Conservation Assessment and Prioritization System (CAPS) Highlands Communities Initiative PHASE 1

Final Report June 9, 2005

Landscape Ecology Program, Department of Natural Resources Conservation, University of , Amherst.

Kevin McGarigal, Associate Professor, Department of Natural Resources Conservation. University of Massachusetts, Amherst, MA 01003. Phone: (413) 577-0655; email: [email protected]

Bradley W. Compton, Research Associate, Department of Natural Resources Conservation. University of Massachusetts, Amherst, MA 01003. Phone: (413) 577-2179; email: [email protected]

Scott D. Jackson, Program Director, UMass Extension Natural Resources and Environmental Conservation Program. University of Massachusetts, Amherst, MA 01003. Phone: (413) 545-4743; email: [email protected]

Kasey Rolih, Research Associate, Department of Natural Resources Conservation. University of Massachusetts, Amherst, MA 01003. Phone: (413) 577-2179; email: [email protected]

Eduard Ene, Research Associate, Department of Natural Resources Conservation. University of Massachusetts, Amherst, MA 01003. Phone: (413) 577-2179; email: [email protected]

Introduction

Planners and conservationists have long sought better ways to proactively conserve the most significant natural areas before they are lost or irreversibly fragmented, but it is difficult to identify which areas are in the greatest need of protection, or which ones provide the greatest ecological value for the cost of protection. Analyzing a landscape’s ecological/biodiversity value requires integrating vast amounts of site-specific information over varying spatial scales. Municipalities and other participants in the decision-making process simply have not had access to the databases and technical tools that these complex analyses require. Furthermore, they have been hindered by a partial or complete lack of information of what—in terms of biodiversity—is present in the landscape, let alone the distribution and inherent value of each natural community unit or the role of each unit in sustaining a fully functioning landscape. Nevertheless, state agencies, conservation organizations and communities across the Commonwealth are spending millions of dollars on land acquisition each year to protect natural areas.

During the past several years, we have pioneered the development of a method for quantitatively evaluating the ecological integrity of a landscape and its biodiversity* value at any scale--state, region, watershed, town, etc. The Conservation Assessment and Prioritization System (CAPS) is a community-based, coarse-filter approach for assessing the ecological integrity of lands and waters and subsequently identifying and prioritizing land for biodiversity conservation. We define ecological integrity as the ability of an area to support biodiversity, and the ecosystem processes necessary to sustain biodiversity, over the long term. Our approach assumes that by conserving intact, ecologically-defined communities of high integrity, we can conserve most species and ecological processes. Moreover, by identifying the lands most worthy and in need of protection, towns, land trusts and others can target their limited dollars strategically; in short, they can get the most out of every conservation dollar. Our coarse filter is a first step in the process of targeting land for conservation. Field work will be required to verify predictions made by our broad-scale model, and a further, fine-filter approach will be necessary to include habitat for species of concern that slip through the cracks—this includes many threatened and endangered species.

Overview of CAPS

CAPS is a computer modeling approach to prioritizing land for conservation based on the assessment of ecological integrity for various natural communities (e.g., deciduous forest, grassland, shrub swamp, first-order stream) within an area. Beginning with a GIS base map depicting various classes of developed and undeveloped land, we evaluate a variety of landscape- based metrics (or indices) to calculate ecological integrity for every point in the landscape. Each landscape metric evaluates a different aspect of the underlying natural community map. A metric may, for example, take into account the size of a natural community patch, its proximity to streams and rivers, the diversity of soil types in the patch, or the intensity of roads in the vicinity. For each natural community, several metrics are applied to the landscape and then integrated in a weighted linear combination. Weights are supplied by expert teams to reflect the relative importance of each metric for each community. This process results in a final “index of ecological integrity” for each point in the landscape. Intermediate results are saved to facilitate analysis—thus one can examine not only a map of the final indices of ecological integrity, but maps of road intensity, natural community patch area, soil series diversity within forested areas, and so on.

Hierarchical Community Levels – Within CAPS, ecological integrity may be assessed at three hierarchical community levels. At the lowest level are primary communities, such as shrub swamp, or cultural grassland. These communities are aggregated into secondary communities, such as palustrine or grasslands based on wildlife habitat use. Finally, secondary communities are aggregated into three tertiary communities: Forests, Nonforested Uplands, and Wetlands and

*For our purposes, we define biodiversity as the diversity of life at all levels of organization from the gene to the landscape and all the ecological and evolutionary processes and interconnections that support life across levels of organization. In its broadest sense, biodiversity is the variety of life forms and environments that support that life. Here, we adopt a more pragmatic focus on the maintenance of viable populations of all native species (from carnivores to soil bacteria) and ecological communities (hereafter simply referred to as “communities”) found in their natural places, distributed and functioning within their natural range of variability.

