Available via license: CC BY 4.0
Content may be subject to copyright.
Citation: Inácio, M.; Freitas, M.C.;
Cunha, A.G.; Antunes, C.; Leira, M.;
Lopes, V.; Andrade, C.; Silva, T.A.
Simplified Marsh Response Model
(SMRM): A Methodological
Approach to Quantify the Evolution
of Salt Marshes in a Sea-Level Rise
Context. Remote Sens. 2022,14, 3400.
https://doi.org/10.3390/rs14143400
Academic Editors: Qiusheng Wu and
Dehua Mao
Received: 15 February 2022
Accepted: 12 July 2022
Published: 15 July 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
remote sensing
Article
Simplified Marsh Response Model (SMRM): A Methodological
Approach to Quantify the Evolution of Salt Marshes in a
Sea-Level Rise Context
Miguel Inácio 1, 2, *, M. Conceição Freitas 1,2 , Ana Graça Cunha 1,2, Carlos Antunes 2, Manel Leira 2,3,4,
Vera Lopes 1,2, César Andrade 1,2 and Tiago Adrião Silva 5
1Departamento de Geologia, Faculdade de Ciências, Universidade de Lisboa, Building C6, Floor 3,
Campo Grande, 1749-016 Lisbon, Portugal; cfreitas@fc.ul.pt (M.C.F.); agcunha@fc.ul.pt (A.G.C.);
vplopes@fc.ul.pt (V.L.); candrade@fc.ul.pt (C.A.)
2Instituto Dom Luiz, Faculdade de Ciências, Universidade de Lisboa, Building C6, Floor 3, Campo Grande,
1749-016 Lisbon, Portugal; cmantunes@fc.ul.pt (C.A.); m.leira@udc.es (M.L.)
3
BioCost Research Group, Facultade de Ciencias and Centro de Investigacións Científicas Avanzadas (CICA),
Universidade de A Coruña, 15071 Corunha, Spain
4Biodiversity and Applied Botany Research Group, Departamento de Botánica, Facultade de Bioloxía,
Universidade de Santiago de Compostela, 15782 Santiago de Compostela, Spain
5Department F.-A. Forel for Environmental and Aquatic Sciences, and Institute for Environmental Sciences,
University of Geneva, 1205 Geneva, Switzerland; geotsilva@gmail.com
*Correspondence: mfinacio@fc.ul.pt
Abstract:
Salt marshes are highly valued coastal environments for different services: coastline protection,
biodiversity, and blue carbon. They are vulnerable to climate changes, particularly to sea-level rise.
For this reason, it is essential to project the evolution of marsh areas until the end of the century. This
work presents a reduced complexity model to quantify salt marshes’ evolution in a sea-level rise
(SLR) context through combining field and remote sensing data: SMRM (Simplified Marsh Response
Model). SMRM is a two-dimensional rule-based model that requires four parameters: a digital
terrain model (DTM), local tidal levels, a sea-level rise projection, and accretion rates.
A MATLAB
script completes the process, and the output is a GeoTIFF file. Two test areas were selected in Tróia
sandspit (Setúbal, Portugal). Additionally, a sensitivity analysis for each parameter’s influence and
a comparison with SLAMM (another rule-based model) were undertaken. The sensitivity analysis
indicates that SLR is the most relevant parameter, followed by accretion rates. The comparison
of SMRM with SLAMM shows quite similar results for both models. This new model application
indicates that the studied salt marshes could be resilient to conservative sea-level rise scenarios but
not to more severe sea-level rise projections.
Keywords: accretion rates; reduced-complexity models; remote sensing application; GIS
1. Introduction
Salt marshes are vegetated intertidal wetlands with high ecological value because of
their biodiversity. They are one of the most productive ecosystems of the biosphere [
1
].
In addition, salt marshes (and mangroves) are widely recognized as playing a significant
role in coastal defense due to their ability to dissipate wave energy and temporarily store
floodwater. Moreover, they are a key feature for wildlife conservation, carbon sequestra-
tion [
2
], and a significant source of organic material and nutrients essential for marine
communities [3]. Some of the earliest conservation reserves were salt marshes [4].
