Area 4: Integration and Forecasting

Description Objectives Outcomes Show All  

Overview

In Area 4 our objectives are to synthesize information collected in Areas 1-3 to produce an integrated picture of habitat provisioning and carbon flow across the landscape, and evaluate how changes in salinity and inundation may change these services in the future. We will accomplish this with a combination of integrative modeling, empirical observations, and remote sensing.

We use a combination of integrative modeling, empirical observations, and remote sensing to produce an integrated picture of habitat provisioning and carbon flow across the landscape, and evaluate how changes in salinity and inundation may change these services in the future. Major activities include A) develop an integrative model that uses a hydrodynamic model (FVCOM), a soil model, and 3 different semi-empirical plant models to predict salinity and inundation patterns, porewater salinities, and plant responses over different time scales, and B) use combined model output to evaluate habitat provisioning and C flow under different scenarios.

Components

Area 4A. Develop an integrative model

Hydrodynamics play a critical role in the distribution and transport of water and materials across the GCE domain. Consequently, an accurate hydrodynamic model is a necessary first step for our integrative modeling efforts. FVCOM will provide information on salinity and inundation patterns of the water that floods the marsh, which will be used in a soil model to predict porewater salinity and soil water content (described below), and in 3 different semi-empirical plant models. The Spartina productivity model (Area 3A) will provide information on individual plant responses to salinity and inundation; the plant community model (Area 3B) will provide spatial information on vegetation dynamics during transitions; and a modified version of the SLAMM model will predict habitat shifts at the landscape level.

Although these models will exchange information offline, they could inform each other, and our goal in GCE-3 is to lay the groundwork for fully integrated mechanistic models with multiple feedbacks. The effects of parameter sensitivity (e.g. bottom friction on the marsh in the hydrodynamic model, below-ground production parameters in the Spartina productivity model) will be investigated for each model individually and for how sensitivities and uncertainties propagate between models.

Area 4B. Evaluate habitat provisioning and carbon flow under different scenarios

We will use these models to run a series of scenarios to evaluate, through hindcasting and forecasting, how pulses and presses in our major drivers (sea level, river flow, precipitation, temperature, groundwater input, and overland runoff) will affect the domain. The models will be used to predict salinity and inundation patterns, porewater salinities, and plant responses over different time scales. Predictions will be evaluated in terms of habitat provisioning and C flow (see below). Climate change will be examined using bias-corrected, downscaled projections of the IPCC model results (Maurer et al. 2007). Human alterations will be evaluated by simulating potential modifications to shoreline armoring and overland runoff based on build-out scenarios from McIntosh County, as well as modifications in the greater Altamaha watershed (e.g. new reservoirs upstream). We will also consider scenarios with feedbacks to human behavior (e.g. building sea walls as sea levels rise, green developments). Our archeological studies will allow us to consider patterns in the pre-development landscape. We will also be able to modify our scenarios in response to experimental results showing changes in creek geomorphology or the top-down effects of blue crabs on plant production.

To evaluate dynamic habitat, we will use the continuous salinity data from our sonde network to map the locations of fresh, oligohaline (<5 PSU), mesohaline (5-15) and polyhaline (15-30) conditions across the domain, and determine how these habitat locations vary over lunar, seasonal, and annual cycles. These will be compared with results from FVCOM which will also be used to evaluate how these salinity ranges might shift given different scenarios.

Two major questions for global and regional C cycling are 1) whether the coastal zone is a net source or sink of CO2 to the atmosphere and the ocean, and 2) how such C fluxes might change over time. We will, in combination with the habitat analyses, evaluate how the changes that might occur in response to changes in salinity and inundation will affect carbon cycling, with the goal of developing new hypotheses about the implications of climate and human activities for the coastal C budget.

 

Research Objectives

Research Outcomes by Objective

  • 4A.1  Run FVCOM to predict salinity and inundation (yr 3-4)
    • 2 report:

      Activities:  2015: Begins yr 3
      2016: Maps of residence time and connectivity have been produced, and we are investigating how those vary with forcing. Changes in salinity in the estuary in response to sea level rise and storms are also being investigated.
      2017: FVCOM has been implemented for both the Duplin River and the larger GCE domain.
      2018: The merged GCE and Duplin River models allows us to produce highresolution predictions of salinity and inundation throughout the domain.