2 Aquatic Communities. Thus, each point on the landscape is a member of a primary community, a secondary community, and a tertiary community. Analysis may be done at any of these three levels alone or in combination, and the results can be interpreted independently at each level or they can be combined into a single, multi-level assessment. Metrics do not apply to developed land—all cells corresponding to developed land cover types are given an ecological integrity index of zero, even though we recognize that even developed land may contribute to the conservation of biodiversity.

Landscape Metric Groups – Landscape metrics are organized into five groups, and these metrics may be computed at any of the hierarchical community levels:

• Composition metrics evaluate the rarity, richness, or evenness of communities or abiotic values in the focal patch, without regard to the spatial context of the patch.

• Spatial character metrics evaluate the shape or configuration of a patch, without regard to its composition or spatial context.

• Context metrics evaluate the composition and configuration of the neighborhood (i.e., ecological context) surrounding each point in a community.

• Condition metrics evaluate negative effects (usually anthropogenic disturbance) on the ecological integrity of each point in a community, based on the composition and configuration of the neighborhood.

• Aquatic (watershed) metrics evaluate the condition of aquatic communities based on the watershed above each point, rather than its surroundings in all directions. These metrics follow flow upstream and uphill to define an influence zone for each point, based on the modeled rate of flow across different land cover types and slopes. Watershed metrics are weighted by this influence zone, so that, for instance, a potential pollution source near a stream would have a stronger influence than one far upstream, or a distance away from the stream.

Combining Metric Results – Results from the landscape metrics are rescaled, weighted, and then combined into an overall index of ecological integrity. First, the results of each metric are rescaled by percentiles for each community so that, for instance, the best 10% of marshes have values ≥ 0.90, and the best 25% have values ≥ 0.75. This is done to adjust for differences in units of measurement among metrics and to account for differences in the range of metric values for each community. The rescaling by community is done to facilitate identifying the “best” of each community, as opposed to the best overall – which is strongly biased towards the dominant, matrix-forming communities. Next, the rescaled values are weighted (weights assigned by the user), to reflect the relative importance of each metric for each community (Appendix C), and then added together to compute an overall index of ecological integrity. The weighted linear combinations are nested. First, all metrics are combined within each metric group (composition, spatial character, context, condition and watershed). Then, these five groups are combined to represent the index of ecological integrity at the current community level. Thus, the final index

3 of ecological integrity for each cell is a weighted combination of the metric outputs for that cell, based on the community the cell falls in.

Identifying and Prioritizing Land for Conservation – Among its many uses, the index of ecological integrity can be used alone or in combination with other approaches to identify and prioritize lands for conservation. The index can be used, for example, to identify the top 10% or 30% of the land (in terms of ecological integrity) in an area that will hopefully provide the greatest ecological value and therefore provide an effective and credible basis for a land conservation strategy. It is especially important to note that the assessed ecological integrity of land in an area (and therefore the lands identified and prioritized for conservation) depends on the geographic extent of the analysis area. This is so because the rescaling of the metrics is done to identify the best of the available lands, but the “available lands” varies with geographic location and extent. Thus, the best example of a particular community within a certain geographic extent might be a relatively poor example when assessed over a much larger extent. For this reason, the index of ecological integrity can be rescaled to reflect the range of conditions within any sub-landscape or geographic extent less than the entire analysis area. For example, the index might be rescaled within each of several logical ecological units such as watersheds or ecoregions.

Objectives

The overall purpose of this project is to generate credible, reproducible, quantitative output (maps, spreadsheets) on the ecological integrity and biodiversity value of lands within the Highlands focus area that can be integrated into the strategic planning and conservation strategies for the towns and the region. This project is being implemented in two phases. Phase 1 involves the ecological assessment; i.e., the CAPS analysis itself. Once we’ve produced the ecological/biodiversity assessment maps and complementary spreadsheets, in Phase 2 we will work with UMass Extension and the Highlands Community Initiative to use this and other information to actually prioritize sites for conservation. The specific objective for this phase of the project are as follows:

1. Compile a comprehensive land cover map, including developed land cover classes and natural communities. This will involve not only the digital creation of the land cover map but also the development of a CAPS model (to assess ecological integrity) for each natural community type.

2. Conduct the CAPS analysis to produce maps and spreadsheets of the ecological integrity of undeveloped lands throughout the region, specific watersheds, and within individual towns.

