DRAFT A Coarse-scale Assessment of the Relative Habitat Value Marine Shoreline Units for the Conservation Fish and Wildlife in Puget Sound Basin

Transcription

1 DRAFT A Coarse-scale Assessment of the Relative Habitat Value of Marine Shoreline Units for the Conservation of Fish and Wildlife in Puget Sound Basin George F. Wilhere, Timothy Quinn, and Dale Gombert Washington Department of Fish and Wildlife Habitat Program Olympia, Washington February 2012

2 Publication and Contact Information For more information contact: George Wilhere Habitat Program Washington Department of Fish and Wildlife Olympia, WA Phone: Acknowledgments Scott Pearson and Dayv Lowry of WDFW, Curtis Tanner of U.S. Fish and Wildlife Service, Paul Cereghino of National Marine Fisheries Service, Helen Berry of the Department of Natural Resources, Hugh Shipman of the Department of Ecology, and Charles Si Simenstad of the University of Washington reviewed portions of this assessment and provided insightful advice. Scott Campbell of the U.S. Army Corps of Engineers provided assistance and advice for processing the PSNERP GIS data. Andy Weiss of WDFW provided guidance and oversight on numerous GIS tasks. Ken Pierce of WDFW routed the shoreline segments. This work was funded in part by the U.S. Environmental Protection Agency - Aquatic Ecosystems Unit, Wetland Program Development Grant No. CE which was awarded to the Department of Ecology. Preferred citation: Wilhere, G.F., T. Quinn, and D. Gombert A Coarse-scale Assessment of the Relative Habitat Value for Marine Shoreline Units for the Conservation of Fish and Wildlife in Puget Sound Basin, draft report. Washington Department Fish and Wildlife, Habitat Program. Olympia, WA. ii

3 Table of Contents I. Executive Summary II. Introduction III. A Framework for Multi-scale Assessments IV. Conceptual Model V Methods VI. Results VII. Discussion References Appendix A: Miscellaneous Tables iii

4 I. Executive Summary Over the next two decades over 1 million additional people are expected to inhabit the Puget Sound Basin (OFM 2007). Thousands of acres of agricultural and timber lands will be converted to residential and commercial uses in order to accommodate this phenomenal growth. In addition to providing valuable commodities these working lands, both agricultural and timber lands, provide habitats for wildlife. To ensure the health and well being of their citizens, promote orderly and efficient land use, and protect natural resources (including fish and wildlife), city and county governments implement comprehensive plans and regulatory land use zoning. To fully realize smart growth, comprehensive land use plans must be based on scientifically-credible information that indicates the most importance areas for the conservation of fish and wildlife habitats areas where development should be avoided. Our purpose is to provide useful, scientifically-credible information for smarter growth in the Puget Sound Basin. The Puget Sound Watershed Characterization is a set of spatially explicit assessments that provide information for regional, county, and watershed-based planning. It is a coarse-scale decision-support tool that should lead to better decisions regarding land use and more effective protection, restoration, and conservation of our region s natural resources. The assessments cover water resources both water flow and water quality and fish and wildlife habitats in terrestrial, freshwater, and marine shoreline areas over the entire Puget Sound drainage basin. The Department of Ecology is leading the assessments for water resources and the Department of Fish and Wildlife is leading the assessments for habitats. Because of differences in dimensions, scale, data quality, and ecosystem-level processes the fish and wildlife assessment was broken into three separate assessments: terrestrial, freshwater, and marine shoreline. This document describes the assessment of marine shoreline habitats. The Puget Sound Nearshore Ecosystem Restoration Project (PSNERP) was formally initiated in September 2001 to assess ecosystem degradation in the Puget Sound Basin; and to formulate, evaluate, and screen potential solutions to ecosystem degradation. The habitat assessments conducted by WDFW for the Puget Sound Watershed Characterization Project were initiated in In 2011 PSNERP and the authors of this report recognized that each group was assessing complementary aspects of nearshore ecosystems and that the integration of our assessments would provide a more comprehensive perspective with which to make management decisions affecting the nearshore. Parts of this report focus on the relationships between this assessment and PSNERP s assessment and how they can be integrated. Our task is to assess the relative value of places for the conservation of fish and wildlife habitats. The principal challenge we faced were the limitations imposed by the currently available spatial data. Given the quality of the fish, wildlife, plant, and habitat data for the shorelines of Puget Sound, we believed an assessment based on the presence, density, and abundance of species and habitats would provide a credible indicator of conservation value. The overarching assumption of that decision is that the relative value of shorelines for the conservation of fish and wildlife habitats is mostly a function of the presence, density, and abundance of the species and habitats for which we collect occurrence data. In general, we collect occurrence data for certain species or habitat types because 1) humans harvest those species, 2) we are concerned about the status of those species or habitats (e.g. threatened or endangered species), or 3) we are concerned about the management of those species or habitats (species sensitive to human disturbances). In other words, we collect data on those species and habitats we care most about. 1

5 Therefore, an assessment based on these data should indicate those places we should care most about for the conservation of fish and wildlife habitats. The main product of the assessment will be a quantitative index that indicates the relative value of marine shorelines for fish and wildlife habitat conservation. Based on this index, shorelines can be ranked for the entire Puget Sound and within each of the seven oceanographic sub-basins of Puget Sound. The main application of this assessment is land use planning, and land use plans should use the results to direct residential development to places that will minimally impact marine shoreline habitats. If development along shoreline is unavoidable, then the first places to develop or develop more densely are those shoreline segments with the lowest scores for the composite and top-5 indices. Development should be avoided along shoreline segments at the highest end of relative habitat value as indicated by the composite and top-five indices. Development at sites with low relative conservation value must still consider the no net loss principles of the Shoreline Management Act (RCW ). Our assessment shows that relative habitat value or place-based conservation priorities cannot be conveyed by one map. The relative value of a place depends on one s perspective, e.g., preferences for quality or quantity, and therefore, it is wise to examine maps for multiple perspectives composite index, top-5 index, and other perspectives as well. The need to assess shoreline habitat value through multiple perspectives creates some challenges for county governments trying to conserve marine shoreline habitats. One perspective on conservation value indicates some shorelines have high value for a multitude of species and clearly, every effort should be made to protect such places. However, another perspective shows that nearly every shoreline segment has relatively high value for some marine fish or wildlife species. These shorelines should be considered for special management as well. When using this assessment remember the following. First, there is no purely objective conservation value that can be empirically validated. Value is based on one s belief about what is valuable, and therefore, subjective. Furthermore, there is a wide variety of potential credible models of conservation value that could be constructed for this assessment. Our model of conservation value was based on a number of subjective judgments for which there was uncertainty: which factors to include, their relative influence, and how to assemble them. Second, this assessment and all the assessments done for the Puget Sound Partnership s Watershed Characterization project do not constitute all the information necessary and sufficient to address natural resources conservation through land use planning and decision making. Therefore, this assessment should be supplemented with other assessments. In fact, high quality local marine shoreline assessments are currently available that may be more useful to local land use planners. Third, this assessment is intended for sub-basin or landscape-scale planning and decision making, and consequently, does not address habitat issues that are best addressed through smaller-scale assessments, such as prioritizing site-level restoration projects. Each of the Salmon Recovery Lead Entities has done their own assessments to support their recovery plans for Chinook and steelhead (e.g., East Kitsap 2004, Pierce County 2004, Snohomish County 2005). The work done by lead entities serves a particular purpose, is highly attuned to local knowledge, and has involved local stakeholders, and therefore, our assessment is not a substitute for the assessments and plans of the lead entities. Furthermore, we did our assessment with the expectation that finer-scale assessments will be done by 2

6 county governments as the need arises for local comprehensive and sub-area plans. City and county land use plans should evaluate the need for smaller-scale assessments and conduct them where needed. Fourth, conservation value was calculated as a composite index consisting of 41 diverse components. This is effectively a richness approach which assumes that our 41 components can serve as adequate surrogates for the majority of species in marine shoreline habitats of Puget Sound. This approach may not adequately address the habitat needs of those species for which we lacked data, in particular, rare or highly sensitive species. If other rare or highly sensitive species are found to inhabit a local jurisdiction, then the special needs of such species should be specifically addressed in local land use plans. Lastly, as data, technology, and knowledge improve over time better assessments will emerge. Also, other initiatives, separate from those of the Puget Sound Partnership, will reassess habitats in the Puget Sound Basin. For Instance, the Western Governors Association has initiated a project to indentify crucial habitats throughout the western states (WGWC 2011). WDFW is participating in that initiative which could influence future assessments of habitat conservation value in Puget Sound. 3

