Tidal Wetlands after Hurricane Sandy: Baseline restoration assessment and future conservation planning

Final Report, February 2018

Submitted By: Co-PIs Chris S. Elphick1, Brian J. Olsen2, W. Gregory Shriver3 and Jonathan Cohen4

Post-doctoral researchers and staff Maureen D. Correll2, Wouter Hantson2, Brian Klingbeil1, and Elizabeth Tymkiw3

1. Department of Ecology & Evolutionary Biology and Center for Conservation and Biodiversity, University of Connecticut, 75 North Eagleville Road, U-43, Storrs, Connecticut 06269 2. School of Biology and Ecology, Climate Change Institute, University of Maine, 200 Clapp Greenhouse, Orono, Maine 04469 3. Department of Entomology and Wildlife Ecology, University of Delaware, 250 Townsend Hall, Newark, Delaware 19716 4. Department of Environmental and Forest Biology, SUNY College of Environmental Science and Forestry, 1 Forestry Drive, Syracuse, NY 13210

Acknowledgments

This work was primarily funded by a grant from the US Department of the Interior, Fish and Wildlife Service and North Atlantic Landscape Conservation Cooperative through the Disaster Relief Appropriations Act of 2013 (award # F14AC00965) and relied heavily on data collected using a Competitive State Wildlife Grant (U2-5-R-1) via the US Fish and Wildlife Service, Federal Aid in Sportfish and Wildlife Restoration award to the states of Delaware, Maryland, Connecticut, and Maine. We received additional funding that supported our work on tidal marsh bird responses to Hurricane Sandy from the US Fish and Wildlife Service (Cooperative Agreement Award Fl5AC00163) and the National Science Foundation (DEB-1340008). Additional support for portions of the work come from the United States Department of Agriculture (National Institute of Food and Agriculture, Project Number ME0- 21710). The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the funding agencies.

All work conducted by the Saltmarsh Habitat and Avian Research Program (SHARP) is a collaborative venture and we thank our co-PIs Adrienne Kovach and Tom Hodgman; other SHARP postdocs Britt Cline, Chris Field, and Kate Ruskin; and graduate students, especially Sam Apgar, Bri Benvenuti, Logan Cline, Meaghan Conway, Tim Freiday, Laura Garey, Alison Kocek, and Sam Roberts. We also thank the many Saltmarsh Habitat and Avian Research Program (SHARP) field technicians who collected data for this effort, and all participating landowners and land managers that allowed access to their properties or logistic support in the field. In particular, we thank the many points-of-contact for coastal resiliency project sites at which we worked: R. Albers, K. Bennet, J. Burger, P. Castelli, C. Chaffee, P. Comins, D. Curson, A. Drohan, L. Duff, N. Ernst, B. Finn, C. Folsom-O’Keefe, B. Gaspar, D. Golden, S. Guiteras, H. Hanlon, J.M. Hartman, K. Holcomb, E. Jordan, M. Katkowski, M. Larson, J. Lister, J. Luk, G. Macdoland, J. Mattei, L. Mitchell, L. Niles, J. Ong, N. Pau, J. Quispe, C. Salazar, J. Smith, M. Whitbeck, J. White, M. Williams, and R. Wolfe. We also thank Janet Leese for countless hours spent digitizing training polygons in the lab and Erika Lentz for reclassifying coastal response raster layers to reflect IPCC ranges. Comments from E. Legaard, K. Legaard, D. Rosco, N. Hanson, members of the PIs respective lab groups, and the larger SHARP team have substantially improved the work described here.

ii Contents

I. Executive Summary …………………………………………………………………………………………………………………. iv II. Manuscript 1: Measuring elevation in northeastern USA tidal marshes …………………………………… 1 III. Manuscript 2: Fine-scale mapping of coastal plant communities in the northeastern USA ……… 14 IV. Manuscript 3: Using systematic conservation planning to identify a focal species for tidal marsh bird conservation in the northeastern United States ………………………………………………………………. 35 V. Manuscript 4: Characterizing uncertainty to inform tidal marsh conservation in response to sea- level rise ………….……………………………………………………………………………………………………………………… 60 VI. Hurricane Sandy coastal resiliency program: restoration project summaries ……...….……………… 69

Thin-layer sediment additions and vegetation planting at Sachuest Point, RI.

iii I. Executive summary

Hurricane Sandy has fundamentally altered the way in which society thinks about coastal resilience in the mid-Atlantic and New England states. Considerable funds are being invested in restoring coastal resources and the need to systematically plan how we invest is increasingly apparent. A thorough understanding of the effectiveness of Hurricane Sandy restoration projects on tidal marshes and the wildlife they support is thus needed to direct conservation actions where they will have the greatest benefits toward increasing the resilience of green infrastructure.

Many species of conservation concern use tidal marshes and face additional threats from sea-level rise. The Saltmarsh Habitat and Avian Research Program (SHARP) has assembled a large collection of historical and contemporary measures of plant and bird communities in tidal marshes from Virginia to Maine. In the current project, we used this platform to begin to assess the efficacy of restoration activities and to provide planning guidance to enhance the future resiliency of natural coastal assets. Our primary focus was to collected elevation, vegetation, and bird data to quantify baseline conditions at National Fish and Wildlife Foundation (NFWF) funded Hurricane Sandy resiliency sites throughout the Northeast and mid-Atlantic states and at paired control sites. These data have been combined with similar data collected at Department of the Interior-funded resiliency sites across the same region into a common database, and will form the basis for long-term investigations into the efficacy of coastal restoration and management work.

In addition to baseline data collection at resiliency sites, our work has had three specific research goals. First, we investigated whether elevation data collected in the field via real-time kinetic (RTK) GPS methods can be accurately replicated using remote sensing data products. Collection of RTK elevation data is both time consuming and expensive, and thus difficult to obtain for large areas with high spatial resolution. Remote sensing data, by contrast, are available across vast areas and are often publically available for free. We tested a variety of remote sensing data at both local and regional scales and found that the 1/9 arc-second (~1 m resolution) data layer from the National Elevation Dataset (NED) produced high quality predictions of the RTK measurements. These predictions were especially good for high elevation marsh, a habitat of particular interest to land managers. Other resolution NED data, and data collected used an unmanned aircraft system, were far less satisfactory and were poor substitutes for field-collected data in our tests.

Second, we compared several remote sensing techniques to develop a tool that accurately maps high- and low-marsh zonation for use in management and conservation planning. We found that random forests outperformed other classifier tools when applied to the most recent National Agricultural Imagery Program (NAIP) imagery, NAIP derivatives, and elevation data between coastal Maine and Virginia, USA. We then used these methods to classify plant zonation within a 500-m buffer around coastal marsh delineated in the National Wetland Inventory. We found mean classification accuracies of 94% for high elevation marsh, 76% for low marsh zones, and 90% overall map accuracy. The detailed output is a 3-m resolution continuous map of tidal marsh vegetation communities and cover classes that can be used in habitat modeling of marsh-obligate species or to monitor changes in marsh plant communities over time.

Third, we used the systematic conservation planning software Marxan to identify priority areas for the conservation of five species that rely on tidal marsh for breeding: clapper rail Rallus crepitans, eastern

iv willet Tringa semipalmata semipalmata, Acadian Nelson’s sparrow Ammodramus nelsonii subvirgatus, saltmarsh sparrow Ammodramus caudacutus, and seaside sparrow Ammodramus maritimus. We compared the spatial prioritization of sites and cost-effectiveness of alternative protection scenarios that considered individual species, groups of species, and all species simultaneously. Scenarios that prioritized areas for conservation based on different single species targets were poorly correlated with each other. Scenarios based on saltmarsh sparrow conservation were most strongly related (rs = 0.76) to site prioritizations that consider all five tidal marsh specialists simultaneously. These results suggest that although no species is a good surrogate for another single species, saltmarsh sparrow may be a viable focal species for conservation planning to protect tidal marsh birds as a group. When comparing multi- species combinations to prioritizations based on saltmarsh sparrow alone, the estimated costs, area of land protection, and number of individuals of each species on protected land were similar, suggesting that it is the best option considered. We were also able to identify areas where conservation is likely to have little or no effect for the target species. Eliminating these sites from consideration for the conservation of these species would reduce the risk of misdirecting limited conservation funds.

Because sea-level rise is likely to alter where suitable habitat lies, long-term conservation planning should account for future conditions. With that in mind, we assessed the likelihood that each marsh patch considered in our initial conservation planning exercise would persist by 2050. To accommodate uncertainty in future conditions, we used both results from sea-level projections that assume static inundation and those that assume dynamic adaptation by marshes in response to sea-level under alternative carbon emissions scenarios. The results suggest that, at most, ~15% of current tidal marsh is secure (>66% chance of persisting), but that much of the remainder has potential to persist via dynamic adaptation (e.g., marsh building). Substantial declines of all tidal marsh specialist breeding birds are predicted, commensurate with the predicted losses. In light of these projections, we identified marsh patches with the greatest chance of persisting in the future and recommend a conservation planning strategy that accounts for both current conservation value and the probability of long-term habitat persistence.

This report consists of four manuscripts, written for submission to the peer review literature, and dealing with (1) the use of remote sensing to estimate marsh elevation, (2) description of marsh vegetation throughout the northeastern USA, (3) current conservation planning for focal marsh birds, and (4) future projections for marsh area and focal marsh birds. We also include baseline data summaries for seventeen coastal resilience projects distributed throughout the northeastern USA.

v II. Measuring elevation in northeastern USA tidal marshes

Maureen D. Correll, Chris S. Elphick, Wouter Hantson, Brittany B. Cline, Elizabeth L. Tymkiw, W. Gregory Shriver, and Brian J. Olsen

Sea levels are rising worldwide (Church et al. 2013, Chen et al. 2017), raising questions about the future of coastal ecosystems and built infrastructure (Hinkel et al. 2014, Garner et al. 2017). Rates of sea-level rise vary regionally due to differences in such things as prevailing winds and shifting ocean currents, and their impact is further affected by factors such as isostatic rebound following past glaciation, and groundwater and oil extraction, which affect the behavior of the land surface (Rhamstorf 2017). Our ability to predict how the world’s coastlines will change is thus limited, yet vitally important given the concentration of human populations in coastal areas (Kummu et al. 2016) and the wide range of services that coastal ecosystems provide (Barbier et al. 2011). The rate of sea-level rise is particularly high along the Atlantic seaboard of North America, especially in mid-Atlantic and southern New England states, where rates are substantially higher than the global average (Sallenger et al. 2012). Portions of this region also contain extensive salt marshes and the consequences of higher sea levels for these marshes has received considerable attention. Models suggest potential for marshes to build elevation via a combination of increased plant production and greater rates of sediment deposition, and some evidence suggests that this might be happening (Kirwan et al. 2016). Sediment flow, however, is declining in many river systems (Syvitski et al. 2005; Weston 2014) and the accretion rates needed to maintain marshes appear to be insufficient in many areas (Crosby et al. 2016). Accretion rates are higher for lower-elevation marshes, however, and are more likely to be sufficient to match current rates of sea-level rise in these areas (Kirwan et al. 2016). Consequently, even where marshes can persist, their elevation relative to sea-level may decline, shifting the nature of the marsh towards lower-elevations. Evidence for a drop in marsh elevation relative to sea-level comes from studies that show consistent shifts towards plant species that are associated with greater tidal flooding (Field et al. 2016; Raposa et al. 2017). Losses in the overall extent of marshes might also be mitigated, at least somewhat, by migration of marsh vegetation into adjacent upland areas (Kirwan et al. 2016). Current evidence, however, suggests that these changes are not happening as quickly as changes within current marshes (Field et al. 2016). Even if some form of marsh can persist over the long term, higher elevation marsh and any species or ecosystem services that it supports are placed at risk due to the lag time needed for ecosystems to transition (e.g., Field et al. 2017a). This problem is exacerbated in areas where patterns of human land use and attitudes towards land protection impose additional constraints on the ability of marshes to migrate (Field et al. 2017b). Given the importance of marsh elevation relative to sea-level to discussions about the future of coastal marshes, it is important that we have good baseline data on current elevations. Locally, this information can be obtained through the use of surface elevation tables (SETs; Lynch et al. 2015) installed in the marsh. At a local scale, use of Geographic Positioning Systems (GPSs) in the form of Real Time Kinematic positioning (RTK) allows for the manual measurement of elevation using a handheld unit and antenna. Increasingly, local scale information also is being collected using unmanned aircraft systems (UASs), which can provide contiguous, high resolution data, over moderately large patches of marsh (Drummond et al. 2015, Ma et al. 2016). Measurements over larger regional scales are harder to come by without making inferences from remote sensing data, either based on assumptions about relative elevation derived from vegetation data (Hladik et al. 2013; Correll et al. submitted), or via direct measurements. Light Detection and Ranging data (Lidar) collected via aircraft offers a high-resolution (albeit high-cost) and large-scale elevation point cloud which when processed can predict elevation at

1 both terrain and vegetation levels (e.g. Hladik et al. 2013, Buffington et al. 2016). The United States Geological Survey National Elevation Dataset (NED; USGS 2015) also provides a set of elevation data products at several different resolutions (2, 1, 1/3, 1/9 arc-seconds and 1 m) derived from source data of topographic maps, manual profiling through photogrammetry, and compiled Lidar. Ideally we would understand how detailed measurements made at local scales relate to those obtained from these larger- scale data sets that can be applied over large regions. With this information, we would be better placed to develop a baseline understanding of current elevation patterns and to track changes over time as sea- levels rise. With these needs in mind, we set out to compare measurements of elevation made with different methods and at different scales. First, we conducted a local-scale study in an extensively studied coastal wetland near Scarborough, Maine, USA. We compared measurements taken using RTK units with elevation data created through UAS imagery, and at three different resolutions (1, 1/3, and 1/9 arc-second layers) provided by the NED. Second, we scaled analysis up to compare RTK data collected at 12,864 points around 655 survey locations between the Canada-Maine border and the Chesapeake Bay, Virginia, to NED data collected at the same three resolutions previously mentioned. Third, we describe the elevation profile of marshes throughout the Northeast and mid-Atlantic USA to provide a general overview of current baseline conditions that can be used for future comparisons. Finally, we illustrate the relationship between elevation estimates obtained during this study and marsh habitat types obtained during a parallel study designed to describe vegetation classes across the same region.

Methods Data collection To explore the efficacy of different elevation measurement techniques for tidal marsh, we measured ~10 ha plots (n = 4, Figure 1, 2) developed by Ruskin et al. (2017) in and around Scarborough, ME as part of a study monitoring demographic rates of tidal marsh birds (Figure 1). Within each plot delineation, we overlaid a 20 x 20 m grid using ArcGIS 10.3 (ESRI 2016) before navigating to each plot in the field. Once in the marsh, a trained technician measured all grid points that fell within plot boundaries using a Trimble R10 RTK sensor system. In addition to RTK data collection we flew a DJI Phantom 3 Advanced UAS over each marsh plot in August 2016 to collect imagery for use with photogrammetry software. We programmed the UAS using the Pix4D Capture mission planning application to fly a grid over each plot at 100 m altitude. The programmed grid provided images collected with ~80% overlap using a MAPIR 16-megapixel red-green- blue aerial mapping camera affixed to the base of the UAS with the camera pointing directly downward. All photographs were collected at wind speeds of less than 15 mph and within 2 hours of local low tide. We placed 8-9 black and white 20 x 28 cm ground control points on each plot during flights to georeference digital mosaics during processing. We recorded the location of each ground control point using a Trimble GEO7X unit, accurate to the decimeter. All imagery was processed using Pix4D photogrammetry software (Pix4D SA, Switzerland) to create a digital orthomosaic and Digital Surface Model (DSM) for each marsh plot. For our regional effort we selected marsh survey sites based on a two-stage cluster sampling scheme (Wiest et al. 2016) to comprehensively survey tidal marshes for bird populations between Maine and Virginia. We navigated to a subset of bird survey locations visited by Wiest et al. (n = 655, Figure 1) between April and September 2015 to sample elevation in tidal marshes within the same study extent. At each survey location we overlaid a 20 x 20 m point grid on the marsh patch, centering the grid around the survey location as closely as possible given the orientation of the marsh patch. Technicians attempted to measure points resulting in a square grid with 5 points on each side, covering a total of 1 ha around each survey location. If a grid point fell in non-marsh habitat, we measured another point

2 from the surrounding grid instead. When a marsh patch was too small or narrow for a this size grid to fit we used either a 5 x 5 m or 1 x 1 m point grid instead (n = 3004 and 530 elevation measurements, respectively), choosing the resolution that provided the largest grid that could be fit within the marsh patch. In total, we measured elevation at 12,864 points associated with the 655 survey locations in tidal marsh between Maine and Virginia. Once we measured elevation at all points in the field, we overlaid the point for each marsh elevation reading from both our small-scale and regional study on the 1, 1/3, and 1/9 arc-second layers (~30 m, ~10 m, and ~1 m resolution, respectively) from the NED (USGS 2015). We then extracted values from these layers for each sampling point. For our small-scale study we also extracted values from the UAS-derived DSMs for each RTK sampling point. To describe elevation data by community type for our regional and fine-scale elevation data, we used a vegetation layer produced by the Saltmarsh Habitat and Avian Research Program (Correll et al. submitted, available at https://nalcc.databasin.org/galleries/46d6e771dd6f4fdb8aa5eb46efffffa7) describing six marsh cover types (high marsh; low marsh; Phragmites australis, hereafter Phragmites; mudflat; pool/panne; terrestrial border) and two bordering cover types (open water; upland) in tidal marshes from Maine to Virginia. We extracted the cover type for each RTK point using ArcGIS 10.3 (ESRI 2016).

Statistical analysis We analyzed our plot-level data using linear and quantile regression using base R and the lqmm package (Geraci, 2014). Quantile regression allows users to specify the percentage of data points included beneath the regression line; for reference, an ordinary linear regression maintains a tau = 0.5. We varied tau from 0.1 to 0.9 (Table 1) and compared these models to linear regressions to identify the best regression structure to model the data we collected. To account for variation among plots, we included a plot “site” as a variable in all models. To explore the predictive power of our independent variables and marsh cover type, we compared models including additive and interactive terms for marsh cover types with over 10 data points (high marsh, low marsh, pool/panne). We assessed variation explained by our linear models using adjusted R2. Due to the difficulties in assessing R2 for quantile regression models we compared models using Akaike’s Information Criterion (AIC, Burnham and Anderson 2004) instead. We analyzed regional data in a linear mixed-modeling framework (LMM) using the lme4 package (Bates et al 2015) in Program R (R Core Team 2016). We modeled the accuracy with which NED data can be used to predict RTK measurements at each spatial scale. We also modelled the relationship between RTK measurement and latitude using identical modelling methods. Because RTK grids comprised of individual data points were always grouped around pre-established survey locations, we included a random variable for “site” in all models. We ran models with and without zero values after examination of our data because there is an obvious concentration of zero values in the NED dataset. False-zero values are commonly produced in remotely sensed datasets occurring near water or extremely wet soil. To explore the utility of measuring additional RTK points to further strengthen the relationships measured in our regional analysis, we subsetted our data by increments of 10% by (1) removing all points taken around randomly-selected survey locations and (2) randomly selecting and removing values from individual points. To explore the predictive power of our independent variables by marsh cover type, we compared models including additive and interactive terms for marsh cover type using a dataset of all points falling within the marsh layer. We assessed variation explained and directly compared models using marginal R2 (Nakagawa and Schielzeth 2013) and AIC.

Results We measured RTK elevation at a total of 1230 points across the four local plots (Figures 1, 2) in Scarborough Marsh. Measurements ranged from -0.11 m to 2.0 m ASL ( = 1.31 m). Elevation values

3 extracted from the DSM produced from UAS imagery (Figure 2) ranged from -7.2 m to 3.3 m ( = 0.74 m). All points surveyed fell within the marsh cover type layer. We collected 979 overlapping photos for photogrammetry analysis at these four plots (Figure 2). The processed DSM rasters display elevation data at a 2 cm resolution. We measured RTK elevation at a total of 12,864 points centered around 655 survey locations between Maine in Virginia in 2015. Values measured using RTK ranged from -1.14 m to 2.76 m above sea level (ASL,  = 0.85). Values extracted at these locations ranged from -0.67 m to 5.63 m ASL ( = 1.04 m) for the 1 arc-second NED, -0.67 m to 5.63 m ASL ( = 1.04 m) for the 1/3 arc-second NED, and -0.87 m to 3.18 m ( = 0.84 m) for the 1/9 arc-second NED. Of the RTK points surveyed, 11,328 points fell within the marsh cover type layer. For our plot-level data from Maine, we found that RTK elevation measurements were only weakly predicted using the 1 arc-second ( = 0.14, R2 = 0.27) and 1/3 ( = 0.14, R2 = 0.27) arc-second NED. The 1/9 arc-second NED had much better predictive power ( = 0.63, R2 = 0.64, Figure 3). RTK elevation measurements were weakly predicted by DSM values, and fit improved when excluding zero values ( = 0.14, R2 = 0.28), although most of this variation was explained by the site variable in our models. Quantile regression models performed better than linear models, with the model using tau = 0.7 ( = 0.07) performing best (Table 1, Figure 4). In our model comparison including marsh cover type, the best performing model included the additive and interactive term for marsh cover. Generally speaking, DSM values were higher than corresponding RTK measurements, with the discrepancy greater at higher elevations. 2 In our regional analysis, we found that the 1 arc-second ( = 0.49, Rm = 0.24, Figure 5a) and 1/3 2 arc-second ( = 0.20, Rm = 0.11, Figure 5b) NED produced only weak predictions of RTK elevation 2 measurements. This relationship was much improved with the 1/9 arc-second data layer ( = 0.83, Rm = 2 0.85, Fig 5c), especially when excluding zero values ( = 0.87, Rm = 0.90, Fig 5d). We found that this relationship was insensitive to data sub-setting and we obtained similar estimates and explained similar 2 variance when using as little as 30% of the full dataset ( = 0.81, Rm = 0.85). Including both additive and interactive terms for marsh cover produced a model that out-performed all others (AIC = 29.94 compared to the next best model). We found that most zero-value NED measurements belong to the “Upland” cover type (Figure 6a), and that regression slopes differed among cover types (Figure 6b). RTK measurements were generally better predicted by the NED for high marsh and terrestrial border, with the worst predictions for mudflat. Finally, we found a strong relationship between RTK measurement 2 and latitude ( =0.19, Rm = 0.42, Figure 7).