3. Provide an initial prioritization of sites for conservation. Note, this will serve as the starting point for the next phase of this project (to be completed outside the scope of this contract and contingent upon additional funding), which will involve using the ecological assessment, complementary spreadsheets, and other information to prioritize sites for conservation.

4 Project Area

Our analysis was done for the Highlands region plus the Housatonic watershed and the remainder of the Deerfield and Westfield watersheds (Fig. 1). Our initial focus area was the Highlands region, which includes 38 rural towns that lie between the Housatonic and River Valleys, and the and Connecticut borders. This region includes portions of four counties (Franklin, Hampshire, Hampden and Berkshire) as well as four major watersheds: the Deerfield, the Westfield, the Highlands Communities Focus Area Farmington and the Berkshire Franklin* Hampshire* Hampden* Housatonic. It covers Becket Ashfield Chesterfield Blandford approximately 1100 Florida Buckland Cummington Chester square miles (13.5% of Hinsdale Charlemont Goshen Granville Massachusetts). The area Monterey Colrain Huntington Montgomery is still largely rural, the New Conway Middlefield Russell result of geographic Marlborough isolation and slow growth Otis Hawley Plainfield Tolland rates during many of the Peru Heath Westhampton last several decades. The Sandisfield Leyden Williamsburg region is the focus of the Savoy Monroe Worthington citizen-borne Highlands Tyringham Rowe Communities Initiative, Washington Shelburne which is being overseen Windsor by The Trustees of Reservations. We • Italics indicates that all or a portion of the town has been extended our analysis area evaluated in our pilot project in the Housatonic watershed. beyond the boundaries of • An * indicates that all of the towns in these counties are part the Highlands focus area of the Plan for Progress to include the remainder of the logical ecological units (i.e., watersheds and ecoregions) that comprise the Highlands region.

Methods

Input Data

GIS data from a variety of sources were combined to create a base map depicting natural communities, developed land types, and roads. Appendix B describes the GIS data used. All data are mapped in 30 m grids. The final land cover layer depicts all natural communities at the primary level, as well as development and roads. See Appendix A for the land cover classification. Several other layers depict subsets of this final land cover, including the intermediate land cover (with large roads and streams omitted), roads, railroads, and streams layers. Finally, several ancillary layers are used by specific metrics. These include the flow and stream flow grids, the flow resistance grid, Ecological Land Unit (ELU), lithology, and point- source pollution.

5

Fig. 1. Project area. CAPS analysis was done for the Highlands Communities (green shading), as well as the remainder of the Housatonic, Deerfield, and Westfield watersheds (blue shading). Town boundaries are shown in gray; watershed boundaries in cyan.

Data Accuracy

The CAPS analysis for the Highlands was done entirely with available data. These GIS data come from a variety of sources, and the quality of these data are variable. We integrated these data sources into a single land cover map, with several parallel data layers, including a flow grid, watershed resistance, and several abiotic layers such as soils and slope. We put considerable

6 effort into integrating these input layers in ways that maximized the accuracy of available data, while making sure the final map generally makes sense, both visually and to the CAPS metrics. Because input data came from several different sources, we have no estimate of the accuracy of the final map, nor of the effect errors in the base map may have on final CAPS results. Nobody should have any illusions that the base map presents a “true” depiction of the landscape—a comparison of the landcover with aerial photos or with familiar places will turn up errors in classification and position. Furthermore, the classification is fairly coarse, and distinctions between deciduous and mixed forest, or between marsh and shrub swamp, are necessarily arbitrary. Many of these communities change over time, so our snapshot based on data generated over several years may depict today’s beaver pond as yesterday’s forested wetland. These issues are important to understand and to communicate clearly to stakeholders and users of the results—nobody should be surprised when they find data errors in their backyards.

Given these caveats, we believe that the effects of many of the data errors will be relatively small. CAPS operates at fairly broad scales, looking at the effects of the surrounding landscape on any particular point. Small errors in classification and placement (small roads and streams omitted, marshes slightly shifted, small forest patches lost because of the grain of the map) will usually have a small but negligible effect on final results. We plan to evaluate the effects of various kinds of errors on CAPS results in the future.

The coarseness of the classification scheme is perhaps a larger issue. Available data necessitated lumping many different forest communities into a broad deciduous/mixed/ coniferous class; likewise, many rare and small-patch-forming communities are omitted. This leaves CAPS unable to compare patches of rich mesic forest to other patches of rich mesic forests, or to evaluate acidic rocky outcrops.

