The SLAMM Model (Sea Level Affecting Marshes Model) Jonathan Clough 3-18-2014 1

Warren Pinnacle Consulting, Inc. Founded in 2001 Located in Central VT, Environmental Modeling Experts Jonathan S. Clough, Founder, Environmental Consultant since 1994 Dr. Amy Polaczyk, SLAMM modeler, with us since 2010 Dr. Marco Propato, Accretion modeling expert, joined in 2011 warrenpinnacle.com warrenpinnacle.com/quals.pdf

SLAMM Sea Level Affecting Marshes Model Simulates the dominant processes involved in wetland conversions under different scenarios of sea level rise inundation, erosion, accretion, soil saturation and barrier island overwash Uses a complex decision tree incorporating geometric and qualitative relationships to represent transfers among coastal classes Can provide numerical and map-based output with minimal computational time 2009 2100, 1 m SLR

“Uncertainty Cloud” for Selected Region SLAMM Modest data requirements allow application to many sites at a reasonable cost Integrated stochastic uncertainty and parameter sensitivity analyses Provides a range of possible outcomes and their likelihood Users have included US EPA, USGS, The Nature Conservancy, National Wildlife Federation, and the U.S. Fish & Wildlife Service, among others Calibrated to historic SLR in Louisiana The model has been applied to more than 100 National Wildlife Refuges in the US http://warrenpinnacle.com/prof/SLAMM “Uncertainty Cloud” for Selected Region The Range Between 5th and 95th percentiles is graphed along with mean and deterministic results

Model Development Overview Intermittently under development since 1985, Park & Titus Three Year EPA STAR grant (2005-2008) provided funds for significant model development. Simulations of entire Georgia and South Carolina Coastline Model assumptions closely re-examined by team of scientists Survey of recent literature Model tested using LIDAR data Model results linked to ecosystem services National Wildlife Federation Funded Simulations Florida, Puget Sound, Chesapeake Bay, Louisiana (Glick et al, 2013) USFWS funding of refuge simulations Over 100 refuges completed to date in USFWS Regions 4,5,8 TNC / GOMA funding of Gulf of Mexico Simulations Point out the STAR grant gave us a chance to take a step back and reexamine the model predictions given modern data and the most recent literature. 5

Latest Version SLAMM 6.2 64-bit (parallel processing in-house) Dynamic Accretion Open Source Elevation Analyses with histograms Increased Flexibility in Parameterization. Upgrade of Salinity Component (Bathymetry) Users Manual & Technical Documentation Update 6

Coming Soon – SLAMM 6.3 and 6.4 SLAMM 6.3 – USGS Sponsored Salinity linkages SAV model SLAMM 6.4 – USFWS Sponsored Roads and Infrastructure module 7

Ongoing Work on SLAMM Model Gulf Coast Prairie LCC “Gap Analysis:” filling in all holes in the Gulf of Mexico NY State – Application to Long Island, NY City, Hudson River Examine the effects of DEM processing and “hydro enforcement” All CT coasts USGS:OR – SLAMM 6.3. Linkages created to EPA salinity models. SAV predictions Habitat switching based on salinity, model testing and documentation Ducks Unlimited – Pacific Northwest Application of uncertainty analysis in WA & OR, evaluating land parcels for restoration TNC TX – Examine effects on infrastructure given development and restoration scenarios Dike model refined to assess likelihood of overtopping Alternative green/grey infrastructure design.

Model Process Overview Addresses Six Primary Processes (Inundation, Erosion, Saturation, Overwash, Accretion, Salinity) These have been determined to be the dominant processes. Each of these processes are modeled rather simply, though the accretion model is currently undergoing a transformation into a more complex formulation. Underline the fact that SLAMM is a relatively simple model which provides strengths of relative ease of application and capability to apply the model at large scales. Titus and Wang 2008

Model Process Overview Inundation: Calculated based on the minimum elevation and slope of the cell. Erosion: Triggered given a maximum fetch threshold and proximity of the marsh to estuarine water or open ocean. Accretion: Vertical rise of marsh due to buildup of organic and inorganic matter on the marsh surface. Rate differs by marsh-type. Salinity: Optional model or linkage to existing model. Salinity affects habitat switching in areas with significant freshwater flows Overwash: Barrier islands undergo overwash at a fixed storm interval. Beach migration and transport of sediments are calculated. Saturation: Migration of coastal swamps and fresh marshes onto adjacent uplands-- response of the water table to rising sea level.