Anthropic pressure over these delicate systems and climate-driven changes in external
forcing, may threaten the services these environments provide. Worldwide, vast areas of
salt marshes have been converted into agricultural land, industrial use, or port facilities,
despite lesser losses to other uses recently [4].
Remote Sens. 2022,14, 3400. https://doi.org/10.3390/rs14143400 https://www.mdpi.com/journal/remotesensing
Remote Sens. 2022,14, 3400 2 of 25
One of the main consequences of present and future climate changes is the sea-level
rise (SLR), which may negatively impact these ecosystems and severely threaten estuarine
areas [
5
]. Expected impacts include the destruction of habitats and the reduction of the
protection they afford to the adjacent coastline and land. Salt marsh establishment is
intrinsically linked to sea-level and tidal oscillations because flooding is the primary mech-
anism for sediment delivery to the marsh platforms [
6
]. Sea level and tidal range control
immersion time, a parameter to which marsh vegetation is susceptible. Salt marsh areas
are also prone to erosion, generally attributed to decreased sediment supply, and increased
magnitude and frequency of extreme events. Both drivers are also indirectly influenced
by SLR [7].
Remote sensing is essential to project the evolution of wetlands in a climate-change con-
text [
8
,
9
], being a rapid and cost-effective method to obtain relevant data [
10
]. The increas-
ing number of studies that use remote sensing reflects the access to this tool, e.g., [11–14].
Studying the future evolution of wetland environments, including salt marshes, is cru-
cial because their reduction or destruction may increase adverse climate change impacts.
For instance, the increase in CO
2
emissions to atmosphere related to wetlands’ destruction
represents 3–19% of the increase due to deforestation [
15
]. An increase in CO
2
emissions to
the atmosphere is explained by the release of organic carbon sequestered by these environ-
ments (known as blue carbon [
16
]). Furthermore, it is estimated that the carbon stored in
European salt marshes represents 4% of all carbon accumulated in coastal areas [17].
Other factors can impact the behavior of these environments in the future. Salt marsh
characteristics and dynamics result from complex interactions among hydrodynamics,
sediment transport, and biological processes [
18
]. Changes in salinity, partially dependent
on rainfall, affect their stability and local hydrodynamic patterns, ultimately affecting the
distribution of accretion rates over different parcels of the salt marsh. Other significant
parameters are elevation distribution and accommodation space in the land, neighboring
salt marshes. These areas may provide room for landward marsh expansion, which may
compensate seaward erosion; thus, maintaining invariance of the total marsh area. In fact,
when SLR rates are higher than accretion rates, salt marshes may evolve in space and time
following two ways: (1) marsh areas at higher elevations and fringing the coastline will
migrate landwards, providing that the three following conditions are met—absence of
vertical barriers (avoiding coastal squeeze), existence of accommodation space, and rate of
horizontal marsh expansion upland equal or higher than the rate of horizontal progression
of the inundation limit; (2) marsh areas at lower elevations will be more frequently flooded,
experience larger submersion periods, and are not colonized, or remain colonized by low
marsh vegetation [19].
In conceptual terms, if accretion rates closely match SLR rates, salt marshes may
grow or maintain their area. However, the acceleration of SLR rates observed recently and
projected into the near future suggests that this equilibrium is at risk of being reversed [
20
].
It is essential to rely on numerical or rule-based models to project and comprehend how
salt marshes can evolve until the end of the 21st century. These are powerful predictive
tools that incorporate nonlinear feedback between salt marsh ecosystem morphology and
sediment transport and deposition processes [
21
]. Salt marsh evolution models and marsh
response to SLR models can be classified as zero-dimensional (a single marsh point within
a marsh is modeled), one-dimensional (simulating the evolution of a salt marsh transect),
two-dimensional (when applied to a marsh platform), and landscape models (that simulate
processes over an entire estuary) [6].
This work aims to present the Simplified Marsh Response Model, which is a rule-
based model to quantify the evolution of salt marsh areas until the end of the century,
combining remote sensing with data in the field and literature. SMRM considers the balance
between SLR and accretion rates. Several models have been proposed recently to project
the evolution of these vulnerable environments under a climate-change context [
22
–
25
].