7 II. Introduction Over the next two decades over 1 million additional people are expected to inhabit the Puget Sound Basin (OFM 2007). Thousands of acres of agricultural and timber lands will be converted to residential and commercial uses in order to accommodate this phenomenal growth. In addition to providing valuable commodities these working lands, both agricultural and timber lands, provide habitats for wildlife. To ensure the health and well being of their citizens, promote orderly and efficient land use, and protect natural resources, city and county governments implement comprehensive plans and regulatory land use zoning. Natural resources include fish and wildlife. In general, conversion of agricultural and timber lands to residential and commercial uses adversely impacts the habitats of fish and wildlife (Azerrad et al. 2009). Effective land use zoning can result in smart growth that minimizes the loss and degradation fish and wildlife habitats. To fully realize smart growth, comprehensive land use plans must be based on scientifically-credible information that indicates the most important places for the conservation of fish and wildlife habitats places where development should be avoided. Our purpose is to provide useful, scientifically-credible information for smarter growth in the Puget Sound Basin. The Puget Sound Watershed Characterization is a set of spatially explicit assessments that provide information for regional, county, and watershed-based planning. It is a coarse-scale decision-support tool that should lead to better decisions regarding land use and more effective protection, restoration, and conservation of our region s natural resources. The assessments cover water resources both water flow and water quality and fish and wildlife habitats in terrestrial, freshwater, and marine shoreline areas over the entire Puget Sound drainage area. The Department of Ecology is leading the assessments for water resources and the Department of Fish and Wildlife is leading the assessments for habitats. The assessments provide a county, WRIA, or oceanographic sub-basin perspective on the relative value of places (i.e., small watersheds or marine shoreline units) for the protection and restoration of water resources and habitats. The primary products of the assessments are maps that show the relative value of these small watersheds and shoreline units. Their relative values are expressed through quantitative indices that can be used to rank shoreline units across the entire Puget Sound or within a single oceanographic sub-basin. The indices, rankings, and all data used to calculate the indices are stored in a geographic database. The target users of the assessments are land use planners of city or county governments and the Puget Sound Partnership. Because of differences in dimensions, scale, data quality, and ecosystem-level processes the fish and wildlife assessment was broken into three separate assessments: terrestrial, freshwater, and marine shoreline. This document describes the assessment of marine shoreline habitats. The Puget Sound Nearshore Ecosystem Restoration Project The Puget Sound Nearshore Ecosystem Restoration Project (PSNERP) was formally initiated as a General Investigation Feasibility Study in September 2001, through a cost-share agreement between the U.S. Army Corps of Engineers and the State of Washington, represented by the Washington Department of Fish and Wildlife (WDFW). The PSNERP project will complete a feasibility study to assess ecosystem degradation in the Puget Sound Basin; to formulate, evaluate, and screen potential solutions to 4

8 ecosystem degradation; and to recommend a series of actions and projects that have a federal interest and are supported by a local entity willing to provide the necessary items of local cooperation. The habitat assessments conducted by WDFW for the Puget Sound Watershed Characterization Project were initiated in In 2011 PSNERP and the authors of this report recognized that each group was assessing complementary aspects of nearshore ecosystems and that the integration of our assessments would provide a more comprehensive perspective with which to make management decisions affecting the nearshore. Parts of this report focus on the relationships between this shoreline habitats assessment and PSNERP s assessment and how they can be integrated. 5

9 III. A Framework for Multi-scale Assessments The patterns we observe in ecosystems are the result of events occurring at multiple spatial scales of organization (Figure 1). Large-scale drivers (outermost ring in Figure 1), such as climate and ocean dynamics together with such human activities as urbanization and deforestation, operate at a regional scale and directly interact with the controls of watershed processes. Those watershed controls include such physical attributes as geomorphology, geology and soils (turquoise ring in Figure 1); they also include the wide variety of human actions that individually and collectively affect watershed processes. Those processes (inner gray ring) include the movement, delivery, and loss of water, sediment, nutrients and wood. Together, the interaction of these natural and human-induced drivers and controls govern the processes, structure, function and, finally, ecological health (Beechie et al. 1999, 2010; Dale et al. 2000; Gove et al. 2001; Hidding and Teunissen 2002). This expresses the scientific consensus that proper functioning of our most highly valued ecosystems depends on what happens in the larger landscape, not just at the site or reach scale. Spatial scale has two dimensions: extent and grain. The extent of shoreline habitats assessment is the Puget Sound Basin. The assessment utilizes two grain sizes. The smaller grain consists of habitat types defined by shorezones (Berry et al. 2001a) and shoreforms (Shipman 2008). These habitat types divide the shoreline into 10,178 distinct shoreline units with an average length of 0.24 miles. The larger grain consists of littoral drift cells (Simenstad et al. 2006a); these divide the shoreline into 812 overlapping units with an average length of 3.4 miles. The results of the characterization can be directly applied to land use plans or decisions that match the grains of the assessment. For instance, shoreline master programs delineating shoreline designations with a spatial resolution of about ¼ mile are an appropriate direct application of both the smaller and larger grained assessment results. The results may also be used to enrich the interpretation of finer-scale assessments and provide a context for finer-scale project planning. The Puget Sound Characterization incorporates a multi-scale spatial framework, emphasizing integration of abiotic and biotic assessments and data interpreted at the larger scales in support of management needs. In describing the details of the framework, we distinguish between those with an explicitly water resources focus (those of water flow and water quality) and those that are, in part, dependent on the movement of materials and energy across watershed boundaries (particularly those of the terrestrial and marine shoreline environments). Our conceptual model for a multi-scale framework is further described in Table 1. Table 1 describes the types of data and assessment effort that may be available at a range of levels and how the information may be applied to planning efforts. The purpose of defining the levels explicitly is to guide the proper collection and application of information by following the admonition to consider and evaluate the effects (at least) one level of organization above and below the level of the action or effect. Therefore, the conceptual framework of Figure 1 and their associated assessment levels define a nested hierarchical framework. For example, questions regarding stream community structure (a fine level) cannot be answered directly by information acquired and analyzed from coarser levels. However, coarse-level information addresses issues about landscape features that control the movement of water, such as the type of geology and areas of water storage. These landscape features indirectly, but profoundly, influence stream community structure. Because processes at a coarse level influence 6

10 conditions at finer levels, one should always make decisions at a finer spatial scale (for example) only within an understanding of the ecological context of the broader landscape conditions. Figure 1. Relationships between drivers, controls, processes, and habitat structure and function for freshwater, terrestrial and marine shoreline environments. Large-scale drivers (such as climate) regulate the type and amount of precipitation. Controls, such as geology and land cover, in turn govern processes such as the movement of water, sediment, wood, nutrients and other chemicals. The processes shown do not equally affect each of the environment types. Human impacts, such as forest clearing, construction of impervious surfaces, fill in floodplains and wetlands, occur at all scales and are represented most fundamentally in the outer ring. Adapted from the Healthy Watersheds Integrated Assessments Workshop (EPA, in review). 7

11 Table 1. Relationships between the level of information and analysis at different spatial scales to both the type of data, required and type of application of results to planning and permitting. Adapted from the Healthy Watersheds Integrated Assessments Workshop (EPA, in review). Level of Information and Coarse/General Fine/Detailed Analysis Unit of Organization Typical spatial scale (area) Type of Data Acquisition Type of Application at Each Level How the Puget Sound Characterizati on results could be applied Oceanographic Subbasin Drift cell Reaches Shorezones >1000 mi mi 2 1 mi mi 2 <1 mi 2 Existing GIS data layers from Puget Sound Characterization Land-use planning and zoning, such as the location, type, and/or intensity of new development to avoid and buffer existing, mapped watershed features. Water flow, water quality, and terrestrial and freshwater assessments are most applicable at this scale, integrating sub-basin information on conditions of importance to each of these processes. Existing GIS data layers from Puget Sound Characterization Refinements of coarselevel assessment for application to land- use planning and zoning to protect existing, mapped watershed features serving important watershed or landscape process and function. The water-flow, water quality, terrestrial and freshwater habitat assessments provide information at a sub-basin scale Using existing data or field collection of new data on biological and physical conditions at these scales. Reach- and watershedscale strategies for land and water protection & restoration. Reachspecific actions & BMPs to protect and restore conditions. Usually requires field collection of new data on biological and physical at these scales. Adaptive management; feedback and site- and reach-scale project designs for the specific BMPs to remediate stressors to restore and protect healthy ecosystems. The Characterization does not provide results at these scales. However, characterization results should be used to confirm whether actions at these scales are appropriate. For example, installation of wood at the site or reach scale should not be undertaken if upper water delivery and storage processes are highly degraded. 8

12 IV. Conceptual Model A conceptual model is a simplified representation of a complex system that emphasizes the interrelationship of the major elements rather than the details of each element. The conceptual model describes the rationale for components and structure of the quantitative model. Other concepts related to model formulation are covered in Appendix A. Conceptual Foundations We begin with a conceptual model of ecosystems in which processes and structures 1 interact to manifest functions (Figure 2; Goetz et al. 2004, Simenstad et al. 2006b). Both process and structure are essential to the maintenance of high-quality, sustainable nearshore ecosystems. The composition and organization of biological communities in nearshore ecosystems is caused by processes such as wave exposure, sediment resuspension, and freshwater flows, and by structures such as beach topography, beach sediments, and salinity. The structure of nearshore ecosystems is both the consequence of and an influence on the action of ecosystem processes (Goetz et al. 2004). For instance, beach topography is the result of wave action but beach topography also influences wave action. In this assessment we focus on a specific set of nearshore ecosystem functions: the provision of habitat for nearshore flora and fauna (Figure 3). The habitat functions of nearshore ecosystems are highly integrated and hierarchical. For example, the ecosystem function of herring (Clupea pallasi) habitat depends in part on the presence of a particular vegetative structure, eel grass (Zostera marina)(figure 4), and the ecosystem function of eel grass habitat depends largely on the structure provided by beach sediments. Figure 2. Ecosystem processes and structures interact to manifest ecosystem functions such as the provision of habitat (Goetz et al. 2004). Habitat Shaping Processes and Structures The most important physical process along the shorelines of Puget Sound is the movement of sediments. Sediment movement occurs within spatially distinct littoral drift cells. Drift cells are 1 Other ecosystem conceptual models separate structure into structure and composition. We consider composition to be an attribute of structure. 9