Conceptual Model Square “raster” cells with elevation, slope, aspect, estimated salinity, wetland type Cells may contain multiple land-types Cell size flexible given size of study area Dry Land The point here is that SLAMM utilizes a raster grid or “bitmap grid” of output that is projected into three dimensions. Land elevations are tracked at height above MTL (mean tide level). This grid is then adjusted vertically for rises in sea level and projected accretion rates. When wetlands fall below their natural elevation range, they are converted to a different class. (Or if erosion, soil saturation, or overwash of barrier islands is predicted, conversion occurs) Various Wetlands Open Water 2D Representation 3D Representation

SLAMM Inundation Model Equilibrium Approach Elevation Land Elevation Salt Elev. (30 day inundation) Salt Boundary MHHW MTL MLW Tidal Flat Regularly- Flooded Marsh (Often Salt Marsh) Transitional or Irregularly- Flooded Marsh Inland Fresh and Dry Land Distance Inland

SLAMM Inundation Model (Migration of Wetlands Boundaries due to Sea Level Rise) Elevation Old Land Elevation New Land Elevation Salt Elev. (30 day inundation) Salt Boundary MHHW MTL Discuss implications of Dikes and Developed land on marsh migration as well as land elevation. MLW Tidal Flat Regularly- Flooded Marsh (Often Salt Marsh) Inland Fresh and Dry Land Water Irregularly- Flooded Marsh Distance Inland

Feedbacks to Accretion SLAMM 6 Allows for Elevation Feedbacks to Accretion as shown by Morris et al. (2002) Linkage to Morris MEM model “Unstable Zone” High-elevation marsh subject to less flooding

Linkage to Marsh Equilibrium Model Explicitly accounts for physical and biological processes affecting marsh accretion http://129.252.139.114/model/marsh/mem2.asp

Detailed SLAMM Land Categories 26 Categories, often derived from NWI (National Wetlands Inventory) May be specified as “protected by dikes or seawalls” Dry Land: Developed and Undeveloped Swamp: General, Cypress, & Tidal Transitional Marsh: Occasionally Inundated, Scrub Shrub Marsh: Salt, Brackish, Tidal Fresh, Inland Fresh, Tall Spartina Mangrove: Tropical Settings Only Beach: Estuarine, Marine, Rocky Intertidal Flats: Tidal Flats & Ocean Flats Open Water: Ocean, Inland, Riverine, Estuarine, Tidal Creek

Sea Level Rise Scenarios Model incorporates IPCC Projections as well as fixed rates of SLR Global (Eustatic) Rates of SLR are corrected for local effects using long-term tide gauge trends or spatial subsidence Grinsted, 2009 Clim. Dyn. Vermeer and Rahmstorf, 2009, Proceedings of the National Academy of Sciences

Powerful SLAMM Interface – Main Interface (Illustrates 3-D Graphing Capabilities)

Powerful SLAMM Interface – Execution Options

Powerful SLAMM Interface – Uncertainty Analyses

Powerful SLAMM Interface – Landcover Maps

Powerful SLAMM Interface – Elevation Maps

Powerful SLAMM Interface – Depth Profiles

Powerful SLAMM Interface – Elevation Histograms

Connectivity Component Method of Poulter & Halpin 2007 Assesses whether land barriers or roads prevent saline inundation Can be used for levee overtop model with fine-scale DEM Must account for culverts and tide gates Subject to small scale DEM error

Built-in Sensitivity Analyses Marshes most sensitive to accretion rates Beaches and Tidal flats most sensitive to parameters that affect SLR rates, tide ranges, and initial condition elevations Dry land most sensitive to SLR rates.

Uncertainty Module Addresses Two Primary Criticisms Accretion Model Doesn’t account for feedbacks (not true in SLAMM 6) Manner in which feedbacks are accounted for is uncertain Lack of uncertainty evaluation How confident are you of the results? Interpretation of deterministic results difficult What to do if available input parameters are not very good? Decision making difficult since likelihood and outcome variability are unknown

Model Output Distributions Parametric Model Input Distributions Model Output Distributions “Uncertainty Cloud” for Selected Region Examining SLAMM results as distributions can improve the decision making process Results account for parametric uncertainties Range of possible outcomes and their likelihood Robustness of deterministic results may be evaluated

Dike Considerations Traditional SLAMM has “on-off” dike layer Option to model dike elevations is included New: Dike elevations may be input at fine scale Connectivity can be used to calculate dike overtop Dike Removal Dynamic accretion processes following dike removal are not represented

