A Decision-Support Framework for Evaluating Riverine Sediment Influence on U.S. Tidal Wetlands

Abstract

Tidal wetlands are essential for coastal resilience, biodiversity, and carbon storage; yet, many are increasingly vulnerable to sea-level rise due to insufficient sediment supply. This study presents a national-scale, GIS-based model that quantifies riverine inorganic sediment contributions to tidal wetland accretion across over 700,000 coastal catchments in the contiguous United States. By integrating datasets from USGS, NOAA, and USFWS, the model calculates sediment yield, thickness, and accretion balance, enabling comparison with current sea-level rise projections. Results reveal significant regional disparities: the Northeast and Midwest exhibit higher sediment accumulation, while the Pacific and Southeast show widespread sediment deficits. Spatial statistical analyses identified clusters of high and low sediment supply, highlighting areas of resilience and vulnerability. A total of 93 field sites confirmed the model’s ability to distinguish between riverine-dominated and mixed-source sedimentation regimes. These findings underscore the importance of riverine sediment in sustaining wetland elevation and inform where non-riverine sources may be critical. The model’s outputs have been shared with coastal planners and stakeholders to support local decision-making, conservation prioritization, and adaptation strategies. This work demonstrates both the challenges and fruitfulness of harmonizing disparate national datasets into a unified framework for assessing wetland vulnerability and provides a scalable tool for guiding coastal resilience planning in the face of accelerating sea-level rise.

1. Introduction

Flooding is a major global cause of infrastructure damage, loss of life, and social instability. In the United States (US), flooding is the leading cause of property damage, with over 4.3 million homes identified as being at high risk [1]. In 2000, 3.0% of the US population (8.5 million people) lived within the 100-year floodplain [2], and 52% of the US population (147 million people) resided in coastal counties [3]. The average annual cost of flooding in the US, currently USD 2 billion, is projected to rise to USD 7–19 billion by the end of the 21st Century [4,5]. The US Federal Emergency Management Agency (FEMA) monitors flood claims through the National Flood Insurance Program, identifying Texas and Louisiana as areas with the highest requests for federal support. In 2019, more than 7.3 million homes along the US Gulf and Atlantic coasts were vulnerable to storm surge flooding, with replacement costs exceeding USD 1.1 trillion [6]. From 2020 to 2023, FEMA declared 80 major flood disasters, with significant economic damage, and costs reaching billions of dollars annually [7]. In 2023 alone, floods resulted in 78 fatalities across the U.S. (NWS, 2025) [8].

Flooding also threatens natural resources, particularly salt marshes, which provide ecosystem services valued at over 1 trillion USD Int annually [9], and they are increasingly threatened by sea-level rise [10]. The resilience of tidal wetlands to sea-level rise depends on ecological and geophysical processes affecting elevation change. Salt marshes are expected to keep pace with relative sea-level rise through feedbacks in which increased inundation during relative sea-level rise results in an increased supply of fluvial sediment [11]. If salt marshes are unable to accrete at or above the rate of sea-level rise, they drown [12]. Mitigation efforts may include engineering solutions such as runnels [13], thin-layer dredge placement [14], and dam removal to enhance sediment delivery to the estuary [15]. Conservation planners, environmental managers, and researchers use decision support frameworks to guide planning. For example, the “Resist, Accept, Direct” framework provides three state spaces for decision-making [16] and the “Protect, Restore, Monitor, Evaluate” model is specific to tidal wetlands and has been useful for framing initial decision-making [17].

Geospatial models and spatially derived metrics inform the use of these frameworks at specific tidal wetland sites. The National Oceanic and Atmospheric Administration (NOAA), the leading US agency for mapping coastal resources, has developed tools such as The Digital Coast (https://coast.noaa.gov/digitalcoast/ (accessed on 1 July 2025)), the Coastal Flood Exposure Mapper (https://coast.noaa.gov/digitalcoast/tools/flood-exposure.html (accessed on 1 July 2025)), Sea-Level Rise Viewer (https://coast.noaa.gov/digitalcoast/tools/slr.html (accessed on 1 July 2025)), and Climate Mapping for Resilience and Adaptation (https://coast.noaa.gov/digitalcoast/tools/cmra.html (accessed on 1 July 2025)). The US Geological Survey (USGS) has developed the Coastal Change Hazards Portal (https://marine.usgs.gov/coastalchangehazardsportal/ (accessed on 1 July 2025)) and Flood Inundation Mapper (https://fim.wim.usgs.gov/fim/ (accessed on 1 July 2025)).

One of the key parameters for predicting vulnerability is sediment availability for wetland accretion. This is sometimes parameterized using remotely sensed data [18] or by assuming a single vertical accretion rate, such as in the Sea Level Affecting Marshes Model (SLAMM) [19]; however, these site-specific models cannot be applied nationally. The importance of riverine sediment versus remobilized estuarine sediment availability is often questioned but remains challenging to parameterize in decision-support tools. Therefore, the goal here is to develop a model that quantifies riverine sediment accretion rates for all tidal wetlands in the contiguous US, which may be implemented in future iterations of NOAA, USGS, and other online decision support tools.

Building on the established use of Geographic Information Systems (GIS) in site-specific coastal studies and the pressing need to assess wetland vulnerability to inform resiliency planning, this project explores the usefulness of public data that is available nationally to develop a geospatial framework that enables researchers and local planning practitioners to explore the dynamic spatial relationship between fluvially delivered sediment and tidal wetland vulnerability for the contiguous US. In a prior study, we investigated regional patterns in the contribution of river sediment to tidal wetland accretion [20]. This paper examines publicly available data, describes the process for developing the GIS approach to modeling riverine sediment contributions, and discusses the challenges and limitations of building the decision support framework.

2. Materials and Methods

2.1. Data Sources

After evaluating various data options, the team established two criteria for selecting data for the model: (1) the data must be available for the contiguous US, and (2) the geographic and attribute data must be relevant and useful for assessing coastal wetland vulnerability. We identified the following datasets as both available and suitable for calculating coastal wetland vulnerability: (1) USGS sediment load from SPAtially Referenced Regressions On Watershed (SPARROW) models, (2) USGS hydrology data, (3) US Fish & Wildlife Service (USFWS) wetlands data, (4) NOAA sea level elevation and extent data, and (5) NOAA sea-level rise trends and forecasts (

Table 1. Data sources used in the study. All URLs accessed on 1 July 2025.