Such models assume that many data are readily available (e.g., salinity, precipitation, wind,
storm surge, uplift/subsidence, land use, erosion, and waves). However, this is not always
Remote Sens. 2022,14, 3400 3 of 25
possible (diversity, quality, and quantity of data are common drawbacks), and in many
cases, they involve a too high investment regarding the results.
Moreover, the assumption that adding more detail and variables to models always
leads to better results is questionable. Adding too many parameters can result in a situation
where neither the studied systems nor the model are understandable [
26
]. In contrast,
reduced complexity models are distinguished by a high degree of simplification, incorpo-
rating only the fundamental parameters to describe the processes or trends characterizing a
particular coastal system at given time and spatial scales, leaving out as many processes as
possible [
27
,
28
]. The Simplified Marsh Response Model (SMRM) is a reduced-complexity,
two-dimensional rule-based model. The motivation to present this simplified model is that
the complexity of the coastal systems often challenges the ability of complex models to fore-
cast timely and accurately the meso to macro-scale evolution of marsh environments [
29
].
SMRM needs only four parameters to run and departs from a digital terrain model (DTM)
of the studied marsh area. Using a DTM is the basis for modeling the evolution of salt
marshes, a fundamental application of remote sensing to coastal areas. In order to evaluate
its performance, SMRM was applied to two test areas in Portugal. The results were com-
pared to those obtained by modeling marsh changes in the same areas for the same period
of time with SLAMM (Sea Level Affecting Marsh Model) (Warren Pinnacle Consulting, Inc.,
Warren, VT, USA; Washington, DC, USA), a more complex rule-based model. In addition,
a sensitivity analysis was performed to evaluate the influence of each input parameter on
the final results.
2. Test Areas
Almost 50% of Portugal’s mainland perimeter corresponds to the Atlantic coast [
30
]
(950 km) (Figure 1A) and about
3
4
of the population lives in coastal municipalities [
31
].
The Portuguese coast hosts several estuarine/lagoonal areas featuring salt marshes. At-
lantic estuaries occur north of Mondego River (central Portugal), and Mediterranean estu-
aries occur on the south and west (southward of Mondego River) coasts of Portugal [
32
].
The salinity and biodiversity are higher in the latter than the former, particularly in the
high marsh. Although both types are anthropically modified, the Mediterranean estuaries
are the best-preserved [33].
Sado estuary, with an area of 23,560 ha [
34
], is a Mediterranean estuary located 40 km
south of Lisbon at the mouth of a river with the same name (Figure 1B). Sado river’s
headwaters is located in “Serra da Vigia” (Ourique, Portugal) at 230 m of elevation [
35
].
Sado features the largest Portuguese watershed, with an area of circa 764,000 ha [
36
].
The climate (dry–subhumid [
37
] or subtropical [
38
]) can be described as uniform along the
watershed, where the mean annual precipitation is 650 mm [
35
]. The river is classified as
Mediterranean [
39
] and the flow ranges are between 1 m
3
/s in the dry season and 80 m
3
/s
in the wet season [
36
]. The circulation and drainage within the estuary are essentially
controlled by tides [
40
] because wind (which mainly blows from the north) only affects the
surface water layer [
35
] and the river flow is too low. Ocean swell from the Atlantic Ocean
cannot pass through the inlet, so the waves are usually smaller than 1 m inside the estuary.
Sado estuary is separated from the Atlantic Ocean by Tróia sandspit, a 25 km-long
barrier trending NNW–SSE. Five growing phases were identified in Tróia sandspit, dated
from 6500 years to present (Figure 1C) [
41
]. Two test areas featuring salt marshes were
selected on the estuarine margin of this spit (Figure 1C–E) to apply the SMRM and the
SLAMM. They are located in the NNW tip of Tróia sandspit, at Caldeira de Tróia (C. Tróia),
which was further subdivided into two marsh expansions: north (C. Tróia (N)—Figure 1D)
and south (C. Tróia (S)—Figure 1E) sectors. C. Tróia (S) was formed during the fourth
growing phase of the sandspit (1870–1100 years before present), while C. Tróia (N) was
installed during the most recent (fifth) growing phase (1100 years–present).
