Discussion
We modeled the effect of salinity and flooding on carbon uptake
mechanistically by incorporating the relationship between salinity,
flooding, and root water uptake. These modeling updates improved
representation of GPP and were validated by GPP measurements taken at
Plum Island Estuary. At hourly to annual time scales, modeled GPP
aligned well with field observations. Incorporating salinity-induced
inhibition of root water uptake brought simulations from 3x too high to
near the values observed in the field. Incorporating flood-induced
inhibition of carbon uptake decreased the estimated GPP by a smaller
proportion compared to the salinity function, but it more accurately
simulated variations over tidal cycles.
Salt marshes are well known for their strict zonation by elevation;
small deviations from salinity concentration and hydroperiod result in
community changes or loss of marsh habitat in lower elevations. S.
alterniflora is a dominant species in salt marshes along North
America’s temperate east coast, including at the Plum Island Estuary
site included in this study. Although S. alterniflora grows
better at lower salinities, other plants compete with it in fresh and
brackish conditions; S. alterniflora dominates salt marshes
because of its tolerance to salinity concentrations and hydroperiods
that kill most other species (Crain et al., 2004; Hwang and Morris,
1994). We set the optimal salinity parameter at zero to represent
maximum growth in freshwater and tried several tolerance parameters to
determine the value that closely simulated GPP observed in the field.
Although our tolerance parameterization resulted in reasonable
simulations of GPP, our model is too sensitive to salinity. Annual
average salinity fluctuates by 2-3 ppt due to variation in evaporation
and precipitation. Site measurements of salinity from 2020 were an
average of 3 ppt higher than salinity from 2018 and measured GPP
decreased by 6%, whereas the model scenarios showed an average of 5 ppt
higher salinity resulted in a 40% decrease in GPP. We think the
application of the Gaussian function is appropriate for mechanistic
representation of salinity’s limitation of carbon uptake, but the model
currently lacks representation of multiple plant adaptations to salt
exposure. This salinity function targets osmotic inhibition of root
water uptake, but does not yet account for ways vegetation can mitigate
salt exposure after uptake (Munns and Tester, 2008). plants can also
take up salt and excrete it or store it in vacuoles until chronic
exposure exceeds a toxicity threshold (Munns and Tester, 2008; Naidoo et
al., 1992; Vasquez et al., 2006). For example, S. alterniflorauses solutes to adjust osmotic pressure in aboveground tissues and uses
salt glands for excretion (Bradley and Morris, 1991; Naidoo et al.,
1992; Vasquez et al., 2006). However, we have improved model simulations
considerably from the default, and this approach has been effective in
modeling biomass for other marsh species.
This model does not yet account for several important coastal wetland
processes, including sediment trapping, stimulated growth, and
increasing vertical accretion with moderate sea level rise. These
processes are critical for understanding carbon cycling and marsh
stability on decadal scales (Kirwan et al., 2010; Morris et al., 2002).
Because this is the first attempt to incorporate salt marsh-specific
controls on GPP in ELM, we kept our representation of flooding simple:
the proportion of aboveground tissue underwater did not contribute to
photosynthesis in each time step. However, this approach may be too
simplistic to provide meaningful model simulations, since inundation
stimulation of S. alterniflora productivity is a major control on carbon
dynamics in salt marshes (Kirwan et al., 2016; Morris et al., 2013).
Currently, ELM estimates plant height of grasses proportionally to LAI,
which may be appropriate for approximating the impact of salinity stress
on plants, but not flooding. We used Li et al.’s Gaussian approach to
model vegetation response to salinity (2021), but they also used this
Gaussian approach for vegetation response to flooding. In this approach,
plants have an optimal flooding depth and a flooding tolerance
parameter. If we adopt this approach, we will need to parameterize the
salt marsh grass plant functional types for these additional flood
parameters, but this information is available for S. alterniflora(Morris et al., 2013). In either approach, flooding above the marsh
surface is addressed, but the impacts of soil waterlogging are not.
Addressing soil waterlogging will likely require modeling available
oxygen and other more detailed soil biogeochemistry; this work is
progressing (Sulman et al. in prep?, O’Meara et al. 2021?).
Our model developments focused on improving representation of GPP in
salt marshes, but to fully represent coastal wetland carbon cycling, ELM
should be able to model vegetation spanning the gradient from saline to
freshwater wetlands. Our approach is flexible to model response curves
of vegetation adapted to fresh or brackish conditions; even upland
vegetation could be given a low salinity tolerance to model the carbon
dynamics of saltwater intrusion into upland forests. However, the
default model parameterization predicted GPP values 3x the measured
values without the salinity response. Although individual plants may be
much more productive when grown in fresh or brackish water, typically
freshwater tidal wetlands are not 3x more productive than salt marshes
(need to cite here). The default model therefore predicts unreasonably
high values for freshwater wetland productivity, likely because the
model was originally designed for upland systems and assumes that
vegetation productivity is primarily limited by dry soil conditions.
Thus, existing parameterization of graminoid plants likely overestimates
productivity under non-water-limited conditions, so that when ELM is
used to model a wetland, high soil moisture allows plants
unrealistically high photosynthesis rates. As mentioned in the previous
paragraph, another problem is that ELM doesn’t address anoxic soil
conditions, which decrease root metabolism or incur metabolic costs for
flood tolerance traits (Colmer and Voesenek, 2009; Naidoo et al., 1992).
Improvement of vegetation in freshwater saturated environments will be
required before the model can be used to investigate carbon dynamics
along a salinity gradient.
A strength of implementing coastal wetlands into ELM and our flexible
approach to salinity is the potential to scale this model from point
simulations to global scales. In this study, we used salinity
measurements to force model simulations, combined with water levels
modeled using NOAA tide pattern models. Expanding this model framework
to broader scales will require estuary-scale measurements or models of
coupled hydrology and salinity. These could be derived from
estuary-scale measurement networks or estuary models. Simulating coastal
wetland vegetation at continental to global scales within a fully
coupled earth system model will require tidal patterns and salinity to
be supplied by ocean or hydrological models. Current ocean models
operate at resolutions too coarse to directly simulate coastal wetland
tide heights and estuary salinities, but ongoing developments in
high-resolution coastal models are making progress toward coupled
coastal process capabilities.