13 Figure 3. Interaction of processes (blue ovals) and structures (brown ovals) to produce habitat functions (green ovals). Each species has different habitat requirements that are met by different interactions among processes and structures. Figure 4. Hierarchical nature of ecosystem structures (brown ovals). Eel grass depends on the structure provided by sediments and beach profile (i.e. topography). Herring depend on the structure provided by eel grass. Habitats for herring and eel grass, depend on the interaction of processes (blue oval) and structures (brown ovals). 10

14 comprised of sediment sources, typically bluffs, where erosion provides sediment for beaches, sediment sinks, where sand and gravel accumulate, and transport reaches where littoral drift connects sources to sinks (Figure 5). Puget Sound s shorelines are comprised of 812 drifts cells with an average length of 3.7 miles and a maximum of length 60 miles. Within drift cells, variation in wave exposure, sediment sources, and local geomorphology have created a variety of shoreforms, such as bluff-backed beach, barrier estuary, open coastal inlet, and rocky shore, (Shipman 2008). Bluff-backed beaches are sediment sources; barrier beaches, barrier estuaries are sediment sinks; and all beaches play some sort of role in sediment transport. Shoreforms provide a variety of environmental settings for fish and wildlife habitats. At a finer spatial scale, ecosystem structures have been mapped as shorezones (Berry et al. 2001b). Shorezones classify shorelines according to sediment type, wave exposure, and slope (sensu Dethier 1995 and Howes et al. 1994). Figure 5. The physical process of sediment movement within littoral drift cells along the shorelines of Puget Sound (from Simenstad et al. 2006a) Littoral drift is the dominant process for shaping and maintaining shoreline habitats. Therefore, in order to assess the quality of shoreline habitats, the integrity of drift cells must also be assessed. PSNERP (Cereghino et al. 2012) has assessed the integrity (i.e., relative degradation) of drift cells, and their results will be used in conjunction we our shoreline assessment. Littoral drift is not the only ecosystem process affecting habitat functions along the shorelines of Puget Sound. For instance, within the nearshore zone species may be directly affected by wave exposure, and just outside the nearshore, species are affected by upwelling currents. Structures in close proximity to the nearshore zone, may also influence habitat quality along shorelines. For instance, rocky reefs in subtidal areas and patches of large trees in upland areas affect foraging habitat quality for white-winged scoter (Melanitta fusca) and nesting habitat quality for great blue herons (Ardea herodias), respectively. 11

15 Assessing the condition or presence of the myriad processes and structures that affect habitats in the nearshore zone was beyond our capability. Therefore, we assessed the presence of the habitat functions. Habitat Functions A vital ecosystem function is the provision of habitats. Habitats are specific to each species. Although there is considerable overlap in the habitat characteristics of some species, e.g., red sea urchins (Strongylocentrotus franciscanus) and green sea urchins (S. droebachiensis), the full multi-dimensional characteristics of habitat are unique for every species and sometimes unique to a particular population of a species (Morrison et al. 1992). The processes that manifest a species habitats are highly complex: individuals integrate multiple factors when selecting habitat, exhibit a wide range of habitat preferences, respond differently to habitats with different qualities, and populations are adaptable to changing habitat conditions. Hence, for many species our understanding of habitat is simplistic, and consequently, our ability to model species habitats or habitat quality is limited and replete with uncertainty. The mapping of habitat quality is technically challenging, but mapping the presence of a species is not (although is it fiscally challenging), and considerable expense has been invested in the mapping the presence of certain marine species, in particular, harvested and imperiled species. By definition, the presence of a species establishes the presence of that species habitat. In other words, if a species is present at a site, then that site is serving a habitat function. However, habitat quality cannot be determined by species presence alone, and the functions (e.g., breeding, rearing, resting) served by that habitat may not be discernable through species presence. Furthermore, species absence does not establish non-habitat; absence may be due to survey error, patterns of seasonal use, or declining population size. Nevertheless, considering the dearth of habitat models, the presence of a species is our most reliable indicator of habitat. Empirical data on the locations of fish or wildlife species collected by WDFW and other agencies generally focus on imperiled or harvested species. For the vast majority of other species, site-scale location data are based on incidental observations or incomplete surveys. These data have a high rate of omission error, i.e., false negatives. For many vertebrate species comprehensive data on locations are available as range maps (e.g., Wahl et al. 2005), but these can be highly inaccurate at spatial scales of about 1 square mile or less. Conservation Value Our task is to assess the relative value of shorelines for the conservation of fish and wildlife habitats. Certain places in a region are readily identified as valuable or even irreplaceable because they contain rare habitat-types, imperiled species, or abundant wildlife. For instance, in the Puget Sound, Protection Island with its dense colonies of breeding birds, the tidelands at the Nisqually River delta, and the waterfowl over-wintering areas of the Skagit River delta are universally recognized by fish and wildlife biologists as crucial places for habitat conservation. The value of such places is obvious and absolute experts are certain that these places should be protected for their ecological values. Most other places lack rare habitats, imperiled species, or abundant wildlife. Such places may have value for the conservation of wildlife habitats, but they lack those qualities that would make their protection obvious. The value of places with common habitats can be assessed but only in a relative sense, and decisions regarding their protection must be based on relative value. Hence, for the multitude of places that contain only common habitats, our assessment cannot determine whether site A or site B should be 12

16 protected. Our assessment can only determine that site A is relatively more valuable than site B, and therefore, site A should be a higher priority for protection than site B. The relative value of a place for the conservation fish and wildlife habitats can be based on a variety of different factors: the presence of rare species or habitat-types (Prendergast et al. 1993, Kerr 1997), richness of species or habitat-types (Williams et al. 1996), the presence of imperiled species (i.e., listed as threatened or endangered), species endemism (Orme et al. 2005), local abundances of particular species or habitat types (Pearce and Ferrier 2001), metrics of habitat quality (Root et al. 2003), metrics of ecological integrity (Andreasen et al. 2001), or efficiency (Wilhere et al. 2008). These factors quantify different aspects of value, and hence, a truly comprehensive assessment would include all of them, however, the available data preclude accurate estimates for most of them. Even with perfect spatial data for species occurrences and highly reliable models for habitat quality assessing the relative value of places for wildlife habitats would remain challenging because measures of conservation value or importance are normative. There is no purely objective conservation value that can be empirically validated because conservation value is based on one s belief about what is valuable, and therefore, it is influenced by personal values. For example, people will answer the following question differently: what is most important, a place rich with common species, a place with a few rare species, or a place with commercially valuable species? Likewise, how various data should be assembled into an indicator of value may be different for each person, and therefore, a multitude of different credible indicators can be devised. Nevertheless, scientists may reach consensus on what factors should be used to indicate value and on the relative influence of those factors. In the absence of consensus, the subjective uncertainty regarding value can be quantitatively expressed. In summary, practical measures of conservation value are constrained by the types, quantity, and quality of available data. Furthermore, conservation value is not descriptive, it is normative, and hence, transparency regarding normative judgments is essential. The challenge we faced was to develop an assessment that respected the limitations imposed by the currently available spatial data but still served as a useful, credible indicator of relative conservation value. A Model for Relative Conservation Value Of the three habitat assessments conducted for the Puget Sound Partnership s Watershed Characterization Project terrestrial, freshwater, and marine shorelines the fish, wildlife, and habitat data for marine shorelines are the most comprehensive and very likely the most accurate. This can be attributed to the one-dimensional nature of shorelines and their relatively small spatial extent 2,468 miles of marine shoreline in Puget Sound compared to over 50,000 miles of rivers and streams in Puget Sound Basin. Given the quality of the data, we believed an assessment based on the presence, density, and abundance of species and habitats would provide a credible indicator of conservation value. The overarching assumption of that belief is that the relative value of shorelines for the conservation of fish and wildlife habitats is mostly a function of the presence, density, and abundance of the species and habitats for which we collect occurrence data. In general, we collect occurrence data for certain species or habitat types because 1) humans harvest those species, 2) we are concerned about the status of those species or habitats (e.g. threatened or endangered species), or 3) we are concerned about the management of those species or habitat s(species sensitive to human disturbances). In other words, we 13