CAPS Analysis

The full details of the CAPS analysis conducted for this project are beyond the scope of this report. Details on the CAPS software, including detailed descriptions of each landscape metric, can be found at the UMass Landscape Ecology Program website (www.umass.edu/landeco). Briefly, once the input data layers are created, analysis in CAPS requires a model to be defined for each natural community. Each community’s model entails selecting a number of metrics, parameterizing them for that community, and weighting them by importance for that community. This model parameterization was originally done by three expert teams as part of the Housatonic watershed pilot project. Additional parameterization and some necessary modifications were done for the Highlands project by Kevin McGarigal, Scott Jackson, and Brad Compton. The metrics selected for each of the communities and their relative weightings are listed in Appendix C. Detailed parameterizations of each metric may be examined by using the CAPS software to load the model file highhousa.cml.

The parameterized model is run on the input layers using the CAPS software, written at UMass by Eduard Ene. This software produces an output grid for each metric at each of the three community levels. These output grids are then rescaled, weighted, and combined into final index of ecological integrity (IEI) values at each of the three levels. The IEI for each cell is a weighted combination of the metric outputs for that cell, based on the community the cell falls in. Results

7 are rescaled by percentiles, so that, for instance, the best 10% of marshes have values ≥ 0.90, and the best 25% have values ≥ 0.75. A separate analysis allows each cell to be assessed in the context of its watershed or ecoregion. For this analysis, the IEI is rescaled by percentiles within each watershed or ecoregion. For example, if the IEI is rescaled by watershed, a marsh with a value of 0.85 would be interpreted as being in the 85th percentile of marshes for its watershed.

Results

CAPS results are best explored interactively, using a GIS that can display grids (e.g., ArcView or ArcMap). This section discusses the results and gives a few examples. We will work with TTOR and the Highlands Consortium to devise useful summaries of these results as well as a prioritization of sites. The attached DVD includes input data, raw metric output, scaled metric output, and the IEI for each level. The attached CD contains a subset of these data. See Appendix D for details.

The most useful results are the landcover grids and the final result grids. These are included on the DVD along with ancillary data and intermediate results, as well as on the CD. The three landcover grids depict natural communities at each of the three community levels, as well as developed land and roads. The primary landcover (Fig. 2) is the most detailed landcover grid. Landcover types are listed in Appendix A, and ArcView legends are provided with the data.

Primary landcover High-intensity urban Low-intensity urban High-intensity residential Low-intensity residential Agricultural Dam Expressway Primary highway Secondary highway Light-duty road Unpaved road Railroad Bridge & culvert Deciduous forest Mixed forest Coniferous forest Foreseted wetland Powerline shrubland Old field Cultural grassland Cliff and steep slope Shrub swamp Marsh Pond Vernal pool Lake River & stream

Fig. 2. Final landcover for the town of Becket. Note that river and stream types are lumped to secondary levels for clarity.

8 Final results grids give the index of ecological integrity (IEI) for each of the three community levels at three scales: the entire project area, watershed, and ecoregion. Because IEIs are scaled from 0 to 1 by percentiles within each community, images such as Fig. 3 tend to be overwhelmed by values for forest communities, because the landscape of is mostly forest (e.g., three shades of green in Fig. 2).

Fig. 3. Index of ecological integrity (IEI) for town of Becket at the secondary level, scaled to the entire project area. Darker areas denote higher IEI values; white areas are developed land.

Communities may be viewed one at a time for a more accurate assessment of high-valued areas by community (e.g., for palustrine wetlands, Fig. 4). This community-by-community approach is a necessary part of a conservation prioritization to be sure of protecting communities that occur in small patches.

Priority areas can be highlighted by showing only the top x%, for instance the top 30% (IEI ≥ 0.70, Fig. 5), or the top 10% (IEI ≥ 0.90, Fig. 6). Because the IEI is scaled by percentiles within each community, these images show (for instance) the top 30% in each community. Depictions of the top x% may keep the (still meaningful) gradations, or they can just show polygons of the top x% in each community on top of another map, such as the land cover map or a USGS topographic map (Fig. 7).

9

Fig. 4. Index of ecological integrity (IEI) as in Fig. 3, for palustrine wetlands only.

Fig. 5. Index of ecological integrity (IEI), top 30%.

10

Fig. 6. Index of ecological integrity (IEI), top 10%.

Fig. 7. Index of ecological integrity (IEI), top 30% purple outline) and top 10% (red outline) over USGS topographic map.