Discussion In the northeastern United States, tidal marshes dominate much of the coastal landscape. Understanding and measuring change in this ecosystem as sea level rise progresses is critical to the conservation of the system, and the biodiversity and services it supports. In our comparison of elevation measurements in tidal marshes across different methods and scales, we found that elevation of tidal marshes was well predicted by the 1/9 arc-second layer of the NED both locally and regionally, however other data sources (UAS, 1 arc-second NED, 1/3 arc-second NED) provided limited predictive power (Figure 5). This key finding can help researchers explore local and regional change in tidal marshes without the need for manual mapping and sampling using RTK units, which can be costly in the sometimes scarce resources afforded conservation professionals and researchers. We also found that elevation measurements from the 1/9 arc-second NED diverged least from RTK measurements in high marsh at both the local and regional scale, and that the relationship was generally stronger for higher elevation cover types. This result, combined with the improved performance of models when excluding zero-values, suggest that the NED predicts elevation more

4 accurately for higher elevations marsh cover types than those closer to the marine interface. Those wishing to focus study on lower elevations, therefore, may benefit more from investing in specialized equipment to manually measure marsh elevation instead of relying on the NED, although predictions were generally good for most habitat types at a regional scale. It is important to note that there are known limitations to the NED which should be considered before using this dataset in any analysis. We found that the 1/9 arc-second NED under-predicts RTK at both the local and regional scales (regression slope = ~0.85 for both scales). While it is a good predictor of relative elevation, it should be used cautiously to predict absolute elevation above MSL. Further, the elevation data that comprise the NED were collected on different dates (2010-2016 for the 1/9 arc- second layer) and there are no current plans to refresh the 1/9 arc-second layer in the near future. Instead resources are currently allocated for further development of the spatial extent and temporal resolution of the 1 m layer, which currently covers a much smaller area than the 1, 1/3, or 1/9 arc- second layers. Those wishing to measure change over time annually or across 2-3 years may still want to invest in RTK data collection for best results. The elevation profile we produced through shows a strong latitudinal trend in tidal marsh elevation, decreasing from north to south (Figure 7). Tidal marshes occur at higher elevations in the north and at steadily lower elevations as one moves south along the coastline. This pattern may be caused by the similar latitudinal gradient in tidal amplitude along the coast, which is driven overwhelmingly by local bathymetry and geography. The Bay of Fundy in eastern Maine regularly exhibits tides nearing 10 m, where the Chesapeake Bay experiences tides of only 1-2 m. Because zonation in plant communities is directly connected to the timing and duration of inundation by incoming and outgoing tides (Bertness 1991), a larger amplitude tide could result in the once-weekly to once-monthly inundation window being located in a higher elevation relative to MSL. While the primary goal of this work was to compare elevation measurements from the NED to those collected using RTK, we also explored the use of UASs in measuring elevation in tidal marshes. We found that the use of raw, UAS-derived DSMs do not work well to predict ground-level elevation in tidal marshes. We believe this is because of the inherent differences in DSM and DEM datasets; DSMs depict the vegetation surface, while DEMs represent ground-level elevation without consideration of vegetation. As a result, even our top model shows that our UAS-derived DSM predicts elevations equal to or higher than the ground elevation estimated using RTK methods. The recent development of Digital Terrain Models (DTMs) using photogrammetry software such as Pix4D is promising and has the potential to address the shortcomings we encountered with UAS-derived DSMs. The ease with which a localized DSM is produced with a UAS is promising, and we expect more accurate tidal marsh elevation profiles to emerge as the technology and software continue to improve.

Conclusions The effective measurement of ground elevation is an important step in the effective study and monitoring of tidal marshes, yet the methods and technology required to produce this type of data are costly and time-consuming to implement. We found that in tidal marshes of the northeastern USA, elevation measurements available from the publicly accessible 1/9 arc-second NED were effective in predicting RTK elevation measurements, explaining 90% of the variation in our regional dataset. Prediction accuracy was highest for high marsh and terrestrial border, but was good (>80%) for most cover types. This finding is key for scientists and land managers wishing to measure and monitor marsh elevation but who are limited by cost and labor availability. Use of the 1/9 arc-send NED can, with minimal preprocessing, be a suitable stand-in for RTK methods for estimating ground elevation of coastal marsh, particularly when measuring elevation in high marsh, a cover type of particular conservation interest due to the ecosystem services it provides for wildlife and humans alike.

5 Literature Cited Barbier EB, Hacker SD, Kennedy C, Koch EW, Stier AC, Silliman BR (2011) The value of estuarine and coastal ecosystem services. Ecological Monographs 81:169–193. Bates D, Maechler M, Bolker B, Walker S (2015) Fitting linear mixed-effect models using lme4. Journal of Statistical Software 67:1–48. Bertness MD (1991) Zonation of Spartina patens and Spartina alterniflora in New England salt marsh. Ecology 72:138–148. Bertness MD, Ellison AM (1987) Determinants of pattern in a New England salt marsh plant community. Ecological Monographs 57:129–147. Buffington K J, Dugger BD, Thorne KM, Takekawa JY (2016) Statistical correction of lidar-derived digital elevation models with multispectral airborne imagery in tidal marshes. Remote Sensing of Environment 186:616–625. Burnham KP, Anderson RP (2004) Multimodel inference: Understanding AIC and BIC in model sSelection. Sociological Methods and Research 33:261–304. Chambers RM, Meyerson LA, Saltonstall K (1999) Expansion of Phragmites australis into tidal wetlands of North America. Aquatic Botany 64:261–273. Chen X, Zhang X, Church JA, Watson CS, King M, Monselesan D, Legresy B, Harig C (2017) The increasing rate of global mean sea-level rise during 1993–2014. Nature Climate Change 7:492–495. Church JA, Clark PU, Cazenave A, Gregory JM, Jevrejeva S, Levermann A, Merrifield MA, Milne GA, Nerem RS, Nunn PD, Payne AJ, Pfeffer WT, Stammer D, Unnikrishnan AS (2013) Sea level change. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM, eds.]. Cambridge University Press, Cambridge. Chu-Agor ML, Muñoz-Carpena R, Kiker G, Emanuelsson A, Linkov I (2011) Exploring vulnerability of coastal habitats to sea level rise through global sensitivity and uncertainty analyses. Environmental Modelling and Software 26:593–604. DJI (2016) The Phantom 3 Advanced quadcopter Unmanned Aircraft System. Available at: https://www.dji.com/phantom-3-adv Crosby SC, Sax DF, Palmer ME, Booth, HS, Deegan LA, Bertness MD, Leslie HM (2016) Salt marsh persistence is threatened by predicted sea-level rise. Estuarine, Coastal and Shelf Science 181:93– 99. Day JW, Christian RR, Boesch DM, Yáñez-Arancibia A, Morris J, Twilley RR, Naylor L, Schaffner L, Stevenson C (2008) Consequences of climate change on the ecogeomorphology of coastal wetlands. Estuaries and Coasts 31:477–491. Donnelly JP, Bertness MD (2001) Rapid shoreward encroachment of salt marsh cordgrass in response to accelerated sea-level rise. Proceedings of the National Academy of Sciences, USA 98:14218–14223. Drummond CD, Harley MD, Turner IL, A Matheen, A Nashwan, Glamore WC (2015) UAV applications to coastal engineering. Australasian Coasts & Ports Conference 2015: 22nd Australasian Coastal and Ocean Engineering Conference and the 15th Australasian Port and Harbour Conference. Engineers Australia and IPENZ 2015: 267-272. Emery NC, Ewanchuk PJ, Bertness MD (2001) Competition and salt-marsh plant zonation: stress tolerators may be dominant competitors. Ecology 82:2471–2485. ESRI (2016) ArcGIS Desktop: Release 10.3. Redlands, CA: Environmental Systems Research Institute. Ewanchuk PJ, Bertness MD (2004) Structure and organization of a northern New England salt marsh plant community. Journal of Ecology 92:72–85. Field, CR, Gjerdrum C, Elphick CS (2016) Forest resistance to sea-level rise prevents landward migration of tidal marsh. Biological Conservation 201:363–369.

6 Field CR, Bayard T, Gjerdrum C, Hill JM, Meiman S, Elphick CS (2017a) High-resolution tide projections reveal extinction threshold in response to sea-level rise. Global Change Biology 35:2058–2070. Field CR, Dayer AA, Elphick CS (2017b) Landowner behavior can determine the success of conservation strategies for ecosystem migration under sea-level rise. Proceedings of the National Academy of Sciences, USA 114:9134–9139. Garner AJ, Mann ME, Emanuel KA, Kopp RE, Lin N, Alley RB, Horton BP, DeConto R, Donnelly JP, Pollard D (2017) Impact of climate change on New York City’s coastal flood hazard: Increasing flood heights from the preindustrial to 2300 CE. Proceedings of the National Academy of Sciences, USA 114:11861–11866. Geraci M (2014) Linear quantile mixed models: the lqmm package for laplace quantile regression. Journal of Statistical Software 57:1–29. Hinkel J, Lincke D, Vafeidis AT, Perrette M, Nicholls RJ, Tol RSJ, Marzeion B, Fettweis X, Ionescu C, Levermann A (2014) Coastal flood damage and adaptation cost under 21st century sea-level rise. In: Proceedings of the National Academy of Sciences, USA 111: 3292–3297. Hladik C, Schalles J, Alber M (2013) Salt marsh elevation and habitat mapping using hyperspectral and LIDAR data. Remote Sensing of Environment 139:318–330. IPCC (2013) IPCC Fifth Assessment Report (AR5). IPCC. Kirwan ML, Guntenspergen GR (2010) Influence of tidal range on the stability of coastal marshland. Journal of Geophysical Research: Earth Surface 115:F02009. Kirwan ML, Temmerman S, Skeehan EE, Guntenspergen GR, Fagherazzi S (2016) Overestimation of marsh vulnerability to sea level rise. Nature Climate Change 6:253–260. Kummu M, Moel H de, Salvucci G, Viviroli D, Ward PJ, Varis O (2016) Over the hills and further away from coast: global geospatial patterns of human and environment over the 20th–21st centuries. Environmental Research Letters 11:034010. Lynch JC, Hensel P, Cahoon DR (2015) The surface elevation table and marker horizon technique: a protocol for monitoring wetland elevation dynamics. Natural Resource Report NPS/NCBN/NRR - 2015/1078, National Park Service. Ma Y, Zhang J, Zhang J (2016) Analysis of Unmanned Aerial Vehicle (UAV) hyperspectral remote sensing monitoring key technology in coastal wetland. Proceedings volume 9796, Selected Papers of the Photoelectronic Technology Committee Conferences held November 2015, 97962S. doi: 10.1117/12.2229746. Miller W, Egler F (1950) Vegetation of the Wequetequock-Pawcatuck tidal-marshes, Connecticut. Ecological Monographs 20:143–172. Nakagawa S, Schielzeth H (2013) A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution 4:133–142. Pennings S, Callaway R (1992) Salt marsh plant zonation: the relative importance of competition and physical factors. Ecology 73:681–690. Philipp KR, Field RT (2005) Phragmites australis expansion in Delaware Bay salt marshes. Ecological Engineering 25:275–291. R Core Team (2016) R: A language and environment for statistical computing. Rhamstorf, S (2017) Rising hazard of storm-surge flooding. Proceedings of the National Academy of Sciences, USA 114:11806–11808. Ruskin KJ, Etterson MA, Hodgman TP, Borowske AC, Cohen JB, Elphick CS, Field CR, Longenecker RA, King E, Kocek AR, Kovach AI, O'Brien KM, Pau N, Shriver WG, Walsh J, Olsen BJ (2017). Demographic analysis demonstrates systematic but independent spatial variation in abiotic and biotic stressors across 59 percent of a global species range. Auk: Ornithological Advances 134:903–916. Sallenger AHS, Jr, Doran KS, Howd PA (2012) Hotspot of accelerated sea-level rise on the Atlantic coast of North America. Nature Climate Change 2:884–888.

7 Syvitski JPM, Vörösmarty CJ, Kettner AJ, Green P (2005) Impact of humans on the flux of terrestrial sediment to the global coastal ocean. Science 308:376–380. USGS (2015) National Elevation Dataset (NED). Available at: https://nationalmap.gov/elevation.html. Accessed January 2016. Weston NB (2014) Declining sediments and rising seas: an unfortunate convergence for tidal wetlands. Estuaries and Coasts 37:1–23. Wiest WA, Correll MD, Olsen BJ, Elphick CS, Hodgman TP, Curson DR, Shriver WG (2016) Population estimates for tidal marsh birds of high conservation concern in the northeastern USA from a design-based survey. Condor: Ornithological Applications 118:274–288. Wilson CA, Hughes ZJ, FitzGerald DM, Hopkinson CS, Valentine V, Kolker AS (2014) Saltmarsh pool and tidal creek morphodynamics: Dynamic equilibrium of northern latitude saltmarshes? Geomorphology 213:99–115.

Table 1. Model structure, AIC values, and model weights for quantile and linear regressions between elevation measurements in tidal marshes in Scarborough, Maine, USA, collected using real-time kinematic (RTK) and digital surface models (DSMs) produced through photogrammetric analysis of images collected via Unmanned Aircraft System (UAS). model regression type tau AIC model weight rtk ~ dsm + site quantile 0.7 0.00 1.00 rtk ~ dsm + site quantile 0.6 12.66 0.00 rtk ~ dsm + site quantile 0.8 44.04 0.00 rtk ~ dsm + site quantile 0.5 68.03 0.00 rtk ~ dsm + site quantile 0.4 164.46 0.00 rtk ~ dsm + site quantile 0.9 206.77 0.00 rtk ~ dsm + site linear . 237.96 0.00 rtk ~ dsm + site quantile 0.3 316.29 0.00 rtk ~ site (null) linear . 426.09 0.00 rtk ~ dsm + site quantile 0.2 537.81 0.00 rtk ~ dsm + site quantile 0.1 883.36 0.00

8 Figure 1. Study area and data collection locations for elevation measurement in tidal marshes in the northeastern USA using real-time kinematic (RTK, purple points) and unmanned aircraft systems (UAS, yellow polygons, inset).

9 Figure 2. Maps and digital surface models (DSMs) of four marsh study plots and elevation measurement points (green dots) using real-time kinematic (RTK) and unmanned aircraft system (UAS) methods in Scarborough Marsh, Maine.

10 Figure 3. Results of linear regression model comparing tidal marsh ground elevation estimated using real-time kinematic (RTK) measurements and the National Elevation Dataset (NED) 1/9 arc-second layer. Data come from four plots in Scarborough, Maine, USA.

Figure 4. Results of quantile regression (tau = 0.7) visualized through A) scatterplot and B) estimates and confidence intervals of the regression slopes comparing tidal marsh ground elevation estimated using real-time kinematic (RTK) measurements and Digital Surface Models (DSMs) created via an unmanned aircraft system (UAS), stratified by cover type. Data come from four plots in Scarborough, Maine, USA.

11 Figure 5. Results of mixed-effect linear regressions comparing tidal marsh ground elevation estimated using real-time kinematic (RTK) measurements and the National Elevation Dataset (NED) from A) 1 arc- 2 2 second data layer (Rm = 0.24), B) 1/3 arc-second data layer (Rm = 0.11), C) 1/9 arc-second data layer 2 2 (Rm = 0.85), and D) 1/9 arc-second data layer excluding zero values (Rm = 0.90). Data come from sites throughout the northeastern USA, from Virginia to Maine.

12 Figure 6. Results of mixed-effects linear regression visualized through A) scatterplot and B) estimates and confidence intervals of regression slopes comparing tidal marsh ground elevation estimated using real-time kinematic (RTK) measurements and the National Elevation Dataset (NED) stratified by marsh community type. Data come from sites throughout the northeastern USA, from Virginia to Maine.

Figure 7. Results of mixed-effect linear regressions comparing tidal marsh ground elevation estimated using real-time kinematic (RTK) measurements across latitudes between coastal Maine and Virginia, USA.

13 III. Fine-scale mapping of coastal plant communities in the northeastern USA (submitted manuscript)

Maureen D. Correll, Wouter Hantson, Thomas P. Hodgman, Brittany B. Cline, Chris S. Elphick, W. Gregory Shriver, Elizabeth L. Tymkiw, and Brian J. Olsen

Coastal marshes are among the world’s most productive ecosystems and provide significant services to humans across the globe. These marshes serve as a gateway between land and sea for humans and wildlife alike, act as a buffer against coastal storms, and provide critical nutrients to marine food webs (Barbier et al. 2011). Tidal marshes also support and protect biodiversity by providing habitat to marine and estuarine fish, crustacean populations, and migratory birds (Boesch and Turner 1984; Master 1992; Brown et al. 2002). Within the world’s tidal marsh systems, those located along the Atlantic coast of the United States support the highest number of terrestrial vertebrate specialists described worldwide (Greenberg et al. 2006). This suite of species includes herpetofauna and mammals, but the majority of described vertebrate specialists are birds. Several species are limited completely to these marshes during the breeding season, several of which are in decline (Correll et al. 2017), with extinction predicted for the saltmarsh sparrow within 50 years (Correll et al. 2017a; Correll et al. 2017b; Field et al. 2017a; Field et al. in press). These declining species nest predominantly within the high-marsh zone, one of several vegetation communities found within coastal marshes. High marsh differs from other marsh areas in elevation, salinity, and frequency of inundation (Bertness and Ellison 1987; Pennings and Callaway 1992; Ewanchuk and Bertness 2004) and is characterized by flooding during spring tides linked to the lunar cycle. In the northeastern United States, the plant species Spartina patens, short-form S. alterniflora, Distichlis spicata, and Juncus gerardii characterize high-marsh zones, which also include Salicornia spp., Glaux maritima, and Solidago sempervirens (Nixon and Oviatt 1973; Bertness 1991; Emery et al. 2001, Ewanchuk and Bertness 2004). Conversely, low marsh is characterized by daily tidal flooding and is a near monoculture of tall form S. alterniflora. The surrounding terrestrial border experiences infrequent inundation by salt water during extreme tides and storms, and is characterized by a more diverse flora that is often dominated by Iva frutescens and Typha spp. (Miller and Egler 1950; Ewanchuk and Bertness 2004). Introduced Phragmites australis (hereafter Phragmites) also occurs within this ecosystem, especially around the borders of disturbed marshes (Chambers et al. 1999; Philipp and Field 2005). These plant community zones can be quickly altered by both natural and anthropogenic stressors such as sea-level rise, nutrient run-off from adjacent uplands, and the spread of introduced species (Day et al. 2008). The significant increase in sea level during recent decades poses one of the largest threats to these marsh ecosystems. As sea levels encroach on the marshes’ seaward side and upland marsh migration is limited by human-developed coastal infrastructure (Field et al. 2017b) and upland habitats (Field et al. 2016), a “pinching effect” can occur, resulting in marsh loss. Coastal marshes can combat rising sea levels through vertical growth, or accretion (Kirwan et al. 2016), but when the rate of sea-level rise exceeds the rate of accretion, marsh area will decline (Crosby et al. 2016). Rising sea levels can also drive invasion of high marsh areas with flood-tolerant low marsh species (Donnelly and Bertness 2001; Field et al. 2016), causing transition from high to low marsh (Kirwan et al. 2016). This pattern, however, is not ubiquitous to all marshes (Kirwan and Guntenspergen, 2010; Wilson et al. 2014). In addition to sea-level rise, extreme storm events that flood the coastline have been shown to permanently alter marsh structure within days (Day et al. 2008) and can have a lasting effect on plant community structure and saltmarsh degradation.