17 collect data on those species and habitats we care most about. Therefore, an assessment based on these data should indicate those places we should care most about for the conservation of fish and wildlife habitats. Another major assumption is that the relative value of a place for the conservation of fish and wildlife habitats can be accurately quantified. And, more specifically that, relative value can be expressed through a single comprehensive number, an index (or a small set of indices). Furthermore, we assumed that the index is a linear function the weighted linear combination of normalized biological data. Better relationships between relative conservation value and the biological data may exist, but lacking any practical means to determine those relationships we chose the most parsimonious formulation a linear equation. The main product of the assessment will be a quantitative index that indicates the relative value of shorelines for fish and wildlife habitat conservation. Based on this index, shorelines can be ranked for the entire Puget Sound and within each of the seven oceanographic sub-basins of Puget Sound. Integrating Process and Function An assessment based solely on the occurrence on fish and wildlife species, i.e., on habitat functions, will not provide a complete understanding of their conservation value and neglect essential information for guiding management actions. We must also consider the integrity of ecosystem processes. PSNERP (Cereghino et al. 2010) provides information on the integrity of nearshore processes within individual drift cells. Specifically, PSNERP assessed the relative degradation of littoral drift, which is roughly the inverse of integrity. In PSNERP s assessment, degradation reflects the relative loss of historical ecosystem services as indicated by landform change and shoreline modification (Table 2). Particular attention was given to indicators of degradation thought to be important in process dynamics. There are separate PSNERP strategies for deltas, coastal inlets, beaches, and barrier embayments. PSNERP mapped 16 deltas in Puget Sound. Given the relatively small number of deltas, their high degree of hydro-geomorphic complexity, and wide range of degradation, we believe that the integration of the PSNERP strategies and our indices should be done ad hoc. There are 266 coastal inlets. Human activities at coastal inlets affect their habitat functions, but activities within the inlet s entire drainage area may also have a significant impact on habitat functions. Therefore, effective management of coastal inlets must also consider land use activities far removed from the inlets location. This will require integration of this assessment, PSNERP s assessment, and the freshwater and terrestrial habitat assessments that are part of the Puget Sound Partnership s Watershed Characterization Project. That complex process is beyond the scope of this report. The beach and barrier embayment strategies focus on drift cells. The connections between shoreline habitat functions and drift cell processes are probably strongest along beaches and embayments. Therefore, we believe this assessment must be integrated with beach and barrier embayment strategies. These strategies make management recommendations based on the relative degradation of drift cells. Protect is recommended for the least degraded drift cells, enhance for the most degraded drift cells, and restore for those drift cells in between (Table 3). We explain how these two assessments can be used in conjunction in the last chapter of this report. 14

18 Table 2. Metrics used to assess degradation of beach and barrier embayments in the PSNERP assessment (Cereghino et al. 2012). Degradation Index Degradation Metrics Beach Embayment Description Lost Embayment Length Loss of length was calculated as the total length of current embayment landform subtracted from the total length of historical embayment landforms within a site. While some change in length was attributed to mapping error, this metric provided a measure of gross physical change in the system to complement presence of linear stressors and nearshore zone development in barrier embayment sites. Nearshore Impervious Parcel Density Sediment Supply Degradation The percentage of land area within 200 m of the shoreline with impervious surfaces estimated as greater than 10% was used to describe the intensity of development at a site. Development indicated by impervious surface was assumed to indicate the combination of intensive use, chronic pollution, modified hydrology, and loss of native vegetation. The mean number of parcels per 100m in a shoreline process unit was used to characterize both challenges and costs of negotiating protection or restoration of sediment supply and transport under complex parcel ownership, as well as chronic impacts from high density residence on vegetation and drift wood. The sediment input degradation metric was developed by Schlenger et al. (2011) to predict the effect of overlapping stressors on the degradation of sediment input. In shoreline process units, this metric calculated the percentage of bluff-backed beach landforms located in a drift cell component showing either divergence or transport (i.e. DZ, LtR or RtL) that was covered by either fill, armoring, railroads, roads or an artificial landform, all of which awee anticipated to potentially affect sediment supply budgets. Tidal Flow Degradation The tidal flow degradation metric was developed by Schlenger et al. (2011) to predict the effect of overlapping stressors on the degradation of tidal flow in embayments and river deltas. Within shoreline process units, tidal flow degradation was estimated as the percent of embayment landform length with either tidal barrier, fill, railroad, or an artificial landform. Table 3. Relationship between drift cell degradation and management recommendations from the Puget Sound Nearshore Ecosystem Restoration Project (PSNERP; Cereghino et al. 2012). Low Degradation Moderate Degradation High Degradation Protect Restore Enhance Sites with opportunities to protect ecosystem processes, structures, and functions. Site where there may be opportunity to increase ecosystem services. Site where strategic actions may enhance ecosystem services. 15

19 V. Methods Spatial Framework Puget Sound has been divided into 7 oceanographic sub-basins based on bathymetry and circulation patterns, however, these sub-basins also reflect regional patterns in shoreline geology, geomorphology, and wave environment (Shipman 2008), which manifest regional patterns in biotic communities. We avoided comparing the habitat conservation value of dissimilar biotic communities by doing our assessment calculations within sub-basins. Puget Sound has 2,468 miles of marine shoreline. For purposes of analysis, this shoreline must be broken into smaller spatial units. We intersected the shoreform (Shipman 2008) and shorezone (Berry et al. 2001a, 2001b) classification systems to produce a shoreline composed of 10,178 segments. Demarcations between segments correspond to mapped changes in shoreform (e.g., barrier estuary, bluff-backed beach), morphology (e.g. flat, platform, ramp), or substrate (rock, gravel, sand), all of which significantly influence plant, fish, and wildlife habitats. Across the entire Puget Sound, the mean segment length was 0.24 miles and ¾ of segments were less than 0.29 miles (Table 4). Mean segment lengths were different among the seven oceanographic sub-basins, with the Juan de Fuca sub-basin having the longest (0.55 miles) lengths and the San Juan sub-basin having the shortest (0.17 miles). In GIS the shoreline is a one-dimensional line. Shoreline segments were converted to two-dimensional polygons for two reasons. First, for some species, such as Dungeness crab, the spatial extent of each occurrence was represented as a polygon and we wanted to maintain that two-dimensional information. Second, there were fish and wildlife occurrence data, such as those for bald eagle nests, that did not intersect the shoreline but were in close proximity to it and we wanted to associate those data with the shoreline. To accomplish both objectives we buffered the shoreline by 400 m (approximately ¼ mile) in the landward direction and by 400 m or extreme low tide, whichever was farther, in the seaward direction. The National Wetlands Inventory (USFWS 1989, Cowardin et al. 1979) inter-tidal polygons were used to delineate the location of extreme low tide. Biological data outside the 400 m buffer were excluded from the analysis. Table 4. Summary of analysis unit sizes by oceanographic sub-basin. Units in miles. sub-basin N Mean min 1st qtr median 3rd qtr max Juan de Fuca San Juan Hood canal Whidbey North Central South Central South Sound ALL Four-hundred meters was chosen as the buffer width because we believed it would encompass most biological resources that might be directly impacted by shoreline development. Four hundred meters is 16

20 the recommended management zone around bald eagle nests and roosts (Watson and Roderick 2000) and ¼ mile (402 m) is roughly the distance at which shoreline development might disturb seal and sealion haul outs (S. Jeffries, WDFW, pers. commun.). Furthermore, for at least 90% of the Puget Sound shoreline, 400 m encompasses the entire nearshore zone (< 10 m depth, sensu Simenstad et al. 2011) and most shallow subtidal areas. Figure 6. Spatial framework for the shoreline habitats assessment. Blue and green polygons are seaward and landward assessment units, respectively. Totten, Eld, and Budd Inlets (left to right) of Puget Sound are depicted. Biological Data We are limited by breadth, precision, and accuracy of the biological data. The data s breadth, i.e., the variety of species and habitats for which occurrence or abundance data are available, is relatively broad: molluscs and crustaceans of commercial/recreational interest, urchins, forage fish, salmonids, numerous bird species, pinnipeds, kelp, eelgrass, surfgrass, and wetlands. The measurement precision for most of these data is at the level of presence/absence. Only PSAMP bird survey data enable an estimate of local density or abundance. The accuracy of our data is affected by the data s age and the methods of data collection. Some data sets are over 20 years old (e.g., WDF 1992). Most data were collected through field surveys, but the data in certain datasets are based on best professional judgment of the biologist. We reviewed all biological datasets managed by WDFW for their relevance to marine shorelines in Puget Sound and their likely accuracy. Our subjective evaluation of likely accuracy considered the dataset s age, how the data were collected, and the detectability of the taxa surveyed. Occurrence data for fish and wildlife are more prone to false negatives than to false positives, and hence, we were particularly 17