11 A final example shows the effects of scaling by different extents (project area, watershed, or ecoregion). The previous examples, scaled by the entire project area, rate each point in the landscape in comparison to the entire project area. Thus, it would be correct to say “Becket has palustrine wetlands that are among the 10% best IEI across the Highlands-Housatonic project area.” Scaling by watershed or ecoregion would change the scope of the comparison, affecting the IEIs. For instance, a statewide analysis would probably raise the IEI of Becket’s forests, because a forest block that may be good in terms on western Massachusetts would be excellent in terms of the entire state.

An example in Westfield and West Springfield highlights this issue (Fig. 8). Unsurprisingly, these towns have generally low IEIs, because this area is highly urbanized. The only communities in the top 10% are a handful of short stream reaches. When rescaling by ecoregion, however, each point of land is now compared just to other points in the same community within the same ecoregion, in this case the Valley ecoregion (Fig. 9). In this particular analysis, the only parts of the Connecticut River Valley ecoregion that were included are the urban areas at the mouths of the Deerfield and Westfield, so absolute IEIs are low as a whole across this ecoregion. The IEI scaled by ecoregion shows many high-valued areas. This analysis allows CAPS to answer the question, “Of course Westfield and West Springfield rank poorly within the Highlands-Housatonic area—they’re mostly urban. But we’re not going to write them off—we still want to prioritize land for protection in these towns. What’s the best that’s there?” The effects of scaling within the Highlands area will generally be much less extreme, but the effects of scaling are the same.

Fig. 8. Index of ecological integrity (IEI) scaled to entire project area. Black lines indicate boundary of project area.

12

Fig. 9. Index of ecological integrity (IEI) scaled to ecoregion. Black lines indicate boundary of project area.

Future Work

The second phase of the CAPS Highlands project is now underway. This phase includes summarization of the results and application of these results to conservation in the Highlands region. This work will be done by UMass Extension in cooperation with the Highlands Consortium.

13 Appendix A: Land cover classes

Land cover classes used for the Highlands project are listed below, with numeric codes for the primary community level. Tertiary communities are listed in boldface, secondary communities in italics, and primary communities have a numeric code. Forests are the same at the secondary and primary levels.

Natural Communities Forests Riverine* 111 Deciduous forest 411 First order flatwater 121 Mixed forest 412 First order pool-riffle 131 Coniferous forest 413 First order plane-bed 191 Forested wetland 414 First order step-pool 415 First order cascade Nonforested Uplands 421 Second order flatwater Shrubland 422 Second order pool-riffle 211 Powerline Shrubland 423 Second order plane-bed 212 Old field 424 Second order step-pool Grassland 425 Second order cascade 221 Cultural grassland 431 Third order flatwater Cliff and steep slope 432 Third order pool-riffle 231 Cliff and steep slope 433 Third order plane-bed 434 Third order step-pool Wetlands & Aquatic 435 Third order cascade Palustrine 441 Fourth order flatwater 311 Shrub Swamp 442 Fourth order pool-riffle 312 Emergent Marsh 443 Fourth order plane-bed 315 Pond 444 Fourth order step-pool 316 Vernal Pool 445 Fourth order cascade Lacustrine 451 Fifth order flatwater 321 Lake 452 Fifth order pool-riffle 453 Fifth order plane-bed 454 Fifth order step-pool 455 Fifth order cascade 461 Sixth order flatwater 462 Sixth order pool-riffle 463 Sixth order plane-bed 464 Sixth order step-pool 465 Sixth order cascade

*Encoded as 4og, where o = order (41x – 46x) and g = gradient (4x1 – 4x6). Orders: First – Sixth Gradients: Flatwater, Pool-riffle, Plane-bed, Step-pool, Cascade

14 Development & Roads Developed land 11 High-intensity urban 12 Low-intensity urban 13 High-density residential 14 Low-density residential 15 Agricultural/Managed open Dams 16 Large Dam (>1000 acre-ft or >40 ft high) 17 Medium Dam (50-1000 acre-ft or 15-40 ft high) 18 Small Dam (15-50 acre-ft or 6-15 ft high) 19 Non-jurisdictional or Unknown Dam (<15 acre-ft and <6 ft high) Roads 21 Expressway 22 Primary highway 23 Secondary highway 24 Light duty road 25 Unpaved road 26 Railroad 31 Culvert 32 Bridge

15 Appendix B: Input Data Layers

Forests – We used National Land Cover Data (NLCD) for the upland forests. Our comparison of NLCD with other data sources showed that most NLCD classes were wildly inaccurate, with errors ranging from small-scale speckling to gross errors such as large areas of development within October Mountain State Forest, and the depiction of the Montague sand plains as forested wetland. We were able to use sources other than NLCD for all but upland forests, which NLCD coarsely represents as deciduous/mixed/coniferous. These classifications were reasonably accurate (this is easily assessed from aerial photos).