14 Marsh degradation and rapid change is likely to continue into the future due to the paired effects of climate change and human development. Sea levels are expected to rise substantially between 2013 and 2100 (IPCC 2014), and continuing storm events affecting coastal regions are also predicted. The future distribution of high- and low-marsh habitat therefore remains uncertain (Chu-Agor et al. 2011; Kirwan et al. 2016). It is essential to develop tools to identify coastal marsh plant communities, particularly high marsh, on a biologically relevant timescale to protect existing ecosystem services and to inform the adaptive management of coastal wetlands as habitat for high-marsh specialist species. The physical and biological characteristics that differentiate high marsh from low marsh and other marsh plant communities are potentially detectible using remotely-sensed multispectral and hyperspectral imagery. Both types of imagery can record wavelengths of light outside of the visible range for humans, with hyperspectral imagery recording reflectance values in much finer detail and precision (hundreds of individual bands recorded) than multispectral imagery (several wide-ranging bands recorded, e.g. red green blue, or RGB, imagery). Several studies have previously demonstrated distinct spectral differences between tidal marsh species using hyperspectral imagery (Rosso et al. 2005; Belluco et al. 2006; Yang 2009). Such imagery, when combined with elevation data, has previously produced high-accuracy classifications of tidal marsh vegetation communities, albeit at smaller spatial scales (e.g. Hladik et al. 2013). Hyperspectral imagery is costly, however, especially across large landscapes (Adam et al. 2010); Belluco et al. (2006) compared several aerial and satellite sensors with changing spatial and spectral resolution, and although hyperspectral imagery performed slightly better than the multispectral, spatial resolution was the most important factor in classifier performance. Belluco et al. (2006) recommend the use of multispectral satellite imagery for the mapping of marsh vegetation. Beside the visible spectrum (RGB), multispectral imagery should also include infrared (IR) reflectance values to allow differentiation of vegetation types, calculation of vegetation indices (e.g. the Normalized Difference Vegetation Index or NDVI, Rouse, Haas, & Schell 1974) and detection of soil moisture differences (Jin and Sader 2005; Pettorelli et al. 2005), particularly in tidal wetlands (Klemas 2011). The IR spectrum has previously been used as a tool to predict tidal marsh communities both in smaller regions within the northeastern United States (Gilmore et al. 2008; Hoover et al. 2010; Meiman et al. 2012) and elsewhere (Isacch et al. 2006; Liu et al. 2010). An exception to this has been in the classification of invasive Phragmites that often borders tidal marshes. Large-scale classification of this wetland class has met with some success (e.g. Borgeau-Chavez et al. 2015, Long et al. 2017), however success is limited when using RGB and IR inputs alone (Samiappan et al. 2017). A large-scale effort to map coastal Phragmites in the northeastern US has not yet been attempted. Due to the large spatial scale at which northeastern tidal marshes occur, publicly-available and low-cost imagery datasets offer the most promising option for repeatedly delineating large swaths of marsh along the coast. Landsat satellite imagery provided by the National Aeronautic and Space Administration (NASA) is publicly available multispectral imagery including RGB and IR bands provided at a 30 x 30 m resolution and is often used for classifying coarse cover types across large landscapes. In the case of tidal marshes, however, the heterogeneity in tidal marsh vegetation often occurs at scales smaller than 30 m pixels, and large-scale classifications of tidal marsh plant communities using this imagery have so far been unfruitful (e.g. Correll 2015). There is thus a clear need for an alternative path to create a regional classification of tidal marsh vegetation. Recent advances in high-resolution airborne imagery provide new opportunities to develop large-scale classifications of coastal plant communities in the northeast, including Phragmites (e.g. Xie et al. 2015). The National Agricultural Imagery Program (NAIP) from the US Department of Agriculture (USDA 2016) captures 3-band, high-resolution RGB orthophotos during the growing season. Since 2007 most states have added a Near-InfraRed (NIR) band to the image requirements to aid in the accurate classification of vegetative cover. The image resolution is 1 m with 6-m horizontal accuracy and a

15 maximum of 10% cloud cover. The imagery, freely available for governmental agencies and the public, is an affordable alternative to commercial aerial and high-resolution multispectral satellite imagery. Recent applications of NAIP imagery include mapping of tree cover (Davies et al. 2010), forest clearings (Baker et al. 2013), isolated trees (Meneguzzo et al. 2013), land cover classification (Baker et al. 2013) and mining activity (Maxwell et al. 2014). In this study we compare several remote sensing techniques applied to NAIP imagery, elevation data from the National Elevation Dataset (NED) provided by the US Geological Survey (USGS 2015), and local tidal information records from the National Oceanic and Atmospheric Administration (NOAA 2016) to develop an affordable tool capable of repeated classification of high-marsh zones in tidal marshes in the northeastern United States. We then use the best-performing classifier to categorize marsh vegetation communities with a 3-m resolution from coastal Maine to Virginia, USA.

Methods Study site and community types Our marsh-mapping effort encompasses all salt marshes of the Northeast Atlantic coast of the USA, from northern Maine to Virginia. To define our classification extent, we applied a 500-m buffer to all coastal, tidal marsh as delineated by the National Wetland Inventory (NWI, USFWS 2010) estuarine emergent wetland (E2EM) layer. The study site is further split into 8 subzones (Figure 1) to accommodate data management and processing. These coastal marshes vary substantially from north to south. Due to local bathymetric structure, tidal amplitudes in the Gulf of Maine are among the highest in the world (Garrett 1972), while those farther south experience much less variation between high and low tides. Similarly, a preponderance of rocky or highly sloped shorelines in the north limits marshes to small (~10 - 100 ha) patches, while southern marshes form larger patches of marsh along the coast. Across our study area, however, tidal marsh ecosystems can be reliably separated into six distinct cover types, plus two bordering cover types, which we included in our marsh mapping effort:

1. High marsh: Area flooded during spring tides related to the lunar cycle and dominated by Spartina patens, Distichlis spicata, Juncus gerardii, and short form Spartina alterniflora. Other species include Juncus roemerianus, Scirpus pungens, Scirpus robustus, Limonium nashii, Aster tenuifolius, and Triglochin maritima. 2. Low marsh: Area flooded regularly by daily tides and dominated by tall form Spartina alterniflora. 3. Salt pools/pannes: Depressed, bare areas with sparse vegetation cover and extremely high soil salinities. Generally, pools retain water between high tides while pannes do not. 4. Terrestrial border: Area infrequently flooded by storm and spring tides and can include areas of marsh with fresh/brackish water due to a high water table and/or runoff from impervious surfaces. Typical plant species include Typha angustifolia, Iva frutescens, Baccharis halimifolia, Solidago sempervirens, Scirpus robustus, and Spartina pectinata. 5. Phragmites: The exotic invasive form of Phragmites australis (subspecies australis). This subspecies is of considerable management interest (Saltonstall 2002), especially in marshes with freshwater input, upland development, and/or increased nutrients (Dreyer and Nierling 1995; Bertness et al. 2002; Silliman and Bertness 2004). 6. Mudflat: Exposed muddy areas free of vegetation. 7. Open water (bordering cover type): Channels and bays leading to open ocean included within the 500-m buffer. 8. Upland (bordering cover type): All non-marsh terrestrial cover included within the 500-m buffer.

16

Data sources We collected training data for marsh vegetation classes both in the field and remotely using aerial imagery, depending on the cover type. We collected training polygons for high marsh, low marsh, and Phragmites between May and August of 2015 and 2016. Technicians collecting polygon data were collectively trained at the beginning of the season in salt marsh vegetation identification. Phragmites polygons mapped were of the invasive Phragmites australis australis and not of the native North American form Phragmites australis americanus. Tecnnicians used a GEO 7X Trimble GPS (Trimble 2015) without an external antenna for all community delineation. Horizontal accuracy of this unit without the external antenna is estimated at < 1m by the manufacturer. We used a generalized random tessellation stratified (GRTS) sampling framework designed to sample tidal marsh bird communities across all ownership types (Wiest et al. 2016) to select randomly- located delineation sites for training data across our study area. Technicians navigated to bird survey points and located contiguous patches of high marsh, low marsh, and Phragmites larger than 10 x 10 m as they traversed to each bird survey location. At each patch, they placed a stake flag or other highly visible marker on the ground to indicate the beginning of polygon delineation and then delineated the outer boundary of the patch on foot by walking the outer perimeter with the GEO 7X. We collected training data for open water, pools and pannes, and mudflat cover classes using manual digitization of 2014-2015 1-m NAIP imagery using ArcGIS 10.3 (ESRI 2016) since these cover classes are easily identifiable in visible wavelength imagery. We used the most recent digital ortho-photography (RGB and NIR) available from the NAIP collected during the growing season from 2014 or 2015 as imagery predictor data (see Appendix A for acquisition year by state). We resampled raw 1 m NAIP imagery to 3 m resolution to match the spatial scale of the NED, which was used as the digital elevation model (DEM) for this analysis. We also calculated NAIP imagery derivatives using ArcMap 10.3 using the raw band values. We refer to them as ‘pseudo’-vegetation indices because we used the raw band values instead of reflectance values. In total, we used the following data inputs as predictor variables: DEM, Raw NAIP Band 1, Raw NAIP Band 2, Raw NAIP Band 3, Raw NAIP band 4, NDVI, the Normalized Difference Water Index (McFeeters 1996), the Difference Vegetation Index (Richardson and Wiegand 1977), and the first three principle components from a principal components analysis (Fung and Ledrew 1987) of the four NAIP bands, which collectively explained > 95% of the variance. Marsh habitats are often influenced by their elevation and topographic context and therefore can often be successfully mapped using elevation data (e.g. Hladik et al. 2013, Maxwell et al. 2016). Elevational data is particularly helpful in tidal marshes, where topography can drive tidal flooding frequency and thus can influence plant species zonation (Silvestri et al. 2005). We used the NED for all elevation predictor data. The NED is derived from different contributed datasets and then processed by the USGS into a near-continuous DEM at various resolutions across the US. We used 1/9-second (~3 m resolution) data when available for our classification. When no 1/9 arc-second imagery was available, we used 1/3 arc-second data (~10 m resolution). To account for the large differences in tidal inundation across our study area, we collected tidal data for the study area from the closest NOAA tidal gauge station, creating 29 different tidal zones (NOAA 2016). For each of the stations we collected the following tidal datums: HAT, MHHW, MHW, MSL, NAVD88 and MAX (Table 1). We resampled the NED data to an exact 3 m resolution to match the upscaled NAIP imagery and clipped the resulting imagery with the 500 m buffer around all coastal tidal marsh in the NWI. We further clipped the NED by the 29 tidal gauge zones of the study area, and rescaled each zone to the NAVD88 datum using the NOAA tidal amplitude data. To calibrate the DEM across the entire study site we used the Mean High Tide (MHT) divided by Mean Highest High Tide (MHHT) value for each tidal gauge zone as a basis for elevational differences between marsh zone types.

17 We classified the NED of each tidal gauge zone based on the tidal amplitude data for that zone. To do this, we rescaled NOAA tidal amplitude data to match the NAVD88 datum used in the elevation dataset, and defined elevation limits for open water, high marsh, low marsh, terrestrial border, and upland class based on flooding history (Figure 2). We then used these thresholds in conjunction with NAIP imagery reflectance values and derivatives to identify water, high marsh, low marsh, and Phragmites cover types. We used only elevation thresholds to define the terrestrial border and upland cover types. In rare cases in small areas along the coast where there was no NED DEM layer available, we classified the marsh communities without the DEM data input. In these cases, no terrestrial border or upland was defined.

Data analysis We compared three classification methods using training data collected in 2015 to delineate tidal marsh cover types. We conducted this comparison of classifiers on NAIP imagery, imagery derivatives, and NED elevation data from the center of our study area (Delaware Bay, subregion 6, Figure 1) to maximize utility of the resulting method to both the north and south. First, we used classification and regression trees (CART), a fast and flexible rule-based classifier where no statistical data distribution is required (Otukei and Blaschke, 2010). CART methods are particularly useful when integrating environmental variables with different measurement scales and are robust for large datasets. Post-hoc pruning removes nodes with low explanatory power to reduce overfitting. We used the R package rpart (Therneau et al. 2015) which implements the CART methods described by (Breiman et al. 1984). Secondly, we used random forests (RF), which are collections of decision trees that improve the accuracy and stability of a single decision tree (Breiman 2001). RFs perform well with small training sets, uses a random subset of the training data to calculate the variable importance, and similar to cross- validation the out-of-bag (OOB) error estimates delivers a measure of classification accuracy (Breiman 2001). We used the R package RandomForest (Liaw and Wiener 2002) with Ntree and mtry default values for all RF analyses. Finally, we used support vector machines (SVM), which are non-parametric classifiers that use risk minimization to separate classes defined by the ‘support vectors’, or points that occur closest to the splitting threshold. In general, SVM offers high training performance versus low generalizing errors, but is sensitive to over-fitting, especially with noisy and unbalanced data. We used the svm function in the R package e1071 (Dimitriadou et al. 2006) with the default parameter values and polynomial kernel. To select the best remote sensing classifier for the final marsh layer, we classified and validated a randomly-selected independent training and validation subset (66% training, 33% validation) of the first year of polygon training data (2015, n = 36 training polygons) from zone 6 and compared classification accuracies across CART, RF, and SVM methods. We also considered specifications and known strengths of the of the classifiers including robustness, sensitivity to over fitting, and performance with relatively small training datasets. We applied the best performing classifier, RF, to all biogeographic zones from Maine to Virginia using NAIP imagery, imagery derivatives, NED elevation data, and NOAA tidal gauge information. For the final classification, we used the out-of-bag (OOB) error estimates produced by the RF algorithm to measure accuracy of our classification by zone. We then clipped the resulting marsh classification by the DEM-based cover types (upland, terrestrial border), wherever the NED layer was available. Due to variable image quality and/or due to high tide during image acquisition, we sometimes encountered artefacts in the imagery that affected the accuracy of our classification, particularly in Zones 1 and 6. Sun glitter or high mud content in open water sometimes caused misclassification of this cover type as low marsh. We used the RF probability scores for the open water class to better represent the actual water cover and then updated the open water classification in Zones 1 and 6 to improve overall accuracy. All methods and datasets involved in our study are freely available to the public, and our analyses are

18 limited to tools available through ArcGIS, a geographic information system commonly-used by federal, state, and private conservation organizations, or simple open-source Program R code (Appendix B).

Results We collected a total of 2655 field training polygons across all cover types from Maine to Virginia (Table 2). Of these training polygons, 36 were used in our methods comparison in Zone 6 for high marsh (n=18), low marsh (n=14), and Phragmites (n=4). In this comparison we found that RF methods generally outperformed the other two tools in classification of tidal marsh plant communities (Table 3). Classification of high and low marsh cover types returned accuracy rates ranging between 73% and 88% across all classifiers, with RF producing substantially lower error rates for high marsh and all three classifiers producing similar levels of accuracy for low marsh (Table 3). All classifiers produced low accuracy rates for the Phragmites cover type, ranging from 32-55%, but SVM methods were substantially worse when compared with the other two classifiers. Our final data layer produced through RF methods classified a total of 16,014 km2 of tidal marsh and bordering communities within our defined 500-m buffer at a 3-m resolution (Figure 3). This layer is now publicly available for public download at https://nalcc.databasin.org/galleries/46d6e771dd6f4fdb8aa5eb46efffffa7. Mean classification accuracies varied among cover types, ranging from >99% for open water and mudflat to 25% for Phragmites (Table 2). Within geographic zones, open water, mudflat, and pools/pannes were classified with high (>95%) accuracy in almost all cases. Classification of the three vegetation types varied among cover classes. High marsh was classified with a mean accuracy of 94%, but with clear regional variation. In the regions from New Jersey north and in the inner portions of Chesapeake Bay, accuracy was generally greater than 95%. In contrast, accuracy from Delaware Bay to Virginia was lower with 83-87% of high marsh correctly classified. High marsh was most commonly misclassified as low marsh. Classification of low marsh was generally less accurate than high marsh, with an overall accuracy of 77%. Low marsh was regularly confounded with open water and classification accuracy tended to be higher in the southern zones, especially the inner portions of Delaware and Chesapeake Bays. Classification of Phragmites showed the greatest variation across regions, and was often confounded with terrestrial border. Overall, classification accuracy for this cover type was 75%, and in coastal Massachusetts accuracy approached 95%. By contrast, overall Phragmites accuracy in coastal New Jersey was 67%, and in our northernmost region (coastal New Hampshire and Maine) accuracy was 20%. Finally, almost 4000 km2 of the data layer is covered by tidal marsh (i.e., excluding open water or upland, Table 4), with the majority classified as high marsh (36%) followed by low marsh (21%), mudflats (7%), Phragmites (7%), pools and pannes (5%), and terrestrial border (24%). The distribution of cover types varies from north to south with an overall increase in low marsh area and Phragmites to the south. Conversely, the percentage of mudflat cover increases to the north.

Discussion Tidal marshes of the northeastern USA are critical pieces of the coastal landscape, providing key wildlife habitat and ecosystem services to humans (Barbier et al. 2011). Our effort applies well-established methods and data sources for remote sensing of wetlands to classify plant communities within this important ecosystem from Maine to Virginia; the resulting data layer is the first of its kind to classify this marsh ecosystem at such a high spatial resolution (3 m) regionally, with a mean map classification accuracy of 90% and a classification accuracy for high marsh vegetation of 94%. The high classification accuracies in the high marsh zone make this data layer particularly helpful for use by marsh managers, researchers, and planners. Tracking change in marsh vegetation can be used

19 to understand which marshes are most impacted by sea-level rise, predict changes in flooding risk potential for coastal properties (Arkema et al. 2013), and to identify marshes with low rates of change to protect as important habitat for marsh-obligate species. As sea levels continue to rise and human development continues to alter marsh hydrology and accretion (Day et al. 2008), a method to repeatedly classify cover types of coastal marsh will be integral to measuring the amount and distribution of available habitat and change in habitat over time. The small spatial resolution and high horizontal accuracy of our data layer also allows it to be used across varying spatial extents, from local municipalities to multi-state regions. The NAIP dataset used in our classification offers a high-resolution, low-cost set of multispectral imagery with a refresh rate of 3 years. This dataset, however, has limitations when used over large areas (see also Meneguzzo et al. 2013), and potential users should carefully consider the pros and cons of this dataset before setting out on a similar classification effort. Variation in the time and day of image acquisition across the NAIP dataset results in different tidal stages, plant phenology, and illumination across images. The NAIP post-hoc color balancing applied to these images by each contractor (7 contractors across our study area) does not correct for differences in illumination and atmosphere in a standardized way. This results in a radiometric imbalance across the spatial extent of the dataset, limiting the use of the NAIP dataset in large-scale classification efforts to vegetative communities with either large differences in reflectance values across the RGB and NIR bands (as is the case with tidal marsh) or those varying across other characteristics measurable across large areas (e.g. elevation in the case of tidal marshes). Further, the temporal resolution of the NAIP dataset (one set of images per year, flown during “greenup” between June and August) is ideal for marsh classification but limits the use of NAIP data to analyses specific to this time of the year. Since the timing of high and low tide changes daily, this temporal resolution also results in imagery taken at different parts of the daily tidal cycle, and low marsh or mudflat areas were likely partially flooded when most imagery was collected. While these sources of error likely contributed to some noise within our classification effort, multiple contractors collected the imagery included in each analysis zone across different times of tidal inundation, making a quantitative comparison of the combined effects of these shortcomings difficult. Conveniently, the RF methods used in our final classification work well with these described sources of variation with low threat of over-fitting (see comparison of data sources and RF classification in Figure 4). Classification success (Table 2) assessed through OOB error estimate shows high average accuracies for most cover types, although there is variation depending on biogeographic zone. Additional work to collect independent data and compare it to this layer’s predictions would be a valuable next step to identify ways to improve current methods; while OOB error estimates are a commonly used measure of RF accuracy (Breiman 2001), they do not systematically evaluate classification accuracy outside the training polygons (e.g. Fry et al. 2011). Future studies to independently validate our classification outside of OOB error estimation the will directly strengthen support for the long-term use of this tool in monitoring community change in tidal marshes over time. In particular, additional tests to assess on- the-ground accuracy should involve data collected across the entire region due to the OOB differences we find across our analysis zones. Although the primary focus of our study was to distinguish between high and low marsh, the importance of invasive Phragmites to many management decisions led us to also consider this cover type. Unfortunately, Phragmites proved particularly difficult to classify, however we have chosen to keep this cover type in the overall classification because accuracy is relatively high (>70%) in the majority of our analysis zones (zones 2-4,6-8). The nature and amount of training data available for this cover type (n = 148 polygons, Table 2) compared to the other cover classes likely contributed to error in our Phragmites classification; previous work using RF methods suggest that training sample balance in imperative between classes for optimal classification results (Belgiu & Drăgu 2016). Our results may have also been influenced by the size of our Phragmites training polygons, as this rapidly-growing cover

20 type can change patch size quickly, influencing classifier accuracy (Kettenring et al. 2016) and make predicting across large landscapes particularly difficult. Additionally, upon assessment of known marshes, the classification algorithm for Phragmites was regularly confounded with terrestrial border species not included in the training data, particularly with stands of Typha spp., which is similar in structure to Phragmites. This problem combined with the variable ground elevations at which Phragmites can be found likely further confounded the classification. A strong need remains for development of a method for large-scale classification of this invasive species across latitudes, flooding regimes, and imagery sources for use in monitoring and management along the coast.

Conclusions Repeated, large-scale classification of coastal vegetation communities is urgently needed to help inform a variety of conservation and management issues related to this rapidly shrinking ecosystem. We present methods for a repeatable classification at a 3-m resolution of distinct cover types within tidal marshes of the northeastern USA to serve 1) as a vegetation community delineation for use in management and conservation decision-making, 2) as a layer for local and regional analyses of this biological community, and 3) as a base layer against which future comparisons can measure land cover change over time. These actions are all integral for the long-term preservation of tidal marshes and the species they support, especially as climate change and other human influences continue to affect this ecosystem.