Data	Organization	URL	Date (s)
SPAtially Referenced Regression On Watershed (SPARROW)	US Geological Survey (USGS)	https://www.usgs.gov/mission-areas/water-resources/science/sparrow-mappers	1999–2014
National Hydrography Dataset Plus High Resolution (NHDPlusHR *)	USGS	https://www.usgs.gov/core-science-systems/ngp/national-hydrography/nhdplus-high-resolution	2000–2018
National Wetlands Inventory (NWI)	US Fish & Wildlife Service (USFWS)	https://www.fws.gov/program/national-wetlands-inventory/download-state-wetlands-data	1972–2019
Sea-level elevation and extent	National Oceanic and Atmospheric Administration (NOAA)	https://coast.noaa.gov/slrdata/	2016
Sea-level rise forecasts	NOAA	https://oceanservice.noaa.gov/hazards/sealevelrise/sealevelrise-data.html	2022
Sea-level rise trends	NOAA	https://tidesandcurrents.noaa.gov/sltrends/	Minimum 30 years to present

* For further details, please refer to the Supplementary Materials.

).

2.1.1. USGS SPARROW

SPARROW data are distributed as Esri shapefiles for five regions of the United States: Midwest, Northeast, Pacific, Southeast, and Southwest. The geographic unit of measurement is the catchment, which represents the smallest area draining to a single outlet point. SPARROW’s catchment-scale data is smaller than the smallest geographic unit within the Watershed Boundary Dataset (WBD) of the USGS National Hydrography Dataset (NHD). Watersheds in the WBD are classified into multiple levels, known as Hydrologic Unit Codes (HUCs), ranging from the broad HUC-2 to the more detailed HUC-12, following a hierarchical numbering scheme from 2 to 12 digits.

The primary purpose of SPARROW is to model streamflow, total phosphorus, total nitrogen, and suspended sediment. Notably, SPARROW provides the only national assessment of suspended sediment at the catchment scale across the U.S. However, southern Florida is excluded from the model due to water diversions and insufficient data to accurately model flux across these watersheds (Figure 1) [21].

The annual accumulated sediment load, measured in kilograms or metric tons per year, is calculated from all contributing sources within each stream reach and is an average using data from 1999 through 2014. The amount of suspended sediment exported from a stream reach depends on a combination of several factors, including climate (e.g., precipitation and wildfire), soil type, land cover, geology, and hydrology. In many areas, human factors such as urbanization and agriculture are leading land cover types that contribute suspended sediment, and these are accounted for in the SPARROW model. The SPARROW model details and resulting data can be examined using the SPARROW website (Table 1) and reference documentation for each SPARROW region [21,22,23,24,25]. Researchers have assessed the accuracy of SPARROW data, and although local sedimentation models can outperform SPARROW, it remains a reliable product useful for assessing marsh sustainability [26,27,28].

In total, there are 2,642,295 catchments in SPARROW data, of which there are 53 HUC−4 coastal watersheds and 2693 HUC−12 watersheds in this coastal zone (

Table 2. Number of watersheds (catchments and Hydrologic Unit Codes 4 and 12) per SPARROW region.

Region	Catchments	Catchments Within Coastal HUC-4 Watersheds	HUC-12 Watersheds	HUC-4 Watersheds Date (s)
1: Midwest	1,354,229	44,858	3563	17
2: Northeast	191,391	172,082	4837	12
3: Pacific	338,825	151,777	11,083	10
4: Southeast	317,604	283,067	6961	8
5: Southwest	440,236	61,251	8707	8
Total	2,642,285	713,035	35,151	53

).

2.1.2. USGS National Hydrography

The National Hydrography Dataset (NHD) is a comprehensive database of streams, rivers, lakes, and other water bodies, digitized from multiple sources. Attributes are calculated from Digital Elevation Models (DEMs) to derive catchments, hydrologic units at hierarchical watershed scales, and flow directions along stream centerlines. In areas with significant relief, the NHD data effectively portrays water flow through watersheds. However, in regions with minimal slope, calculating flow direction and contributing area can be problematic [29].

In 2016, the USGS released the NHD Plus High Resolution (NHDPlusHR), which integrates NHD with the Watershed Boundary Dataset and elevation data from the 3-D Elevation Program (3DEP) [30]. The Environmental Protection Agency (EPA) manages the maintenance and distribution of NHDPlus Version 2 (NHDPlusV2), an updated version of NHDPlusV1 (released in 2006), which was released in 2012, and is considered the best nationwide medium-resolution dataset for stream hydrography. The development of NHDPlusHR began by adapting tools from NHDPlusV2 to work with the new hydrologic data model in NHDPlusHR, then production of catchments, flow direction, flow accumulation, and the Value-Added Attribute (VAA) tables. Next, a community review and collection of feedback was conducted. Lastly, the framework tools incorporated the most up-to-date data [30]. When states or the USGS implement corrections/updates, the data moves to the last phase, phase 5, where the data are refreshed by HUC-4 watershed. A map showing the status of each HUC-4 watershed is available here: https://www.usgs.gov/media/images/nhdplus-hr-status-map-0 (accessed on 1 July 2025).

A comparison of NHDPlusHR and NHDPlusV2 revealed that the HR product offers more spatial detail in estuaries, while the V2 data lacks some smaller tributaries. Consequently, given the better spatial detail and inclusion of small tributaries, the NHDPlusHR product was used to derive stream centerlines that connect with the coast. Source dates for NHDPlusHR range from 2000 to 2018 and are, therefore, comparable with the source dates for the SPARROW data.

NHDPlusHR GIS data is complex, with many relational attribute tables that contain extensive data describing hydrology. For more information about this robust dataset, please visit: USGS NHDPlus HR Value-Added Attribute (VAA) Navigator (https://www.usgs.gov/media/videos/nhdplus-hr-value-added-attribute-vaa-navigator (accessed on 1 July 2025)). For each HUC-4 watershed (N = 53), NHDFlowline data (line features) were related to the external VAA table to select a subset of streams that flow to the ocean (Figure 2):

(1) All stream features were included except for coastlines (FTYPE = 566), which form the boundary between land and water, cross at bays and river mouths, and outline coastal islands (USGS NHD Data Dictionary, https://www.usgs.gov/ngp-standards-and-specifications/national-hydrography-dataset-nhd-data-dictionary-feature-classes (accessed on 1 July 2025)).

(2) Streams with a Stream Level greater than 0 (or no NULL values) were included in the analysis. Stream Level identifies the stream connection hierarchy, where level 1 lines flow directly to the ocean, level 2 lines connect with level 1, etc. Features with a NULL stream level are not connected to other streams and, therefore, do not flow to the ocean and were excluded from the analysis.

(3) Streams with Divergence Codes of 0 or 1, but not 2, were included in the analysis. Divergence Codes of 0 represent single streams with no alternative paths, while codes of 1 identify streams with alternative paths, with value 1 indicating the main path. Codes of 2 represent minor paths. Only single paths (value = 0) and primary stream paths (value = 1) were included, excluding minor paths (value = 2) to ensure the model considered one stream per coastal stream network.

2.1.3. USFWS National Wetlands Inventory

There are several sources for wetlands geospatial data, including the National Land Cover Dataset (NLCD) available from the USGS and the Multi-Resolution Land Characteristics (MRLCs) Consortium, the National Wetlands Inventory (NWI) from the US Fish & Wildlife Service (US FWS), and NOAA’s Coastal Change Analysis Program (C-CAP). Although the NWI data can be somewhat outdated in certain regions, it provides the only hierarchical classification system that enables the identification of tidal wetlands, including both saline tidal wetlands (e.g., emergent marsh) and freshwater tidal habitats.