Nonforested Uplands – Three of these communities came from the UMass Resource Mapping Unit’s 1999 Land Use: cultural grassland (from pasture), powerline shrubland (from powerlines/pipelines), and old field (from open land). A fourth community, cliffs and steep slopes, was defined as any forested land with a slope > 60%.

Wetlands – We used three different sources for wetlands, because none of the three were complete: Massachusetts DEP/Wetlands Conservancy, National Wetlands Inventory (NWI), and our modeled wetlands for the Housatonic watershed. DEP wetlands were photo-interpreted, and are generally of high quality; unfortunately they have not been completed for the entire Highlands area. DEP wetlands were used for about half of the Highlands area. NWI wetlands are also photo-interpreted, but are of fairly low quality (they were done from 1:40,000 monochrome photos back in the late 1970’s). We used NWI for about 1/4 of the Highlands area, where nothing else was available. As part of the CAPS prototype in the Housatonic watershed, we built a wetlands map based on extensive field sampling, Landsat images, and terrain data. The accuracy of these data are intermediate between DEP wetlands and NWI. We used these wetlands for the Housatonic watershed and about 1/4 of the Highlands.

Lakes and Ponds – We used MassGIS 1:25k hydrography to represent lakes and ponds. Ponds were defined as being waterbodies smaller than 5 ha, lakes as those larger than 5 ha. This is based on a logistic regression of sizes of lakes and ponds in areas where NWI falls within the Highlands, because NWI distinguishes between lakes and ponds, whereas DEP wetlands depict all open water as one class.

Vernal Pools – We used Potential Vernal Pools from MassWildlife’s Natural Heritage and Endangered Species Program. Potential vernal pools that fell within a larger wetland (up to 1 ha) identified the wetland as a vernal pool; others were treated as a single pixel pool (30 m × 30 m).

Streams and Rivers – Streams and rivers are based on our work for Natural Heritage and Endangered Species Projgram’s Living Waters project. MassGIS 1:25k stream centerlines were used to define streams. Streams are classified by order and gradient. Order is calculated from the stream centerline data; and gradient is based on the digital elevation model. We identified rivers that flow into the state to correct the order of these stream networks. For rivers wider than 30 m, the open water class from Land Use was used to represent the entire river basin, and the class based on order and gradient was applied to the entire width.

16 Developed Land – Developed land comes directly UMass Resource Mapping Lab’s 1999 Land Use. Developed land is lumped into five types: high-intensity urban, low-intensity urban, high- density residential, low-density residential, and agricultural/managed open.

Dams – Dams (in four size classes) were developed in collaboration with DEP and Mass Riverways as part of Natural Heritage’s Living Waters project. Dams were derived from a MassDEP point shapefile and digitized as lines over stream centerlines overlaid on the MassGIS 1 meter, 1:5000 black and white orthophotos. Dams are treated as a developed type.

Roads and Railroads – Roads are railroads are from MassGIS’s 1:25k roads and trains layers. Roads were reclassified into five types based on original road classes as well as surface type (for unpaved roads).

Elevation – A digital elevation model (DEM) was created by David Goodwin of the UMass Resource Mapping Unit from MassGIS digital terrain model (DTM) elevation contours, elevation points, and topographic breaklines as part of the Living Waters project.

Flow – A flow grid (giving the direction of expected water flow for each cell) based on a digital elevation model was created for all of mainland Massachusetts by our lab as part of the Living Waters project. This flow grid conforms to MassGIS centerline data. We used this flow grid directly.

Aquatic Resistance – We modified the approach of Randhir et al. 2001 (Forest Ecology and Management 143:47-56) to build a time-of-travel grid for each cell in the project area, based on land cover, slope, flow, and stream gradient. This grid was used to define the influence area within the watershed of each point for our watershed metrics.

Ecological Land Units (ELUs) – We used The Nature Conservancy’s Ecological Land Units in our abiotic richness and evenness metrics. ELU’s seek to describe ecologically significant units of land, based on several abiotic layers: elevation, bedrock geology, distribution of deep glacial sediments that mask bedrock’s geochemical effects, moisture availability, and landform.

Lithology – We used MassGIS’s near-surface bedrock lithology layer (from USGS) in our abiotic richness and evenness metrics.