References Adam E, Mutanga O, Rugege D (2010) Multispectral and hyperspectral remote sensing for identification and mapping of wetland vegetation: a review. Wetlands Ecology and Management, 18(3): 281-296. Arkema KK, Guannel G, Verutes G, Wood SA, Guerry A, Ruckelshaus M, Kareiva P, Lacayo M, Silver JM (2013) Coastal habitats shield people and property from sea-level rise and storms. Nature Climate Change 3:913-918. Baker B, Warner T, Conley JF, McNeil BE (2013) Does spatial resolution matter? A multi-scale comparison of object-based and pixel-based methods for detecting change associated with gas well drilling operations. International Journal of Remote Sensing 34(5): 1633–1651. Barbier EB, Hacker SD, Kennedy C, Koch EW, Stier AC, Silliman BR (2011) The value of estuarine and coastal ecosystem services. Ecological Monographs 81(2); 169–193. Belgiu and Dragut (2016) Random forest in Remote Sensing: A review of applications and future directions. ISPRS Journal of Photogrammetry and Remote Sensing 114: 24-31. Belluco E, Camuffo M, Ferrari S, Modenese L, Silvestri S, Marani A, Marani, M (2006) Mapping salt-marsh vegetation by multispectral and hyperspectral remote sensing. Remote Sensing of Environment 105(1): 54–67. Bertness MD (1991) Zonation of Spartina patens and Spartina alterniflora in New England salt marsh. Ecology 72(1): 138–148. Bertness MD, Ellison, AM (1987) Determinants of pattern in a New England salt marsh plant community. Ecological Monographs 57(2): 129–147. Bertness MD, Ewanchuk PJ, Silliman BR (2002) Anthropogenic modification of New England salt marsh landscapes. Proceedings of the National Academy of Sciences, USA 99(3): 1395–1398. Boesch DF, Turner RE (1984) Dependence of fishery species on salt marshes: the role of food and refuge. Estuaries 7(4): 460. Bourgeau-Chavez L, Endres S, Battaglia M, Miller ME, Banda E, Laubach Z, Marcaccio J (2015) Development of a bi-national Great Lakes coastal wetland and land use map using three-season PALSAR and Landsat imagery. Remote Sensing, 7(7): 8655-8682. Breiman, L (2001) Random forests. Machine Learning 45(1): 5–32.

21 Breiman L, Friedman JH, Olshen RA, Stone CJ (1984) Classification and Regression Trees. The Wadsworth Statistics Probability Series (Vol. 19). Brown S, Harrington B, Parsons K, Mallory E (2002) Waterbird use of northern Atlantic wetlands protected under the North American wetlands conservation act. Waterbirds 25: 106–114. Chambers RM, Meyerson LA, Saltonstall K (1999) Expansion of Phragmites australis into tidal wetlands of North America. Aquatic Botany 64(3–4): 261–273. Chu-Agor ML, Muñoz-Carpena R, Kiker G, Emanuelsson A, Linkov I (2011) Exploring vulnerability of coastal habitats to sea level rise through global sensitivity and uncertainty analyses. Environmental Modelling & Software 26(5): 593–604. Correll MD (2015) The biogeography and conservation of tidal marsh bird communities across a changing landscape. Dissertation, University of Maine, Orono Maine USA. Correll MD, Wiest WA, Hodgman TP, Kelley JT, McGill BJ, Elphick CS, Olsen BJ (in press) A Pleistocene disturbance event describes modern diversity patterns in tidal marsh birds. Ecography. Correll MD, Wiest WA, Hodgman TP, Shriver WG, Elphick CS, McGill BJ, O'Brien K, Olsen BJ. (2017) Predictors of specialist avifaunal decline in coastal marshes. Conservation Biology 31(1): 172–182. Crosby SC, Sax DF, Palmer ME, Booth HS, Deegan LA, Bertness MD, Leslie HM (2016) Salt marsh persistence is threatened by predicted sea-level rise. Estuarine, Coastal and Shelf Science 181: 93– 99. Davies KW, Petersen SL, Johnson DD, Davis DB, Madsen MD, Zvirzdin DL, Bates JD (2010) Estimating juniper cover from National Agriculture Imagery Program (NAIP) imagery and evaluating relationships between potential cover and environmental variables. Rangeland Ecology & Management 63(6): 630–637. Day JW, Christian RR, Boesch DM, Yáñez-Arancibia A, Morris, J, Twilley RR, Stevenson, C (2008) Consequences of climate change on the ecogeomorphology of coastal wetlands. Estuaries and Coasts 31(3): 477–491. Dimitriadou E, Hornik K, Leisch F, Meyer D (2006) e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly E1071), TU Wien. R package version 1.6-8. https://CRAN.R- project.org/package=e1071 Donnelly JP, Bertness MD (2001) Rapid shoreward encroachment of salt marsh cordgrass in response to accelerated sea-level rise. Proceedings of the National Academy of Sciences, USA 98(25): 14218– 14223. Dreyer GD, Niering WA (1995) Tidal Marshes of Long Island Sound: Ecology, History and Restoration. Bulletins 34. Emery NC, Ewanchuk PJ, Bertness MD (2001) Competition and salt-marsh plant zonation: stress tolerators may be dominant competitors. Ecology 82(9): 2471–2485. ESRI (2016) ArcGIS Desktop: Release 10.3. Redlands, CA: Environmental Sysstems Research Institute. Ewanchuk PJ, Bertness, MD (2004) Structure and organization of a northern New England salt marsh plant community. Journal of Ecology 92: 72–85. Field CR, Gjerdrum C, Elphick CS (2016) Forest resistance to sea-level rise prevents landward migration of tidal marsh. Biological Conservation 201: 363-369. Field CR, Bayard TS, Gjerdrum C, Hill JM, Meiman S, Elphick, CS (2017a) High-resolution tide projections reveal extinction threshold in response to sea-level rise. Global Change Biology 23(5): 2058–2070. Field CR, Dayer AA, Elphick CS (2017b) Landowner behavior can determine the success of conservation strategies for ecosystem migration under sea-level rise. Proceedings of the National Academy of Sciences 114: 9134-9139. Field CR, Ruskin KJ, Benvenuti B, Borowske A, Cohen JB, Garey L, Hodgman TP, Kern RA, King E, Kocek AR, Kovach AI, O’Brien KM, Olsen BJ, Pau N, Roberts, SG, Shelly E, Shriver WG, Walsh J, Elphick CS (in press) Quantifying the importance of geographic replication and representativeness when

22 estimating demographic rates, using a coastal species as a case study. Ecography. Fry J, Xian G, Jin S, Dewitz J, Homer CG, Yang L, Wickham JD (2011) Completion of the 2006 National Land Cover Database for the conterminous United States. Photogrammetric Engineering and Remote Sensing 77: 858–566. Fung T, Ledrew E (1987) Application of principal components analysis to change detection. Photogrammetric Enginnering and Remote Sensing 53(12): 1649–1658. Garrett C (1972) Tidal resonance in the Bay of Fundy and Gulf of Maine. Nature 238: 441–443. Gilmore MS, Wilson EH, Barrett N, Civco DL, Prisloe S, Hurd JD, Chadwick C (2008) Integrating multi- temporal spectral and structural information to map wetland vegetation in a lower Connecticut River tidal marsh. Remote Sensing of Environment 112(11): 4048–4060. Greenberg R, Maldonado JE, Droege S, McDonald MV (2006) Terrestrial vertebrates of tidal marshes: Evolution, ecology, and conservation. Studies in Avian Biology (32). Hladik C, Schalles J, Alber M (2013) Salt marsh elevation and habitat mapping using hyperspectral and LIDAR data. Remote Sensing of Environment 139:318-30. Hoover M, Civco D, Whelchel A (2010) The development of a salt marsh migration tool and its application in Long Island Sound. ASPRS 2010 Annual Conference Proceedings. San Diego, CA USA. IPCC (2014) Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II, and III to the FIfth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri RK and Meyer LA (eds)]. IPCC, Geneva, Switzerland. Isacch JP, Costa CSB, Rodriguez-Gallego L, Conde D, Escapa M, Gagliardini D, Iribarne OO (2006) Distribution of saltmarsh plant communities associated with environmental factors along a latitudinal gradient on the south-west Atlantic coast. Journal of Biogeography 33(5): 888–900. Jin S, Sader S (2005). Comparison of time series tasseled cap wetness and the normalized difference moisture index in detecting forest disturbances. Remote Sensing of Environment 94(3): 364–372. Kettenring KM, Mock KE, Zaman B, McKee M (2016) Life on the edge: reproductive mode and rate of invasive Phragmites australis patch expansion. Biological Invasions 18(9): 2475-2495. Kirwan ML, Guntenspergen GR (2010) Influence of tidal range on the stability of coastal marshland. Journal of Geophysical Research 115(F2): 1-11. Kirwan ML, Temmerman S, Skeehan EE, Guntenspergen GR, Faghe S (2016) Overestimation of marsh vulnerability to sea level rise. Nature Climate Change 6(3): 253–260. Klemas V (2011) Remote sensing of wetlands: case studies comparing practical techniques. Journal of Coastal Research 27(3): 418–427. Liaw A, Wiener M (2002) Classification and regression by randomForest. R News 2(3): 18-22. Liu C, Jiang H, Hou Y, Zhang S, Su L, Li X, Wen Z (2010) Habitat changes for breeding waterbirds in Yancheng National Nature Reserve, China: a remote sensing study. Wetlands 30(5): 879–888. Long AL, Kettenring KM, Hawkins CP, Neale CM (2017) Distribution and drivers of a widespread, invasive wetland grass, Phragmites australis, in wetlands of the Great Salt Lake, Utah, USA. Wetlands 37(1): 45-57. Master TL (1992) Composition, structure, and dynamics of mixed-species foraging aggregations in a southern New Jersey salt marsh. Colonial Waterbirds 15(1): 66-74. Maxwell AE, Strager MP, Warner TA, Zégre NP, Yuill CB (2014) Comparison of NAIP orthophotography and RapidEye satellite imagery for mapping of mining and mine reclamation. GIScience & Remote Sensing 51(3): 301–320. Maxwell AE, Warner TA, Strager MP (2016) Predicting palustrine wetland probability sing Random Forest machine learning and Digital Elevation Data-derived terrain variables. Photogrammetric Engineering & Remote Sensing 82(6): 437-447. McFeeters SK (1996) The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. International Journal of Remote Sensing 17(7): 1425–1432.

23 Meiman S, Civco D, Holsinger K, Elphick CS (2012) Comparing habitat models using ground-based and remote sensing data: Saltmarsh Sparrow presence versus nesting. Wetlands 32(4): 725–736. Meneguzzo, DM, Liknes GC, Nelson MD (2013) Mapping trees outside forests using high-resolution aerial imagery: A comparison of pixel- and object-based classification approaches. Environmental Monitoring and Assessment 185(8): 6261–6275. Miller W, Egler F (1950) Vegetation of the Wequetequock-Pawcatuck tidal-marshes, Connecticut. Ecological Monographs 20(2): 143–172. National Oceanic and Atmospheric Administration 2016. Tides and Currents. Available at: https://tidesandcurrents.noaa.gov. Accessed February 2017. Nixon SW, Oviatt CA (1973) Ecology of a New England salt marsh. Ecological Monographs 43(4): 463– 498. Otukei JR, Blaschke T (2010) Land cover change assessment using decision trees, support vector machines and maximum likelihood classification algorithms. International Journal of Applied Earth Observation and Geoinformation 12(1): 27–31. Pennings S, Callaway R (1992) Salt marsh plant zonation: the relative importance of competition and physical factors. Ecology 73(2): 681–690. Pettorelli N, Vik JO, Mysterud A, Gaillard JM, Tucker CJ, Stenseth, NC (2005) Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution 20(9): 503–510. Philipp KR, Field RT (2005) Phragmites australis expansion in Delaware Bay salt marshes. Ecological Engineering 25(3): 275–291. Richardson, AJ, Wiegand, CL (1977) Distinguishing vegetation from soil background information. Photogrammetric Engineering and Remote Sensing 43(12): 1541–1552. Rosso PH, Ustin SL, Hastings A (2005) Mapping marshland vegetation of San Francisco Bay, California, using hyperspectral data. International Journal of Remote Sensing 26(23): 5169–5191. Rouse JW, Haas RH, Schell JA (1974) Monitoring the vernal advancement and retrogradation (greenwave effect) of natural vegetation. Progress Report. NASA Goddard Space Flight Center, College Station Texas, USA. Samiappan S, Turnage G, Hathcock L, Casagrande L, Stinson P, Moorhead R (2017). Using Unmanned Aerial Systems for high-resolution remote sensing to map invasive Phragmites australis in coastal wetlands. International Journal of Remote Sensing 38(8-10): 2199-2217. Saltonstall K (2002) Cryptic invasion by a non-native genotype of the common reed, Phragmites australis, into North America. Proceedings of the National Academy of Sciences, USA 99(4): 2445– 2449. Silliman BR, Bertness MD (2004) Shoreline development drives invasion of Phragmites australis and the loss of plant diversity on New England salt marshes. Conservation Biology 18(5): 1424-1434. Silvestri S, Defina A, Marani M (2005) Tidal regime, salinity and salt marsh plant zonation. Estuarine, Coastal and Shelf Science 62(1–2): 119–130. Therneau T, Atkinson B, Ripley B, Ripley MB (2015) rpart: Recursive Partitioning and Regression Trees. R Package Version 4.1-10. Trimble 2015. GEO 7X Ground Positioning System. Available at: https://www.trimble.com/mappingGIS/geo-7-series US Department of Agriculture 2016. National Agriculture Imagery Program accessed through the Geospatial Data Gateway. Available at: http:// datagateway.nrcs.usda.gov. Accessed February 2016. US Geological Survey 2015. National Elevation Dataset (NED). Available at: https://nationalmap.gov/elevation.html. Accessed January 2016.

24 US Fish and Wildlife Service. National Wetland Inventory. Available at: https://www.fws.gov/wetlands/index.html. Accessed January 2016. Wiest WA, Correll MD, Olsen BJ, Elphick CS, Hodgman TP, Curson DR, Shriver WG (2016) Population estimates for tidal marsh birds of high conservation concern in the northeastern USA from a design-based survey. The Condor 118(2): 274–288. Wilson C, Hughes ZJ, FitzGerald DM, Hopkinson CS, Valentine V, Kolker AS (2014) Saltmarsh pool and tidal creek morphodynamics: dynamic equilibrium of northern latitude saltmarshes? Geomorphology 213: 99–115. Xie Y, Zhang A, Welsh W (2015) Mapping wetlands and Phragmites using publically available remotely sensed images. Photogrammetric Engineering & Remote Sensing 81(1):69-78. Yang J (2009) Mapping salt marsh vegetation by integrating hyperspectral imagery and LiDAR remote sensing. In Y Wang (Ed) Remote Sensing of Coastal Environments. CRC Press Boca Raton, FL 173– 186.

Table 1. Summary and definitions of datums collected by the National Oceanic and Atmospheric Administration (NOAA) and used to classify marsh cover type from Virginia to northern Maine, USA.

Datum NOAA explanation MAX Maximum: Highest observed water level. (No datum) HAT Highest Astronomical Tide: The elevation of the highest predicted astronomical tide expected to occur. MHHW Mean Higher High Water: Average of the higher high water height of each tidal day. MHW Mean High Water: Average of all the high water heights. MSL Mean Sea Level: The arithmetic mean of hourly heights observed NAVD88 North American Vertical Datum of 1988

25 Table 2. Percent classification error for cover classes of tidal marsh vegetation communities from Virginia to Maine, USA. Coverage created through random forest classification, with estimates organized by regional zones that extend from the Gulf of Maine (zone 1) to the eastern Chesapeake Bay (zone 8, Figure 1). Error percentages were derived from out-of-bag (OOB) errors. Zone 1 Zone 2 Zone 3 Zone 4 # # # #

OOB polygons OOB polygons OOB polygons OOB polygons High Marsh 1.97% 223 1.35% 37 4.89% 70 1.67% 74 Low Marsh 26.07% 136 23.99% 16 21.35% 32 27.63% 28 Mudflat 0.09% 57 0.05% 46 0.09% 63 2.83% 9 Phragmites 80.20% 36 5.82% 8 16.02% 29 13.66% 12 Pool/Panne 4.54% 67 3.91% 45 5.88% 35 12.92% 40 Open Water 0.09% 91 0.44% 62 0.03% 148 0.31% 66

Zone 5 Zone 6 Zone 7 Zone 8 Mean () # # # # # OOB polygons OOB polygons OOB polygons OOB polygons OOB polygons High Marsh 5.99% 103 12.23% 65 16.28% 81 2.29% 16 5.83% 669 Low Marsh 29.61% 39 19.01% 31 23.56% 41 13.65% 10 23.11% 333 Mudflat 0.80% 29 1.30% 64 0.53% 24 0.32% 2 0.75% 294 Phragmites 33.09% 14 19.68% 14 17.22% 22 14.79% 13 25.06% 148 Pool/Panne 4.49% 37 3.45% 74 0.82% 43 0.00% 19 4.50% 360 Open Water 0.26% 119 0.07% 190 0.23% 134 0.08% 41 0.19% 851

26 Table 3. Average classification error (%) for three different remote sensing techniques for classifying tidal marsh vegetation communities in Delaware Bay, USA using National Agricultural Imagery Program (NAIP) imagery from 2015.

High Marsh Low Marsh Phragmites Random Forests (RF) 11.86% 15.86% 48.73% Classification and Regression Trees (CART) 27.42% 18.78% 45.61% Support Vector Machines (SVM) 26.51% 13.81% 68.29%

Table 4: Total marsh area (km2) and percentage of total marsh for each cover class throughout the northeastern USA, from Virginia to Maine. Coverage created through random forest classification, with estimates organized by regional zones that extend from the Gulf of Maine (zone 1) to the eastern Chesapeake Bay (zone 8, Figure 1).

Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Zone 8 Total Total Marsh (km2) 398 177 214 243 653 515 542 914 3656 High Marsh (%) 38 36 33 36 54 26 40 27 36 Low Marsh (%) 17 9 6 8 21 44 32 13 21 Mudflat (%) 28 21 10 5 3 4 1 1 7 Phragmites (%) 1 7 4 9 5 7 8 12 7 Pool/Panne (%) 8 3 3 3 6 5 3 6 5 Terrestrial Border (%) 7 24 44 38 10 15 16 41 24

27 Figure 1. Geographic extent of our study site from Maine to Virginia, USA, divided into 8 biogeographic regions used in the remote sensing classification of tidal marsh cover types.

28

Figure 2. Graphic representation of the tidal elevation limits for tidal marsh community classification effort in the northeastern USA based on flooding history are represented by the National Oceanic and Atmospheric Administration (NOAA) vertical datums.

29

Figure 3. Visualization of a detailed random forest classification of vegetation communities within tidal marsh habitat in the Northeastern USA. Maps shows example portions of the entire coverage in A) northern Massachusetts and B) Delaware Bay, New Jersey; entire coverage is available at https://nalcc.databasin.org/galleries/46d6e771dd6f4fdb8aa5eb46efffffa7

30

Figure 4. Comparison of A) a random forest classification of tidal marsh vegetation communities in Scarborough Marsh, Maine with source data from B) the National Agriculture Imagery Program and C) the National Elevation Dataset (1/9 arc-second layer).

31

Appendix A. Acquisition dates and technical image information for National Agriculture Imagery Program imagery used to classify coastal marshes in the northeastern US.

Imagery characteristic Description Platform Aircraft Spatial resolution 1m (rescaled to 3m) Blue (400–580 nm) Green (500–650 nm) Spectral resolution Red (590–675 nm) NIR (675–850 nm)/(675-940 nm) State of Maine year of acquisition 2015 State of New Hampshire year of acquisition 2014 State of Massachusetts year of acquisition 2014 State of Rhode Island year of acquisition 2014 State of Connecticut year of acquisition 2014 State of New York year of acquisition 2015 State of New Jersey year of acquisition 2015 State of Maryland year of acquisition 2015 State of Delaware year of acquisition 2015 State of Virginia year of acquisition 2014

32

Appendix B. Source R code for Random Forest classification of coastal plant communities using National Agriculture Imagery Program, National Elevation Dataset, and National Oceanic and Atmospheric Administration regional data for the northeastern USA.