Other products, such as NLCD and C-CAP, provide generalized classes. For example, NLCD includes categories such as woody wetlands and emergent herbaceous wetlands, while C-CAP includes palustrine and estuarine forested, scrub/shrub, and emergent wetlands. The most recent NLCD data is from 2023, while the latest C-CAP data is from 2016. In contrast, the NWI uses a very detailed wetland classification system, the Cowardin classification system, to identify many types and detailed characteristics of wetlands [31]. The hierarchical classification system is composed of the following components, in this order: system (e.g., marine, estuarine, riverine, lacustrine, and palustrine); sub-system (e.g., 1 = subtidal and 2 = intertidal for estuarine, varying by system); class (varying by system and sub-system); subclass (varying by class); split class (used when more than one class is present); split subclass; water regime; modifier 1; and modifier 2. Importantly, there are four groups of modifiers, such as the water regime modifiers, which classify the wetland by the degree of tidal influence, an essential factor for this study. Complete definitions of this system can be found in Cowardin et al. (1979) [31]. This level of tidal influence is not available in the classifications used for NLCD and C-CAP (Figure 3). For example, E1UBLx is decoded as follows: E (estuarine system), 1 (subtidal subsystem), UB (unconsolidated bottom class), L (saltwater tidal water regime), and x (excavated modifier). Therefore, the NWI data was the preferred source of wetland data due to the ability to identify the head of tide, which is critical to identifying all tidal wetlands across the United States.

As the primary goal of the project was to identify sediment accumulation on tidal wetlands, the NWI data were used to identify coastal habitats and tidal wetlands (Figure 4). The location of coastal habitats (i.e., estuarine, marine, or tidal riverine) was used to identify the NHD streams that are within coastal areas by defining the locations of coastal habitats and, within these areas, the locations of tidal wetlands. Tidal wetlands were used to link the amount of suspended sediment from the SPARROW data to the wetlands. The two queries to select the coastal and tidal habitats within the NWI data were:

Coastal Habitats = ATTRIBUTE begins with E, M, or R1

(1)

Tidal Wetlands = ATTRIBUTE contains E2AB, E2EM, E2SS, E2FO, R1EM, R1AB, ends with Q, Qb, Qd, Qm, Qh, R, Rb, Rd, Rm, Rh, Rx, S, Sb, Sd, Sm, Sh, Sx, T, Tb, Td, Tm, Th, Tx, V, Vb, Vd, Vm, Vh, Vx, does not begin with R1RB, R1UB, R1SB, R1RS

(2)

Since the NWI data uses the Cowardin nested hierarchical classification system attributed in a single field, ATTRIBUTE, we must use a selection/query that accounts for the beginning and end of attribute values (

Table 3. Tidal wetlands from the National Wetlands Inventory (NWI), where AB is Aquatic Bed, EM is Emergent, FO is Forested, and SS is Scrub/Shrub.

System	Subsystem	Class	Subclass	Modifier
Estuarine (E)	Intertidal (2)	AB, EM, FO, SS	AB: Algal, Rooted Vascular, Floating Vascular EM: Persistent, Nonpersistent FO and SS: Deciduous, Evergreen	All
Riverine (R)	Tidal (1)	AB, EM	AB: Algal, Aquatic Moss, Rooted Vascular, Floating Vascular EM: Nonpersistent	All
Marine (M)	Intertidal (2)	AB	Algal, Rooted Vascular	Regularly flooded–fresh-tidal Seasonally flooded–fresh-tidal Temporarily flooded–fresh-tidal Semipermanently flooded–fresh-tidal Permanently flooded–fresh-tidal
Palustrine (P)	N/A	AB, EM, FO, SS	AB: Algal, Floating Vascular, Rooted Vascular EM: non-persistent, persistent FO and SS: deciduous, Evergreen	Permanently flooded–tidal Semipermanently flooded–tidal Seasonally flooded–tidal Temporarily flooded–tidal

).

2.1.4. NOAA Sea Level

Several data sources from NOAA were used in the model: (1) spatial extent (polygons) for predicted sea level extent at various elevations of sea-level rise (SLR), (2) predicted rates of SLR [32,33], and (3) local relative sea-level trends using tide gauges (NOAA Sea-Level Trends https://tidesandcurrents.noaa.gov/sltrends/ (accessed on 1 July 2025)). The spatial extent of SLR data is available by state, with some states divided into regions. For the coastal contiguous US, 64 geodatabases were downloaded from NOAA SLR Data (https://coast.noaa.gov/slrdata/ (accessed on 1 July 2025)). The NOAA SLR polygons depict flood inundation at 1-foot increments (from 1 to 10 feet). Given that NWI data vary by image date, we used the NOAA SLR polygons at the 10-foot level to limit the tidal wetlands to these areas (Figure 5A). Although NOAA has predicted flood inundation for the contiguous US, large sections of New Jersey are missing data. Therefore, for the Northeast region, we did not use the SLR polygons to limit the tidal wetlands from NWI.

Using the NOAA predictions of SLR from the NOAA Global Change Report (https://www.ncei.noaa.gov/access/monitoring/monthly-report/global/201713 (accessed on 1 July 2025)), our model employed a conservative, medium-rate measure of SLR based on 2020 predictions [33]. This approach estimates the impact of current SLR predictions with modeled rates of sediment accumulation on tidal wetlands. Using Matlab version 23.2 (R2023b), the SLR predictions were converted from X, Y locations to a gridded (1-degree by 1-degree) layer (Figure 5B) and spatially joined with the NHD stream centerlines (Figure 5C).

2.2. Model Workflow

Given that each data source has different geographies, spatial extents, and attributes, as described above, this required a thorough examination of each data source to determine the optimal approach to combining the data for the purpose of computing sediment accumulation to coastal wetlands. After many trials and changes, we ultimately built a model that integrated each data source and calculated sediment thickness and accretion balance for all coastal streams and tidal wetlands in the contiguous US (Figure 6). Importantly, as we developed the model, we tested the output at each step to ensure we obtained results that were expected, and when the model did not produce expected results, we modified the data processing and tested the output again. This process was repeated until we created a model that was thoroughly tested.

Recalling that the objective of the model was to map sediment load delivered to all tidal wetlands, the model was limited to the spatial extent of the SPARROW data, as these data contain the starting point, or the amount of accumulated riverine sediment load per catchment. Propagation of error can be a critical issue in modeling when combining multiple geographies with differing spatial accuracies. Given this potential problem, we decided to use the highest spatial resolution data, the NHDPlusHR, as the geography for the resulting model and pulled attributes from overlaying datasets to this geography. The foundation of the model workflow is the transfer of SPARROW catchment attributes to the streams (NHDPlusHR) and wetlands (NWI). Therefore, there was no alteration of any geography, and therefore no propagation of error, as the model computed these outputs. Additionally, even though the NWI data can be relatively outdated in some areas, these data were used to (1) identify the extent of the tidal zone and (2) final data visualization of calculated TP and LP sediment yield, thickness, and accretion balance. At no time is the NWI data used to derive sedimentation metrics.