      Results:  2017: Wang et al. (submitted) used the FVCOM model to evaluate salinity variability, residence times, and connectivity in the GCE domain (Fig. 12).
      2018: Run FVCOM to predict salinity and inundation The model was upgraded to FVCOM4.0 and is running stably.

    • 2014 report:

      Activities:  Begins yr 3

    • 2015 report:

      Activities:  Maps of residence time and connectivity have been produced, and we are investigating how those vary with forcing. Changes in salinity in the estuary in response to sea level rise and storms are also being investigated.

    • 2016 report:

      Activities:  FVCOM has been implemented for both the Duplin River and the larger GCE domain.

      Results:  Wang et al. (submitted) used the FVCOM model to evaluate salinity variability, residence times, and connectivity in the GCE domain (Fig. 12).

    • 2017 report:

      Activities:  The merged GCE and Duplin River models allows us to produce highresolution predictions of salinity and inundation throughout the domain.

      Results:  Run FVCOM to predict salinity and inundation The model was upgraded to FVCOM4.0 and is running stably.

    • 2019 report:

      Activities:  Aerial photographs (Obj 2B.2) were analyzed using random forest to generate classified maps of habitat distributions for 2017 and 2018.

      Results:  Habitat maps delineating 11 tidal habitats were generated with overall classification accuracies ranging from 0.83 (2017) to 0.88 (2018). These are being used for change detection analyses.

    • 2020 report:

      Activities:  We have begun a large-scale effort to track wrack disturbance on the landscape via regular drone flights. During the first year, we identified a test site that we used to optimize flight conditions, develop a workflow for image processing and classification, and standardize protocols for ground-truthing. We also used the test site to obtain ground-truth data for field validation of NDVI. In Dec. 2019 we began regular monthly flights at a 23 ha site (Airport Marsh). For each flight we identify perturbations (wrack packets) using PCA and machine learning techniques and then compare them with previous imagery in order to quantify patch size, frequency, and longevity. Drone imagery is used to guide selection of patches for field monitoring of plants, invertebrates, porewater, decomposition, and wrack characteristics, and we have also set up motion-activated cameras in the field to capture visits from mammals and birds. New wrack patches are added to the monitoring effort each quarter. Initial sampling of areas affected by wrack show that plant densities decreased and stem height increased in wrack patches, and that densities of the herbivorous marsh crab Sesarma reticulatum increased significantly in wrack-covered areas relative to controls.

    • 2021 report:

      Activities:  We continued monthly drone flights at the Dean Creek site, which are used to guide selection of wrack patches for field sampling. We are currently sampling a suite of variables (plants, invertebrates, porewater, temperature, decomposition) in 16 wrack patches (plus paired controls).

      Results:  We have not seen differences to date in areas affected by wrack in terms of decomposition, soil organic matter, or ammonium concentrations. However, there is a significant decline in Spartina, snail, and crab density; initial observations suggest they take 7-11 months to recover.

  • 4A.2  Run the soil model to predict porewater salinity (yr 4-5)
    • 2 report:

      Activities:  2015: Begins yr 3
      2016: Model simulations have been performed to assess seasonal patterns and inter-annual variations. We also performed a sensitivity analysis to quantify the role of different external forcings (e.g. precipitation), or model parameters (e.g. soil hydraulic conductivity).
      2017: We have run the soil model over several years to assess seasonal and interannual variability, and have begun a comparison of the model with patterns seen in Landsat data.
      2018: The porewater model is being used to relate soil conditions to satellite observations and fall monitoring data, and to provide contextual data for the Spartina model (Obj. 4A.3).

      Results:  2017: A sensitivity analysis conducted on the soil model shows that porewater salinity in the Spartina zone is controlled by tidal salinity whereas high marsh plants are sensitive to changes in ET and precipitation.
      2018: Run the soil model to predict porewater salinity Model simulations show spatial and interannual variability of porewater salinity.

    • 2014 report:

      Activities:  Begins yr 3

    • 2015 report:

      Activities:  Model simulations have been performed to assess seasonal patterns and inter-annual variations. We also performed a sensitivity analysis to quantify the role of different external forcings (e.g. precipitation), or model parameters (e.g. soil hydraulic conductivity).