“Hindcasting” Capability Run the model with historical data for validation and calibration Results will be imperfect Historical elevation data with high vertical resolution unavailable Historical land-cover data are spotty and changes in NWI classification have occurred Model will not predict land-use changes, beach nourishment or shoreline armoring For many sites, hindcasting is not possible due to insignificant RSLR “signal” In GOM, land subsidence amplifies SLR signal enough to make hindcasting possible

SLAMM Infrastructure Module Grant funded by US Fish and Wildlife Service Integrates predicted SLR and tide-ranges with roads and infrastructure databases Predicts effects on infrastructure Better captures infrastructure effects on surrounding wetlands

Original Inundation (No Roads)

Inundation with Roads

Vulnerability of Alligator River Roads to 1M of SLR by 2100

Initial Condition

2025 under 1M by 2100

2050 under 1M by 2100

2075 under 1M by 2100

SLAMM Erosion Model Erosion assumed a function of wave action Maximum Fetch calculated at each cell based on previous land-changes. When threshold of 9km is exceeded horizontal erosion rates are implemented. 9 km threshold based on visual inspection of maps Value verified within literature (Knutson et al., 1981) Tidal Flats have different assumptions

Overwash Assumptions Barrier islands of under 500 meter width are identified and assumed to be affected Frequency of “large storms” is user input Assumed effects are professional judgment based on observations of existing overwash areas (Leatherman and Zaremba, 1986). Effects editable in SLAMM 6

Water Table Rise Near Shore, Based on Carter et al., 1973 Soil Saturation SLAMM estimates (fresh) water table from the elevation nearby swamps or fresh-water wetlands As sea levels rise, this applies pressure to fresh water table (within 4km of open salt water) Model results could include “streaking” as a result of soil saturation predictions. Water Table Rise Near Shore, Based on Carter et al., 1973

SLAMM Salinity Model Required as marsh-type is more highly correlated to salinity than elevation when fresh-water flow is significant (Higinbotham et. al, 2004) Simple steady-state salinity model; not hydrodynamic Adds complexity to model development Requires additional model specifications Estuary Geometry Freshwater flow and projections Linkages to external salinity models are already built in to SLAMM 6.3

SLAMM 5 Salinity Component (For cells defined as “in-estuary”) Elevation Salinity calculated as a function of estuary width, tide range, fresh-water flows, and bathymetry FWH, fn of River Discharge Fresh Water Tide Range + SLR Salt Water SaltMarsh Brackish Tidal Fresh Tidal Swamp Estuary Area, Moving Inland salinity decreasing, but not linearly

Salinity Calibration Successfully calibrated to 5 GA estuaries Good match of salinity to river mile vs LMER data Publication pending Spatially calibrated to salinity data in Port Susan Bay, WA

Next Steps in Model Development Make “flow-chart” of habitat switching and land-categories modeled completely flexible (international applications) Linkage to hydrodynamic, sediment transport models More salinity testing Wider testing of SAV module Model evaluation and refinement – erosion, overwash, soil saturation Seeking collaborative partners

Galveston Bay Hindcast and Initial Forecast Results Meeting with Stakeholders in TX Incorporation of & Response to Stakeholder Comments Final results available at GOMA and SLAMMView website http://www.slammview.org/slammview2/reports/Galveston_Report_6_30_2011_w_GBEP_reduc.pdf

Complicating Factor: Subsidence Spatial maps used in hindcasting Used to convert eustatic to local SLR Held constant over simulation Gabrysch and Coplin 1990

Accretion Marsh Type Study Location Accretion Rate Measured (mm/y) Average Accretion (applied to model, mm/yr) Freshwater Yeager et al, 2007 North of Trinity River Dam 1.3 2.9 Williams, 2003 2.5 White et al., 2002 4.9 Saltwater Ravens et al., 2009 West Bay 2.0 4.7 or 7.75 Trinity Bay, South of Dam 10.2 5.3 Slide showing areas of high and low sediment supply is located at end of slideshow

1979 2009 Different Footprint Fresh Marsh Expansion Conversion from reg flooded (salt) to fresh marsh – poses a problem for hindcast metrics – is this really happening? Salt to fresh marsh conversion is not happening due to SLR and therefore SLAMM won’t predict it. Different Footprint Fresh Marsh Expansion