### Fine-scale mapping of coastal plant communities in the northeastern USA###

# Is the randomForest package installed? if (!require("randomForest",character.only = TRUE)) { install.packages("randomForest",dep=TRUE) } library(randomForest) library(rgdal) library(raster)

# Input: data @ 3m resolution, derived from NAIP and DEM datasets ### Layers: DEM, PCA(1-3), NDVI, NDWI, DVI, IMG(1-4) # Input: Training data: pixel info derived from training polygons ### Classes: HM, LM, Phrag, Mudflat, Open Water, Pool/Panne # Study site was separated in 8 zones

###################################################################### ######################################################

# Loop through the different zones for (i in 1:8){ #Select zone zone <-i #12345678 # Set workspace for each zone wd <- paste0(".../RData/NAIP_",zone,"_3m") setwd(wd)

#Read DEM, VI & IMG files from working directory files_VI <- list.files(getwd(), pattern="tif$", full.names=FALSE) stack_files_VI <- stack(files_VI[1:11])

#Read training, based on the training polygons,from working directory # the 'class' column stores the different classes for the classification TrainData <- read.csv(" .csv")

#Or read training shp-file Tname<- paste0("T_Zone",zone,"_2017") TrainingData <- readOGR(".../RData/NAIP_",zone,"_3m", Tname)

#Plot training data on the first raster layer plot(stack_files_VI[1])

33

plot(Trainingdata, add=TRUE)

# Extract training Pixel values for each class # based on code from http://amsantac.co/blog/en/2015/11/28/classification-r.html Vtype <- "Type" Tdata = data.frame(matrix(vector(), nrow = 0, ncol = length(names(stack_files_VI)) + 1)) for (i in 1:length(unique(TrainingData[[Vtype]]))){ category <- unique(TrainingData[[Vtype]])[i] categorymap <- TrainingData[TrainingData[[Vtype]] == category,] dataSet <- extract(stack_files_VI, categorymap) dataSet <- sapply(dataSet, function(x){cbind(x, class = rep(category, nrow(x)))}) tdata <- do.call("rbind", dataSet) Tdata <- rbind(tdata, df) } Tdata$class<-factor(Tdata$class) table(Tdata$class)

# Write csv of the training data to the working directory write.csv(Tdata, 'T_Data.csv')

# RF classifications RF_DEM_NAIP <- randomForest(class ~ ., data=Tdata, na.action = na.omit, importance = TRUE, confusion=TRUE)

# Plot variable importance OOB RF_DEM_NAIP RF_DEM_NAIP$confusion varImpPlot(RF_DEM_NAIP) plot(RF_DEM_NAIP)

# Create the RF classification maps Using all PC cores beginCluster() RF_Map_DEM_NAIP <- clusterR(stack_files_VI, predict, args = list(model = RF_DEM_NAIP, datatype="INT1U", type="response", overwrite=TRUE)) endCluster() # Plot the RF classification plot(RF_Map_DEM_NAIP) # Save as geotiff writeRaster(RF_Map_DEM_NAIP, filename = ".../Output/RF_Map_DEM_NAIP_2017.tif", overwrite=TRUE)

}

################################################# #### END #### #################################################

34

IV. Using systematic conservation planning to identify a focal species for tidal marsh bird conservation in the northeastern United States (submitted manuscript)

Brian T. Klingbeil, Jonathan B. Cohen, Maureen D. Correll, Christopher R. Field, Thomas P. Hodgman, Adrienne I. Kovach, Brian J. Olsen, W. Gregory Shriver, Whitney A. Wiest, and Chris S. Elphick

Systematic conservation planning was developed specifically to identify priority area networks that ensure the representation and persistence of biodiversity and is widely considered the standard for identifying spatial priorities for conservation investment (Margules & Pressey 2000; Kukkala & Moilanen 2013). Effective use of systematic conservation planning and associated software (e.g., ConsNet, C-Plan, Marxan and Zonation; reviewed in Moilanen et al. 2009) has traditionally required detailed information on the distribution of biodiversity at broad spatial scales. Insufficient time and funding, however, remain major obstacles to collecting data on the distribution of most species at the scales where conservation planning occurs, leading conservation biologists and resource managers to base many decisions on surrogates for which data already exists (Brooks et al. 2006; Lindenmayer & Likens 2011). Surrogate species can take many forms (e.g., flagship, focal, indicator, keystone, landscape and umbrella; reviewed in Favreau et al. 2006; Caro 2010) that differ in their intended purpose and application. The potential costs and benefits of using these shortcuts or proxies for conservation, as well as the tradeoffs inherent in selecting a particular type of surrogate have been extensively reviewed and debated (e.g., Caro 2010; Branton & Richardson 2011). Despite their widespread use, little consensus exists and considerable challenges remain regarding best practices to decide among potential species once the decision to use a surrogate has been made (Favreau et al. 2006; Rodrigues and Brooks 2007). Testing the efficacy of biodiversity surrogates and the use of surrogates to identify conservation areas with systematic conservation planning are frequently intertwined (e.g., Larsen et al. 2009; Nicholson et al. 2013; Jones et al. 2016). Studies generally select surrogate species a priori based on predefined criteria (e.g., rarity, broad spatial distribution, large area requirements, low fecundity or dispersal limitation), data availability, or management needs, and then measure overlap in distributions to quantify representation. This can be an effective approach, but there remains considerable debate around how well conservation planning based on surrogate species also acts to conserve other species, suggesting there may be ways to improve upon the process (Larsen et al. 2009; Granthem et al. 2010). Specifically, we suggest that systematic conservation planning could be used to select species for spatial prioritization of protected areas without relying on a priori criteria, (e.g., in instances where expert opinion or sufficient published material is lacking) and to quantitatively assess their efficacy as surrogates for additional species in real-world situations that minimize cost associated with land acquisition. Tidal marshes are a type of coastal wetland restricted to a narrow strip along temperate coasts (Chapman 1977). Many are situated near areas of high human development, making them among the most economically important yet vulnerable ecosystems, with this vulnerability exacerbated by sea-level rise (Arkema et al. 2013). The eastern coast of North America is home to over one-third of the global extent of tidal marsh and the highest level of vertebrate endemism of any tidal marsh region worldwide (Greenberg & Maldonado 2006). This combination of high vulnerability and concentration of unique species means conservation action needs to be prioritized quickly and suggests that a conservation shortcut may be beneficial (Brooks et al. 2006). Tidal marshes in the Northeast support a wide range of habitat specialist and generalist avian species throughout the year but not all birds are equally vulnerable to habitat loss and degradation

35

(Greenberg & Maldonado 2006; Correll et al. 2016). A negative relationship between population trend and degree of tidal marsh specialization exists, indicating that the more specialized a species is to tidal marsh habitat, the less likely it is to persist in this ecosystem over time (Correll et al. 2016). This relationship exemplifies the tradeoffs between specialist and generalist life history strategies where specialists often reach higher densities than generalists within their defined niche space (Dennis et al. 2011), but habitat generalists outperform specialists when these landscapes are degraded or fragmented and no longer maintain specialist environmental requirements. Over 100 species were observed in a recent survey of northeastern tidal marshes, yet five species in particular, the Clapper Rail (Rallus crepitans), Eastern Willet (Tringa semipalmata semipalmata), Acadian Nelson’s Sparrow (Ammodramus nelsonii subvirgatus), Saltmarsh Sparrow (Ammodramus caudacutus), and Seaside Sparrow (Ammodramus maritimus) were identified to have extremely high levels of a marsh specialization index (Correll et al. 2016). These taxa use coastal marshes almost exclusively as breeding habitat, usually nesting within a few centimeters of the ground (Cornell Lab of Ornithology 2015), making them particularly vulnerable to habitat loss or degradation. Indeed, long-term population declines have recently been identified in 3 of the 5 species (Correll et al. 2017), suggesting that increased attention from the conservation community is necessary. Current efforts to prioritize action for bird conservation in North America are at least partly determined with a focal species strategy (USFWS 2011; ACJV 2017). The focal species approach (Lambeck 1997) attempts to incorporate processes that threaten species viability, such as fragmentation or loss and degradation of habitat into conservation assessments. Focal species are generally selected on the basis of life history characteristics that make them vulnerable to key threats, are often of high conservation concern, and their inclusion into conservation plans is expected to confer protection to other co-occurring species facing similar threats (Fleishman et al. 2000; Favreau et al. 2006; Nicholson et al. 2013; Lindenmayer et al. 2014). To enhance the likelihood of success and engagement of stakeholders, the role of the species as a potential unifier for partnerships and the likelihood that factors affecting its status can be realistically addressed may also be considered as secondary factors (Lindenmayer et al. 2002; USFWS 2011). Based on these criteria, each of the five species characterized by a high marsh specialization index (Correll et al. 2016) have potential to guide tidal marsh bird conservation because their fates are closely linked to that of the habitat. Considering secondary factors, Saltmarsh Sparrow and Clapper Rail, appear to be especially strong candidates, and potentially work as a pair because they are representative of higher elevation and lower elevation marshes, respectively. Both species are of management interest for distinct reasons, with the sparrow facing near-term extinction (Field et al. 2017) while the rail is a more abundant, but still declining popular game species. Together, they represent the full range of marsh habitat and could bring together different stakeholders needed to achieve conservation success at regional scales. Despite this reasoning, a danger exists in relying principally on value-laden choices or pre-defined criteria rather than systematically evaluating all potential possibilities (Fleishman et al. 2000; Branton & Richardson 2011). The effectiveness of a focal species approach is contingent upon the assumption that management interventions aimed at conserving those focal species will confer protection on a number of co-occurring species and will not compromise the protection of others (Lambeck 1997; USFWS 2011; Lindenmayer et al. 2014). Consequently, the use of focal species to delineate areas for conservation may be ineffective if focal species do not reliably co-occur with other species of interest (Andelman & Fagan 2000). The assumption that other species are receiving protection as a result of the protection of a focal species, therefore, needs to be validated (Lindenmayer et al. 2014). In this study, we used systematic conservation planning methods to inform critical aspects of the focal species approach to planning for the protection of multiple co-occurring species. We addressed two main questions: 1) Can spatial prioritization methods provide an effective approach to

36 quantitatively select focal species? 2) Can approaches based on a single focal species match or outperform approaches based on multiple focal species? We addressed these questions by prioritizing areas for protection of each species individually and for all five species simultaneously to determine how well each species can represent each of the others (i.e., if one species is a good surrogate for another species) and how well each species can represent all species (if one species is a good surrogate for all species). To address our second question, we quantify the area, number of individuals protected and cost-effectiveness of planning options focused on potential focal species. We compared 2 single species strategies with 2 multi-species strategies and a strategy that uses all 5 species. By exploring these two questions we were able to identify ways in which spatial prioritization can inform aspects of focal species identification in a novel manner. By applying this process to a case study, we show that implementation is feasible for real world decision-making and provide recommendations that support the use of focal species in the management of tidal marsh birds throughout northeastern USA.

Methods We used the software program Marxan (version 2.43; Ball et al. 2009) to identify priority areas for protection of Clapper Rail, Eastern Willet, Acadian Nelson’s Sparrow, Saltmarsh Sparrow, and Seaside Sparrow. We prioritized 8,405 saltmarsh patches between Maine and Virginia (Fig.1) based on scenarios (i.e., species and their associated population target) described below. Saltmarsh patches between Maine and Virginia were delimited by creating a 50-m buffer around Estuarine Intertidal Emergent Wetland polygon features (Cowardin et al. 1979) identified in the National Wetlands Inventory (USFWS 2010). Polygons with buffers that intersected were joined into a single patch. Full details of saltmarsh patch creation appear in Wiest et al. (2016). Saltmarsh patches were preferred over a regular grid for spatial prioritization because they are biologically relevant spatial units restricted to the habitat of interest that provide a direct link between conservation prioritization and the location and extent of a particular marsh and its inhabitants. Each saltmarsh patch was associated with estimates of density for Clapper Rail, Eastern Willet, Acadian Nelson’s Sparrow, Saltmarsh Sparrow, and Seaside Sparrow derived from a comprehensive regional marsh bird survey in 2011 and 2012 (Wiest et al. 2016). A Bayesian network model incorporated a suite of environmental covariates and survey results that were adjusted to account for imperfect detection in order to estimate population density within 8,405 saltmarsh patches throughout the region (Wiest 2015). Density estimates indicated that Clapper Rails, Eastern Willets and Seaside Sparrows had population size > 100,000 in the northeastern USA, whereas Saltmarsh (< 60,000) and Acadian Nelson’s (< 7,000) Sparrows had much lower population sizes in the region (Wiest 2015; Wiest et al. 2016). To incorporate current protected areas into our spatial prioritization, we used the Protected Areas Database (USGS, Gap Analysis Program 2012; GAP) and our saltmarsh patch layer (Wiest et al. 2016) to identify the overlap between existing protected areas and saltmarsh patches. We considered only areas with permanent protection from conversion of natural land cover and a mandated management plan in operation to maintain a primarily natural state (i.e., GAP status 1 or 2). Areas with a GAP status of 3 or 4 were not included as they may be subject to extractive uses. Protected areas and saltmarsh patches rarely overlapped perfectly, so the percent of each saltmarsh patch protected was quantified. We estimated the cost of each patch from county-level asset values (USD/ha) of agricultural land including buildings (US Department of Agriculture 2012). Although the cost associated with purchasing saltmarsh patches will likely differ from the average county value, it provides a good index of relative costs across the region. Patches that spanned multiple counties received an average of the

37 county estimates. Final cost estimates incorporated protected area information and reflected only the cost of adding unprotected land within a patch.

Can spatial prioritization methods provide an effective approach to quantitatively select focal species? Marxan was used to identify priority areas for protection based on percent-based targets that range from 10% to 100% of the current population estimate, at 10 percent intervals. We examined these targets for each species individually and for all five species simultaneously. Marxan offers a number of heuristic algorithms to identify a reserve system that minimizes the total cost of sites in a network, while meeting a set of targets for conservation features (Ball et al. 2009). We evaluated all scenarios using a simulated annealing algorithm followed by a two-step iterative improvement algorithm. An optional parameter, the boundary length modifier, promotes selection of contiguous planning units by attempting to reduce the total boundary length of the full reserve network. We set this parameter equal to zero because we had no reason to favor selection of spatially clustered sites. Marxan also includes the option to force planning units into or out of a solution a priori based on information available to the user (e.g., to reflect currently protected areas or areas that are known to be unavailable for selection). We used this option to force all saltmarsh patches with ≥ 99% of their area overlapping current protection to be represented in solutions. Patches with only partial protection were not forced into solutions because doing so can lead to more costly reserve systems that do not support larger populations of target species (Supporting Information). We constrained all analyses so that each species had to be represented by at least the minimum population target in all runs. To force targets to be met, a species penalty factor was incorporated to make potential solutions that did not meet targets more costly than solutions for which targets were represented. The species penalty factors were set to 100 in individual species scenarios and 1,000 for Saltmarsh Sparrow in all multi-species analysis. These penalty factors were the lowest possible values that allowed Marxan flexibility to find an efficient solution, while ensuring targets were met in full in all runs for all species (Ball et al. 2009, Andron et al. 2010). These penalty factors ensure that species are protected at equivalent numbers of individuals despite differences in land cost in their distribution. Marxan output provides a selection frequency that shows the number of times each planning unit was selected in the optimal solution of a run, out of all runs of a given scenario. To quantify selection frequency, we ran each percent-based scenario 10,000 times with 1 million simulated annealing algorithm iterations in each run. Sensitivity analysis showed that scenarios run with at least 1 million iterations and 10,000 runs would be highly similar to those run with more iterations yet still provide flexibility to explore alternative solutions, but that fewer iterations could lead to substantially different results (Supporting Information). We used Spearman rank correlations and the selection frequency from each scenario to determine if scenarios that prioritize conservation of different species ranked sites in a similar way, which would suggest they are interchangeable from the perspective of site selection. For example, we compared the selection frequency of Saltmarsh Sparrow with a target of protecting 50% of the current population with the selection frequency of Clapper Rail with a target of protecting 50% of the current population. These comparisons allowed us to determine how well each species represented each other species at a particular population target, as well as the maximum strength of association between each species pair. We also quantified the average correlation of each species pair across all 10 population targets (i.e., 10%-100%). To determine how well a single species represents the suite of five species, we used the same process to quantify the maximum and average Spearman rank correlations between the selection frequency of sites prioritized by each species and that based on all species combined. To identify patterns relevant for species conservation in the Northeast we used the selection frequency to characterize the relative importance of particular patches and determined the corresponding area in each sub-region (Fig.1) that was required to achieve a

38 particular conservation target. We classified saltmarsh patches in the region by the number of times they were selected in 10,000 optimal solutions. We distinguished patches that were (a) selected in all 10,000 optimal solutions, (b) selected in at least 1, but not all, optimal solutions, and (c) not in any optimal solution.

Can approaches based on a single focal species match or outperform approaches based on multiple focal species? To determine how single species approaches perform compared to multi-species approaches we used Marxan to estimate the number of birds represented, number of saltmarsh patches, area protected and relative cost (in USD) of the lowest cost solutions for five scenarios with predetermined combinations of species. We identify a scenario by its constituent species because all comparisons were made with a population target of 50,000 individuals for each species: 1) Saltmarsh Sparrow (SALS), 2) Clapper Rail (CLRA), 3) Saltmarsh Sparrow and Clapper Rail (SALS + CLRA), 4) Saltmarsh Sparrow, Clapper Rail, Eastern Willet and Seaside Sparrow (4TMB, i.e., tidal marsh birds), and 5) Saltmarsh Sparrow, Clapper Rail, Eastern Willet, Seaside Sparrow and Acadian Nelson’s Sparrow (5TMB). This represents a reasonable population target for conservation planning in the region that provides a clear buffer above that likely to be needed to ensure long-term viability allowing for strong protection to ensure that subpopulations can be spread across the region to provide geographical representation (Tear et al. 2005; Trail et al 2010). The five scenarios were selected based on identification of focal species a priori and the results of analyses to answer the first question of the study. We explore scenarios with and without Acadian Nelson’s Sparrow because it has a distinctly northern geographic distribution relative to the other species and it hybridizes with Saltmarsh Sparrow within a portion of our study region (Walsh et al. 2015), but it is still a species of management concern. In order to include Acadian Nelson’s Sparrow, we use a population target equal to 99.9% of the current population estimate, in addition to 50,000 individuals for each of the other four species, because the current population size is much smaller than 50,000 individuals within the study region. Marxan was used to identify a single “best” solution that minimizes cost compared to all other solutions identified from each run. The lowest cost solutions were more sensitive to the number of iterations than the number of runs, and a greater number of iterations generally identified a more efficient reserve system, often with a significantly lower cost (Supporting Information). Therefore, we used 100 million iterations of the simulated annealing algorithm and 100 repeat software runs to identify the lowest cost solution for each of the five management scenarios with a population target of 50,000 for each species (except Acadian Nelson’s Sparrow, see above). To quantify uncertainty, we repeated this process (100 million iterations of the simulated annealing algorithm, and 100 repeat software runs) 10 times for each scenario and identified the lowest cost solution from each of the 10 groups of 100 runs. We take this approach, rather than use the selection frequency from a single set of 100 runs, to ensure that we are using a distribution of the best possible scenarios, given the data. We calculated the minimum, mean and 95% confidence intervals based on the 10 lowest cost solutions for the number of birds represented, number of saltmarsh patches, area protected, and estimated cost (based on land values in USD) of the lowest cost solutions. We compare these 4 characteristics among the 5 scenarios to determine if a single focal species (scenarios 1 or 2) provides equivalent representation as scenarios that evaluate multiple focal species (scenarios 3 and 4) or all species (scenario 5).

Results Correlations of single-species site prioritizations suggest that Clapper Rail and Eastern Willet were most similar, although the strength of this association was low as were all other single-species comparisons

39

(maximum rs < 0.535; Table 1). Of the five species, site prioritization based on Saltmarsh Sparrow conservation was most strongly associated with that based on all five species combined (maximum rs = 0.759; Table 1). Comparison of alternative suites of focal species targets to be used in conservation plans further indicated that site selection decisions based on Saltmarsh Sparrow alone have high potential to serve as an effective surrogate for tidal marsh bird conservation (Fig. 2). For example, the optimal solution for protecting 50,000 Saltmarsh Sparrows will also protect at least that many Clapper Rails and Eastern Willets, as well as over 175,000 Seaside Sparrows. Protection for Nelson’s Sparrow was less adequate, but still ensured over 45% of the US population is in protected patches. The best scenario for Clapper Rail, by contrast, failed to protect more than 27,000 individuals of any of the other species, although the cost was substantially lower. Comparing the scenario for Saltmarsh Sparrows to the three multi-species scenarios shows few additional benefits in terms of the numbers of birds protected or difference in cost (Fig. 2). The proportion of marshland identified as most valuable varied among species, regions, and with the level of the population goals. When the goal was to protect 90% of a population, relatively large areas (up to half, or more) were selected in every model run for several scenarios (Fig. 3). For all species except Nelson’s Sparrow, such areas were especially common in the southern regions, such as Delaware Bay and coastal New Jersey. For lower population goals (e.g., protecting only 50% of the population), there was more flexibility as to which sites were selected, with very few appearing in every model run (Fig. 3). Across the different scenarios there was considerable variation in the proportion of marsh area that never contributed to the chosen solution (Fig. 3). Areas were more likely to be selected at least sometimes in the more southern sub-regions. Nonetheless, there were patches that were never selected in any solution, regardless of the target species (Fig. 4). Sub-regions differed in the percent of saltmarsh patches that never contributed to an optimal solution, with few in southern New England, on Long Island, or in Delaware Bay or coastal Delmarva.