The geographic unit for data processing was the HUC-4 watershed, which is the unit of geography for distributing the NHDPlusV2 data. For computational efficiency and model development testing, it was most effective to test the model iteratively for several HUC-4 watersheds across all the SPARROW regions to address specific nuances in different parts of the country. Once the model was finalized, it was run for all HUC-4 watersheds (N = 53 iterations). This iterative approach to model development resulted in a workflow that has been tested and modified at multiple locations throughout the US coastal zone. Consequently, each step in the model development process underwent multiple assessments and tests for alternative data processing approaches.

Integrating diverse data sources presented the following challenges:

• Geographic Variability: Each data source had different geographic extents and units, requiring careful database management across the study area.
• Attribute Differences: The attributes varied significantly between datasets, necessitating detailed examination and harmonization to ensure accurate and meaningful integration.
• Temporal Differences: The datasets were collected at different times, which may affect the accuracy of integrating disparate data. We cannot quantify the potential impact of these temporal offsets on model outputs because there are no alternative data sources on a national scale for comparison. National-scale modeling of overlapping terrestrial and coastal processes is inherently limited by the necessary use of disparate data sources produced across timescales that may not fully align. This manuscript highlights this limitation and attempts to provide a first-ever solution within the context of sediment transport from terrestrial to coastal systems.
• Resolution Differences: The spatial resolution of the datasets varied, which can impact the precision of an integrated geospatial model. Therefore, higher resolution data, such as the NHD stream centerlines, provided more spatial detail, so these data formed the basis for the model, and then other data, such as the rates of SLR, were joined to these higher resolution data.
• Computational Efficiency: Processing large volumes of data across multiple regions requires efficient computational strategies. Iterative testing and optimization were essential to manage the computational load, and then each HUC-4 within each region was processed through the model.
• Data Gaps: Some regions, like parts of New Jersey, had missing data, which required alternative approaches or exclusion from analyses. This required that each SPARROW region in the study had a version of the model to ensure proper data processing.
• Model Validation: Ensuring the reliability of the integrated model involved extensive testing and validation across different regions and conditions.

Despite these challenges, the iterative approach to model development and thorough testing at multiple locations helped create a robust workflow and reliable results.

2.2.1. NHD Level Path and Terminal Path

The Coastal Streams, selected from the NHD data, were overlaid with the Coastal Habitats, selected from the NWI data, to create a new data layer containing streams within the coastal zone. Next, we investigated several methods for quantifying the amount of sediment accretion per stream. The SPARROW data contains the average annual delivered aggregated sediment load for each catchment. Along the coastal US, there were incongruities between the SPARROW catchments and USGS watersheds, as well as the resulting hydrologic network of streams. We examined each unique stream segment within the SPARROW sediment accretion areas and found too many inconsistencies with the values for contributing drainage areas. Therefore, the alternative to using the drainage area from SPARROW was to use the NHD stream data, where each segment can be related to the Value-Added Attributes (NHDPlusFlowline VAA table). Each record in the VAA table has a unique Terminal Path (TP) and Level Path (LP) [34]. The TP field in the VAA Table (“TerminalPathID”) contains the same value for all streams within a drainage area, and the LP field (“LevelPathI”) contains unique values for each stream within a drainage area. Therefore, the workflow to calculate the sediment accretion rates was tabulated for both the entire drainage area (TP) and the individual streams within each drainage area (LP) (Figure 7).

2.2.2. Sediment Yield, Thickness, and Accretion Balance

To calculate the rate of SLR for each stream segment, the following equation was used:

SLR (cm yr⁻¹) = (RSL2020 − RSL2005)/15

(3)

where RSL2020 and RSL2005 are predicted relative sea levels in 2020 and 2005 [33], and 15 is the difference between 2020 and 2005 (the base year).

To calculate sediment yield for each TP and LP stream, this equation was used:

Yield (kg km⁻² yr⁻¹) = (MAX_accl (MT yr⁻¹) × 1000)/MAX_CumArea (km²)

(4)

where MAX_accl is the maximum value for the annual accumulated load (in metric tons) for each catchment in the SPARROW data that intersected with the NHD stream centerlines. The value of 1000 converts metric tons to kilogram, and MAX_CumArea (km²) is the maximum cumulated upstream watershed area from the SPARROW data. This computation was performed separately for both Terminal Paths (TPs) and Level Paths (LPs).

Due to small spatial differences between the boundaries for catchments in the SPARROW data and the hydrologic network of the NHDPlusHR data, sediment loads for each TP and LP were calculated by multiplying yield by the total drainage area according to the NHDPlusHR:

TP Load (kg/yr) = Yield (kg km⁻² yr⁻¹) × TotDA (km²)

(5)

LP Load (kg/yr) = Yield (kg km⁻² yr⁻¹) × (TotDA (km²) − TotDivDA (km²))

(6)

where the calculated load for the TP streams used the total drainage area (TotDA) for the entire watershed, and LP streams used the total DA (TotDA) minus the area of divergence stream drainages (TotDivDA).

Next, the calculation of the thickness of sediment accretion for each TP and LP assumes that the sediment load is distributed uniformly across the wetland area using the following equation:

Thickness (mm/yr) = (Load (kg yr⁻¹)/1560 (kg m⁻³)/NWI_AREA (m²)) × 1000

(7)

where 1560 kg m⁻³ is the assumed bulk density of freshly deposited coastal sediment [35], NWI_AREA is the total area (m²) of tidal wetlands within 75 m of the stream centerline, and the value 1000 converts meters to millimeters. We chose a bulk density that was the average of two extreme endpoints: quartz sand (2650 kg m⁻³) and highly organic sediments (470 kg m⁻³). Lastly, using the calculated fields described above, the total sediment accretion balance was calculated by comparing sediment thickness with the rate of SLR in 2020 (Figure 8):

Accretion Balance (mm yr⁻¹) = Thickness (mm) − (RSL2020(cm yr⁻¹) × 10)

(8)

2.2.3. Model Assessment Process

Once quality assurance checks were completed, the final data were compiled, combining the individual layers for each HUC-4 within each SPARROW region. To visualize the calculated LP and TP stream segments, the attributes from the stream centerlines were linked to adjacent tidal wetland polygons. Given that the stream centerlines may not be topologically adjacent to tidal wetlands, we conducted a series of tests to determine an ideal distance from the centerline to the nearest tidal wetlands. We ran a series of distance tests on HUC watersheds throughout all SPARROW regions, starting with 10 m through 200 m at 10 m increments.