Remote Sens. 2022,14, 3400 4 of 25
Remote Sens. 2022, 14, x FOR PEER REVIEW 4 of 27
fourth growing phase of the sandspit (1870–1100 years before present), while C. Tróia (N)
was installed during the most recent (fifth) growing phase (1100 years–present).
C. Tróia (N) salt marsh extends over 4.3 ha and consists mainly of a high marsh, with
small areas of a low marsh at lower heights. Adjacent to the salt marsh, there are two
different environments: dunes of moderate slope (landwards) and tidal mudflats
(seawards). C. Tróia (S) has similar characteristics to the northern sector, even though it
displays a larger area (10.5 ha) and a larger proportion of low marsh vegetation.
Figure 1. Test area locations. (A) Location in the Iberian Peninsula; (B) Sado estuary and Tróia
sandspit. The red dot indicates the location of Setúbal-Tróia tide gauge; (C) Tróia sandspit evolution
phases [41]; (D) C. Tróia (N) (Google satellite); (E) C. Tróia (S) (Google satellite).
3. Models
3.1. Simplified Marsh Response Model (SMRM)
The SMRM is a reduced complexity and two-dimensional rule-based model of marsh
response to SLR that only requires the input of four parameters: (1) a high-resolution
DTM, representing both the target marsh area and the adjacent surfaces (e.g., tidal flats,
backbarrier flats, dunes, alluvial plains), with an acceptable spatial range, resolution, and
quality to allow the simulation of marsh expansion; (2) critical tidal levels, representing a
transition between main intertidal environments and vegetation sensitive to submersion
time; in the test-cases addressed herein, the model considers mean high-water neap
elevation (MHWN) as the critical boundary between the tidal flat and low marsh, mean
high-water (MHW) as the transition between the low and high marsh, and mean high-
water spring (MHWS) as the upper growth limit of the high marsh [42]; (3) at least one
SLR projection, including an initial rate and an optional SLR acceleration value; (4)
representative accretion/sedimentation rates for high marsh, low marsh, and tidal flat
domains, which can remain invariant over time or change throughout the century (in this
last case, acceleration values of accretion rates must be introduced). The SMRM works
with several data sources, joining the results obtained from fieldwork and laboratory
(accretion rates), monitoring equipment (tidal levels and SLR), physical models (SLR), and
remote sensing (DTM).
The model was built considering the premises that present-day tidal amplitudes will
not change significantly in future decades and that the salt marsh surface’s signal and rate
of evolution result from the balance between SLR rates and accretion rates (Table 1).
Table 1. Possible results for the evolution of salt marsh areas, considering the balance between SLR
rates and accretion (Acc.) rates.
Situation Possible Result
SLR rates < Acc. rates Expansion
Figure 1.
Test area locations. (
A
) Location in the Iberian Peninsula; (
B
) Sado estuary and Tróia
sandspit. The red dot indicates the location of Setúbal-Tróia tide gauge; (
C
) Tróia sandspit evolution
phases [41]; (D) C. Tróia (N) (Google satellite); (E) C. Tróia (S) (Google satellite).
C. Tróia (N) salt marsh extends over 4.3 ha and consists mainly of a high marsh,
with small areas of a low marsh at lower heights. Adjacent to the salt marsh, there are
two different environments: dunes of moderate slope (landwards) and tidal mudflats
(seawards). C. Tróia (S) has similar characteristics to the northern sector, even though it
displays a larger area (10.5 ha) and a larger proportion of low marsh vegetation.