    • 2016 report:

      Activities:  We have run the soil model over several years to assess seasonal and interannual variability, and have begun a comparison of the model with patterns seen in Landsat data.

      Results:  A sensitivity analysis conducted on the soil model shows that porewater salinity in the Spartina zone is controlled by tidal salinity whereas high marsh plants are sensitive to changes in ET and precipitation.

    • 2017 report:

      Activities:  The porewater model is being used to relate soil conditions to satellite observations and fall monitoring data, and to provide contextual data for the Spartina model (Obj. 4A.3).

      Results:  Run the soil model to predict porewater salinity Model simulations show spatial and interannual variability of porewater salinity.

    • 2019 report:

      Activities:  We extended our studies of Spartina biomass (O'Donnell & Schalles 2016) by applying our algorithm to Landsat 8 data and expanding our analysis to the Georgia coast. We are also optimizing our algorithm to generate estimates of below-ground biomass, and continue to collect monthly samples of Spartina above- and below-ground biomass for groundtruthing.

      Results:  Our expanded analysis of long-term trends in Spartina biomass was applied to 7 USGS HUCs along the Georgia coast and showed declines in all but the Altamaha HUC (which has the most freshwater input). These data were presented at several meetings and are being included in a cross-coastal LTER climate synthesis manuscript. A manuscript describing below-ground biomass trends is also in prep.

    • 2020 report:

      Activities:  The headward erosion of tidal creeks has been linked to grazing by Sesarmid crabs (Vu et al. 2017; Vu and Pennings 2018). We have evidence that creeks affected by this phenomenon have become increasingly prevalent over the past few decades, which has implications for invertebrate communities and predator-prey interactions on the marsh platform (Crotty et al., 2020; key findings). We took advantage of a space-for-time substitution to evaluate marsh disturbance and recovery due to this phenomenon (Wu et al., in press). We found multiple patterns of disturbance magnitude and recovery trajectory for the various response variables that belied any simple univariate understanding of "disturbance and recovery." We are also calculating rates of creekbank slumping/accretion to assess the relative importance of different geomorphic features in controlling erosion.

    • 2021 report:

      Activities:  We plan to use the monthly drone flights (Obj. 4A1) and aerial photos (Obj. 2B2) to evaluate changes in creek configuration and creek slumping over time.

      Results:  Wu et al. (2021) found sharp differences in the effects and recovery trajectories of different variables in response to headward-eroding creeks.

  • 4A.3  Run the plant models to predict vegetation response yr (2-6)
    • 2 report:

      Activities:  2015: We have developed scenarios and the model infrastructure for the Spartina model. However, biomass predictions in initial runs diverged from observations after 18-24 mo, partly because resource allocation was not incorporated into the model. (See Obj. 3A7).
      2016: The Spartina model is currently being validated against biomass data that was collected in the field as part of the GCE LTER and also against literature data from other locations.
      2017: The plant model is being revised to include mechanistic transport from above to below ground tissues based on field observations. We have also been working on developing links to the GCE soil model.
      2018: We have incorporated a salinity and inundation response into the plant model in order to be able to make predictive simulations of high and low salinity years as well as high and low discharge.

      Results:  2018: Run the plant models to predict vegetation response We are currently running the hydrodynamic and soil models to see if historical patterns of porewater salinity as predicted by the model explain changes in plant community composition in longterm monitoring plots.

    • 2014 report:

      Activities:  We have developed scenarios and the model infrastructure for the Spartina model. However, biomass predictions in initial runs diverged from observations after 18-24 mo, partly because resource allocation was not incorporated into the model. (See Obj. 3A7).

    • 2015 report:

      Activities:  The Spartina model is currently being validated against biomass data that was collected in the field as part of the GCE LTER and also against literature data from other locations.

    • 2016 report:

      Activities:  The plant model is being revised to include mechanistic transport from above to below ground tissues based on field observations. We have also been working on developing links to the GCE soil model.

    • 2017 report:

      Activities:  We have incorporated a salinity and inundation response into the plant model in order to be able to make predictive simulations of high and low salinity years as well as high and low discharge.

      Results:  Run the plant models to predict vegetation response We are currently running the hydrodynamic and soil models to see if historical patterns of porewater salinity as predicted by the model explain changes in plant community composition in longterm monitoring plots.