Assessing the output of the model involved the following steps:

(1)

Iterative Testing: The model was tested iteratively, with adjustments based on feedback and results from each iteration. The team used their first-hand knowledge of specific estuaries, wetlands, and rivers across the East, Gulf, and West coasts to explore the model output. This process helped refine the model and improve its accuracy.

(2)

Visualization and Mapping: We used visualization tools to map the model’s output and disseminate the maps to the public via a website (https://tinyurl.com/SedimentPancake (accessed on 1 July 2025)), making it easier to identify patterns and discrepancies. Interactive maps allowed for detailed examination of the results.

(3)

Data Comparison: We conducted a data literature review to identify published sediment accumulation rates across the US and compared these data with our model results.

(4)

Regional Comparisons: We conducted regional and watershed comparisons to ensure the model performed consistently across different geographic areas. This involved testing the model in various HUC-4 regions and adjusting parameters as needed.

(5)

Expert Review: We sought feedback from experts in geomorphology, geology, and hydrology to validate the model’s assumptions and results. Their insights were crucial in identifying potential issues and areas for improvement.

(6)

Workshops and Stakeholder Feedback: We conducted workshops to gather feedback from stakeholders, including coastal managers and researchers. Their input helped us understand the practical utility of the model, and we used this feedback to make edits to the published website (https://tinyurl.com/SedimentPancake (accessed on 1 July 2025)).

To visualize the calculated LP and TP stream segments, the attributes from the stream centerlines were linked to adjacent tidal wetland polygons. Given that the stream centerlines may not be topologically adjacent to tidal wetlands (where there is open water), we conducted a series of tests to determine an ideal distance from the centerline to the nearest tidal wetlands. We ran a series of distance tests, starting with 10 m through 200 m at 10 m increments, on randomly selected HUC watersheds throughout all SPARROW regions. In future work, it would be ideal to have a distance measurement based on the local geography, but for the development of this national framework, we selected 75 m as the best distance for transferring the stream centerline attributes to the nearest tidal wetland polygons. Again, this data transfer is for data visualization purposes and can be seen on the maps in the accompanying website (https://tinyurl.com/SedimentPancake (accessed on 1 July 2025)).

Surface Elevation Tables (SETs) are installed at many locations across the coastal US, including marker horizon measurements of sediment accretion over annual timescales, which are more comparable with our predictions [36]. Other sources of wetland sediment include the National Estuarine Research Reserve System (NERRS), the Long Term Ecological Research Network (LTER), and the Coastal Carbon Research Network (CCRN). We conducted a literature review and identified 18 reports containing SET and other sediment accretion data for coastal wetland locations across the contiguous US. We used these field measurements of accretion rates to compare with our modeled prediction of riverine sediment accretion rates. We did not expect our predictions to directly correspond with field measurements because measured rates of sediment accretion reflect more sediment sources than our riverine model incorporated. However, a comparison is useful to investigate where our model is similar or different, which indicates locations where the sediment source is dominated by riverine versus marine. The information within the peer-reviewed sources was merged using a script written in R to compile the various formats into a single CSV file, which was then converted to a geodatabase point feature class (N = 93). The script and resulting data are available for download (https://hydroshare.org/resource/fa0611211f9e461daf0576966ac62412/ (accessed on 1 July 2025)).

2.2.4. Identification of Regional Spatial Patterns

A website was created using Esri’s Experience Builder, which includes background information on the data sources, processing methods, results, and interactive maps for each SPARROW region (https://tinyurl.com/SedimentPancake (accessed on 1 July 2025)). Several workshops were conducted to gather feedback on the website, data, and its utility for coastal management. Building on the results presented in Ensign et al. [20], we further explored the relationship between location, sediment thickness, and accretion balance using spatial statistics. Spatial statistics are useful for identifying significant clusters in space, which is useful as another data visualization tool to assist planners and coastal managers. Therefore, we used this technique to identify clusters of riverine sediment thickness and accretion balance for LP and TP streams. Several steps were taken to identify clusters:

(1)

Mean LP and TP thickness and accretion balance were computed for each HUC-12 watershed within each SPARROW region.

(2)

Global Moran’s I measured spatial autocorrelation based on the location of each watershed and the LP and TP attributes.

(3)

Incremental Spatial Autocorrelation (ISA) identified distances with significant clustering.

(4)

Getis-Ord General G identified significant clustering of high and low values.

(5)

Using results from the global statistics (Moran’s I, ISA, and General G), Anselin Local Moran’s I identified statistically significant hot and cold watersheds for both LP and TP. Each computation returns z-scores and p-values, which indicate the amount of statistical significance.

3. Results

3.1. GIS Data Considerations for Coastal Management

The large volume of data from several federal agencies, including USGS, NOAA, and USFWS, resulted in spatial discrepancies among the varying sources. Although most of the data sources were at the same spatial scale (1:24,000), the primary sources used to create these datasets varied across the country. For example, while the geography between the USGS NHD and the USFWS NWI is complementary, there are many locations where stream centerlines and wetland boundaries do not match. Therefore, modeling the spatial relationships between these disparate datasets required spatial decision-making to ensure the model results produced the most spatially accurate and geomorphically consistent mapping results.

Most local studies use data editing to fix slivers and other topological problems, but with Big Data, such as this national model, it is prohibitive and incorrect to edit public data sources. Consequently, the project team investigated various approaches to data integration, carefully considering the differences in geomorphology, geology, and riverine sedimentation across the continental US. This exploration included regional and watershed comparisons and statistical analyses to identify problems, adjust the GIS parameters and tools, re-run the model, and reassess the model output until we reached a point when slight adjustments to the model produced no new results, and we exhausted all options given the best available data.

3.2. Model Predictions Versus Measured Field Data

It is important to remember that our model uses SPARROW data to predict the average annual inorganic sediment accretion from riverine and terrestrial sources to all coastal streams, while the reference sites (N = 93) measure short-term (1–5 years) organic and inorganic sediment accumulation from all sources, which include riverine, terrestrial, marine, and estuarine waters. Therefore, the measurements could not be used for validating the model predictions or calibrating the model.

Comparison of the predicted versus measured outcomes reveals insights into tidal wetland processes (Figure 9). At locations where the predicted accretion is equal to or greater than measured, riverine sediment is sufficient to account for all the measured accretion. Where predicted sediment accretion is less than measured accretion, this means that other sources of sediment, such as marine and estuarine, contribute to the sediment supply.

In the Northeast U.S., the majority of field measurements have sediment deposition coming from a combination of marine, estuarine, and riverine sources. In the Gulf of Mexico, the Mississippi Delta has a surplus of riverine sediment, and the western and eastern Gulf of Mexico have multiple sources of sediment. Along the Pacific coast, the north is dominated by riverine sediment, while the south is a mixture of riverine, marine, and estuarine sediment.

3.3. Regional Patterns and Trends

Summary statistics for LP and TP (

Table 4. Summary statistics for each SPARROW region, where TP is Terminal Path and LP is Level Path.