21 concerned about the potential frequency of false negatives in each dataset. We settled on 41 data sets (Tables 5 and 6). Most data sets mapped the occurrences of single species (i.e., Dungeness crab, herring). Some data sets mapped the simultaneous occurrence of multiple species (i.e., shorebird and waterfowl concentrations). Table A4 gives examples of data sets that were excluded from the assessment. For some data sets which were likely to have a high rate of false negatives we developed models for probability of occurrence (explained below). Figure 7. Biological data used in shoreline habitats assessment. Different polygon colors and fill patterns and different line colors represent different plant, fish, or wildlife species. White dots are observations of PSAMP bird surveys. Discovery bay and Port Townsend are depicted. With a few exceptions the fish and wildlife species included in the assessment were priority species as designated by WDFW s Priority Species and Habitats program (PHS; WDFW 2008). Priority species require protective measures for their survival due to their population status, sensitivity to habitat alteration, and/or recreational, commercial, or tribal importance. Priority species include State Endangered, Threatened, Sensitive, and Candidate species; and animal aggregations considered vulnerable (e.g., heron colonies, shorebird concentrations). We also included a subset of the bio-band data from DNR s shorezone database (Berry et al. 2001b). These data are referred to as bio-bands because certain plants and animals create a well-defined series of cross-shore color bands. Each bio-band is named for the most prominent species in the band or by the general description of the species assemblage. The abundance of each band is recorded as either absent, patchy or continuous, which we translated to 0, 1, or 2. We included bio-bands for species of concern (e.g., eel grass, kelp) and excluded common species (e.g., barnacles, sand dollars). 18

22 The highest quality data utilized in our assessment was that collected by the Puget Sound Ambient Monitoring Program (PSAMP; Nysewander et al. 2005). PSAMP has conducted highly systematic aerial surveys of birds on Puget Sound since The complete data set contained 381,214 observations; over half the observations are of multiple birds. We removed records that were older than 1995, summer surveys, non-marine birds (e.g., common raven, northern flicker), or extremely abundant birds (e.g., glaucous-winged gull). This filtering process reduced the data set to 196,312 observations and included 65 bird species (Table A1) and observations for 31 categories of partially indentified birds (e.g., Unidentified Diving Duck ; Table A2). We summarized the data by calculating two indices. First, for the years 2000 to 2009, the most recent year for which data were available, we calculated the median density of birds for each shoreline polgygon. Second, for species of concern (Table A3) we calculated average density per shoreline polygon over the years 2005 to Table 5. Summary of plant and wetland data used in the index. Taxon Description Units Source dune grass salt-tolerant grasses, dominated by Leymus mollis brackish/ freshwater wetlands sedges assemblages; found at freshwater streams and river mouths high salt marsh low salt marsh surfgrass eelgrass brown kelp chocolate brown kelp bull kelp giant kelp brackish/ freshwater wetlands assemblages; Triglochin/Salicornia/ Deschampsia/Distichlyus dominated by Salicornia Phyllospadix spp. of lower intertidal Zostera marina and introduced Z. japonica large bladed Laminaria spp. Laminaria setchellii, Eisenia and/or Pterygophora, Hedophyllum, includes Egregia Nereocystis spp. Macrocystis spp. Amount = shoreline length bioband density Density 0 = Absent 1 = 0-50% cover 2 = % cover wetlands (NWI) all wetlands except marine sub-tidal square feet DNR Shorezone (Berry et al 2001a, 2001b) USFWS

23 Table 6. Summary of fish and wildlife data used in the index. Prob. Taxon PHS model Description Units Source Northern abalone X X documented occurrences WDF 1992 Clams; intertidal hardshell X X beds that could be commercially Clams; subtidal X X harvested or have significant WDF 1992 Crab; red rock X recreational usage Crab; Dungeness X X Square Pacific oyster X X non-native Crassostrea gigas feet WDF 1992 beds that could be commercially Geoduck X X WDF 1992 harvested Pandalid shrimp X X pink, coonstripe, and spot shrimp WDF 1992 Sea Urchins X X documented occurrences of red and green sea urchins WDF 1992 Herring Holding Areas X where adults congregate each winter prior to spawning Square WDFW 1992 Herring Spawning Areas X regular surveys over XX years feet WDFW Surf smelt X X Sand lance X X Bull Trout Chinook Salmon Chum Salmon Coastal Cutthroat Coho Salmon Pink Salmon Sockeye Steelhead Trout Bald Eagle Communal roosts Bald Eagle nest Great Blue Heron colonies X X X X X X X X X X X data represent more than 30 years of spawning beach surveys number of stream mouths inhabited by species that intersect shoreline segment zone around roost site; radius = 400 m zone around nest site; radius = 200 m zone around occurrence point; radius = 1000 ft feet count Square feet Black Oystercatcher nests survey data from 2010 count Shorebird X large regular concentrations Waterfowl X large regular concentrations Square Important Bird Areas support species of concern or high feet densities of birds Bird Density median density of all birds from 2000 to 2009 (Tables A3 & A4) birds / At Risk Bird density density of at risk birds from 2005 to 2009 (Table A5) km 2 count Seal/sea lion haul-out X both natural (e.g., islands) and within artificial (e.g., buoys) haul outs 400 m of shore 1 WDSM: WDFW s Wildlife Survey Data Management 2 PSAMP: Puget Sound Ambient Monitoring Program WDFW WDFW WDFW Fishdist WDFW WSDM 1 Audubon 2001 WDFW PSAMP 2 WDFW WSDM 20

24 Probability of Occurrence We believed the spatial data for some species likely had a high rate of false negatives. We were particularly concerned about data collected by the Washington Department of Fisheries in the 1980s (WDF 1992). To compensate for the shortcomings of these data we developed simple probability of occurrence (PO) models. The shoreform-shorezone classification system were treated as habitat types and we calculated P(S H), the probability that a species is present given the presence of a particular habitat type. Table 7. Data used in the calculation of P(S H), the probability that a species is present given the presence of a particular habitat type. Each n represents a count and N is the total number of shoreline segments. Species Present Absent Habitat Type Present n pp n ap n pp + n ap Absent n pa n aa n pa + n aa n pp + n pa n ap + n aa N For each species the PO model begins with a data table like Table 7. The probability of a particular shoreform-shorezone habitat type, H, occurring in a randomly selected shoreline segment is: P(H) = ( n pp + n ap ) / N (1) The probability of species, S, occurring near a randomly selected shoreline segment (i.e., within the 400 m buffer) is: P(S) = ( n pp + n pa ) / N (2) And, the probability that a species is present given the presence of a particular habitat type is: P(S H) = n pp / ( n pp + n ap ) (3) We developed PO models for 11 species (Table 6). The probability of occurrence was based on only one variable the habitat types created through the intersection of shoreforms (Shipman 2008) and the habitat types (Detheir 1990) in the DNR shorezone data (Berry et al. 2001b). Habitat preferences of fish and shellfish species cannot be accurately described with such a simplistic model, however, we believed that an inaccurate model was better than inaccurate data with a high rate of false negatives. The model results were merged with the occurrence data, i.e., empty records in the occurrence data were substituted with the probability of occurrence and data records indicating presence were set to 1. We also calculated group-equalized indices of association between species and habitat types (De Caceres and Legendre 2009). The indices of association were perfectly correlated (correlation = 1) with P(S H), and so we used P(S H) in the conservation value indices. Data Normalization Our species data come in numerous forms: linear units, areal units, counts, density, presence/absence, and categorical (Table 6). To combine these various data we must convert them to commensurate units. 21

25 We adopted the rank-sum method of Cereghino et al. (2012). In effect, we converted data with were originally in nominal, interval, and ratio scales to a common form of ordinal scale. The first conversion was to density data in linear or areal units were divided by the length or area, respectively, of their corresponding shoreline segment. Salmon occurrence data were not converted to density; they remained as counts. All data for each shoreline segment were then converted to ranks relative to other segments in the same oceanographic sub-basin. Ranks were normalized from 0 to 1 within sub-basins. The segment with the smallest rank (i.e., smallest density, probability, count, etc.) was set to zero and the largest rank was set to 1. Indices of Conservation Value We want to quantify the relative conservation value of marine shorelines. There are myriad formulations for a quantitative index, each with their own particular advantages and disadvantages. We limited our assessment to two simple formulations based on two perspectives of relative conservation value that reflect the quantity versus quality dichotomy. One perspective holds that value should be based on the number of ecological functions or habitats at a place. The other perspective holds that value should be based on the highest quality functions or habitats at a place. The first perspective, more functions are better, can be implemented by summing the amount of habitats at each shoreline segment. That is, summing the normalized ranks for each species or habitat. The composite index of habitat conservation value for a shoreline segment j is: H AV j w hnrhj h 1 here w h are subjective weights that determine the relative contribution of species or habitats to the index, H is the number of species or habitats included in the assessment, and NR hj is the normalized rank for that species or habitat h at shoreline segment j. The weights are normalized so that they sum to 1, and therefore, the index is effectively a weighted average. The resulting average score was renormalized within sub-basins so that the maximum value equaled 1. All weights in equation 4 were set to 1. We could have assigned greater weights to species or habitats that we thought were more important, such as federally listed salmon species or eelgrass, but that involves making value judgments that we wished to avoid in this assessment. Such value judgments should be informed by the opinions of stakeholders and policy makers. (4) 22