    • 2019 report:

      Activities:  As part of an ROA supplement awarded in 2019, we sampled 33 stations during successive neap and spring tides to obtain high-temporal-resolution analysis of carbonate chemistry. We also continued processing cores from all 11 GCE sites for C content and radioisotope dates.

      Results:  (which has the most freshwater input). These data were presented at several meetings and are being included in a cross-coastal LTER climate synthesis manuscript. A manuscript describing below-ground biomass trends is also in prep.

    • 2020 report:

      Activities:  Dieback is an important drought-associated disturbance in the GCE domain (Silliman et al. 2005, McFarlin et al. 2015, Angelini et al. 2016). If we experience a drought or see other signs of perturbation (i.e., plants thinning due to increased inundation), these should be captured by our drone flights. Our plan is to use a similar field protocol to that being used for the wrack assessment to follow these areas in the field. To date, however, no droughts or dieback have been observed during GCE-IV.

    • 2021 report:

      Activities:  We identified several incipient dieback areas (areas where the plants turned prematurely brown) using the drone flights over Dean Creek and have set up plots in two areas where we are sampling the same variables as those being followed in the wrack patches (Obj. 4A1).

      Results:  Hensel et al. (2021) found that hogs maintain large disturbed patches in marshes by feeding on both plants and mussels. This paper, which was in Nature Communications, has had significant news coverage, as it showed that megafauna can reduce the resilience of salt marshes.

  • 4A.4  Ecosystem Properties - Carbon budgets
    • 2020 report:

      Activities:  Visualization of the cumulative distribution of wrack shows that it is concentrated at channel edges and is quite dynamic over time (Fig. 1). Our goal is to develop a cradle-to-grave view of wrack and other perturbations and to use them to produce scaled-up estimates of the system-wide importance of these events for ecosystem properties such as NPP.

    • 2021 report:

      Activities:  This past year we streamlined the workflow for processing drone imagery using PCA, and also collected temperature data to calibrate the drone’s thermal band to aid in identifying wrack patches.

      Results:  Regular drone flights over the airport marsh site (7 flights from Jul-2019 through May-2020) showed that most wrack is found close to the water’s edge, but it persists longer at higher elevations. Although wrack only affected ~5% of the site, wrack patches can be persistent or re-occur in the same spot repeatedly: 1/3 of the wrack-affected pixels were covered in more than 1 image (Fig. 1).

  • 4B.1  Develop scenarios (yr 3)
    • 2 report:

      Activities:  2015: Begins yr 3
      2016: Calendar year daily statistics of Altamaha River discharge over the period 1932-2014 were calculated and binned into 3 levels of discharge in order to define periods of wet, dry, normal and variable years for use in modeling efforts (Activities Fig. 4).
      2017: The hydrodynamic model for the Duplin is currently running for the time period Aug 2012 to Dec 2015 which will simulate dry, wet, and normal years of river discharge effects. The model for the entire domain has been run for these years as well and is being used to run simulations representing different levels of sea level rise.
      2018: Completed yr 4.

      Plans:  2016: We plan to develop further scenarios that can be used in model runs.

    • 2014 report:

      Activities:  Begins yr 3

    • 2015 report:

      Activities:  Calendar year daily statistics of Altamaha River discharge over the period 1932-2014 were calculated and binned into 3 levels of discharge in order to define periods of wet, dry, normal and variable years for use in modeling efforts (Activities Fig. 4).

      Plans:  We plan to develop further scenarios that can be used in model runs.

    • 2016 report:

      Activities:  The hydrodynamic model for the Duplin is currently running for the time period Aug 2012 to Dec 2015 which will simulate dry, wet, and normal years of river discharge effects. The model for the entire domain has been run for these years as well and is being used to run simulations representing different levels of sea level rise.

    • 2017 report:

      Activities:  Completed yr 4.

    • 2019 report:

      Activities:  We launched a large-scale effort to track disturbances via regular drone flights. We acquired a Matrice 200 drone with a Micasense RedEdge Altum camera, obtained appropriate permits and FAA licenses, performed tests to optimize flight conditions, began monthly flights over an initial test site, and obtained ground-truth data for field validation of disturbances. We have developed a work flow for processing imagery and are optimizing our algorithm to detect disturbances.