Statistic	Midwest	Southwest	Pacific	Southeast	Northeast
Total Area (km2)	104,628	368,558	428,832	606,200	406,025
Total Wetland Area (km2)	50,084	34,531	48,106	156,251	79,879
Percent Wetland	47.9%	9.4%	11.2%	25.8%	19.7%
Total Tidal Wetland Area (km2)	5957	2296	1044	6356	3582
Percent Tidal	5.7%	0.6%	0.2%	1.0%	0.9%
Total Tidal Wetland Area TP (km2)	3925	1574	661	5271	3016
Total Tidal Wetland Area LP (km2)	3925	1574	661	5271	3016
Percent of Tidal Used in Model	65.9%	68.6%	63.3%	82.9%	84.2%
Average RSL 2020 1M (cm/yr)	1.09	0.78	0.31	0.66	0.93
Number of TP streams (N = 16,970)	552	3600	4226	7712	880
Number of LP streams (N = 171,902)	8214	82,534	11,429	66,679	3046
Average TP Watershed Size (km2)	471,866.64	217.28	729.10	477.92	2397.87
Average LP Watershed Size (km2)	382.8	8.51	134.84	10.58	274.74

) were computed by region (Midwest, Northeast, Pacific, Southeast, and Southwest). A two-sample t-test revealed a significant difference in average watershed size between TP and LP (p < 0.01), as TP represents the entire estuary, while LP represents sub-estuary tributaries. Within each region, there were no significant differences between TP and LP for mean riverine sediment (ACCL), mean sediment yield, mean sediment load, mean sediment thickness, and mean accretion balance.

Comparing watershed size and accretion balance across regions showed a significant difference (p = 0.02). For instance, the Midwest and the Pacific have the largest average watershed size, but the Midwest has the lowest accretion balance for LPs, while the Pacific has high accretion balances. Similarly, the Southwest (western Gulf of Mexico) has the highest accretion balance but average watershed size. There are stark differences in accretion balance by region, with the Pacific region generally having a much lower density of tidal wetlands per area in comparison to the other regions. For additional summary comparisons among the regions, including an assessment of the SPARROW regression analysis uncertainty, please see Ensign et al. [20].

3.4. Spatial Statistical Analysis

Spatial statistical analysis was conducted to compare the model results across the SPARROW regions to identify spatial patterns. Results from the global spatial statistics were all statistically significant (

Table 5. Global spatial statistics, z-score * and p-value, for each SPARROW region.

Attribute	Global Moran’s I z-Score	Global Moran’s I p-Value	ISA ** Peak Distance (m)	ISA Peak Distance z-Score	General G z-Score	General G p-Value
Northeast
LP Thickness	8.56	0.00	100,000	9.55	8.97	0.00
LP Accretion Balance	8.56	0.00	150,000	7.00	9.51	0.00
TP Thickness	71.45	0.00	250,000	221.26	111.72	0.00
TP Accretion Balance	71.45	0.00	250,000	205.51	118.77	0.00
Midwest
LP Thickness	2.24	0.03	15,000	3.74	1.78	0.07
LP Accretion Balance	2.15	0.03	15,000	3.69	1.60	0.11
TP Thickness	2.74	0.01	10,000	4.88	2.66	0.01
TP Accretion Balance	2.74	0.01	10,000	4.88	2.66	0.01
Pacific
LP Thickness	1.71	0.09	260,000	4.22	2.41	0.02
LP Accretion Balance	1.71	0.09	250,000	4.22	2.38	0.02
TP Thickness	3.14	0.00	60,000	8.21	5.83	0.00
TP Accretion Balance	3.14	0.00	60,000	8.21	5.83	0.00
Southeast
LP Thickness	4.42	0.00	110,000	3.47	3.84	0.00
LP Accretion Balance	4.42	0.00	250,000	2.88	4.09	0.00
TP Thickness	32.31	0.00	15,000	33.95	29.00	0.00
TP Accretion Balance	32.31	0.00	15,000	33.95	28.99	0.00
Southwest
LP Thickness	1.79	0.07	10,000	7.23	0.91	0.36
LP Accretion Balance	1.79	0.07	10,000	7.23	0.93	0.35
TP Thickness	1.52	0.13	10,000	7.01	1.50	0.13
TP Accretion Balance	1.52	0.13	10,000	7.01	1.53	0.13

* z-scores with values between 1.65 and 1.96 are significantly clustered (p = 0.10), 1.96 to 2.58 are more significantly clustered (p = 0.05), and greater than 2.58 are the highest significantly clustered (p = 0.01). ** ISA refers to Incremental Spatial Autocorrelation.

). There was very little difference between thickness and accretion balance within each region. This is to be expected since the accretion balance is computed using predicted sediment thickness; however, the rate of sea-level rise did not have an impact on these results. In comparing the regions, all z-scores for Global Moran’s I, Incremental Spatial Autocorrelation (ISA), and General G were all higher for TP in comparison with LP.

The distances for peak z-scores are greatest for the largest regions. For example, the peak distances for the Northeast and Pacific were much larger (250,000 m and 260,000 m, respectively) than the other smaller regions, such as the Midwest and Southwest (15,000 m and 10,000 m, respectively). The southeast had varying peak distances, which indicates differences in significant clustering from local to regional distances, and this region also had the most geographically complex shape that spanned both the Atlantic and Gulf coasts, whereas the other regions were generally linear coastal geographies.

Considering Moran’s I z-scores for TP thickness, the regions from largest to smallest were Northeast (71.45, p = 0.00), Southeast (32.31, p = 0.00), Pacific (3.14, p = 0.00), Midwest (2.74, p = 0.01), and Southwest (1.52, p = 0.13).

Using the significant results for each region, Getis-Ord Gi* local hot spot analysis identified watersheds with significant clusters of high and low values of thickness for both LP and TP (Figure 10).

Riverine sediment thickness varies from region to region, where the Northeast, Midwest, and Southwest having substantially more sediment in comparison with the Pacific and Southeast. In the Northeast, the LP and TP streams vary considerably, where there are clusters of low sediment thickness for LP and clusters of high thickness for TP. This is particularly noteworthy in Penobscot Bay, where there is a high cluster of LP thickness, and in Chesapeake Bay, where there is a high cluster of TP thickness. TP clusters of low values are along most of the Atlantic coast, where the model predicts less than expected riverine sediment. In the Pacific region, where the watersheds are dominated by high topographic relief and smaller wetlands, the region is dominated by no significant clusters. There are a few exceptions, such as a cluster of low LP thickness in San Francisco Bay and high thickness in Pelican Bay in Northern California. In the southeast, the region has low riverine sediment where both LP and TP have clusters of low thickness. For example, the Georgia Embayment on the Atlantic and the Big Bend region of Florida on the Gulf Coast have significantly low clusters. In the Midwest region, the LP and TP have vastly different predictions of sediment thickness and resulting clusters. For example, the Mississippi Delta has high TP sediment thickness, but significant clusters of less than expected thickness on the western section of the coast. For example, LP and TP results are similar maps of thickness and clusters where watersheds near Houston, in Galveston Bay, have high clusters of sediment and the opposite, low clusters, in Aransas Bay between Victoria and Corpus Christi.