3. Models
3.1. Simplified Marsh Response Model (SMRM)
The SMRM is a reduced complexity and two-dimensional rule-based model of marsh
response to SLR that only requires the input of four parameters: (1) a high-resolution DTM,
representing both the target marsh area and the adjacent surfaces (e.g., tidal flats, backbar-
rier flats, dunes, alluvial plains), with an acceptable spatial range, resolution, and quality
to allow the simulation of marsh expansion; (2) critical tidal levels, representing a transi-
tion between main intertidal environments and vegetation sensitive to submersion time;
in the test-cases addressed herein, the model considers mean high-water neap elevation
(MHWN) as the critical boundary between the tidal flat and low marsh, mean high-water
(MHW) as the transition between the low and high marsh, and mean high-water spring
(MHWS) as the upper growth limit of the high marsh [
42
]; (3) at least one SLR projection,
including an initial rate and an optional SLR acceleration value; (4) representative accre-
tion/sedimentation rates for high marsh, low marsh, and tidal flat domains, which can
remain invariant over time or change throughout the century (in this last case, acceleration
values of accretion rates must be introduced). The SMRM works with several data sources,
joining the results obtained from fieldwork and laboratory (accretion rates), monitoring
equipment (tidal levels and SLR), physical models (SLR), and remote sensing (DTM).
The model was built considering the premises that present-day tidal amplitudes will
not change significantly in future decades and that the salt marsh surface’s signal and rate
of evolution result from the balance between SLR rates and accretion rates (Table 1).
Table 1.
Possible results for the evolution of salt marsh areas, considering the balance between SLR
rates and accretion (Acc.) rates.
Situation Possible Result
SLR rates < Acc. rates Expansion
SLR rates ≈Acc. rates Stability
SLR rates > Acc. rates Inundation
Remote Sens. 2022,14, 3400 5 of 25
The process starts by comparing the elevation of each DTM cell with the elevation
thresholds (as previously defined) at the reference time (2020). This step automatically
classifies each cell as terrestrial environment, high marsh, low marsh, or tidal flat (subtidal
areas are not discriminated). Next, the accretion rate characterizing each type of intertidal
environment for each year is applied to the discretized surface and a new DTM is generated,
corresponding to the next year’s (2021) morphology. This process is repeated each year,
and the transition levels are also adjusted yearly according to projected changes in sea
level (Figure 2). The iteration sequence ends in the selected year as an output (2100 is the
example shown herein, although decadal iterations were also performed).
Remote Sens. 2022, 14, x FOR PEER REVIEW 5 of 27
SLR rates ≈ Acc. rates Stability
SLR rates > Acc. rates Inundation
The process starts by comparing the elevation of each DTM cell with the elevation
thresholds (as previously defined) at the reference time (2020). This step automatically
classifies each cell as terrestrial environment, high marsh, low marsh, or tidal flat (subtidal
areas are not discriminated). Next, the accretion rate characterizing each type of intertidal
environment for each year is applied to the discretized surface and a new DTM is
generated, corresponding to the next year’s (2021) morphology. This process is repeated
each year, and the transition levels are also adjusted yearly according to projected changes
in sea level (Figure 2). The iteration sequence ends in the selected year as an output (2100
is the example shown herein, although decadal iterations were also performed).
The script to execute this set of steps was written in MATLAB R2018a (and updated
to MATLAB R2020a) [43,44]. The output is a GeoTIFF file, which preserves the
georeference of the original DTM. The final step is the generation of final maps and the
determination of the new areas (high marsh, low marsh, and total area), based on tidal
levels for the considered output year, which can be completed through any GIS software.
In this work, this step was undertaken using ArcGIS Pro [45]. The SMRM script is
available online [46].
Figure 2. Flowchart showing relations among the parameters of SMRM, including inputs, outputs,
intermediate parameters and processes.
In addition to the previous metrics, it is possible to obtain a parameter to estimate the
maturity of the marsh, which depends on the ratio between high and low marsh areas and
indicates the robustness of the salt marsh as a whole. In agreement with [42], a salt marsh
is classified as young if the low marsh area is larger than the high marsh; intermediate, if
both areas are broadly equivalent; and mature, if the high marsh is dominant over the low
marsh area [42]. In order to obtain a quantitative indicator, an HM/LM (High Marsh/Low
Marsh) ratio was established by the authors (HM/LM < 0.9 → young; HM/LM = [0.9–1.1]
→ intermediate; HM/LM > 1.1 → mature).
3.2. Sea Level Affecting Marsh Model (SLAMM)
The SLAMM model (version 6.7, Warren Pinnacle Consulting, Inc., Warren, VT, USA;
Washington, DC, USA) [47,48] is also a two-dimensional rule-based model. It was initially
Figure 2.