Discussion Most stated objectives for conservation, (e.g., maximize biodiversity), are too vague to be useful within a decision-making framework and often less effective at garnering support from the public than a single “flagship” species (Tear et al 2005; Caro 2010; Thomas-Walters and Raihani 2016). Management that only considers a single species-specific outcome (e.g., game species) however, may be too restrictive and inefficient to achieve the broad conservation goals of many organizations and can reduce return on conservation investment (Laitila and Moilanen 2012; Gallo and Pechjar 2016). A compromise is to find a species that can be used to garner public support, while also functioning as a focal species, such that management interventions aimed at conserving the focal species will confer protection on a large number of co-occurring species (Lambeck 1997; Lindenmayer et al. 2014). We show that spatial prioritization software that is already used to evaluate the relative effectiveness of alternative focal species identified from pre-defined criteria (e.g., Larsen et al. 2009; Nicholson et al. 2013; Jones et al. 2016) may be perform equally well to select the best candidate among a set of species, ignoring pre- defined criteria. This may offer a streamlined and unbiased approach to incorporate into future conservation planning. For tidal marsh bird conservation in the Northeast, pre-defined criteria suggested that Saltmarsh Sparrow and Clapper Rail were ideal candidates for focal species. Comparison of single-species site prioritizations with a comprehensive prioritization based on all species identified Saltmarsh Sparrow, but not Clapper Rail, as the best option considered for a representative focal species to prioritize land. However, prioritizing for the conservation of one species necessarily limits the resources for

40 conservation of other species, so it is critically important to understand how a focus on Saltmarsh Sparrow would affect planning for other species of interest. We found that the costs, area of land protection, and number of individuals of non-focal species on protected land were similar when planning for Saltmarsh Sparrow as compared to any of the three scenarios that evaluated multi-species combinations of saltmarsh birds. These results suggest that if increased attention and funds are focused on protecting and managing Saltmarsh Sparrow populations, other species are unlikely to suffer. In contrast, planning focused solely on Clapper Rail, the other species identified by pre-defined criteria, failed to protect population targets for any additional species, although the lower cost of this scenario may allow for additional protection. Identification of a focal species, like Saltmarsh Sparrow that can act as a surrogate for a group of species has advantages and limitations (Rodrigues and Brooks 2007; Caro 2010; Branton and Richardson 2011). In the current conservation climate where time and resources are in short supply, a focal species may be a useful starting point for decision making that can facilitate rapid and efficient means of identifying and prioritizing land management. For example, the Atlantic Coast Joint Venture – a collaboration among conservation agencies focused on bird conservation along the US East Coast – has identified Saltmarsh Sparrow as a focal species for coastal marsh conservation (ACJV 2017). Moreover, it may be cost-effective to focus time and effort on monitoring a single species rather than a suite of species. However, a focal species approach can be ineffective if the species fails to capture the conservation needs of the species group of interest (Andelman and Fagan 2000; Lindenmayer et al. 2002). Populations of all specialist marsh species face similar threats (e.g., human modification, sea-level rise). We show that Saltmarsh Sparrow is likely to be the best focal species for protecting avian tidal marsh specialists, but it may also be an effective focal species for conservation of tidal marsh species in general. Spatial prioritization that incorporates additional species distributions with the saltmarsh patches currently associated with bird density and cost estimates could be used to evaluate if this is the case and extend conservation planning to a broader suite of species. The flexibility of a quantitative approach analyzed with freely available software enables frequent updating and recalculating as new or improved data becomes available with limited investment of time and money. This approach can facilitate data sharing and coordination between local, regional and national efforts to guide land purchases that maximize efficiency. Deciding when to purchase land and when to hold on to limited resources is one of the most important conservation decisions. Availability of land is often unpredictable and decisions may need to be made quickly (see Alagador et al. 2016 for an approach that incorporates scheduling of priority area acquisition). Spatial prioritization and systematic conservation planning were developed in response to the criticism that protecting biodiversity in an “ad hoc” fashion was an inefficient use of limited conservation funding (Pressey 1993). In general, however, the approach has been used largely to identify the very best sites. Analyses can also be used to identify sites that are least likely to contribute to conservation planning goals, and that generally should be viewed as poor conservation investments. Our analysis, for example, identified a substantial number of saltmarsh patches that were never selected in scenarios that protect each species at their current population size. Identification of areas where conservation for tidal marsh birds is likely to have little or no effect could be as important as identifying the best sites. Knowing which areas do not advance conservation of a project’s target eliminates the risk associated with spending limited conservation funds on sites that provide little or no benefit, yet this information is often unavailable to decision makers. Our results highlight the importance of incorporating a quantitative approach into the decision- making process and a potentially valuable but underused aspect of systematic conservation planning. However, our study is not without limitations and hurdles to widespread implementation. First, our study relied on high quality species distribution data to estimate the density of each species in every

41 patch. This allowed us to quantify differences in the number of birds protected in alternative site prioritizations. These data are rarely available for many taxa or regions of conservation interest (Pimm et al. 2014) and this is a common critique of focal species approaches (e.g., Lindemayer et al. 2002). Second, our study was narrowly focused on five species from a restricted taxonomic subset in a single habitat. Many studies assess how well focal species represent large subsets of diverse groups of species in multiple habitats or even biodiversity as a whole (e.g., Favreau et al. 2006; Rodrigues and Brooks 2007; Larsen et al. 2009). It would be difficult to assess all possible combinations of species with the approach we undertook, however, advances in cloud computing and parallel processing (e.g., Marxan.net; Watts and Possingham 2013) may soon eliminate this limitation.

Supporting Information Sensitivity analyses are provided (Appendix 1) and ArcGIS shapefiles of spatial prioritization scenarios will be uploaded to a freely accessible online database (Data Basin) in accordance with funding requirements.

Literature Cited Alagador D, Cerdeira JO, Araújo MB. 2016. Climate change, species range shifts and dispersal corridors: an evaluation of spatial conservation models. Methods in Ecology and Evolution 7:853–866. Andelman SJ, Fagan WF. 2000. Umbrellas and flagships: efficient conservation surrogates or expensive mistakes? Proceedings of the National Academy of Sciences 97:5954-5959. Ardron JA, Possingham HP, Klein CJ. 2008. Marxan good practices handbook. Pacific Marine Analysis and Research Association, Vancouver. Arkema KK, Guannel G, Verutes G, Wood SA, Guerry A, Ruckelshaus M, Kareiva P, Lacayo M, Silver JM. 2013. Coastal habitats shield people and property from sea-level rise and storms. Nature Climate Change 3:913-918. Atlantic Coast Joint Venture. 2017. Flagship species initiative. Hadley, Massachusetts. http://acjv.org/flagship-species-initiative/ (Accessed October 2017). Ball IR, Possingham HP, Watts M. 2009. Marxan and relatives: software for spatial conservation prioritisation. Pages 185-195 in Moilanen A, Wilson KA Possingham HP editors. Spatial conservation prioritisation: quantitative methods and computational tools. Oxford University Press, Oxford. Branton M, Richardson JS. 2011. Assessing the Value of the Umbrella-Species Concept for Conservation Planning with Meta-Analysis. Conservation Biology 25:9-20. Brooks TM, Mittermeier RA, Fonseca GAB da, Gerlach J, Hoffmann M, Lamoreux JF, Mittermeier CG, Pilgrim JD, Rodrigues ASL. 2006. Global Biodiversity Conservation Priorities. Science 313:58–61. Caro T 2010. Conservation by proxy: indicator, umbrella, keystone, flagship, and other surrogate species. Island Press. Chapman VJ. 1977. Introduction. Pages 1–20 in Chapm VJ editors Wet coastal ecosystems. Elsevier, New York. Cornell Lab of Ornithology. 2015. The Birds of North America Online. Available from http://bna.birds.cornell.edu/bna/ (accessed August 1, 2016). Correll MD, Wiest WA, Hodgman TP, Shriver WG, Elphick CS, McGill BJ, O'Brien KM, Olsen BJ. 2017. Predictors of specialist avifaunal decline in coastal marshes. Conservation Biology 31:172–182. Correll MD, Wiest WA, Olsen BJ, Shriver WG, Elphick CS, Hodgman TP. 2016. Habitat specialization explains avian persistence in tidal marshes C7 - e01506. Ecosphere 7:e01506. 10.1002/ecs2.1506

42

Cowardin LM, Carter V, Golet FC, LaRoe ET 1979. Classification of wetlands and deepwater habitats of the United States. US Fish and Wildlife Service. Dennis RLH, Dapporto L, Fattorini S, Cook LM. 2011. The generalism–specialism debate: the role of generalists in the life and death of species. Biological Journal of the Linnean Society 104:725– 737. Favreau JM, Drew CA, Hess GR, Rubino MJ, Koch FH, Eschelbach KA. 2006. Recommendations for assessing the effectiveness of surrogate species approaches. Biodiversity & Conservation 15:3949–3969. Field CR, Bayard T, Gjerdrum C, Hill JM, Meiman S, Elphick CS. 2017. High-resolution tide projections reveal extinction threshold in response to sea-level rise. Global Change Biology 23: 2058–2070. Fleishman E, Murphy DD, Brussard PF. 2000. A new method for selection of umbrella species for conservation planning. Ecological Applications 10:569-579. Gallo T, Pejchar L. 2016. Improving habitat for game animals has mixed consequences for biodiversity conservation. Biological Conservation 197:47–52. Grantham HS, Pressey RL, Wells JA, Beattie AJ. 2010. Effectiveness of Biodiversity Surrogates for Conservation Planning: Different Measures of Effectiveness Generate a Kaleidoscope of Variation. PLOS ONE 5:e11430. Greenberg RG, Maldonado JE. 2006. Diversity and endemism in tidal marsh vertebrates. Pages 32–53 in Greenberg R, Droege S, Maldonado J, and McDonald MV editors Terrestrial Vertebrates of Tidal Marshes: Ecology, Evolution, and Conservation. The Cooper Ornithological Society, Camarillo. Jones KR, Plumptre AJ, Watson JEM, Possingham HP, Ayebare S, Rwetsiba A, Wanyama F, kujirakwinja D, Klein CJ. 2016. Testing the effectiveness of surrogate species for conservation planning in the Greater Virunga Landscape, Africa. Landscape and Urban Planning 145:1–11. Kukkala AS, Moilanen A. 2013. Core concepts of spatial prioritisation in systematic conservation planning. Biological Reviews of the Cambridge Philosophical Society 88:443-464. Laitila J, Moilanen A. 2012. Use of many low-level conservation targets reduces high-level conservation performance. Ecological Modelling 247:40–47. Lambeck RJ. 1997. Focal species: a multi-species umbrella for nature conservation. Conservation Biology 11:849-856. Larsen FW, Bladt J, Rahbek C. 2009. Indicator taxa revisited: useful for conservation planning? Diversity and Distributions 15:70–79. Lindenmayer DB, Lane PW, Westgate MJ, Crane M, Michael D, Okada S, Barton PS. 2014. An empirical assessment of the focal species hypothesis. Conservation Biology 28:1594-1603. Lindenmayer DB, Likens GE. 2011. Direct Measurement Versus Surrogate Indicator Species for Evaluating Environmental Change and Biodiversity Loss. Ecosystems 14:47–59. Lindenmayer DB, Manning AD, Smith PL, Possingham HP, Fischer J, Oliver I, McCarthy MA. 2002. The focal-species approach and landscape restoration: a critique. Conservation Biology 16:338-345. Margules CR, Pressey RL. 2000. Systematic conservation planning. Nature 405:243-253. Moilanen A, Wilson KA, Possingham H. 2009. Spatial conservation prioritization: Quantitative methods and computational tools. Oxford University Press. Nicholson E, Lindenmayer DB, Frank K, Possingham HP. 2013. Testing the focal species approach to making conservation decisions for species persistence. Diversity and Distributions 19:530–540. Pressey RL, Humphries CJ, Margules CR, Vane-Wright RI, Williams PH. 1993. Beyond opportunism: key principles for systematic reserve selection. Trends in Ecology & Evolution 8:124-128. Rodrigues ASL, Brooks TM. 2007. Shortcuts for biodiversity conservation planning: the effectiveness of surrogates. Annual Review of Ecology, Evolution, and Systematics 38:713–737.

43

Tear TH, Kareiva P, Angermeier PL, Comer P, Czech B, Kautz R, Landon L, Mehlman D, Murphy K, Ruckelshaus M. 2005. How much is enough? The recurrent problem of setting measurable objectives in conservation. BioScience 55:835-849. Thomas-Walters L, J Raihani N. 2017. Supporting conservation: the roles of flagship species and identifiable victims. Conservation Letters 10:581–587. Traill LW, Brook BW, Frankham RR, Bradshaw CJA. 2010. Pragmatic population viability targets in a rapidly changing world. Biological Conservation 143:28–34. United States Department of Agriculture (USDA). 2012. Data generated by National Agricultural Statistics Service. https://www.nass.usda.gov. Accessed November 2015. United States Fish and Wildlife Service (USFWS). 2010. Data generated by National Wetlands Inventory. http://www.fws.gov/wetlands/Data/State-Downloads.html. Accessed November 2010. United States Fish and Wildlife Service (USFWS). 2011. Migratory Bird Program. https://www.fws.gov/birds/management/managed-species/focal-species.php. Accessed May 2016. United States Geological Survey (USGS), Gap Analysis Program (2012). Protected areas database of the United States (PADUS). http://gapanalysis.usgs.gov/padus/. Accessed August 2015. Walsh J, Shriver WG, Olsen BJ, O'Brien KM, Kovach AI. 2015. Relationship of phenotypic variation and genetic admixture in the Saltmarsh-Nelson's sparrow hybrid zone. The Auk 132:704-716. Watts ME, Possingham HP. 2013. Marxan.net: Cloud infrastructure for systematic conservation planning. URL http://marxan.net Wiest WA. 2015. Tidal marsh bird conservation in the Northeast USA. PhD dissertation. http://gradworks.umi.com/37/30/3730240.html University of Delaware, Newark. Wiest WA, Correll MD, Olsen BJ, Elphick CS, Hodgman TP, Curson DR, Shriver WG. 2016. Population estimates for tidal marsh birds of high conservation concern in the northeastern USA from a design-based survey. The Condor 118:274-288.

44

Table 1. Mean (bottom-left triangle) and maximum (upper-right triangle) of ten Spearman rank correlation coefficients for association between site prioritizations of percent based targets (10 - 100% of current population) that differ in the target (species) identity but not the target amount. Saltmarsh Sparrow + Clapper Rail + Saltmarsh Seaside Nelson's Target Clapper Rail Eastern Willet Nelson's Sparrow + Sparrow Sparrow Sparrow Seaside Sparrow + Eastern Willet

Clapper Rail 0.202 (50%)† 0.383 (100%) 0.093 (10%) 0.535 (20%) 0.497 (10%)

Saltmarsh Sparrow 0.192* 0.304 (80%) -0.010 (30%) 0.438 (100%) 0.759 (90%)

Seaside Sparrow 0.359 0.295 0.027 (50%) 0.427 (100%) 0.579 (10%)

Nelson's Sparrow 0.072 -0.026 0.011 0.005 (20%) 0.276 (10%)

Eastern Willet 0.511 0.415 0.382 -0.048 0.657 (80%)

Saltmarsh Sparrow + Clapper Rail + Nelson's Sparrow + Seaside 0.411 0.719 0.438 0.261 0.642 Sparrow + Eastern Willet

*Associations are based on selection frequency of Marxan output with 10,000 runs and 1 million iterations.

† Value in parentheses indicates species target amount (10 - 100% of current population) where highest correlation coefficient occurred.

45

Figure 1. Delineation of 8 tidal marsh sub-regions in northeastern North America (after Wiest et al. 2016). Location of the study region in the United States is in the upper left corner. Table headings identify: Patches (the total number of patches), Cost (mean cost per hectare in US dollars) and Area (total area in hectares) for each sub-region.

46

Figure 2. Level of protection for each of five management scenarios, with two single species scenarios (SALS = Saltmarsh Sparrow; CLRA = Clapper Rail) and three multi-species combinations (SALS + CLRA = Saltmarsh Sparrow and Clapper Rail; 4TMB = Saltmarsh Sparrow, Clapper Rail, Eastern Willet (WILL), and Seaside Sparrow (SESP); and 5TMB = Saltmarsh Sparrow, Clapper Rail, Eastern Willet, Seaside Sparrow, and Acadian Nelson’s Sparrow (NESP). In each scenario, Marxan was used to ensure representation of 50,000 individuals of each target species, except for NESP for which the goal was set at 99.9% of current population size. For each protection scenario, we present mean and 95% confidence interval of (A) the number of individuals, (B) saltmarsh area, and (C) the number of saltmarsh patches protected, as well as (D) the minimum cost index (see text and supporting information regarding interpretation of cost) based on 100 runs in Marxan using 10 million simulated annealing iterations, repeated 10 times.

47

Figure 3. Relative importance of marsh area in each sub-region for conserving each of five tidal marsh birds (SALS: Saltmarsh Sparrow; CLRA: Clapper Rail; NESP: Acadian Nelson’s Sparrow; SESP: Seaside Sparrow; and WILL: Eastern Willet) at three population targets (A = 90% of the current population; B = 50% of the current population; C = 10% of the current population). Percent of area in each sub-region identified as: Always selected (patches occurred in solutions for all 10,000 Marxan software runs – black), Selected (patches occurred in solutions for 1 - 9,999 Marxan software runs – dark gray), and Not Selected (patches did not occur in a solution for any of 10,000 Marxan software runs – light gray).

48

Figure 4. Number of selected (black) and unselected (light gray) saltmarsh patches in each of 8 sub- regions (mapped in Fig.1). Patches were classified as unselected if they were never included in any solution from 10,000 Marxan runs of 6 scenarios designed to protect 100% of the current population of each of 5 tidal marsh specialists individually or all 5 species combined.

49

Supporting Information: Appendix 1

Sensitivity Analyses Prior to using Marxan (version 2.43; Ball et al. 2009) to identify priority areas for tidal marsh bird conservation we conducted a suite of sensitivity analyses that explore the effects of: 1) quantifying conservation targets with a range of percent based (10% increments of current population sizes) and minimum population sizes (10,000, 25,000, 50,000 individuals), 2) comparing scenarios that add to existing protected areas with those that do not, and 3) changing the computational effort. We used a simulated annealing algorithm followed by a two-step iterative improvement algorithm for all analyses. An optional parameter, the boundary length modifier, promotes selection of contiguous planning units by attempting to reduce the total boundary length of the full reserve network. We set this parameter to zero because we had no a priori reason to favor selection of spatially clustered sites. Consequently, the likelihood of any planning unit being selected was only a function of its cost and the targets it contained. We constrained all analyses so each species had to be represented by at least the minimum population target in all runs. To ensure we met our targets, we incorporated a species penalty factor to make potential solutions that did not meet targets costlier than solutions for which targets were represented. We set the species penalty factor to 100 for individual species scenarios and 1000 for Saltmarsh Sparrow in all combined species analysis. These penalty factors were the lowest possible values that allowed Marxan flexibility to find an efficient solution while ensuring targets were met in full in all runs for all species (Ball et al. 2009; Andron et al. 2010).

1) Conservation Targets We prioritized sites based on four potential management scenarios that represented species at fixed population-size targets of 10,000, 25,000 or 50,000 individuals. We selected scenarios designed to protect: 1) Saltmarsh Sparrow, 2) Clapper Rail, 3) Saltmarsh Sparrow + Clapper Rail, and 4) Saltmarsh Sparrow + Clapper Rail + Eastern Willet + Seaside Sparrow at each fixed-population target. In addition, we prioritized sites based on each of the four species combinations using percent-based targets set at 10% increments of the current population of each species. We ran each scenario 10,000 times with 1 million simulated annealing iterations followed by two-step iterative improvement to quantify the selection frequency of sites. We used Spearman rank correlations and the selection frequency from each scenario to determine if scenarios with percent-based targets ranked sites in a similar way to scenarios that used population size targets.

Results Strong associations existed between site prioritizations based on percent-based targets and those identified when using fixed population-size targets (Table S1). Relationships were similar regardless of the number or identity of target species or the size of the population target.

50

Table S1. Mean of ten Spearman rank correlation coefficients for association between percent based (10-100%) site prioritizations and each of three fixed population size based site prioritizations (50,000, 25,000 and 10,000) with the same target identity.*

Target 50,000 25,000 10,000

0.982 0.981 0.977 Saltmarsh Sparrow 0.974 0.970 0.968 Clapper Rail 0.948 0.978 0.971 Saltmarsh Sparrow + Clapper Rail

Saltmarsh Sparrow + Clapper Rail + 0.856 0.952 0.928 Seaside Sparrow + Eastern Willet *Associations are based on selection frequency of Marxan output with 10,000 runs and 1 million iterations.

2) Protected Areas We used the Protected Areas Database (USGS, Gap Analysis Program 2012; GAP) and our saltmarsh patch layer (Wiest et al. 2016) to identify the overlap between existing protected areas and saltmarsh patches in the region. We considered only areas with permanent protection from conversion of natural land cover and a mandated management plan in operation to maintain a primarily natural state (i.e., GAP status 1 or 2). Areas with a GAP status of 3 or 4 were not included as they may be subject to extractive uses. Protected area boundaries and saltmarsh patch boundaries rarely correspond so the percent of each saltmarsh patch protected was quantified. Marxan includes the option to force planning units into or out of a solution a priori based on information available to the user. We used Marxan to evaluate (a) differences in spatial prioritization of sites when requiring scenarios to include patches containing areas that were already protected versus not, and (b) the relative importance of current protected areas. We evaluated scenarios ranging from forced inclusion of patches with at least 25% of their area protected to only those that are 100% protected. This is an important consideration because adding area to existing protected areas potentially has fewer legal, logistical and financial constraints than establishing and managing new protected areas. All analyses used a simulated annealing algorithm followed by a two-step iterative improvement with 1 million iterations and repeated 10,000 times. We used two strategies to evaluate separate considerations for how current protection may influence conservation decisions. First, we compared scenarios that require the algorithm to add to existing protected areas with those that have no constraints and only seek to minimize cost). We used Spearman rank correlations to quantify associations between the selection frequencies of a single optimal scenario that had no required patches and nine different scenarios that required patches with protection ranging from 100% to 25% to be included in the final protected area network. We evaluated this relationship with the same four species combinations described above to represent management possibilities for tidal marsh bird conservation with fixed population-size targets of 10,000, 25,000 and 50,000 individuals. Second, we evaluated the importance of current protected areas as a function of the population target or the target identity. To do this, we initially identified saltmarsh patches that were selected in the optimal solutions of (a) 10,000 runs (always selected), (b) 1 – 9999 runs (selected), or (c) zero runs

51

(never selected). Then we determined the proportion of the currently protected area in each of these three groups of patches. We evaluated the importance of protected areas for each of the four species individually, and for all species simultaneously. In each case, we evaluated the numeric population targets at 10% increments spanning the current population size.

Results Generally, target population level had more influence than target species’ identity in determining how much a solution differed from the optimal protected area network, when patches with protection were forced into it. Scenarios that required patches with protection to be included were highly similar to those that did not when population targets were high (e.g., 50,000; Figure S1), regardless of the mix of species targeted. Similarity declined, however, as patches with a smaller proportion of their area protected were forced into the solution, especially for lower population targets (e.g., 10,000 individuals; Figure S1). Frequently, there was a clear threshold beyond which, the number of patches forced into the solution caused a substantial divergence from the optimal solution. For low population targets, patches with protection ranging from 25-70% were sufficient to capture most species targets, but the selected patches were not the most cost efficient network (Table S2).