4. Discussion

The goal of this research was to investigate the contribution of river sediment to the coastal environments of the contiguous US. To develop the model, we explored many data sources and methods for utilizing these data. Through a collaborative and multidisciplinary approach, we developed a model that works for this large geography with vastly different coastal geomorphologies. To our knowledge, this is the first product that comprehensively assesses the contribution of river sediment to the varying rates of sediment accretion in tidal wetlands of the contiguous US. The results have served coastal conservation professionals and researchers through a variety of products, presentations, and applications that guide decision-making related to coastal wetland management. The current study has explored the challenges of linking statistical models of watershed sediment flux with coastal geospatial models of tidal wetlands. Below, we summarize the challenges and future research that are needed within the context of data availability, coastal modeling, and resiliency.

4.1. Data and Modeling Challenges

We explored many federal public data sources, and through iterative data exploration, we settled on SPARROW, NHDPlus HR, NWI, and NOAA inundation areas. For example, given that SPARROW is the only national product with sedimentation rates, we first investigated joining all data to this product, but the catchment polygon geography does not accurately represent hydrologic units, or watersheds, and the goal was to identify potentially vulnerable wetlands. There is substantial evidence that mapping the elevation of wetlands is prone to difficulties when using remote sensing versus RTK field instruments [37,38,39]. In the approach we have taken in developing this framework, there is no need to know the elevation of each tidal wetland in the contiguous US. Instead, we have removed this known problem by using the amount of sediment accretion from river sources (SPARROW) and applied these data to each coastal stream segment. This accretion rate is then compared with the local rate of SLR to determine the accretion balance. Using this approach, there is no need to know the explicit elevation of each wetland surface, and therefore, the problems of DEMs in the coastal zone are not an issue because the elevation surface is not needed.

The riverine suspended sediment bulk density is a parameter in our model that warrants investigation by the user when investigating a particular estuary. Noting that measurements of bulk density of suspended sediment (not deposited sediment that contains a mixture of organic and inorganic sources) are rarely measured or reported, our assumption of 1560 kg m⁻³ is a general value helpful for standardizing model output. Users with particular knowledge of a river sediment source may substitute a more accurate value in Equation (7) to calculate accretion thickness. Using the website (https://tinyurl.com/SedimentPancake (accessed on 1 July 2025)), users can investigate the data at specific estuaries and customize parameters such as bulk density and sea-level rise rate with their own information. The sediment thickness and accretion balance calculations (Equations (7) and (8)) are easy to recalculate on a case-by-case basis, and all attributes that are needed to perform the calculations are provided. Importantly, our modeling approach incorporated uncertainty in river sediment load using the reported root mean square error of the SPARROW model.

Sea-level rise rate is another parameter that can be customized with more current or site-specific information that a user may have. The sea-level rise rate used in each estuary calculation is provided to the user, and the user may substitute their own value using Equation (8) to recalculate the accretion balance.

4.2. Status of Public Environmental Data in the United States

As geographers and spatial data scientists are well aware, all cartographic data is inherently outdated. Even the most current data, such as real-time sensors, has a delay from recording to dissemination. So, care must be taken when deriving conclusions from data that may no longer be representative of the current landscape. For example, the data used in this project were selected because they met the first criterion, which was that the data were available for the entire study area. It is likely that there are more current data sources available from local or published research results, but the drawback would be inconsistency across the United States, which would necessitate a separate model for each original data source.

The US Fish & Wildlife Service National Wetland Inventory (NWI) data are the most detailed wetland data, but they vary in collection dates, where some coastal areas have old wetland maps from the 1970s, while others are more recent. Given the difficulty and cost to create very detailed wetland classification maps, the need to map both vegetation type and wetland function, several agencies are collaborating to utilize newer satellite products rather than field-based mapping, which is a dilemma that has plagued the wetland mapping community for many decades [40]. Given that map products age quickly, especially in areas that are rapidly urbanizing, the need for updated maps is exceeding the rate at which wetland maps are being produced.

The USGS is developing several new geospatial wetland products, such as the National Unvegetated to Vegetated Ratio (UVVR), to address coastal management needs (https://geonarrative.usgs.gov/us_coastal_wetland_collection/ (accessed on 1 July 2025)). The UVVR is a calculation of the average annual fraction of vegetation derived from 30 m Landsat imagery from 2014 through 2018 [41,42]. Although it has been calculated for the contiguous US, the spatial extent of UVVR is limited to tidal wetlands sourced from NWI data, which is outdated in many parts of the country, and the raster product does not represent polygonal marsh areas well [43].

The USGS is also developing standardized marsh units based on the NWI wetlands footprint. Marsh units, areas between high-elevation ridge lines, are calculated based on flow accumulation and surface slopes determined from relative elevations using 1 m digital elevation models [44]. This product may be useful for generalized coastal resiliency planning, but it does not overlay well with other existing data, such as the USGS National Hydrography Dataset.

4.3. Future Directions for Data to Support Coastal Models

The contribution of different land uses to coastal sedimentation is an important factor that is considered within the SPARROW multivariate regression model. For details regarding each variable, we recommend reading the documents found on the SPARROW Mappers website (https://www.usgs.gov/mission-areas/water-resources/science/sparrow-mappers (accessed on 1 July 2025)). In the SPARROW model, urban, agricultural, and other land use is considered when computing the rate of sedimentation. However, other sources of data are updated more frequently and could be considered in future research. For example, two primary data sources that are useful for supporting coastal habitat models but have yet to be investigated holistically for the contiguous US are the National Land Cover Dataset (NLCD) and terrain models using Light Detection and Ranging (LiDAR).

The NLCD is available from the Multi-Resolution Land Characteristics (MRLC) Consortium (https://www.mrlc.gov/data (accessed on 1 July 2025)). This national effort provides a comprehensive temporal perspective on general land cover classes from 2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019, 2021, and 2023 for the contiguous US, while Hawaii and the Pacific Islands are available for 2001, 2005, 2011, and Alaska has data for 2001, 2011, and 2016. Future research can investigate the usefulness of the NLCD and the NWI data to determine tidal wetlands by comparing these newer data with the historic NWI and with a probability ranking for the likelihood of the MRLC wetlands being tidal or non-tidal. If data integration is successful, these new tidal wetland data could be used to calculate sediment accretion rates for the highest-ranking tidal wetlands. At this time, our model has been developed using vector NWI data, which is the highest topologically correct type of data to compare with the USGS NHD data. In future work, a comparison can be created using the latest raster-based land cover product to determine if a combined vector (NHD) and raster (NLCD) model sufficiently estimates sediment accretion in tidal wetlands.