Flowchart showing relations among the parameters of SMRM, including inputs, outputs,
intermediate parameters and processes.
The script to execute this set of steps was written in MATLAB R2018a (and updated to
MATLAB R2020a) [
43
,
44
]. The output is a GeoTIFF file, which preserves the georeference
of the original DTM. The final step is the generation of final maps and the determination
of the new areas (high marsh, low marsh, and total area), based on tidal levels for the
considered output year, which can be completed through any GIS software. In this work,
this step was undertaken using ArcGIS Pro [45]. The SMRM script is available online [46].
In addition to the previous metrics, it is possible to obtain a parameter to estimate the
maturity of the marsh, which depends on the ratio between high and low marsh areas and
indicates the robustness of the salt marsh as a whole. In agreement with [
42
], a salt marsh
is classified as young if the low marsh area is larger than the high marsh; intermediate,
if both areas are broadly equivalent; and mature, if the high marsh is dominant over the low
marsh area [
42
]. In order to obtain a quantitative indicator, an HM/LM (High Marsh/Low
Marsh) ratio was established by the authors (HM/LM < 0.9
→
young; HM/LM = [0.9–1.1]
→intermediate; HM/LM > 1.1 →mature).
3.2. Sea Level Affecting Marsh Model (SLAMM)
The SLAMM model (version 6.7, Warren Pinnacle Consulting, Inc., Warren, VT, USA;
Washington, DC, USA) [
47
,
48
] is also a two-dimensional rule-based model. It was initially
developed to project the evolution of North American salt marsh areas until the end of the
21st century. However, SLAMM has not been widely applied to the Mediterranean–Atlantic
salt marshes due to marked differences in vegetation, extent, and structure, although
Remote Sens. 2022,14, 3400 6 of 25
its application seems acceptable [
49
]. SLAMM works in a complex relation among six
processes: inundation, accretion, erosion, overwash, saturation, wind, and salinity.
In order to apply SLAMM to Portuguese test areas, spatial and site-specific parameters
(which are detailed further below) are mandatory: digital terrain model (DTM), slopes,
wetland category maps, tidal levels, diurnal tidal range, direction offshore, salt elevation,
accretion rates, and at least one SLR scenario. To compare the obtained results with the
results provided by SMRM, no more than the mandatory parameters were considered.
4. Parameters
The parameters considered to project the evolution of the test areas are local tidal
levels, DTM, SLR, and accretion rates. Other parameters such as wind and waves are not
used in SMRM and were not considered in SLAMM. Although sometimes these parameters
seem relevant [
50
], they are not significant in the Sado estuary, as explained in Section 2of
this work. In addition, salinity can be considered uniform in a small area such as C. Tróia.
4.1. Local Tidal Levels
The first parameter considers the threshold limits to automatically classify each DTM
cell as dune, high marsh, low marsh, and tidal flat. This work considers tidal levels because
they correctly identify the transitions among the environments. However, other data can
be considered for other areas, which can be chosen by the model’s user.
Local tidal levels used in this study were modeled using harmonic analysis and based
on data acquired by the Setúbal-Tróia tide gauge (years 1977, 1978, 1993, 1998, and 2005),
located 2 km north of the study site (Figure 1B). Tide tables of 2000–2016 were computed
from the harmonic model, mean values were evaluated, and a correction for the 2020 local
mean sea level was applied (NMM1938, Cascais—the datum considered for all the heights
presented in this study) [51,52].
SMRM and SLAMM use different tidal levels. SMRM considers the MHWN, MHW,
and MHWS, whereas SLAMM uses heights of 30/60/90 days annual inundation periods.
Furthermore, SLAMM also requires the introduction of the great diurnal tidal range (equiv-
alent to the difference between mean highest high water and mean lowest low water levels)
and the salt elevation (often defined as the elevation that is inundated by saltwater less
than every 30 days in a year) [
48
]. These parameters were computed from the harmonic
tidal model (Table 2).
Table 2. Tidal levels for SMRM and SLAMM in 2020.
