (e.g. watershed, EMAP hexes) would be important.

49. Integration Methods One purpose of ReVA is to develop and evaluate integration methods. Based on a preliminary analysis of integration methods performed on a complete suite of variables in the Mid-Atlantic region, we have classified the methods into several different groups listed below. The methods that rank condition appear to have similar results, however this may not be the case in other regions or for subsets of data. It is recommended a suite of integration methods be employed to explore different aspects of environmental condition and vulnerability.

50. Integration Methods The integration methods can be organized into several different groups: Methods that rank by condition - these all give similar maps: –Quintiles - Best/Worst –Simple Sum –Analytical Hierarchy Process (AHP) Methods that rank by distance to some reference condition –State-Space Analysis –Principle Component Analysis (PCA) –Criticality Analysis

51. Integration Methods Methods that rank by vulnerabilities –Stressor/resource overlay –Stressor/resource matrix Methods that group by similar characteristics –Cluster analysis –Self-organizing maps (SOM) Visualizations and trade-offs –User-specified weightings –Radar plots

52. Quintiles This quintile summary is widely used for spatial analysis and an example is provided in the Landscape Atlas of the Mid-Atlantic Region (Jones et al. 1997, Wickham et al. 1999). This approach divides the total numeric range of each variable into five equal subdivisions. The highest quintile represents the largest values for the resources and the lowest values for the stressors. The summary map for best quintile color codes the number of variables in the best quintile for a given watershed. Those with the most variables in the best quintiles are colored in green. Similarly, the worst quintile color codes the number of variables in the worst quintile for a given watershed. Those with the most variables in the worst quintiles are colored in red.

53. Quintiles

54. Simple Sum This approach uses a weighted average method to weight all of the variables. This particular instance of the weighted average method weights all of the variables the same and does not allow the user to choose specific weights.

55. Simple Sum

56. Analytical Hierarchy Process The Analytic Hierarchy Process (Saaty 1980) is a method intended to use expert judgements to cluster variables into groups and and weight those groups by their relative importance or stakeholder input. In use, AHP may utilize more politics than expert judgement. An outcome score for each watershed would be generated based on the user choices for groups and weights. In ReVA's application of AHP, the variables are clustered automatically into groups by principal component analysis, and weighted by the principal component scores. Those automated scores are translated into an outcome score for each watershed. Those watersheds scoring best under AHP are colored in green, while those scoring poorly are colored in red.

57. State-Space The State Space method uses the correlation matrix directly to transform the space to near independent axes. The transformation differs from PCA, as the State Space uses the correlation matrix directly, and PCA uses the correlation indirectly. The relative condition of each watershed is determined by measuring the Euclidean distance between them. As in the other methods, the watersheds in the best condition come out green, while those in poor condition appear red.

58. State-Space

59. Principal Component Analysis This method transforms the data so that all variables are roughly independent of one another. Then, the distance from the transformed scores to the ideal score are calculated. The map displayed goes from green to red, representing closest to farthest from ideal conditions.

60. Principal Component Analysis

61. Criticality Analysis This method involves measuring the distance between each watershed and a natural state, representing the historical conditions under which the system evolved. The concept is that the further the system has been removed from the conditions under which it evolved the existing adaptive and recovery mechanisms, the greater the probability that it will go critical and change state (Dubois 1979, Gatto and Renaldi 1987). Those watersheds that have moved the smallest distance from their "natural" state are colored in green. Those watersheds farthest from their "natural" state are colored in red.

62. Criticality Analysis

63. Stressor/Resource Overlay This method is a modification of Parkhurst et al. (1997). Each resource is regressed on the stressor variables using multivariate regression. The resulting regression coefficients relate each stressor to each resource creating a stressor/resource coefficient matrix. The rows are summed, with the largest resulting sums showing which resources are affected most by stress. The columns are also summed to show which stressor has the most impact on resources. This is not a graphable method, but the variables of most influence can be summarized in a table.

64. Cluster Analysis This method clusters variables based on the similarity of their variables. This method groups units together based on similar values for variables, different from SOM. This particular clustering does not make judgments on whether one cluster is better or worse than another cluster. It only seeks to put similar groups together. Groups in SOM are created based on some relation to an objective, i.e. this group is bad/good compared to the objective. The clusters in the cluster analysis are just based on similarity of variables with no judgment made on whether the cluster is bad or good. The coloring for the cluster analysis is intentional to make a distinction between the green-to-red ordering of the other methods.

65. Self-Organizing Map A Self-Organizing Map (SOM) groups watersheds into quality categories based on their environmental variables. The emphasis of SOM is to associate watersheds with similar variables. Since they are grouped into quality categories, they are color-coded from green to red to show the progression to groups with high environmental qualities to groups with low environmental qualities. The clusters in the cluster analysis are just based on similarity of variables with no judgment made on whether the cluster is bad or good. Groups in SOM are created based on some relation to an objective, i.e. this group is bad/good compared to the objective.

66. User-specified Weighted Average This is a simple approach that adds resources and subtracts stressors to arrive at an estimate of average or overall condition. This approach allows the users to determine the weights they wish to assign to particular groups of variables for summary. For instance, if forest and water variables are more highly valued by stakeholders, all variables in that group can be weighted very high in relation to other variables. Those variables considered unimportant can be omitted by giving zero weight. The resulting scores are influenced by the weighting, and are displayed on a green to red (best to worst) scale. The default summary given in the ReVA application is for each variable to have equal weight.

67. Radar Plots A radar plot is a pie chart for each watershed that allows the user to look at all the variables for a watershed at once to get a sense of the overall condition of that watershed thus allowing a "visual integration." Each pie slice represents one variable, with the slice being filled in with green to match the level of resource for that variable. A variable at the optimal level will have a completely green pie slice, and a variable with the least optimal level will have a slice that appears all grey. The accompanying figure provides an example of a this visualization method. The area of the radar plot could be used to determine how a watershed ranks relative to other watersheds. In the interactive version, you can highlight dots at the end of a pie slice to see which variable is being measured. Right-clicking on this dot will give you options to see a table of all variable values for the watershed, or to see that variable plotted for all watersheds.

68. Radar Plots