Figure S1. Spearman rank correlation coefficients between the optimal scenario that does not force any patches into the final solution and scenarios (A – D) that require inclusion of patches with at least a certain level of protection. The required level of protection was set at several levels ranging from 100% to 25%. All scenarios were assessed at 3 population target levels (50,000; 25,000; 10,000). Poor correlation suggests that forcing patches with existing protection into the final solution, weakens the solution. A = Saltmarsh Sparrow; B = Clapper Rail; C = Saltmarsh Sparrow + Clapper Rail; D = Saltmarsh Sparrow + Clapper Rail + Seaside Sparrow + Eastern Willet.

52

Table S2. Example of relative costs of conservation for 4 scenarios (indicated by column headings) at three population targets (50,000; 25,000; 10,000) that differ as a function of percent of protected area that is required for a patch to be forced into the Marxan solutions. The % protected refers to the percent of saltmarsh patch area with protection status of GAP 1 or 2 in the Protected Areas Database (U.S. Geological Survey, Gap Analysis Program 2012; GAP). NA indicates that current protected area was ignored and no patches were required to be represented in solutions. Costs are for comparative purposes and do not reflect the minimum cost possible (see Table S4). Shading indicates situations for which no additional patches without protection are needed to meet the minimum number of individuals for a target (i.e., target can be achieved just by purchasing unprotected portions of partially-protected patches). Saltmarsh Sparrow Saltmarsh Saltmarsh Sparrow + Clapper Rail % Protected Clapper Rail Sparrow + Clapper Rail + Seaside Sparrow + Eastern Willet Population Target: 50,000 NA $1,638,187,070 $285,510,065 $1,635,140,889 $1,646,869,725 100% $1,621,355,836 $285,753,571 $1,646,954,278 $1,640,268,901 90% $1,591,771,117 $294,075,035 $1,611,850,052 $1,616,014,462 80% $1,631,993,352 $307,706,226 $1,635,376,675 $1,628,318,249 70% $1,628,262,995 $432,178,561 $1,627,136,954 $1,636,533,405 60% $1,639,271,285 $509,754,840 $1,650,162,143 $1,630,065,255 50% $1,678,145,536 $572,630,411 $1,657,795,148 $1,682,466,086 40% $1,745,529,860 $723,807,881 $1,740,331,179 $1,747,327,682 30% $1,839,139,774 $993,891,162 $1,832,217,912 $1,838,483,384 25% $1,928,632,280 $1,173,206,300 $1,929,120,229 $1,927,150,540 Population Target: 25,000 NA $281,816,958 $127,120,779 $334,256,819 $334,269,487 100% $284,473,016 $126,553,408 $332,494,404 $337,628,633 90% $291,214,990 $136,951,288 $344,343,529 $343,178,237 80% $324,769,164 $173,122,173 $364,715,648 $367,192,458 70% $364,458,697 $298,671,499 $402,909,690 $404,403,688 60% $454,577,098 $378,711,788 $453,359,619 $454,769,812 50% $520,793,034 $453,258,471 $521,671,462 $520,562,631 40% $676,115,618 $628,967,856 $674,710,317 $675,560,979 30% $935,326,558 $930,003,257 $935,449,155 $935,233,149 25% $1,111,371,539 $1,111,371,539 $1,111,371,539 $1,111,371,539 Population Target: 10,000 NA $54,178,096 $56,998,310 $82,384,725 $87,477,141 100% $55,033,754 $57,192,852 $86,277,578 $82,728,956 90% $57,422,758 $68,357,947 $82,801,823 $80,180,956 80% $98,173,960 $93,568,724 $100,975,568 $100,663,234 70% $235,139,397 $235,139,397 $235,139,397 $235,139,397 60% $370,153,330 $370,153,330 $370,153,330 $370,153,330 50% $453,258,471 $453,258,471 $453,258,471 $453,258,471 40% $628,967,856 $628,967,856 $628,967,856 $628,967,856 30% $930,003,257 $930,003,257 $930,003,257 $930,003,257 0% $1,111,371,539 $1,111,371,539 $1,111,371,539 $1,111,371,539 53

3) Simulated Annealing and Number of Software Runs We evaluated sensitivity to software using the same four species combinations described above to represent different management strategies for tidal marsh bird conservation. We used fixed population-size targets of 25,000 or 50,000 individuals to represent the most realistic conservation targets. We evaluated the number of simulated annealing algorithm iterations (100,000/10,000,000/100,000,000) and the number of software runs (100/1,000/10,000) at three orders of magnitude that are representative of published accounts using Marxan. To quantify variation as a consequence of the number of runs or iterations, we executed each scenario 10 times and calculated the minimum, mean and 95% confidence intervals for the lowest cost solution. We used Spearman rank correlations to estimate the degree of association in the mean site selection frequency. We examined associations between scenarios that differed in the number of iterations to determine whether those with few iterations prioritize sites in a similar fashion to those with many iterations. We also quantified similarity (1 - Bray-Curtis dissimilarity index) among repeated software executions of scenarios with the same number of iterations to determine if replication of an analysis would produce similar results. Finally, we compared similarity for solutions obtained with different iteration frequencies to determine how variable potential reserve networks may be as a consequence of the number of iterations used to identify them.

Results The number of simulated annealing iterations strongly influenced the cost estimate for optimal reserve systems. Moreover, the strength of this relationship differed among conservation scenarios, although there were generally diminishing returns after a million iterations (Figure S2). Given the geographic scope of our analysis and the high land costs in our study area, even small percentage improvements could translate into large financial benefits. For example, increasing simulations from 1 million to 10 million iterations reduces the cost estimates by millions of dollars, regardless of target species, when using target population sizes of 50,000 individuals (Table S3). In contrast, the number of software runs had little influence on cost estimates. Changing this parameter resulted in less than 2.5% difference in the cost of an optimal solution, regardless of the number or identity of conservation targets and their respective values (Table S4). In general, increasing the number of simulated annealing iterations also increased the similarity among the relative frequencies with which sites are prioritized in a reserve network, although similarity was generally high (> 93%) regardless of the number of iterations assessed (Table S5). Whereas similarity among scenarios with the same number of iterations was high, increasing the number of iterations caused prioritizations to diverge, suggesting that site prioritizations based on few iterations are not equivalent to those based on many (Table S6).

54

Figure S2. Savings achieved by increasing the number of simulated annealing iterations for 4 scenarios (SALS = Saltmarsh Sparrow; CLRA = Clapper Rail; SALS + CLRA = Saltmarsh Sparrow + Clapper Rail; SALS + CLRA + SESP + WILL = Saltmarsh Sparrow + Clapper Rail + Seaside Sparrow + Eastern Willet) at two population targets (black bars: 50,000; gray bars: 25,000). Percent decrease in minimum cost was calculated relative to the equivalent scenario with one order of magnitude fewer iterations. The minimum cost for each scenario was estimated as the lowest observed cost from 10 Marxan scenarios of 100 runs that incorporated the specified number of iterations.

55

Table S3. Percent change, minimum and mean cost to protect saltmarsh specialist birds under 4 species target scenarios, each with two population targets (50,000 and 25,000 individuals) as a function of the number of simulated annealing iterations for Marxan scenarios. Mean cost and 95% confidence interval are based on executing a scenario in Marxan 10 times that incorporates 100 runs and simulated annealing iterations that range from 100 thousand to 100 million. Percent change in minimum cost was calculated relative to the equivalent scenario with one order of magnitude fewer iterations. The minimum cost for each scenario was estimated as the lowest observed cost from 10 Marxan scenarios of 100 runs that incorporated the specified number of iterations. Note that costs are based on county-level asset values (USD/ha) of agricultural land including buildings (USDA 2012), rather than actual cost of saltmarsh purchasing, but are assumed to reflect relative costs across the region and provide a good index for comparison. 50,000 25,000 Iterations % Change Minimum Cost Mean Cost +/- 95% CI % Change Minimum Cost Mean Cost +/- 95% CI Saltmarsh Sparrow 100,000 - $1,768,529,827 $1,882,217,116 $32,278,284 - $444,877,243 $499,052,279 $21,393,701 1,000,000 8.05% $1,626,112,373 $1,704,094,995 $37,494,779 35.32% $287,748,611 $292,958,173 $2,409,825 10,000,000 2.66% $1,582,806,227 $1,596,630,242 $6,869,583 6.88% $267,944,228 $270,461,363 $1,212,637 100,000,000 2.33% $1,545,988,942 $1,549,224,805 $1,802,522 3.01% $259,875,990 $261,450,081 $469,261 Clapper Rail 100,000 - $311,253,673 $375,289,464 $28,584,862 - $129,618,455 $134,400,831 $2,098,878 1,000,000 7.55% $287,751,145 $292,221,852 $2,026,137 1.38% $127,828,343 $142,165,496 $8,772,878 10,000,000 1.21% $284,257,150 $285,462,130 $508,081 0.57% $127,102,414 $137,881,015 $6,400,422 100,000,000 0.39% $283,148,271 $283,837,488 $312,750 0.22% $126,816,494 $138,263,151 $5,924,417 Saltmarsh Sparrow + Clapper Rail 100,000 - $1,886,973,770 $1,948,258,132 $17,959,827 - $494,516,152 $572,194,957 $28,830,043 1,000,000 9.35% $1,710,519,023 $1,726,860,060 $5,448,203 30.75% $342,427,767 $349,692,858 $2,521,198 10,000,000 7.28% $1,586,054,532 $1,610,761,524 $12,151,810 7.70% $316,072,166 $321,117,153 $1,991,255 100,000,000 2.56% $1,545,478,364 $1,549,609,443 $1,701,973 2.42% $308,409,680 $309,887,428 $681,029 Saltmarsh Sparrow + Clapper Rail + Seaside Sparrow + Eastern Willet 100,000 - $1,848,600,017 $1,932,633,218 $38,679,310 - $338,917,790 $350,870,163 $4,614,850 1,000,000 10.56% $1,653,425,717 $1,694,487,460 $17,538,923 -1.00% $342,318,939 $351,618,564 $3,264,881 10,000,000 4.19% $1,584,198,578 $1,601,739,759 $9,283,078 7.08% $318,069,503 $319,880,541 $731,559 100,000,000 2.87% $1,538,678,384 $1,551,909,249 $3,453,807 2.93% $308,759,970 $309,977,942 $781,908 56

Table S4. Percent change, minimum and mean cost to protect saltmarsh specialist birds under 4 species target scenarios, each with two population targets (50,000 and 25,000 individuals) as a function of the number of repeated runs for a Marxan scenario. Mean cost and 95% confidence interval are based on executing a scenario in Marxan 10 times that incorporates 1,000,000 iterations of the simulated annealing algorithm and a number of repeated runs that range from 100 to 10,000. Percent change in minimum cost was calculated relative to the equivalent scenario with one order of magnitude fewer runs. The minimum cost for each scenario was estimated as the lowest observed cost from 10 Marxan scenarios of 100 runs that incorporated the specified number of iterations. Note that costs are based on county-level asset values (USD/ha) of agricultural land including buildings (USDA 2012), rather than actual cost of saltmarsh purchasing, but are assumed to reflect relative costs across the region and provide a good index for comparison.

57

Table S5. Mean percent similarity (1 - Bray-Curtis Index) and 95% confidence interval between the selection frequency of patches as a function of the number of simulated annealing iterations for Marxan scenarios that differ in the species and population target (50,000 or 25,000) of interest.*

Saltmarsh Sparrow + Saltmarsh Sparrow + Clapper Rail Iterations Saltmarsh Sparrow Clapper Rail Clapper Rail + Seaside Sparrow + Eastern Willet

Population Target: 50,000 100,000 94.63 ± 0.03 93.51 ± 0.03 94.42 ± 0.02 94.42 ± 0.07 1,000,000 94.90 ± 0.02 94.52 ± 0.03 94.84 ± 0.02 94.79 ± 0.02 10,000,000 96.36 ± 0.01 95.76 ± 0.04 96.26 ± 0.02 96.25 ± 0.02 100,000,000 97.09 ± 0.02 96.07 ± 0.04 97.03 ± 0.02 97.01 ± 0.02 Population Target: 25,000 100,000 93.56 ± 0.02 92.93 ± 0.03 93.05 ± 0.04 94.35 ± 0.06 1,000,000 94.81 ± 0.02 94.66 ± 0.03 94.56 ± 0.02 94.28 ± 0.06 10,000,000 96.00 ± 0.02 95.68 ± 0.04 95.81 ± 0.02 95.49 ± 0.09 100,000,000 96.96 ± 0.02 95.97 ± 0.04 96.75 ± 0.04 96.56 ± 0.05 *Mean similarity and 95% confidence interval are based on all possible pairwise comparisons (n = 45) of 10 Marxan scenarios that consist of 100 runs and simulated annealing iterations that range from 100 thousand to 100 million.

58

Table S6. Matrix of Spearman rank correlations (rho) for association among selection frequency of patches in scenarios that differ in the number of simulated annealing iterations.*

50,000 individuals 25,000 individuals Iterations 100,000 1,000,000 10,000,000 100,000 1,000,000 10,000,000 Saltmarsh Sparrow 1,000,000 0.967 ─ ─ 0.968 ─ ─ 10,000,000 0.915 0.950 ─ 0.918 0.948 ─ 100,000,000 0.888 0.923 0.970 0.880 0.909 0.958 Clapper Rail 1,000,000 0.929 ─ ─ 0.924 ─ ─ 10,000,000 0.769 0.827 ─ 0.746 0.810 ─ 100,000,000 0.675 0.725 0.876 0.672 0.730 0.897 Saltmarsh Sparrow + Clapper Rail 1,000,000 0.936 ─ ─ 0.943 ─ ─ 10,000,000 0.880 0.942 ─ 0.882 0.939 ─ 100,000,000 0.852 0.913 0.965 0.827 0.887 0.943 Saltmarsh Sparrow + Clapper Rail + Seaside Sparrow + Eastern Willet 1,000,000 0.938 ─ ─ 0.986 ─ ─ 10,000,000 0.880 0.944 ─ 0.941 0.940 ─ 100,000,000 0.852 0.914 0.968 0.892 0.892 0.949 *Comparisons are based on the mean frequency with which patches were selected in 10 Marxan outputs that incorporate 100 runs and the number of iterations indicated.

Literature Cited Ardron JA, Possingham HP, Klein CJ. 2008. Marxan good practices handbook. Pacific Marine Analysis and Research Association, Vancouver. Ball IR, Possingham HP, Watts M. 2009. Marxan and relatives: software for spatial conservation prioritisation. Spatial conservation prioritisation: quantitative methods and computational tools. Oxford University Press, Oxford. United States Department of Agriculture (USDA). 2012. Data generated by National Agricultural Statistics Service. https://www.nass.usda.gov. Accessed November 2015. United States Geological Survey (USGS), Gap Analysis Program (2012). Protected areas database of the United States (PADUS). http://gapanalysis.usgs.gov/padus/. Wiest WA. 2015. Tidal marsh bird conservation in the Northeast USA. PhD Dissertation, University of Delaware, Newark.

59

V. Characterizing uncertainty to inform tidal marsh conservation in response to sea-level rise

Brian T. Klingbeil, Jonathan B. Cohen, Maureen D. Correll, Christopher R. Field, Thomas P. Hodgman, Adrienne I. Kovach, Brian J. Olsen, W. Gregory Shriver, Whitney A. Wiest, and Chris S. Elphick

Although the consequences of climate change are broadly understood, there is much uncertainty as to exactly how conditions will change. This uncertainty represents one of the major challenges to the development of effective adaptation strategies. Conservation planning, for example, would ideally prioritize not only land that is important now, but also land that will remain or become important in the future. Accounting for uncertainty in planning exercises is thus a critical problem that conservation practitioners must overcome. Coastal environments are among the most economically important yet vulnerable ecosystems (Barbier et al. 2011; Arkema et al. 2013), covering just 9% of the Earth’s surface but supporting over 25% of humans (Kummu et al. 2016). People have long relied on coastal ecosystems for highly productive farmland and easy access to the sea for food, travel and commerce. Coastal wetlands sequester carbon, filter pollutants, buffer against storms, support fisheries, and provide recreational opportunities (Gedan et al. 2009). Despite their importance, at least 25% of the world’s coastal wetlands have been lost through conversion for human use (McLeod et al. 2011) and much of the remaining area is vulnerable to sea-level rise. Sea-level rise is projected to accelerate dramatically with profound effects on coastal regions, though much uncertainty regarding the magnitude of future rates remains (McGranahan et al. 2007; Nicholls and Cazenave 2010; Sallenger et al. 2012; Kirwan and Megonigal, 2013). Tidal marshes are dynamic; already vulnerable due to their frequent proximity to human development, their persistence in the face of sea-level rise depends on a suite of biological and physical processes (French 2006; Kirwan and Murray 2007). Historically, these marshes have been able to avoid inundation or transition to open water or mud flats by building vertically at rates similar to or exceeding, sea-level rise (Cahoon et al. 2006; French 2006) or by migrating inland (Smith 2013). Historical responses to sea-level rise are an imperfect model for the future, however, because factors such as climate, water quality, sediment delivery rates, primary productivity, and space for marsh migration continue to change with human activity (Parris et al. 2012; Kirwan and Megonigal 2013). Eastern North America contains over one-third of the world’s tidal marsh. Northeastern USA tidal marshes experience roughly twice the average global sea-level rise (Sallenger et al. 2012), with even higher rates in the past decade (Goddard et al. 2015). These marshes support the highest level of vertebrate endemism of any tidal marsh region worldwide (Greenberg and Maldonado 2006) and provide many additional species, particularly birds, with important habitat throughout the year (Correll et al. 2016). Uncertainty remains a major hurdle to conservation planning for sea-level rise (Nicholls & Cazenave 2010) and can stem from differences between carbon emissions scenarios (e.g. representative concentration pathways; van Vuuren et al. 2011), or differences in model assumptions (e.g., static vs. dynamic process-based models; Hawkins and Sutton 2008; Kirwan et al. 2016). Static inundation models have dominated larger-scale assessments of marsh vulnerability (e.g. Nicholls 2004; Cooper et al. 2008), but assume that marshes do not respond to increased rates of sea-level rise through biophysical feedback that accelerates soil building, or through landward marsh migration. Most static models predict significant or catastrophic marsh losses this century (e.g., Cooper et al. 2008; Craft et al. 2009), but there are concerns that they underestimate marsh resilience (e.g., Kirwan et al. 2016). Dynamic

60 models generally have been restricted to smaller spatial scales because they incorporate biophysical feedbacks that may strongly depend on local conditions (Passeri et al. 2015) and require data derived from field measurements (e.g., Morris et al. 2002; Kirwan and Guntenspergen 2010). They generally predict smaller marsh losses than static models, but may overestimate marsh resilience to sea level rise (Parkinson et al. 2017). To assess the long-term consequences of climate change for high priority conservation targets in coastal marshes, we compared the effects of sea-level rise predicted under both static and dynamic models that use multiple emissions scenarios. We quantified the likely effects on the populations of five high priority conservation targets under each scenario, determined the temporal and spatial patterns of population change, and examined how uncertainty affects the practical planning decisions that should be made for the target species.

Methods We evaluated the effects of sea-level rise on tidal marsh area in the Northeastern USA using results from a decision support tool designed to guide planning activities (https://woodshole.er.usgs.gov/project- pages/coastal_response/data.html; Lentz et al. 2015). The tool predicts land elevation with respect to projected mean high water in 2030 and 2050 based on sea-level rise estimates from the Intergovernmental Panel on Climate Change’s (IPCC) representative concentration pathways (RCP) scenarios 4.5 (radiative forcing trajectories stabilize shortly after 2100) and 8.5 (radiative forcing trajectories continue to increase over time) (Stocker et al. 2014), vertical land movement rates, and current elevation data (Lentz et al. 2015). This data layer is similar to a static inundation model that ignores potential dynamic capabilities of the landscape, but differs from many such models in that it integrates uncertainty associated with all inputs, including multiple RCP scenarios, rather than developing separate estimates for each scenario. The probability associated with the predicted marsh elevation at each location is provided as a separate layer to quantify uncertainty (Lentz et al. 2015). Finally, the tool provides a third data layer that describes the likelihood that the coastal landscape will respond dynamically to sea level rise and remain above water (Lentz et al. 2015). We reclassified the adjusted elevation layer (hereafter ‘static layer’) for each target date from the initial five categories into a binary raster in which pixels are predicted to be above or below sea level given no dynamic response. We also used a reclassified layer describing the ability of marshes to respond dynamically (hereafter ‘dynamic layer’), in which a continuous probability of avoiding inundation was converted to a categorical classification that reflects the IPCC scenario ranges (Mastrandrea et al. 2010): 1 - unlikely (<33%), 2 - as likely as not (33-66%), 3 - likely (66%-90%), and 4 - very likely (>90%). We delimited saltmarsh patches between Maine and Virginia by creating a 50-m buffer around all Estuarine Intertidal Emergent Wetland polygons (Cowardin et al. 1979) identified in the National Wetlands Inventory (USFWS 2010) following Wiest et al. (2016). Polygons with buffers that intersected were merged. The 100-m distance between patches was based on home range size and movement estimates for Nelson’s and Saltmarsh Sparrows (Shriver et al. 2010), two priority species for saltmarsh conservation in the region. We extracted estimates of adjusted elevation and likelihood of a dynamic response to sea level rise by 2030 and 2050 using the extract by mask tool in ArcGIS 10.4 (ESRI 2016) with the saltmarsh patch layer (Wiest et al. 2016) as a mask. This procedure provided static and dynamic inundation predictions for areas corresponding to 8,129 (97%) of saltmarsh patches for each decade for which sea level predictions were available. We classified the remaining 276 patches (< 0.08% of the original saltmarsh area) that lacked clear relationships to data in the sea-level rise layers due to differences in land cover classifications, as having insufficient data.