26 Figure 8. Process for calculating the index value for a single shoreline segment. Using species (S) and habitat (H) data a probability of occurrence model (P) was generated for some species and a density (D) was calculated for other species and habitats. All data are ranked (R) relative to the same data in other shoreline segments in the same sub-basin. All ranks are normalized (N) from 0 to 1 relative to other segments in the same sub-basin. The normalized ranks are summed and then renormalize so that the maximum sum equals 1. Figure 9. Data used in the calculation of conservation value indices. Forty-one types of data contribute to the indices. Salt marsh box includes sedges, high slat march, and low salt marsh (Tables 5 and 6). Different data come in different forms: green = probability of occurrence models, orange = counts; blue = density; black = amount. 23

27 The average value produced by equation 4 can obscure sites that are highly important for a single species or habitat but unimportant for most other species. Planners need to be aware of such sites, and hence, a second index addresses this shortcoming of the average value. The second perspective, higher quality function is better, can be implemented by taking the maximum value of NR hj for each segment j. Because many of the data are presence/absence, 78% percent of the maximum NR hj were equal to 1. Hence, to obtain a more informative index that would more clearly discriminate among shoreline segments we chose to average the five highest normalized ranks at each segment. The resulting average was renormalized within sub-basins so that the maximum value equaled 1. We called this the top-5 index. Index Properties Before relying on any model for planning or decision making, the model should be evaluated. We evaluated the index s sensitivity to each component (i.e., the normalized ranks) and we examined the index through various statistical and graphical analyses. The index has 41 components. A sensitivity analysis was done to understand the relative influence of each component on the index. Sensitivity analysis reveals which parameters or variables in a model are most influential. If a model is highly sensitive to a variable, then a small change in the variable s value causes a large change in the model s output. The sensitivity of a model s output, Y, to a variable X is defined as: S = Y / X (5) Sensitivity analysis was done by calculating the composite index for all shoreline segments with the weights set to 1, recalculating the composite index after altering a single weight by a small amount (e.g., 5%), and applying equation 5 to each shoreline segment. The process was repeated for each variable. The index value of each shoreline segment is effectively a separate model output, and hence each segment has its own sensitivity to each variable. A mean sensitivity was calculated for each component by averaging over the separate sensitivities of all segments. We also examined the contribution of each component to the composite index. The contribution of a component to the composite index value of shoreline segment j is: C j wh NRhj AV j (6) here w h are subjective weights that determine the relative contribution of species or habitats to the composite index, AV j, and NR hj is the normalized rank for that species or habitat h at shoreline segment j. The weights were all set to 1. Verification and Validation Model verification checks that the model does what it s supposed to do. Verification looks for errors in data processing and verifies that index calculations were done correctly. For complex models processing multiple large data sets this is an essential task. We verified the model output by doing all index 24

28 calculations in both R (RDCT 2005) and Excel. GIS tasks were verified by having one analyst perform the task and a second analyst review the results. Another important form of verification is comparing the output of the model against one s knowledge of the Basin do the places the model shows to be relatively more valuable agree with experts beliefs about places that are relatively more valuable and do the places the model shows to be relatively less valuable agree with experts beliefs about places that are relatively less valuable. This is verification and not validation because the expertise assessing the model output is the same expertise that developed the model. Model validation entails testing the accuracy of model predictions. In other words, given fresh data, i.e., data not used to develop the model, will the empirical response fall within the model s prediction interval. This type of rigorous validation is purely objective. We cannot do a rigorous validation because indices of conservation value are normative. There is no purely objective conservation value that can be empirically validated. Value is based on one s belief of about what is valuable, and therefore, it is influenced by personal values. Statistical models are developed with one data set and validated with a separate data set. Our model was not based on data but on professional expertise, which includes knowledge of ecological concepts, conservation principles, mathematics, modeling, and the relevant technical literature. Hence, our model can only be validated through peer-review by others with the appropriate professional expertise. 25

29 VI. Results Index Components Probability of Occurrence Models We constructed PO models for 11 species. The models were simplistic, but nevertheless, the results generally conformed to our knowledge of these species habitat associations (Dethier 2006). The probabilities of occurrence for abalone and urchin (Figure 10C), for instance, are highest on boulder, bedrock, and cobble substrates. The probabilities of occurrence for surf smelt and sand lance (Figure 10A) are highest for sand and gravel and lowest for bedrock and boulder substrates. However, the results for Dungeness and red rock crabs (Figure 10B) probably reflect a bias in the data toward commercially valuable species. In the GIS data the mapped area of occurrence in Puget Sound for Dungeness crab was 20 times greater than the area for red rock crab. Hence, we suspect a high rate of false negatives in the GIS data for red rock crab, and consequently the PO model is likely to be very inaccurate. Furthermore, Dungeness crab are more abundant in subtidal than intertidal zones, and therefore, the associations with shoreline habitat types are probably weak. Normalized Ranks The composite index has 41 components (species, wetlands, wildlife concentrations, etc.), and the properties of the components varied. For instance, commonness varied among components. For components without a PO model, the percent of shoreline segments with non-zero values reflects the commonness of that component in Puget Sound. For instance, eagle nests occur near 19 percent of segments but black oystercatchers nests occur on only 1% of segments. Components with a PO model had many more shoreline segments with non-zero values than components without PO models. According to the PO model for Dungeness crab, that species could occur (probability > 0) on 95% of shoreline segments (Figure 11). In contrast, according to the GIS data, bald eagle communal roosts occur near only 1 % of segments. The distributions of normalized ranks were also different among components. For instance, for shoreline segments with non-zero values, normalized ranks for median bird density were uniformly distributed but normalized ranks for at-risk bird density were right-skewed, i.e., more segments with low values than with high values (Figure 12B). Also, for shoreline segments with non-zero values, normalized ranks for wetlands were approximately uniformly distributed but normalized ranks for eelgrass were left-skewed, i.e., more segments with high values than with low values (Figure 12C). Correlations among normalized ranks (Table A8) were mostly low (ρ <0.2) to moderate( 0.2 ρ < 0.6), which indicates that all components add unique information to the index. The highest correlations (ρ> 0.75) were among salmon species because many species co-occur in the same streams. 26

30 A B C Figure 10. Results of probability of occurrence models for two species of forage fish (A) which prefer sandy-gravelly substrates, two species of crab (B), and two species known to prefer rocky substrates (C). 27

31 Figure 11. Percent of shoreline segments with non-zero values for each component of the composite index. All bars are greater than zero. Blue bars indicate components with a probability of occurrence (PO) model. The Indices The basic result of the shoreline habitats assessment is a map (Figure 13). The map shows the relative value of every shoreline segment. In many places the index values conform to our expectations. For instance, the relatively intact mouths of the Nisqually and Skokomish rivers have high index values and the degraded shorelines of Olympia and Shelton have low index values (Figure XX). This pattern is repeated throughout the Puget Sound the shorelines along large urban areas (Tacoma, Seattle, Bremerton, Everett) have low scores and shorelines along areas known to have high ecological value (e.g., Freshwater Bay, Nooksack River mouth) have high scores. Between these extremes, however, the vast majority of segments have moderate scores 70 percent of shoreline length has scores between 0.3 and 0.7 and there is no discernable pattern in shoreline habitat value. Values for the composite index were roughly normally distributed (Figure 14) with a mean of 0.48 and about 1.5 percent of shoreline length being above 0.9 and 1.5 percent being below 0.1. The distribution of index values varied by sub-basin (Figure 17). For instance, the distribution was skewed right in the San Juan sub-basin (mean = 0.44), but skewed left in the Whidbey sub-basin (mean=0.55). The composite index and top-5 index were highly correlated, ρ = 0.83, but there are obvious differences between the two indices (Figures 13 and 15). The distribution of the top-5 index is highly skewed with 54 percent of shoreline length having index value greater than 0.9. This indicates that over half the shoreline segments have high scores for at least 4 components of the index. 28

32 A B C Figure 12. Distribution of normalized rank values for various index components: A) shellfish, B) birds, and C) wetlands and shoreline vegetation. Probability of occurrence models were used for all shellfish species. Zero means the component is not present in a shoreline segment. Ranks normalized from 0 to 1 within sub-basins. IBA means important bird areas. 29

33 Figure 13. The Composite Index. The composite index is the mean normalized rank of all 41 components. Highest conservation value is dark green and lowest value is dark red. Map depicts South Sound sub-basin and part of the Hood Canal sub-basin (upper left hand corner). Index values were normalized within sub-basins. Figure 14. Distribution of composite index values for all Puget Sound shorelines. Mean equals

34 Figure 15. The Top-5 Index. The top 5 index is the mean normalized rank of the five highest components in each shoreline segment. Highest conservation value is dark green and lowest value is dark red. Map depicts South Sound sub-basin and part of the Hood Canal sub-basin (upper left hand corner). Index values were normalized within sub-basins. Figure 16. Distribution of top-5 index values for all Puget Sound shorelines. Mean equals

35 Figure 17. Distribution of composite index values for the San Juan and Whidbey sub-basins. Figure 18. Comparison of composite and top 5 indices. For easier interpretation continuous index values were converted to 20 quantiles (i.e., vigintiles). Highest conservation value is dark green and lowest value is dark red. Map depicts South Sound sub-basin and part of the Hood Canal sub-basin (upper left hand corner of each map). Index values were normalized within sub-basins. 32