Point-source Pollution – Point-source pollution was defined by Massachusetts Natural Heritage and Endangered Species Program as part of their Living Waters project. These data are based on an assessment of pollution risk compiled from six DEP and EPA data layers: TRI (Toxic Release Inventory), RCRIS (Resource Conservation and Recovery Information), PCS (Permit Compliance System), MINES (Mineral Industry Locations), IFD (Industrial Facility Discharge Sites), and CERCLIS (Superfund National Priority List Sites) from the EPA Basins 3.0 website (http://www.epa.gov/waterscience/basins/metadata.htm). UST (Underground Storage Tank Locations), GRWTR (Ground Water Discharge Permits), and DEP Solid Waste Facilities point sources are available from MassGIS (http://www.state.ma.us/mgis/laylist.htm). See Heritage’s Living Waters Technical Report for details.

17 Appendix C: Metric Parameterizations

This table gives relative weights within and among metric groups for each metric by community at each community level.

i Condition Edge effects Road intensity Development intensity Aquatic Metrics WS Dev. intensity WS Road density WS Upstream road WS Dam intensity WS Percent impounded WS Pt-source pollution Composition Community rarity Community richness Community evenness Abiotic richness Abiotic evenness Spatial Character Patch area Core area Context Similarity Distance to water Connectedness Tertiary Communities

Forests 1 1 0 121 1 11 1 11 1 1 112 0 000000 Nonforested uplands 3 00011 2 11 2 1 0 1 2 112 0 000000 Wetlands & aquatic 1 0 2100 0 00 2 00 1 2 211 2 110001 Secondary Communities

Deciduous forest 1 00021 1 11 1 11 1 1 112 0 000000 Mixed forest 1 00021 1 11 1 11 1 1 112 0 000000 Coniferous forest 1 00021 1 11 1 11 1 1 112 0 000000 Forested wetland 1 00021 1 11 1 11 1 1 112 1 110001 Shrubland 3 00011 2 11 2 1 0 1 2 112 0 000000 Grassland 3 00011 2 11 2 1 0 1 2 112 0 000000 Cliff and steep slope 3 00011 2 11 2 1 0 1 2 112 0 000000 Palustrine 1 0 2100 1 1 0 2 1 0 3 1 111 1 110001 Lacustrine 0 00000 2 1 0 1 00 1 3 311 3 110001 Riverine 2 22110 1 1 0 6 00 1 3 211 5 111111

18

i Composition Community rarity Community richness Community evenness Abiotic richness Abiotic evenness Spatial Character Patch area Core area Context Similarity Distance to water Connectedness Condition Edge effects Road intensity Development intensity Aquatic Metrics WS Dev. intensity WS Road density WS Upstream road WS Dam intensity WS Percent impounded WS Pt-source pollution Primary Communities

Deciduous forest 1 0 0 0 11 1 11 1 11 1 1 112 0 000000 Mixed forest 1 0 0 0 11 1 11 1 11 1 1 112 0 000000 Coniferous forest 1 0 0 0 11 1 11 1 11 1 1 112 0 000000 Forested wetland 1 0 0 0 11 1 11 1 11 1 1 112 1 110001 Powerline shrubland 3 0 0 0 11 2 1 0 2 1 0 1 2 112 0 000000 Old field 3 0 0 0 11 2 11 2 1 0 1 2 112 0 000000 Cultural grassland 3 0 0 0 11 2 11 2 1 0 1 2 112 0 000000 Cliff and steep slope 3 0 0 0 11 2 1 0 2 1 0 1 2 112 0 000000 Shrub swamp 0 0 0 00 1 1 0 2 1 0 3 1 111 1 110001 Emergent marsh 0 0 n/a 0 00 1 1 0 2 1 0 3 1 111 1 110001 Pond 0 0 0 0 00 1 1 0 1 1 0 3 1 111 1 110001 Vernal pool 0 0 0 0 00 0 00 1 1 0 2 1 111 0 000000 Lake 0 0 0 0 00 2 1 0 1 00 1 3 311 3 110001 First order streams 2 0 0 0 1 0 0 00 6 00 1 4 211 4 111111 Second order streams 2 0 0 0 1 0 0 00 6 00 1 4 211 4 111111 Third order streams 2 0 0 0 1 0 0 00 6 00 1 3 211 5 111111 Fourth order streams 2 0 0 0 1 0 0 00 6 00 1 3 211 5 111111 Fifth order streams 2 0 0 0 1 0 0 00 6 00 1 2 211 6 111111 Sixth order streams 2 0 0 0 1 0 0 00 6 00 1 2 211 6 111111

19 Appendix D: GIS Data Directory

This appendix lists all GIS data provided on DVD. The Landcover grids and Final results are also provided on a CD. All data are Arc/Info grids unless otherwise noted.