61

We estimated the percent of salt marsh area remaining by each future date using two approaches that differ in how they address the potential loss of information when using irregularly shaped polygons to extract data from a raster comprising square pixels. First, we assumed that the values of pixels extracted from the raster layers were representative of the polygon as a whole. For each raster, we multiplied the proportion of pixels in each class by the original area of the saltmarsh patch. For the static model, this resulted in an estimate of the area of saltmarsh expected to be submerged or unflooded. For the dynamic model, we obtained an estimate of the area of each patch with a given chance of remaining as marsh based on the IPCC likelihood categories. Second, we used only the area of the pixels extracted by the mask to estimate the proportion of the entire area that is predicted to fall into a given category. Both approaches produced similar results and we present only those based on the first here. To evaluate conservation consequences under different scenarios we took estimates of the current abundance of five bird species that are the focus of conservation efforts – Clapper Rail (Rallus crepitans), Eastern Willet (Tringa semipalmata semipalmata), Acadian Nelson’s Sparrow (Ammodramus nelsonii subvirgatus), Saltmarsh Sparrow (Ammodramus caudacutus), and Seaside Sparrow (Ammodramus maritimus) – within each of the 8,405 saltmarsh patches (Wiest 2015; Wiest et al. 2016). We predicted future population sizes for each species under the static model by multiplying the original density estimates for each patch by the proportion of the patch’s original area that is expected to remain above sea level by a given year, and then summing across patches. To estimate change assuming dynamic responses to sea-level rise, we prorated the original density estimates for each patch based on the proportion of the original patch area classified as falling within each IPCC likelihood category.

Results Regardless of whether we assume a static inundation model or a dynamic response to sea level rise, dramatic declines in the area of marsh are predicted (Table 1). The static model predicted almost half the saltmarsh area in the northeastern USA will be lost by 2030 and only ~10% will avoid inundation by 2050 (Table 1). Uncertainty in the likelihood of a dynamic response of salt marsh to sea level rise was high and differed little between 2030 and 2050. Despite differences in model specifications, our best estimates suggest that only ~15% of the current area is likely or highly likely to adapt to sea-level rise (Table 1). Substantial reductions in the population sizes of all five tidal marsh specialist birds are consequently predicted, under both static and dynamic models (Figure 2). Most species are projected to undergo losses of >85% under both scenarios, with Acadian Nelson’s Sparrow faring only slightly better. Models differ slightly in the severity of predicted losses, with the dynamic model showing somewhat more optimistic projections for all species except Acadian Nelson’s Sparrow. Differences, however, are small, ranging from 2.7% - 6.0%. Models differed in the spatial distribution of salt marsh predicted to be lost in the Northeast by 2050 (Figure 3). Fewer than 10% of saltmarsh patches shared similar predictions for the entire patch area (i.e. 328 patches predicted to be completely inundated or not likely to respond dynamically, and 300 patches predicted to remain above sea level or likely to respond dynamically). In general, the dynamic model suggests that salt marshes from coastal New Jersey south to the Eastern Chesapeake Bay are likely to retain a greater percentage of their current extent than is predicted by the static model. In contrast, the static model suggests that coastal Maine, southern New England and Long Island will fare better in response to rising sea levels than the dynamic model. Between two-thirds and >90% of current area is projected to be lost in all regions by both models, however.

62

Discussion Sea-level rise predictions are sensitive to underlying climate change scenarios, model specifications, and temporal scale (Hawkins and Sutton 2008; Stocker et al. 2014), all of which generate uncertainty in future projections. Potential changes, however, are sufficiently severe that it is important that the uncertainty not prevent action, making it necessary that we identify conservation and management decisions with a high chance of success across a range of future scenarios. The static and dynamic models examined here differed markedly in their predictions for 2030, but both predicted with reasonable certainty that as little as 10-15% of the current tidal marsh in the northeastern US will remain by 2050 (Lentz et al. 2016; Table 1). This predicted loss is expected to have devastating effects on the species that rely on this unique ecosystem, with over 75% of the current population of each specialist marsh bird species lost from the region by 2050, regardless of model. In the face of potentially catastrophic losses to tidal marsh in the Northeast, where availability of land to protect is limited and unpredictable (Field et al. 2017), we need to understand the uncertainty associated with all available conservation options. Our models suggest that at least 15% of the regional area, and a similar amount in most sub-regions, is likely to remain potential habitat for tidal marsh birds in 2050. Much of the remaining area may respond dynamically to sea level rise, however, this is much less certain. Given this situation, we propose the following planning approach to address the uncertainty associated with marsh loss due to sea level rise, and suggest that it is likely to be generalizable and scalable to address uncertainty associated with other aspects of climate change:

1. Identify areas of overlap that are predicted to be inundated in static models and unlikely (<33% chance) to adapt based on dynamic models, and exclude those from further consideration. 2. Prioritize remaining patches to maximize representation of conservation benefits relative to costs based on current conditions (see Klingbeil et al. in review). 3. Identify areas of overlap with those predicted both to remain above sea level in static models and likely or very likely (>66% chance) to adapt in dynamic models. Prioritize the best of these sites based on the results from (2) because they are most likely to have value in both the short- and long-term. Sites in this group that are not ranked high in (2) should be evaluated to determine whether their value could be increased via management. 4. Next, prioritize sites that are predicted to be inundated in static models but likely or very likely (>66%) to adapt in dynamic models based on the ranking in (2). 5. Finally, for areas identified as likely as not (33-66% chance) to adapt in dynamic models, prioritize based on (2), and focus work to improve predictions on those sites that are the highest priorities. The best of these sites are potentially also targets for management efforts to improve the chance of long-term persistence.

Climate change predictions are often provided for a range of future dates, whereas management and policy implementation requires information that is expected to be robust to fluctuations over time. This is problematic when projections differ significantly as a function of time, as is likely to be true in the many cases where climate change will cause a progressive shift from one land cover type to another. Prioritizing areas that are both currently important, and considered likely to be stable over time under a variety of scenarios and modelling approaches, is likely to be the best way to identify where holdout populations (sensu Hannah et al. 2014) will persist. Clearly identifying sites that are unlikely to support conservation goals under any circumstance is also likely to help avoid situations where local interests might distort global priorities. Finally, folding information on the likelihood of a dynamic response into planning decisions might help distinguish sites where management is most likely to be effective.

63

Literature Cited Arkema KK, Guannel G, Verutes G, Wood SA, Guerry A, Ruckelshaus M, Kareiva P, Lacayo M, Silver JM. 2013. Coastal habitats shield people and property from sea-level rise and storms. Nature Climate Change 3:913-918. Barbier EB, Hacker SD, Kennedy C, Koch EW, Stier AC, Silliman BR. 2011. The value of estuarine and coastal ecosystem services. Ecological Monographs 81:169–193. Brook BW, Sodhi NS, Bradshaw CJA. 2008. Synergies among extinction drivers under global change. Trends in Ecology & Evolution 23:453–460. Cahoon DR, Hensel PF, Spencer T, Reed DJ, McKee KL, Saintilan N. 2006. Coaastal wetland vulnerability to relative sea-level rise: wetland elevation trends and process controls. Pages 271–292 in Verhoeven JTA, Beltman B, Bobbink R, and Whigham D editors Wetlands and Natural Resource Management: Ecological Studies. Springer-Verlag, Berlin. Cooper MJP, Beevers MD, Oppenheimer M. 2008. The potential impacts of sea level rise on the coastal region of New Jersey, USA. Climatic Change 90:475. Cornell Lab of Ornithology. 2015. The Birds of North America Online. Available from http://bna.birds.cornell.edu/bna/ (accessed August 1, 2016). Correll MD, Wiest WA, Olsen BJ, Shriver WG, Elphick CS, Hodgman TP. 2016. Habitat specialization explains avian persistence in tidal marshes. Ecosphere 7:e01506. Cowardin LM, Carter V, Golet FC, LaRoe ET 1979. Classification of wetlands and deepwater habitats of the United States. US Fish and Wildlife Service. Craft C, Clough J, Ehman J, Joye S, Park R, Pennings S, Guo H, Machmuller M. 2009. Forecasting the effects of accelerated sea-level rise on tidal marsh ecosystem services. Frontiers in Ecology and the Environment 7:73–78. ESRI. 2016. ArcGIS 10.3. Environmental Systems Research Institute, Redlands, CA. Field, CR, Dayer, AA, Elphick, CS. 2017. Landowner behavior can determine the success of conservation strategies for ecosystem migration under sea-level rise. Proceedings of the National Academy of Sciences 114:9134–9139. French J. 2006. Tidal marsh sedimentation and resilience to environmental change: Exploratory modelling of tidal, sea-level and sediment supply forcing in predominantly allochthonous systems. Marine Geology 235:119–136. Gedan KB, Silliman BR, Bertness MD. 2009. Centuries of human-driven change in salt marsh ecosystems. Annual Review of Marine Science 1:117–141. Goddard PB, Yin J, Griffies SM, Zhang S. 2015. An extreme event of sea-level rise along the Northeast coast of North America in 2009–2010. Nature Communications 6:1–9. Greenberg RG, Maldonado JE. 2006. Diversity and endemism in tidal marsh vertebrates. Pages 32–53 in Greenberg R, Droege S, Maldonado J, and McDonald MV editors Terrestrial Vertebrates of Tidal Marshes: Ecology, Evolution, and Conservation. The Cooper Ornithological Society, Camarillo. Hawkins E, Sutton R. 2009. The potential to narrow uncertainty in regional climate predictions. Bulletin of the American Meteorological Society 90:1095–1107. Hannah L, Flint L, Syphard AD, Moritz MA, Buckley LB, McCullough IM. 2014. Fine-grained modeling of species’ response to climate change: holdouts, stepping-stones, and microrefugia. Trends in Ecology and Evolution 29:390-397. Kerr RA. 2011. Vital details of global warming are eluding forecasters. Science 334:173–174. Kirwan ML, Guntenspergen GR. 2010. Influence of tidal range on the stability of coastal marshland. Journal of Geophysical Research: Earth Surface 115:F02009.

64

Kirwan ML, Megonigal JP. 2013. Tidal wetland stability in the face of human impacts and sea-level rise. Nature 504:53–60. Kirwan ML, Murray AB. 2007. A coupled geomorphic and ecological model of tidal marsh evolution. Proceedings of the National Academy of Sciences 104:6118–6122. Kirwan ML, Temmerman S, Skeehan EE, Guntenspergen GR, Fagherazzi S. 2016. Overestimation of marsh vulnerability to sea level rise. Nature Climate Change 6:253. Kummu M, Moel H de, Salvucci G, Viviroli D, Ward PJ, Varis O. 2016. Over the hills and further away from coast: global geospatial patterns of human and environment over the 20th–21st centuries. Environmental Research Letters 11:034010. Lemos MC, Rood RB. 2010. Climate projections and their impact on policy and practice. Wiley Interdisciplinary Reviews: Climate Change 1:670–682. Lentz EE, Stippa SR, Thieler ER, Plant NG, Gesch DB, Horton RM. 2015. Coastal landscape response to sea-level rise assessment for the northeastern United States (ver. 2.0., December 2015): U.S. Geological Survey data release, http://dx.doi.org/10.5066/F73J3B0B. Lentz EE, Thieler ER, Plant NG, Stippa SR, Horton RM, Gesch DB. 2016. Evaluation of dynamic coastal response to sea-level rise modifies inundation likelihood. Nature Climate Change 6:696. Mastrandrea M et al. (2010) Guidance note for lead authors of the IPCC fifth assessment report on consistent treatment of uncertainties. Intergovernmental Panel on Climate Change (IPCC), URL https://www.ipcc-wg1.unibe.ch/guidancepaper/ar5_uncertainty-guidance-note.pdf McGranahan G, Balk D, Anderson B. 2007. The rising tide: assessing the risks of climate change and human settlements in low elevation coastal zones. Environment and Urbanization 19:17–37. Mcleod E, Chmura GL, Bouillon S, Salm R, Björk M, Duarte CM, Lovelock CE, Schlesinger WH, Silliman BR. 2011. A blueprint for blue carbon: toward an improved understanding of the role of vegetated coastal habitats in sequestering CO2. Frontiers in Ecology and the Environment 9:552–560. Morris JT, Sundareshwar PV, Nietch CT, Kjerfve B, Cahoon DR. 2002. Responses of Coastal Wetlands to Rising Sea Level. Ecology 83:2869–2877. Moss RH et al. 2010. The next generation of scenarios for climate change research and assessment. Nature 463:747. Nicholls RJ. 2004. Coastal flooding and wetland loss in the 21st century: changes under the SRES climate and socio-economic scenarios. Global Environmental Change 14:69–86. Nicholls RJ, Cazenave A. 2010. Sea-level rise and its impact on coastal zones. Science 328:1517–1520. Parkinson RW, Craft C, DeLaune RD, Donoghue JF, Kearney M, Meeder JF, Morris J, Turner RE. 2017. Marsh vulnerability to sea-level rise. Nature Climate Change 7:756. Parris AS, Bromirski P, Burkett V, Cayan DR, Culver ME, Hall J, Horton RM, Knuuti K, Moss RH, Obeysekera J, Sallenger AH. 2012. Global sea level rise scenarios for the United States National Climate Assessment. National Oceanic and Atmospheric Administration Technical Memo OAR CPO-1. Passeri DL, Hagen SC, Medeiros SC, Bilskie MV, Alizad K, Wang D. 2015. The dynamic effects of sea level rise on low-gradient coastal landscapes: A review. Earth’s Future 3:159-181. Sallenger Jr AHS, Doran KS, Howd PA. 2012. Hotspot of accelerated sea-level rise on the Atlantic coast of North America. Nature Climate Change 2:884. Shriver WG, Hodgman TP, Gibbs JP, Vickery PD. 2010. Home range sizes and habitat use of Nelson's and Saltmarsh sparrows. The Wilson Journal of Ornithology 122:340-345. Smith JAM. 2013. The role of Phragmites australis in mediating inland salt marsh migration in a mid- Atlantic estuary. PLOS ONE 8:e65091. Smith LA, Stern N. 2011. Uncertainty in science and its role in climate policy. Phil. Trans. R. Soc. A 369:4818–4841.

65

Stocker TF, Qin D, Plattner GK, Tignor MMB, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM. 2014. Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of IPCC the Intergovernmental Panel on Climate Change. Cambridge University Press. Cambridge. Travis JMJ. 2003. Climate change and habitat destruction: a deadly anthropogenic cocktail. Proceedings of the Royal Society B: Biological Sciences 270:467–473. United States Fish and Wildlife Service (USFWS). 2010. Data generated by National Wetlands Inventory. http://www.fws.gov/wetlands/Data/State-Downloads.html. Accessed November 2010. Van Vuuren DP, Edmonds J, Kainuma M, Riahi K, Thomson A, Hibbard K, Hurtt GC, Kram T, Krey V, Lamarque J-F, Masui T, Meinshausen M, Nakicenovic N, Smith SJ, Rose SK. et al. 2011. The representative concentration pathways: an overview. Climatic Change 109:5. Wiest WA. 2015. Tidal marsh bird conservation in the Northeast USA. PhD dissertation. http://gradworks.umi.com/37/30/3730240.html University of Delaware, Newark. Wiest WA, Correll MD, Olsen BJ, Elphick CS, Hodgman TP, Curson DR, Shriver WG. 2016. Population estimates for tidal marsh birds of high conservation concern in the northeastern USA from a design-based survey. The Condor 118:274-288.

Table 1. Percentage of northeastern North American tidal marsh present in 2010 that is predicted to remain above sea level (a.s.l.) with a static model or adapt to sea level rise (SLR) with a dynamic model in 2030 and 2050. See methods for procedure used to estimate area and description of insufficient data. See Lentz et al. (2016) for a detailed description of the models. Model 2030 2050 Static Elevation predicted to be greater than 0 meters a.s.l. 51.05% 10.52% Insufficient data 0.08% 0.08% Dynamic Unlikely to adapt to SLR increases (0% to 33%) 0.06% 0.07% As likely as not to adapt to SLR increases (33% to 66%) 84.88% 85.34% Likely to adapt to SLR increases (66% to 90%) 14.73% 14.28% Very likely to adapt to SLR increases (90% to 100%) 0.25% 0.23% Insufficient data 0.08% 0.08%

66

Figure 1. Delineation of 8 tidal marsh sub-regions in North America (after Wiest et al. 2016). Location of study area in North America indicated by box in top left corner of map.

67

Figure 2. Percent of 2010 tidal marsh area and population estimates for 5 tidal marsh birds predicted to remain in 2050. Estimates derived from static and dynamic models of coastal response to sea level rise.

Figure 3. Percent of 2010 tidal marsh area in each of 8 sub-regions in northeastern North America predicted to remain in 2050. Estimates derived from static and dynamic models of coastal response to sea level rise.

68

VI. Hurricane Sandy coastal resiliency program: restoration project summaries

Baseline elevation, vegetation, and bird data were collected at 237 points within restoration projects and at 153 control points using standard SHARP protocols (https://www.tidalmarshbirds.org/). Through separate funding, another 646 points at restoration sites and 476 at control sites were surveyed in association with resiliency projects funded by the Department of Interior. Overall, six different types of resiliency projects were sampled: living shorelines (9 sites), sediment deposition (15 sites), hydrological alteration (20 sites), marsh migration (1 site), invasive species removal (8 sites), vegetation planting (9 sites), and post removal (1 site) (multiple resiliency activities were involved at some sites). Over 1700 baseline sites also were surveyed in 2010-14 using funding from other sources. These older data have been combined with all data collected since 2014 into a single unified database. Abundance of tidal marsh obligate birds was determined using the standard SHARP point count survey protocol, which uses a 5-minute passive period followed by a broadcast sequence of secretive marsh bird calls. Full details of this protocol are available on-line at http://bit.ly/2oQzb3M and in Wiest et al. (2016). From these point counts, we estimated abundance using data from the passive portion of the surveys and the unmarked package in program R (Fiske and Chandler 2011). For each site we provide three types of vegetation data. We collected information on percent cover of different cover classes and dominant species within 50 m of each bird survey point. Additionally, we estimated the abundance of major vegetation types along 100 m transects that pass through each bird survey point. These data were collected using the standard SHARP vegetation protocol, available on-line at http://bit.ly/2FmKCY4. Finally, we collected real-time kinematic (RTK) elevation data by placing a 25-point square grid over each survey point. Usually grids were 100 x 100 m, with RTK points spaced 20 m apart. When a grid of that size would not fit within a marsh patch, we reduced the grid size to ensure that at least 10 points were surveyed, with a minimum spacing of 5 m between points. For more details, see our sampling protocol, online at http://bit.ly/2oOhD9g, and section II: Measuring elevation in northeastern USA tidal marshes.

In the remaining pages of this report, we provide summary information for all sites investigated as part of this project. Data summaries from other sites are available through other SHARP reports. Restoration projects included in this report are:

 Reusing dredged materials to enhance salt marsh in Ninigret Pond (RI) (NFWF 41739)  Coastal resiliency planning and ecosystem enhancement for Northeastern Massachusetts (NFWF 41766)  Strengthening Sachuest Bay’s coastal resiliency (NFWF 41795)  Strengthening Sunken Meadow State Park’s resiliency (NY) (NFWF 42442)  Restoring Spring Creek Park’s salt marsh and upland habitat (NY) (NFWF 42958)  Rejuvenating Sunset Cove’s salt marsh and upland habitat (NY) (NFWF 42959)  Wetland restoration in Suffolk County (NY) (NFWF 43006)  Reusing dredge material to restore salt marshes and protect communities (NJ) (NFWF 43095)  Restoring Delaware Bay’s wetlands and beaches in Mispillion Harbor Reserve and Milford Neck Conservation Area (NFWF 43281)  Liberty State Park saltmarsh and upland enhancement (NFWF 43290)

69

 Enhancing Wampanoag Tribe of Gay Head’s land resiliency in Martha’s Vineyard (MA) (NFWF 43322)  Creating a resilient Delaware Bay shoreline in Cape May and Cumberland Counties (NJ) (NFWF 43429)  Strengthening Marshes Creek through green and grey infrastructure (NJ) (NFWF 43931)  Repairing infrastructure and designing wetland and beach restoration plans along the Central Delaware Bayshore (NFWF 44157)  Protecting North Beach’s salt marsh and emergency route (MD) (NFWF 44167)  Rocky Neck State Park thin-layer deposition projects  Stratford Point living shoreline project

Each summary includes information on restoration techniques employed, timing of restoration work, maps of restoration and control survey points, and abundance of tidal marsh obligate birds, relative abundance of different vegetation cover types, and elevation data from restoration and control sites. Information about restoration activities and timing was provided by the relevant point-of-contact for each project; in some cases, this information was not provided to us and is thus omitted. Timing of management activities at restoration sites limited our ability to collect post- restoration data because only one project was completed before 2016, and some are still not completed (see site summaries for details). Nonetheless, we have developed a plan via other funding that will ensure continued monitoring of birds, vegetation, and elevation through 2022, thereby taking advantage of the detailed baseline data collected during this project.

Literature cited

Fiske I, Chandler R. 2011. unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1-23. Wiest WA, Correll MD, Olsen BJ, Elphick CS, Hodgman TP, Curson DR, Shriver WG (2016) Population estimates for tidal marsh birds of high conservation concern in the northeastern USA from a design-based survey. Condor: Ornithological Applications 118:274-288.

70