SMRM
MSL MHWN MHW MHWS
0.19 m 0.90 m 1.26 m 1.61 m
SLAMM
MSL 90 days 60 days 30 days
0.19 m 1.32 m 1.47 m 1.62 m
SLAMM site-specific parameters
Great Diurnal Tidal Range Salt Elevation
2.14 m 1.62 m
4.2. Digital Terrain Model
The colonization of new marsh areas depends on the annual inundation periods, gov-
erned mainly by topography. The latter is a fundamental parameter since it influences many
processes determining coastal change [
53
]. To evaluate the uncertainty of the inundation,
it is essential to consider a remote sensing technique that accurately represents the studied
area: a high-resolution DTM.
A DTM was obtained by interpolating a 2011 LiDAR survey performed by Agência
Portuguesa do Ambiente (APA) and Direção Geral do Território (DGT), both Portuguese
Remote Sens. 2022,14, 3400 7 of 25
public institutions. The horizontal resolution of the LiDAR varies between 1 and 2 m (2 m
for intertidal environments such as salt marshes), and its vertical precision varies from 8 to
24 cm [
30
]. The LiDAR survey used topographic and bathymetric sensors, guaranteeing
the data quality regardless of the surface type. Significant factors were also considered
(topography and hydrography, temperature, humidity, precipitation, wind, water turbidity,
pollution, flora, and fauna) to produce the final DTM [30].
The studied salt marshes display many tidal channels less than 2 m wide, matching
the horizontal resolution of the considered DTM. Consequently, some of these channels
are not well represented and generate sinks. Considering that tidal channels were not
separately represented in simulations produced by both models, these artifacts (sinks) were
leveled with the surrounding surface elements using the fill tool provided by ArcGIS Pro.
Heights of homologous points measured in the field using a DGPS-RTK (Leica Viva
NetRover GS08) and yielded by the DTM were compared to validate the accuracy of
the terrain model. In total, 2722 points distributed over the high marsh, low marsh,
and tidal flat of both C. Tróia (N) (1844 points) and C. Tróia (S) (878 points) were measured
(Figure 3, Table A1) between 2016 and 2021. Although there is a gap of at least five years
between the LiDAR survey and DGPS records, the difference should only correspond to the
sedimentation that occurred in this period, because the erosion of marshes is not laminar
and usually occurs at the edges [54].
Remote Sens. 2022, 14, x FOR PEER REVIEW 8 of 27
Figure 3. Boxplots showing the distribution of heights obtained through DGPS and DTM. (A) C.
Tróia (N); (B) C. Tróia (S). Blue lines indicate the heights of local critical tidal levels: MHWS—mean
high water spring; MHW—mean high water; MHWN—mean high water neap.
A statistical evaluation of the DTM accuracy was performed in addition to the
comparison between the DGPS and DTM data. The root mean square error (RMSE) was
determined, considering both data sources (Equation (1), Table 3). The RMSE is a metric
widely used to express the vertical accuracy of elevation datasets [55].
𝑅𝑀𝑆𝐸=∑𝑍 −𝑍
𝑛 (1)
ZDTM is the height of each point in the DTM, ZDGPS is the height of each homologous
point measured with the DGPS, n is the number of points being checked, and i is an integer
from 1 to n. Besides the RMSE, the linear error (L.E.) was also determined at a confidence
level of 95% (the metric used by the National Standard for Spatial Data Accuracy (Federal
Geographic Data Committee, 1998)) (Equation (2), Table 3).
𝐿.𝐸.%=1.96×𝑅𝑀𝑆𝐸 (2)
Table 3. RMSE and L.E. of C. Tróia (N) and (S) salt marshes DTM.
C. Tróia (N)
Global Dune High marsh Low marsh Tidal flat
RMSE 17 cm 29 cm 13 cm 18 cm 14 cm
L.E. 33 cm 57 cm 25 cm 36 cm 27 cm
C. Tróia (S)
Global Dune High marsh Low marsh Tidal flat
RMSE 22 cm 39 cm 21 cm 15 cm 31 cm
L.E. 43 cm 76 cm 41 cm 29 cm 60 cm
A global RMSE of 17 cm for C. Tróia (N) was obtained, which is compatible with the
vertical accuracy of the DTM. The error is higher on dunes, probably because the DGPS
data were measured between 2016 and 2021 and the DTM corresponds to a survey of 2011.