36 Figure 19. Mean composite index of habitat conservation value for shoreforms. Figure 20. Mean composite index of habitat conservation values for Dethier (1990) habitat types. Habitat types arranged from largest to smallest mean index value. 33

37 Table 8. Shorelines with the highest habitat value in each sub-basin according to the composite index. Where shorelines did not have an official place name we succinctly describe the location. Sub-basin Places with Highest Index Value mouth of Dungeness River Freshwater Bay (includes mouth of Elwha River) Juan De Fuca Clallam Bay (includes mouth of Clallam River) Green Point (includes mouth of Siebert Creek) beach from mouth of Lyre River to mouth of Field Creek Semiahmoo Point mouth of Nooksack River San Juan Birch Bay north end of Orcas Island east side of March Point north end of Indian Island Kilisut Harbor North Central Scow Bay shoreline from Old Fort Townsend State Park to Kala Point east side of Useless Bay cove at Dewey Sundis Beach Whidbey mouth of Tom Moore Slough on south fork of Skagit River southern tip of Camano Island north end of Tulalip Bay and Hermosa Point mouth of Quilcene River and Quilcene Bay mouth of Dosewallips River Hood Canal mouth of Skokomish River and Annas Bay mouth of Duckabush River coves east and west of Sun Beach shoreline from Agate Pass bridge to Sandy Hook spit south of Illahee South Central mouth of Wilson Creek shoreline along Lakota, Adelaide, and Buenna shoreline along Dockton area on Maurey Island Nisqually National Wildlife Refuge (mouth of Nisqually River) shoreline from Nisqually River to Sequalitchew Creek South Puget Bayshore shoreline from Mill Bight to Baird Cove mouth of Kennedy Creek The previous results are for the magnitude of the index. In many practical applications we only need to know the rank of shoreline segments. That is, we only need to know which segments have higher or lower index values than other segments. For easier interpretation we converted the continuous index value to 20 quantiles (i.e., vigintiles). Think of each quantile as a binning of ranks. Each vigintile 34

38 contains 5 percent of the shoreline length in its respective sub-basin. In other words, 5 percent of the shoreline will have the highest rank and 5 percent will have the lowest rank, and the remaining 90 percent of the shoreline is evenly distribution among the other 18 vigintiles. When comparing the quantized composite and top 5 indices (Figure 18) we see that places in extreme low (Olympia, Shelton) or extreme high (mouths of Nisqually and Skokomish rivers) quantiles of the composite index are also in low and high quantiles, respectively, for the top 5 index. Places in more moderate quantiles for the composite index may land in a different quantile for the top 5 index. When the top 5 quantile is greater than the composite quantile, that indicates that that shoreline segment may have lower overall habitat value for all components but has high habitat value for a subset of the 41 components. The assessment also enables us to compare the relative habitat value of shoreforms and habitat types (Figures 19 and 20). Barrier beach and bluff-backed beach have the highest mean values for the composite index, and closed lagoon marsh and barrier lagoon have the lowest values. The low mean index values for lagoon landforms may be due to their shallow depths. Several of the shellfish (Dungeness crab, shrimp, geoduck, subtidal clams) are mainly located in sub-tidal areas, but these subtidal areas are within 400 m of the shoreline. Lagoons are often more than 400 m from sub-tidal areas. Among the Dethier (1990) habitat types, mixed coarse, partially exposed; gravel, partially exposed; mixed fine, exposed; mixed fine, partially exposed; sand, partially exposed had the highest mean composite index values, and artificial, protected; boulder, protected; gravel, semi protected; hardpan, semi-protected; gravel, protected had the lowest values. We used the composite index to identify the shorelines with the highest habitat value in each sub-basin (Table 8). Bays and river mouths where commonly the highest value places in each sub-basin. This is especially true in the Juan de Fuca and Hood Canal sub-basins where river mouths and their respective drainage basins are still relatively undeveloped. Index Properties The sensitivity analysis shows that the composite index is most sensitive to the species that have PO models (Figure 21). This sensitivity is caused by the high proportion of shoreline segments that have non-zero values for PO models. The sensitivity of the index to a component was highly correlated (ρ = 0.80) with the component s average contribution to the index. Sound wide, the components that made the largest contribution to the composite index were urchins, Dungeness crab, surf smelt, and wetlands (Figure 22). Twenty-two components each comprised, on average, 1 of the composite index, and of those, 9 components each comprised 0.1 percent of the composite index. The components contributions varied by sub-basin (Figure 23). For instance, the biggest contributors in the Juan de Fuca sub-basin were urchins, Dungeness crab, wetlands, and important bird areas (IBA). In contrast, the biggest contributions to the composite index in the South Puget Sub-basin were intertidal clams, oysters, wetlands, and smelt. 35

39 Figure 21. Results of the sensitivity analysis. Blue bars correspond to species with probability of occurrence (PO) models. Vertical dashed lines demark groups with similar taxonomy or functional characteristics. Figure 22. Mean percent contribution of components to composite index of habitat conservation value. There is a radial axis for each of the 41 components of the composite index. NWI and IBA mean national wetlands inventory and important bird areas, respectively. 36

40 Figure 23. Mean percent contribution of components to composite index of habitat conservation value for South Puget and Juan de Fuca sub-basin. There is a radial axis for each of the 41 components of the composite index. NWI and IBA mean national wetlands inventory and important bird areas, respectively. 37

41 VII. Discussion The main application of this assessment is land use planning, and land use plans should use the results to direct residential development to places that will minimally impact marine shoreline habitats. If development along shorelines is unavoidable, then the first places to develop or develop more densely are those shoreline segments with the lowest scores for composite and top-5 indices. Development should be avoided along shoreline segments at the highest end of relative habitat conservation value as indicated by the composite and top-5 indices. City and county governments have regulatory authority over land use along marine shorelines. The allowed land uses they designate through shoreline master programs and comprehensive plans may be the most important actions affecting the health of marine shoreline habitats. Maintaining shoreline habitat functions while accommodating human population growth will require sophisticated (and possibly as yet unknown) approaches to land use planning and residential development. In many places the index values conform to our expectations. For instance, the relatively intact mouths of the Nisqually and Skokomish rivers have high index values and the degraded shorelines of Olympia and Shelton have low index values. This pattern is repeated throughout the Puget Sound the shorelines along large urban areas (Tacoma, Seattle, Bremerton, Everett) have low scores and shorelines along areas known to have high ecological value (e.g., Freshwater Bay, Nooksack River mouth) have high scores. Between these extremes, however, there is no discernable pattern in shoreline habitat value. Our assessment shows that relative habitat value or place-based conservation priorities cannot be conveyed by one map. The relative value of a place depends on one s perspective, e.g., preferences for quality or quantity, and therefore, it is wise to examine maps for multiple perspectives composite index, top-5 index, and other perspectives as well. The need to assess shoreline habitat value through multiple perspectives creates some challenges for county governments trying to conserve marine shoreline habitats. One perspective on conservation value (the composite index) indicates some shorelines have high value for a multitude of species and clearly, every effort should be made to protect such places. However, another perspective (the top-5 index) shows that nearly every shoreline segment has relatively high value for some marine fish or wildlife species. These shorelines should be considered for special management as well. The Indices We developed an index to indicate conservation value for fish and wildlife of marine shorelines in Puget Sound. Multi-species measures of conservation value or importance are normative. There is no purely objective single conservation value that can be empirically validated because value is based on one s belief about what is valuable, and therefore, it is influenced by personal values. Nevertheless, given the limitations imposed by the comprehensiveness and accuracy of the available GIS data, we believe our index is a reasonable indicator of conservation value. There is more than one way to construct an index of conservation value for marine shorelines of Puget Sound. We have created two the composite index and the top-5 index, and there are many more ways to quantitatively assess conservation value. For instance, we assigned equal influence to every component of the index. Others might believe that certain components are better indicators of conservation value or that some species are more valuable than others. Also, our index had a flat 38