      Results:  Our initial drone imagery is already yielding important insights into patch dynamics in the salt marsh, with evidence of shifting areas of wrack from month to month that leave residual signals on the landscape (Fig. 7).

    • 2020 report:

      Activities:  As sea-level rise causes salt water to intrude further upstream, there is the potential that there will be upstream shifts in intertidal vegetation (e.g., brackish marsh converting to salt marsh). In the GCE domain this would manifest most clearly along the salinity gradient of the Altamaha River estuary. We therefore have multiple efforts underway to detect these changes: We conduct an annual survey of the distribution of bankside Spartina cynosuroides (characteristic of brackish marshes) and sample permanent plots at four sites with mixed vegetation as part of our fall monitoring effort. In 2018 we added observations of vegetation on Broughton Island, which is located in the middle of the estuary and has a dynamic mix of oligohaline and mesohaline species and also established an annual photo survey of bankside trees in the tidal fresh forest. We developed a random forest classifier to generate maps delineating 11 tidal habitat types based on orthoimagery collected in 2017 and 2018, which we are using for change detection analysis associated with the passage of Hurricane Irma. This past year we acquired Sentinel-2 satellite imagery to map marsh and forest distributions along the corridor and we are also collecting ground-truth observations in the forest. We are particularly interested in evidence for tree mortality as the result of salt encroachment.

    • 2021 report:

      Activities:  We collected ground reference data in the tidal fresh forest to improve and validate our habitat classification of Sentinel-2 imagery. We also continue annual bankside surveys along the Altamaha River salinity gradient, as well as monitoring of mixed vegetation on Broughton Island.

      Results:  Hierarchical clustering using ground reference data were used to create a dataset that will be used for classification of Sentinel-2 data. An initial classification using the MLC classifier was quite promising (Fig. 2).

  • 4B.2  Evaluate C flow (yr 3-6)
    • 2 report:

      Activities:  2015: Begins yr 3
      2016: To better understand the factors controlling seasonal CO2 fluxes and the extent of autotrophy/respiration in the coastal South Atlantic Bight (SAB), we measured pCO2 from the GCE domain in each season (April, July, Sept., Dec.) as part of the series of oceanographic cruises conducted this past year. Underway pCO2 was measured in cross-shelf transects, and discrete samples were also collected for DIC,Total Alkalinity (TA), and pH measurements. These samples are currently being processed. We also collected and dated 4 cores at GCE monitoring sites to examine carbon sequestration rates.
      2017: Samples for DIC, Total Alkalinity, and pH collected during cruises conducted in the South Atlantic Bight during year 3 are being processed.
      2018: Samples for DIC, Total Alkalinity, and pH collected during cruises conducted in the South Atlantic Bight have been processed and are awaiting QA/QC.

      Results:  2016: High resolution maps of sea surface pCO2 over the region suggest that while the estuarine zones are a strong source of CO2 to the atmosphere, the SAB shelf is a net sink of atmospheric CO2 during all seasons (Results Fig. 8). These data will also be used to evaluate the extent of in-situ DIC generation and export from estuarine zones to the coastal ocean.
      2017: Observations of DIC, TA, and pH collected during GCE cruises are being used to validate the NOAA Gray's Reef National Marine Sanctuary CO2 time series.
      2018: Evaluate C flow Reimer et al. (2017) found that a longterm increase in pCO2 occurred in the coastal ocean despite the fact that neither discharge nor salinities had changed significantly (Fig. 11). They suggest that the trend is potentially due to increases in both riverine DIC concentration and flux from intertidal areas.

    • 2014 report:

      Activities:  Begins yr 3

    • 2015 report:

      Activities:  To better understand the factors controlling seasonal CO2 fluxes and the extent of autotrophy/respiration in the coastal South Atlantic Bight (SAB), we measured pCO2 from the GCE domain in each season (April, July, Sept., Dec.) as part of the series of oceanographic cruises conducted this past year. Underway pCO2 was measured in cross-shelf transects, and discrete samples were also collected for DIC,Total Alkalinity (TA), and pH measurements. These samples are currently being processed. We also collected and dated 4 cores at GCE monitoring sites to examine carbon sequestration rates.