Landcover grids \landcover\primary Landcover for primary level, including roads and streams \landcover\secondary Landcover for secondary level, including roads and streams \landcover\tertiary Landcover for tertiary level, including roads and streams \landcover\*.avl ArcView legend files for each landcover grid

Final results \results\final\primary\iei Final primary Index of Ecological Integrity, entire region \results\final\primary\iei_w Final primary Index of Ecological Integrity, by watershed \results\final\primary\iei_e Final primary Index of Ecological Integrity, by ecoregion

\results\final\secondary\iei Final secondary Index of Ecological Integrity, entire region \results\final\secondary\iei_w Final secondary Index of Ecological Integrity, by watershed \results\final\secondary\iei_e Final secondary Index of Ecological Integrity, by ecoregion

\results\final\tertiary\iei Final tertiary Index of Ecological Integrity, entire region \results\final\tertiary\iei_w Final tertiary Index of Ecological Integrity, by watershed \results\final\tertiary\iei_e Final tertiary Index of Ecological Integrity, by ecoregion

\results\final\...\iei.avl ArcView legend files for CAPS results \results\final\...\watershed Arc/Info coverage of watersheds \results\final\...\ecoregion Arc/Info coverage of ecoregions

Auxiliary grids & coverages (source for most: MassGIS) \auxil\elevation Elevation grid, in meters \auxil\hillshade Hillshading grid, for display \auxil\slope Slope grid, in percent slope \auxil\slopeln Logarithm of slope, for prettier viewing \auxil\aspect Aspect grid, in degrees \auxil\litho Near-bedrock lithology (coverage) \auxil\soils Soils data from SCS (coverage) \auxil\openspace Protected open space (coverage) \auxil\biomap NHESP’s BioMap & and supporting landscape (coverage) \auxil\watershed Watersheds (coverage) \auxil\ecoregion Ecoregions (coverage) \auxil\highhousa Outline of entire project area (coverage) \auxil\highlands Outline of Highlands region (coverage) \auxil\massachusetts Outline of Massachusetts (coverage)

20 \auxil\towns Massachusetts towns (coverage) \auxil\*.avl ArcView legend files for each auxililiary grid and coverage

Scaled landscape metric* results \results\scaled\primary\... Rescaled landscape metric results for primary level \results\scaled\secondary\... Rescaled landscape metric results for secondary level \results\scaled\tertiary\... Rescaled landscape metric results for tertiary level \results\scaled\aquatic\... Rescaled landscape metric results for aquatic metrics

Raw landscape metric* results \results\raw\primary\... Raw (unscaled) landscape metric results for primary level \results\raw\secondary\... Raw (unscaled) landscape metric results for secondary level \results\raw\tertiary\... Raw (unscaled) landscape metric results for tertiary level \results\raw\aquatic\... Raw (unscaled) landscape metric results for aquatic metrics

Input grids \caps\flc Final land cover \caps\ilc Intermediate land cover \caps\roads Roads \caps\rails Railroads \caps\streams Streams \caps\flow Flow grid \caps\strflow Streamflow grid \caps\elu Ecological land units \caps\lithology Lithology \caps\lithodry Lithology for uplands \caps\lithowet Lithology for wetlands \caps\resist Watershed resistance grid \caps\slope Slope \caps\point Point-source pollution

CAPS software & parameterization \caps\ht.exe CAPS software \caps\highhousa.cml CAPS model file for Highlands

* For a list of landscape metrics and their corresponding grid names, see CAPS Landscape Metrics, below.

21

CAPS Landscape Metrics The following grids are supplied in the \results folder, both in raw and scaled forms. Raw metrics are the original, unscaled results. Scaled metrics are rescaled by percentiles within each community, thus values of cconn ≥ 0.90 represent the 10% best locations for connectedness for each community. These scaled metrics were combined using the weights listed in Appendix C to create the final Indices of Ecological Integrity.

Grid name Landscape metric comp Composition metrics crarity Community rarity crich Community richness cceven Community evenness crabio Abiotic richness caeven Abiotic evenness spchar Spatial Character metrics cparea Patch area cpcarea Core area context Context metrics csim Similarity cdwater Distance to water cconn Connectedness cond Condition metrics cedge Edge effects crint Road intensity cdint Development intensity aqua Aquatic metrics wdevint Watershed development intensity wrdens Watershed road density wurcross Watershed upstream road crossings wdamint Watershed dam intensity wperimp Watershed percent impounded wpspoll Watershed point-source pollution

22