To assist with the conversion from MRLC land cover to tidal wetlands, regional mapping efforts, such as the Pacific Marine and Estuarine Fish Habitat Partnership (PMEP), can potentially provide accurate coastal wetlands maps (https://www.pacificfishhabitat.org/data/ (accessed on 1 July 2025)) for this region of the US. If data framework standards are adopted, then regional data could be combined with other data to produce a better contiguous US map; however, this requires extensive collaborations and agreed-upon standardized data sources and classification schemes across many regional programs to be successfully integrated for a comprehensive assessment of tidal wetlands.

Linked with the USGS NHD is the Water Quality Portal, which provides sampling data for the US and elsewhere (https://www.waterqualitydata.us/ (accessed on 1 July 2025)). The Portal, a collaboration among USGS, EPA, and the National Water Quality Monitoring Council, provides access to a variety of data sources such as streamflow from USGS National Water Information System (NWIS) and EPA Storage and Retrieval Water Quality Exchange (STORET). Although the Portal does not provide data visualization or geospatial analysis, it provides access to over 290 million records across 2.7 million sites, an unprecedented achievement that illustrates agency collaboration and potential for growth in big data dissemination [45].

LiDAR data is becoming more widespread for data collection via satellite, airborne, and UAS platforms, and these elevation data are being utilized in many coastal studies. For example, LiDAR-derived Digital Terrain Models (DTMs) have been widely used to classify coastal geomorphology/landforms [46,47,48,49,50,51,52,53,54,55,56,57]. LiDAR point clouds and the derived topographic surface (DEM) have been used to assess global flood risk [58,59] and improve map accuracy for coastal wetlands [60]. These substantial achievements in remote sensing will ultimately make the development of large area mapping at high spatial resolution feasible. Importantly, elevation is a primary factor in assessing wetland vulnerability to SLR. Therefore, developing LiDAR mapping standards along the coast and incorporating these products into vulnerability assessments will be a key milestone in the future development of the tidal wetland vulnerability framework. The framework presented here is the first step to providing a national assessment of tidal wetland vulnerability due to riverine sedimentation rates; however, there is much future work that can take place to address burgeoning GIS and remote sensing data sources, as well as identifying and quantifying additional factors that influence tidal wetland sustainability.

4.4. Future Directions for Large-Scale Models to Support Rapid Assessment of Flood Risk

Several models exist that predict the extent of storm surge and pluvial flooding, but they are time-consuming to generate [61]. A recent review of several models illustrates their potential for mapping the impacts of sea-level rise, but there are limits to the models, as some are best suited for local site-specific assessment, some are regional/global with coarse resolution and broad assumptions, and although some are supported by NOAA, there is little consistency and interoperability between models [62].

From a human and environmental health perspective, there is a need for rapid assessment of changing conditions to inform decision-making and emergency response. Synthetic Aperture Radar (SAR), such as Sentinel-1, is a data source that can be quickly processed to identify water inundation even when weather conditions are not amenable to other satellite sensors. The key factor in the analysis is to be fast so that emergency management can inform evacuation and rescue efforts. Therefore, research has developed reliable and fast methods for processing Sentinel-1 imagery for flood identification [63]. Uncrewed Aerial Vehicles (UAVs) are also being used to collect SAR imagery for rapid local mapping of water level changes, even in riparian areas covered by dense vegetation canopy and coastal marshes that are difficult to map because of changing shallow water environments [64]. Another approach to rapidly assessing flood conditions is to use the USGS stream gauges to compare the peak water level with the 10-year average to create a flood ratio [65]. However, the spatial distribution of USGS gauges is insufficient to support emergency management.

Approaches to assessing coastal flooding are quickly evolving. The combination of spatial statistics, storm surge models, predictions of sea-level rise, and high spatial resolution topography enables users to visualize flood risk [66,67]. The Federal Emergency Management Agency (FEMA) relies on the 100-year floodplain as the limit that requires flood insurance, but far too often, flood losses occur outside the floodplain [68]. Floodplain mapping has developed over many decades, and FEMA’s Flood Insurance Rate Maps (FIRMS) are available nationally, but the mapping techniques are focused on hydrologic models of fluvial flooding, not the increasingly problematic pluvial flooding. Although research has identified alternative approaches to modeling flood risk, the site-specific examples are not widely adopted and have little support from the planning and policy decision-makers. For example, flood inundation hydrologic models have been correlated with post-storm damage assessment data, but these maps have not been used to change flood mitigation policy [69,70]. Research should continue to (1) test new flood mapping techniques that incorporate wetlands, geomorphology, soils, and geospatial techniques (e.g., topography, distance, adjacency, etc.) [71,72,73,74] and (2) further develop communication tools for flood mapping, flood frequency, the importance of pluvial flooding, and related land cover to encourage community involvement and decision-making [75,76,77].

Lastly, real-time or near-real-time precipitation and hydrologic data should utilize geospatial technologies, including remote sensing and GIS, to provide flood alert warning systems [78]. Research is progressing towards an integrated flood forecasting framework that integrates the NHD and National Water Model (NWM), but it has only been tested in selected areas [79] and is currently on hold while agencies decide the best methods for updating NHD and NWM.

5. Conclusions

There were two primary objectives in this study: (1) to test multiple public data sources to build a framework to map coastal sedimentation in tidal wetlands, and (2) to utilize these data to build a framework that is implemented for the entire contiguous US coast. When embarking on such lofty objectives, there are challenges along the way. We used an iterative approach that tested model output at every step of data processing, in many geographic locations, and with input from many stakeholders. As we discovered nuances and difficulties in the source data, we regrouped and reflected on the overarching goals and adapted over time. This study has used an interdisciplinary collaborative approach to iteratively develop a model that has computed river sediment accumulation for all tidal wetlands of the contiguous US.

To our knowledge, this is the first continental GIS model that predicts the contribution of riverine sediment to the rate of accumulation on tidal wetlands and compares this with current rates of sea-level rise. Importantly, these calculations do not include wetland surface accumulation from marine sediments and organics. Therefore, wetlands that persist through time and have low sediment accumulation rates from riverine sources, even accounting for changing sea level, must be maintaining the marsh surface elevation through non-riverine sources. This distinction is important as it informs the science and coastal policy communities to investigate these alternative sources so that the management of these resources is monitored and protected. Additionally, research has shown that modeled coastal wetlands are slow to change with rising sea levels; however, there is a tipping point when marsh degradation quickly accelerates in low-slope coastal areas, where the coastal change moves from slow to suddenly rapid erosion [80]. Biodiversity, carbon storage, and flood resiliency will dramatically change. To address this pressing need, coastal communities worldwide are assessing risk, identifying adaptation strategies, and developing resiliency plans [81]. The coastal framework presented here is an additional tool, one of many, that planners and elected officials can choose to integrate into their decision-making processes.