      Results:  High resolution maps of sea surface pCO2 over the region suggest that while the estuarine zones are a strong source of CO2 to the atmosphere, the SAB shelf is a net sink of atmospheric CO2 during all seasons (Results Fig. 8). These data will also be used to evaluate the extent of in-situ DIC generation and export from estuarine zones to the coastal ocean.

    • 2016 report:

      Activities:  Samples for DIC, Total Alkalinity, and pH collected during cruises conducted in the South Atlantic Bight during year 3 are being processed.

      Results:  Observations of DIC, TA, and pH collected during GCE cruises are being used to validate the NOAA Gray's Reef National Marine Sanctuary CO2 time series.

    • 2017 report:

      Activities:  Samples for DIC, Total Alkalinity, and pH collected during cruises conducted in the South Atlantic Bight have been processed and are awaiting QA/QC.

      Results:  Evaluate C flow Reimer et al. (2017) found that a longterm increase in pCO2 occurred in the coastal ocean despite the fact that neither discharge nor salinities had changed significantly (Fig. 11). They suggest that the trend is potentially due to increases in both riverine DIC concentration and flux from intertidal areas.

    • 2019 report:

      Activities:  We used Spartina biomass from clip plots to calibrate a biomass algorithm based on drone reflectance, and we are now working to pair these high-resolution observations with satellite imagery (WorldView, Sentinel, Landsat, MODIS).

      Results:  Spartina biomass maps were successfully produced from the drone imagery flown at the Airport Marsh test site (Fig. 8). Spartina biomass estimates were barely affected (<1.3%) when scaled up using MODIS. This will be presented at the first NOAASECOORA Drones in the Coastal Zone Workshop in March, 2020.

    • 2020 report:

      Activities:  In GCE-III we used data from Landsat5 period of record (1984 to 2011) to evaluate changes in S. alterniflora biomass at a site on Sapelo Island. We have now extended these analyses both temporally (to include Landsat8) and spatially (to the entire Georgia coast). The spatial analysis was made possible by habitat maps developed for the GA coast as part of a leveraged project, following the approach we first took for the Duplin River (Hladik et al. 2013). Our expanded analysis showed long-term declines in S. alterniflora biomass along all of the GA coast except in the Altamaha River estuary (which has the most freshwater input). Over the coming years our goal is to derive similar algorithms to assess biomass patterns for brackish and fresh marsh species. We are also poised to get scaled-up GPP estimates of the domain from MODIS. We have parameterized a Light Use Efficiency (LUE) model for S. alterniflora within the flux tower site, which is an important step towards developing a model that we can use on a per-pixel basis for all of the marshes within the GCE domain. Eventually, we plan to use this framework to create a time-series of GPP estimates for the study area at 8-day intervals from 2000 to 2021.

    • 2021 report:

      Activities:  We collected ground observations of Juncus biomass, which we are using to calibrate biomass estimates derived from Sentinel-2 imagery. This will complement our existing estimates of Spartina biomass. We are also poised to get scaled-up GPP estimates of the domain from MODIS based on our parameterized Light Use Efficiency model for S. alterniflora.

      Results:  Hawman et al. (2021) evaluated the annual cycle of GPP and light use efficiency measured at the flux tower and found that the cloudiness index and daily maximum tide height were the primary factors that explained deviation in S. alterniflora light use efficiency.

  • 4B.3  Evaluate habitat provisioning (yr 3-6)
    • 2 report:

      Activities:  2015: Begins yr 3
      2016: We are starting to use FVCOM to evaluate how salinity ranges (and hence dynamic habitat) will vary with sea level rise (Activities Fig. 5).
      2017: C. Hladik led an effort to correct tidal marsh digital elevation models for salt, brackish, and tidal fresh marshes based on field observations obtained with an RTK GPS.
      2018: We combined a Landsatbased Spartina alterniflora aboveground biomass algorithm with vegetation mapping to study large scale (620 km2) seasonal phenology and interannual variation of Spartina along the Georgia coast since 1984.