2009 Pred. 2009 Obs. Fresh Marsh Expansion Anthropogenic Actions

2100, Scenario A1B Mean Irreg. Flooded Marsh Streaks – soil saturation Undeveloped Dry Land Inland Fresh Marsh Developed Dry Land Irregularly Flooded Marsh Irreg. Flooded Marsh Regularly Flooded Marsh Swamp Tidal Swamp Tidal Fresh Marsh Inland Shore Open Ocean Estuarine Open Water Riverine Tidal Inland Open Water Streaks – soil saturation

2100, Scenario A1B Maximum Irreg. Flooded Marsh Undeveloped Dry Land Inland Fresh Marsh Developed Dry Land Irregularly Flooded Marsh Irreg. Flooded Marsh Regularly Flooded Marsh Swamp Tidal Swamp Tidal Fresh Marsh Inland Shore Open Ocean Estuarine Open Water Riverine Tidal Inland Open Water

2100, 1 Meter Irreg. Flooded Marsh Undeveloped Dry Land Inland Fresh Marsh Developed Dry Land Irregularly Flooded Marsh Irreg. Flooded Marsh Regularly Flooded Marsh Swamp Tidal Swamp Tidal Fresh Marsh Inland Shore Open Ocean Estuarine Open Water Riverine Tidal Inland Open Water

2100, 1.5 Meters Irreg. Flooded Marsh Undeveloped Dry Land Inland Fresh Marsh Developed Dry Land Irregularly Flooded Marsh Irreg. Flooded Marsh Regularly Flooded Marsh Swamp Tidal Swamp Tidal Fresh Marsh Inland Shore Open Ocean Estuarine Open Water Riverine Tidal Inland Open Water

2100, 2 Meters Irreg. Flooded Marsh Undeveloped Dry Land Inland Fresh Marsh Developed Dry Land Irregularly Flooded Marsh Irreg. Flooded Marsh Regularly Flooded Marsh Swamp Tidal Swamp Tidal Fresh Marsh Inland Shore Open Ocean Estuarine Open Water Riverine Tidal Inland Open Water

SLAMM Strengths Open source Relatively simple model Ease and cost of application Relatively quick to run (enables uncertainty analysis) Contains all major processes pertinent to wetland fate Provides information needed by policymakers

Strengths (cont.) Detail oriented flow chart Relatively minimal data requirements Designed in poor data environment -- has assumptions to work through those conditions. Internal uncertainty & sensitivity analyses

SLAMM Model Limitations Not a Hydrodynamic Model Conceptual model captures these sites initial conditions well; future changes in hydrodynamics may not be properly represented. Spatially Simple Erosion Model Could be modified or replaced with more sophisticated model Beach erosion is ephemeral and difficult to quantify anyway

Model Limitations No Mass Balance of Solids i.e. accretion rates not affected by bank sloughing Storms do not mobilize sediment Overwash component is subject to considerable uncertainty Timing and size of storms is unknown Based on observations of barrier islands after large storms

Mcleod, Poulter, et al., 2010 SLAMM 5 Advantages Ocean & Coastal Management “SLR impact models and environmental conservation, a review of models and their applications” SLAMM 5 Advantages Can be applied at wide range of scales Provides detailed information about coastal habitats and shift in response to SLR Can be used to identify potential future land-use conflicts Integrates numerous driving variables “Provides useful, high-resolution, insights regarding how SLR may impact coastal habitats.” “elevation, habitat type, slope, sedimentation, accretion, erosion rates, existing seawalls”

Mcleod, Poulter, et al., 2010 SLAMM 5 Disadvantages Lacks feedback mechanisms between hydrodynamic and ecological systems Changes in wave regime from erosion not modeled Note wave setup is recalculated on basis of land loss Lacks feedback between salinity and accretion rates in fresh marshes SLAMM 6 does include feedbacks between frequency of inundation and accretion rates and links to mechanistic modeling. Does not include a socioeconomic component to estimate costs; not useful for adaptation policies

Considerations and Costs of Implementation GIS expertise required to produce raster inputs Tidally-coordinated LiDAR elevation data highly beneficial NWI data often out-of-date Alternative data sources have often been used Crosswalk process time consuming Salinity model requires additional support Model QA tests can be time-consuming

To Stay In Touch with Future Model Developments SLAMM webpage http://warrenpinnacle.com/prof/SLAMM Includes brief model overview, bibliography Updated with latest projects and results Technical documentation with full model specs Model executable available at this site Model code is “open source” available for review or modification Email me -- jclough@warrenpinnacle.com