42 structure (Figure 9). Others might prefer a hierarchical structure in which similar components are grouped together, the groups are assigned weights that determine their relative influence, and then the relative values of groups are summed to yield an index. In both cases, relative influence of components and index structure, we chose to minimize subjective judgments, which led to equal influence and the flat structure. The 41 components of the index are not a comprehensive accounting of the many animal and plant species and habitat or biotic community types that are found along marine shorelines in Puget Sound. We lack occurrence data for the majority of species. However, in general, we collect occurrence data for certain species or habitat types because 1) humans harvest those species, 2) we are concerned about the status of those species or habitats (e.g. threatened or endangered species), or 3) we are concerned about the management of those species or habitat s(species sensitive to human disturbances). In other words, we collect data on those species and habitats we care most about. Therefore, an assessment based on these data should indicate those places we should care most about for the conservation of fish and wildlife habitats. For most of the components that comprise the indices, we believe that error rates in the occurrence data are acceptable. Lower error rates in the data could be achieved but the cost of data collection could be prohibitive. For a subset of the components we believed that the error rates were likely to be unacceptable, and for these we developed PO models. These models may overestimate the probability of occurrence, but we believed using a model with a high rate of false positives was more precautionary, and hence, preferable, to using data with a high rate of false negatives. Those components for which we did not develop a PO model were assumed to be equally accurate. This is unlikely to be true. Some datasets are regularly updated through annual systematic surveys (e.g., PSAMP bird surveys), while other datasets rely on the reporting and recording of incidental observations (e.g., bald eagle nests, great blue heron colonies). We could have compensated for these differences in accuracy by weighting some data sets more heavily than others, but we chose not do this because evaluating relative accuracy and assigning weights would entail numerous subjective judgments. The extent of our analysis covers the entire Puget Sound, however, we split the Sound into oceanographic sub-basins. The spatial extent over which an assessment is conducted affects one s interpretation of relative conservation value. For instance, a shoreline segment could have high relative value in the Whidbey sub-basin but have only a moderate relative value in Puget Sound. A land use plan for Island County might target that shoreline for protection, but a conservation plan for the entire Puget Sound might not. A shoreline with high regional value should be considered more valuable to regional authorities than a shoreline that has high local value but only low or moderate regional value. On the other hand, an shoreline with low regional value could be the most valuable shoreline within a local jurisdiction. Our assessment in its current form does not enable Basin-wide comparisons. Regional authorities should keep that in mind when using this assessment to identify the most valuable shorelines within Puget Sound. The assessment could be restructured and reformulated to provide Sound-wide comparisons. A favorite saying amongst statisticians is all models are wrong. What they really mean is all models have uncertainty, and our assessment is no different. The information depicted in Figure 13 is incomplete because it lacks an expression of uncertainty. However we can make generalizations about where we can be confident of the relative values and where we should be wary. For assessments of this type, relative values in the top quantile and bottom quantile tend to have the smallest uncertainty, and 39

43 places with more moderate values tend to have the largest uncertainty. A set of places with different relative values but all in the moderate range may effectively have the same value. Greater uncertainty does not mean that an shoreline segment has lesser value than that estimated by this assessment. Greater uncertainty means that the actual conservation value could be larger or smaller than the estimated value. The composite index is a continuous value from 0 to 1 (or 0 to 100). For the purposes of mapping, the continuous values were divided into categories via equal intervals or statistical quantiles. Interval widths or quantile sizes are somewhat arbitrary, and therefore, one must be cautious when interpreting maps (Figure 24). For instance, a shoreline segment with a score of 0.78 would be in the highest value category when using 4 equal intervals but be in a lesser category when using 5 equal intervals. Furthermore, categories obscure some quantitative relationships between segments. For instance, when using 5 equal intervals, an segment with a score of 0.61 is in the same category as a segment with a score of 0.79 but 0.61 is actually closer to a segment with a score of 0.59 which is in a different category. Figure 24. Interpretation of maps. The composite index is a continuous value from 0 to 1. To facilitate the interpretation of relative conservation value, the continuous values are divided into categories via equal intervals or statistical quantiles. Quartile and quintiles, for instance, are quantiles in which each group contains 25% and 20% of the shoreline segments, respectively. These simplifications warrant two caveats. First, within categories segments do not have the same conservation value. Second, the categories are somewhat arbitrary and can obscure relationships between segments. Our indices could be improved several ways. First and foremost, more up-to-date and accurate occurrence data are needed. Some of the occurrence data, in particular, those data described in WDF (1992), are decades old. Inaccurate data can result in the mischaracterization of high value shorelines as low value and low value shorelines as high value. Both errors lead to an inefficient allocation of resources for protection and restoration of shorelines. Second, a process of validating the index should be explored. The current index is based on best professional judgment. The conservation value expressed by our index might be compared to other measures of conservation value. Third, an index 40

44 that integrates process, structure, and function could be developed. Our index of conservation value is based on habitat functions. Habitat functions are emergent properties of ecosystem processes and structures, and hence, the relationship between functions and processes or structures is sometimes obscure. An quantitative model built on these relationships would provide a fuller understanding of why a place is important and insights about how to manage that place. Until we have such a model, we will integrate the results of this assessment, which emphasizes function, with those of PSNERP s assessment (Cereghino et al. 2012) which emphasizes processes. Integrating Assessments of Function and Process We have two large-scale assessments for Puget Sound shorelines one that assesses shoreline habitat functions and another (Cereghino et al. 2012) that assesses the major nearshore process that affects shoreline habitats. Management decisions affecting shoreline must utilize both assessments. The connections between shoreline habitat functions and drift cell processes are probably strongest along beaches and barrier embayments. There are 812 drift cells in Puget Sound, so some system is needed to simplify and thereby facilitate the integration of the PSNERP strategies with the results of our assessment. The beach and barrier embayment strategies make management recommendations based on the relative degradation of drift cells (Table 2). Protect is recommended for the least degraded drift cells, enhance for the most degraded drift cells, and restore for those drift cells in between (Table 3). To simply the PSNERP assessment we combined the beach and barrier embayment strategies by comparing the two recommendations for each drift cell and taking the recommendation with the minimum level of degradation. To facilitate integration of the PSNERP strategies with our assessment, we calculated the mean composite index for each drift cell and then divided the drift cells into three groups by terciles. The three groups correspond to three levels of mean conservation value high, medium, and low. The integration scheme uses PSNERP s three management recommendations and the three levels of conservation value (Figure 25). The combination of the PNSERP strategy and conservation value should help to further refine management priorities within sub-basins. For instance, a drift cell recommended for protection with high habitat conservation value should be a higher priority for protection than drift cell recommended for protection with a medium or low conservation value. Likewise, a drift cell recommended for restoration with high habitat conservation value should be a higher priority for protection than drift cell recommended for restoration with a medium or low conservation value. Where conservation value and management recommendations align (i.e., protect-high, restore-medium, enhance-low), which occurs for about one-third of drift cells, then site-level management decisions should be straightforward (Figure 26). That is, a drift cell with low degradation (i.e. protect recommendation) and high habitat conservation value is an obvious candidate for protection, a drift cell with moderate degradation (i.e., restore recommendation) and medium habitat conservation value make sense to restore, and drift cell with high degradation (i.e. enhance recommendation) and low habitat conservation value makes sense to enhance. When the assessments do not align, what is the management recommendation? For instance, how should we manage shorelines that have an enhance recommendation (high degradation) but high habitat conservation value (which describes about 8 percent of drift cells)? Enhance signifies a low priority for protection or restoration but high habitat conservation value contraindicates that recommendation. Site-level management decisions for these drift cells will require further analysis. 41

45 Local information is important for all site-level decisions, but will be especially important under these circumstances. Why would the assessments not align? For instance, why would a drift cell have high degradation and high habitat conservation value or low degradation and low habitat conservation value? There are several potential explanations. First, there may be time lags in the responses of fish, wildlife, and plant species to the degradation of nearshore processes. In other words, it may take some time for a degraded drift cell to lose its habitat conservation value. Second, certain fish and wildlife species may be responding to ecosystem structures and processes other than those related to littoral drift. For instance, they may be responding to the proximity of nearby rocky reefs, upwelling currents, local fetch, or human disturbances. Third, species could be responding to structures or processes that occur at spatial scales different than those of our assessment. And finally, the two assessments may not align because one or both of them are wrong. Even the best models are occasionally wrong, and hence, there will be portions of the Puget Sound shoreline (hopefully very small portions) where the assessments are wrong. Figure 25. Overlay of mean composite index onto the PSNERP strategies for beaches and embayments. The composite index was averaged over each drift cell. Terciles for mean composite index (thinner center line) were calculated for each sub-basin. Beach and embayment strategies (thicker line) were combined by comparing strategies for each drift cell and taking the recommendation with the minimum level of degradation: enhance > 42

46 restore > protect. Green is highest conservation value and red is lowest conservation value. Management recommendations for deltas (Nisqually, Deschutes, and Skokomish) not shown. Figure 26. Using PSNERP s assessment (Cereghino et al. 2012) and the composite index to guide shoreline management. Prot, rest, and enha refer to protect, restore, and enhance management recommendations of the PSNERP beach and barrier embayment strategies. High, med, low, refer to high, medium and low levels of conservation value and correspond to the third, second, and first terciles of the composite index, respectively. Where the two assessments align (i.e., prot-high, rest-med, enha-low), management decisions should be more straightforward than where the assessments do not align (e.g., prot-low, enha-med). Other Assessments This assessment and all the assessments done for the Puget Sound Partnership s Watershed Characterization project do not constitute all the information necessary and sufficient to address natural resources conservation through land use planning and decision making. Therefore, this assessment should be supplemented with other assessments. In fact, high quality shoreline assessments are currently available that may be more useful to local land use planners (e.g., May and Peterson 2003, Diefenderfer et al. 2009). This assessment is intended for sub-basin or landscape-scale planning and decision making, and consequently, does not address habitat issues that are best addressed through smaller-scale assessments, such as prioritizing site-level restoration projects. Each of the Salmon Recovery Lead Entities has done their own assessments to support their recovery plans for Chinook and steelhead (e.g., East Kitsap 2004, Pierce County 2004, Snohomish County 2005). The work done by lead entities serves a particular purpose, is highly attuned to local knowledge, and has involved local stakeholders, and 43