      Results:  2016: McFarlin et al. (2015) evaluated the effects of the loss of Spartina alterniflora on habitat provisioning for benthic epifauna, macroinfauna and meiofauna. In the GCE domain, abundances of all invertebrate groups and the diversity of macroinfauna were lower in bare plots, with clear separation between infaunal assemblages in bare and reference (=vegetated) plots (Results Fig. 9). In contrast, there was overlap between the assemblages and the abundance of some groups (i.e. meiofauna) increased in bare plots in Louisiana, suggesting that the role of S. alterniflora is context-dependent.
      2017: Hladik et al. produced improved habitat classifications for the tidal fresh marshes in the GCE domain (Fig 13).
      2018: Evaluate habitat provisioning We have extended the spatial and temporal ranges of our remote sensing study of Spartina biomass. We found substantial differences in Spartina biomass across seven tidal watersheds that demonstrate the role of freshwater discharge in promoting productivity.

    • 2014 report:

      Activities:  Begins yr 3

    • 2015 report:

      Activities:  We are starting to use FVCOM to evaluate how salinity ranges (and hence dynamic habitat) will vary with sea level rise (Activities Fig. 5).

      Results:  McFarlin et al. (2015) evaluated the effects of the loss of Spartina alterniflora on habitat provisioning for benthic epifauna, macroinfauna and meiofauna. In the GCE domain, abundances of all invertebrate groups and the diversity of macroinfauna were lower in bare plots, with clear separation between infaunal assemblages in bare and reference (=vegetated) plots (Results Fig. 9). In contrast, there was overlap between the assemblages and the abundance of some groups (i.e. meiofauna) increased in bare plots in Louisiana, suggesting that the role of S. alterniflora is context-dependent.

    • 2016 report:

      Activities:  C. Hladik led an effort to correct tidal marsh digital elevation models for salt, brackish, and tidal fresh marshes based on field observations obtained with an RTK GPS.

      Results:  Hladik et al. produced improved habitat classifications for the tidal fresh marshes in the GCE domain (Fig 13).

    • 2017 report:

      Activities:  We combined a Landsatbased Spartina alterniflora aboveground biomass algorithm with vegetation mapping to study large scale (620 km2) seasonal phenology and interannual variation of Spartina along the Georgia coast since 1984.

      Results:  Evaluate habitat provisioning We have extended the spatial and temporal ranges of our remote sensing study of Spartina biomass. We found substantial differences in Spartina biomass across seven tidal watersheds that demonstrate the role of freshwater discharge in promoting productivity.

    • 2020 report:

      Activities:  One of the most important questions we would like to address is whether the GCE marshes will be resilient to future changes such as sea-level rise. Burns et al. (2020 a, b) used historical aerial photos to evaluate changes in marsh features over approximately 70 years (see key findings). Although the marshes were dynamic, losses (primarily due to channel widening) were largely offset by gains in other areas (primarily channel contraction, with some migration into the upland). Langston et al. (2020) used a vertical accretion model to predict that the GCE marshes will be stable in the vertical direction through 2100 due to existing “elevation capital”, in contrast to marshes elsewhere that are perched lower in the tidal range (i.e., VCR). However, the long-term stability of salt marshes remains an area of active research as there are many unknowns regarding feedbacks between flooding depth, plant growth and sediment accretion.

    • 2021 report:

      Activities:  We continued analyzing soil cores collected along elevation, salinity, and disturbance gradients, to provide insight into how shifts in vegetation and changes in creek morphology affect accretion rates and carbon accumulation.

      Results:  O’Connell et al. (2021) found that belowground biomass is declining in some interior marsh areas over time, perhaps indicating areas of low marsh resiliency to sea-level rise. We also participated in a cross-site effort (Zinnert et al. 2021) that described the state changes expected in coastal wetlands in response to long-term changes in temperature and sea level (Fig. 3).

  • 4C.1  Driver-Response Relationships - Data synthesis
    • 2019 report:

      Activities:  We have analyzed our long-term salinity data from the Altamaha River estuary to evaluate the conditions under which high salinity events occur (Fig. 6).

      Results:  We identified 79 high salinity events, ranging from 1 to 51 days, in the Altamaha River estuary over the last 16 years. These events could be explained by river flow that was objectively low based on the historical record, strong up-estuary winds, or (to a lesser extent) unusually high tides (Sheldon & Alber 2019).

    • 2020 report:

      Activities:  One of our major accomplishments in GCE-III was the implementation of a hydrodynamic model (FVCOM) in both the GCE domain and the Duplin River (McKnight 2016, Wang et al. 2017). We are now in the process of switching to the Delft3D modeling framework because of the added flexibility and additional functions available. Comparisons with observations from GCE hydrographic moorings indicate that the heat and salt balances are realistically represented. We are currently working on implementing a water quality model.

    • 2021 report:

      Activities:  We have successfully implemented the hydrodynamic, heat flux, and water quality modules of Delft3D in the GCE domain.

      Results:  Output from the Delft3D hydrodynamic model indicate that variability in temperature and salinity are realistically represented (Fig. 4).

  • 4C.2  Driver-Response Relationships - Dynamical system models
    • 2019 report:

      Activities:  We plan to use the results from the SALTEx experiments (Obj. 3B.2) as a test case for characterizing ecological responses from field data with a known perturbation (i.e. saltwater intrusion).

    • 2020 report:

      Activities:  Our soil model (Miklesh & Meile 2018) predicts porewater salinity based on hydrology and evapotranspiration. We are now developing a dynamic model of soil temperature based on a radiation balance coupled with a one-dimensional heat propagation model in the subsurface to study the effect of changes in external drivers, inundation patterns, and potential porewater mixing (Fig. 21).

    • 2021 report:

      Activities:  We are using data collected at the flux tower (Obj. 1A2) to validate and revise our soil temperature model.

      Results:  The soil model simulates radiative forcing and heat propagation in the marsh subsurface.

  • 4C.3  Driver-Response Relationships - Simulations of salinity and inundation
    • 2019 report:

      Activities:  The hydrologic monitoring data from groundwater wells will be used to investigate links between major disturbances (high rainfall events, very high tides), and vegetation

    • 2020 report:

      Activities:  We are working on both empirical and mechanistic models of plant production. Our Belowground Ecosystem Resilience Model (BERM) uses extreme gradient boosting to predict below-ground biomass of S. alterniflora based on a suite of environmental (i.e., elevation, temperature) and biological (i.e., foliar N, plant phenology) variables (O’Connell et al., in review). One of the exciting things about this effort is that it is based on above-ground proxies and can be scaled with readily available remote sensing data to evaluate spatiotemporal patterns (Fig. 3). We are also developing a whole-plant balanced growth model that we can use to evaluate the time-course of recovery from disturbance.

    • 2021 report:

      Activities:  Our Belowground Ecosystem Resilience Model uses extreme gradient boosting to predict below-ground biomass of S. alterniflora

      Results:  The BERM model (O’Connell et al. 2021) is based on above-ground proxies and can be scaled with readily available remote sensing data to evaluate spatiotemporal patterns in belowground biomass (Fig. 5).

  • 4C.4  Modeling - Develop driver-response models
    • 2020 report:

      Activities:  We are using a variety of approaches to link external drivers to response variables, particularly in the context of our disturbance framework. We have analyzed our long-term salinity data from the Altamaha River estuary and found that high salinity events could be explained by low river flow, strong up-estuary winds, or (to a lesser extent) unusually high tides (Sheldon & Alber 2019). We are currently using several agnostic statistical approaches (wavelet analysis, empirical mode decomposition) to evaluate hydrological drivers in comparison with satellite-derived biomass estimates (Objective C2), and we are also using data on both response and recovery from the SALTEx experiment to generate driver-response curves for multiple parameters (pore water nutrients, plant community composition, sediment elevation).

    • 2021 report:

      Activities:  We are using empirical mode decomposition and wavelet coherence analysis to investigate patterns in hydrological forcing to the GCE domain and assess the correspondence between them. We have also developed a series of nonlinear driver-response models in which the driver and response obey different mathematical forms.

      Results:  We have analyzed the timeseries of satellite-derived marsh productivity data (Objective 2B4) and flux tower-derived net ecosystem exchange fluxes (Objective 1A2) using entropy-based approaches and methods in dynamical systems analysis to identify causal connections to timeseries reflecting environmental forcing (Fig. 6).
 
LTER
NSF

This material is based upon work supported by the National Science Foundation under grants OCE-9982133, OCE-0620959, OCE-1237140 and OCE-1832178. Any opinions, findings, conclusions, or recommendations expressed in the material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.