The RMSE of the C. Tróia (S) salt marsh is globally higher. However, the global value is
within the accuracy limits of the DTM. An RMSE of 17 cm was considered in the sensitivity
analysis of the model performed in C. Tróia (N).
In this study, the DTM is considered static until the end of the century due to
negligible vertical land motion (either isostatic or tectonic driven) compared with the
Figure 3.
Boxplots showing the distribution of heights obtained through DGPS and DTM. (
A
) C.
Tróia (N); (
B
) C. Tróia (S). Blue lines indicate the heights of local critical tidal levels: MHWS—mean
high water spring; MHW—mean high water; MHWN—mean high water neap.
In general, each environment is distributed between the theoretical tidal limits, so the
SMRM will correctly classify each pixel of the DTM. On average, heights extracted from the
DTM are higher than those obtained in the field, which can be justified by the salt marsh
vegetation or water, even though these parameters were considered during the survey and
data processing. C. Tróia (N) high marsh is higher than the high marsh in the southern
sector. Although there are natural differences between the two data sources, the DTM is
considered a reasonable morphologic representation of the study areas.
Besides the DTM, SLAMM requires a slope file and a *.txt file with each pixel’s
SLAMM wetland categories. The slope file was obtained from the original DTM, and the
categorization was completed manually according to the classification provided by this
model’s Technical Documentation [
47
]. SLAMM does not clearly differentiate high and low
marsh. Instead, it considers “regularly flooded marsh” and “irregularly flooded marsh”,
which was associated with low and high marsh, respectively, in agreement with other
Remote Sens. 2022,14, 3400 8 of 25
authors [
49
]. Dunes were classified as undeveloped dry land since they are vulnerable to
inundation and colonization by marsh vegetation.
A statistical evaluation of the DTM accuracy was performed in addition to the compari-
son between the DGPS and DTM data. The root mean square error (RMSE) was determined,
considering both data sources (Equation (1), Table 3). The RMSE is a metric widely used to
express the vertical accuracy of elevation datasets [55].
RMSE =s∑(ZDTM i −ZDGPS i )2
n(1)
Table 3. RMSE and L.E. of C. Tróia (N) and (S) salt marshes DTM.
C. Tróia (N)
Global Dune High marsh Low marsh Tidal flat
RMSE 17 cm 29 cm 13 cm 18 cm 14 cm
L.E. 33 cm 57 cm 25 cm 36 cm 27 cm
C. Tróia (S)
Global Dune High marsh Low marsh Tidal flat
RMSE 22 cm 39 cm 21 cm 15 cm 31 cm
L.E. 43 cm 76 cm 41 cm 29 cm 60 cm
Z
DTM
is the height of each point in the DTM, Z
DGPS
is the height of each homologous
point measured with the DGPS, nis the number of points being checked, and iis an integer
from 1 to n. Besides the RMSE, the linear error (L.E.) was also determined at a confidence
level of 95% (the metric used by the National Standard for Spatial Data Accuracy (Federal
Geographic Data Committee, 1998)) (Equation (2), Table 3).
L.E.(95%)=1.96 ×RMSE (2)
A global RMSE of 17 cm for C. Tróia (N) was obtained, which is compatible with the
vertical accuracy of the DTM. The error is higher on dunes, probably because the DGPS
data were measured between 2016 and 2021 and the DTM corresponds to a survey of 2011.
The RMSE of the C. Tróia (S) salt marsh is globally higher. However, the global value is
within the accuracy limits of the DTM. An RMSE of 17 cm was considered in the sensitivity
analysis of the model performed in C. Tróia (N).
In this study, the DTM is considered static until the end of the century due to negligible
vertical land motion (either isostatic or tectonic driven) compared with the magnitude of
sea level changes. However, changes can occur due to land motion, and if that is the case,
they can be added to the DTM error or to SLR.
A combination of data on absolute mean sea level derived from satellite observations
with high-precision GPS data of vertical ground motion measured near tide gauges and
relative mean sea level data measured at those tide gauges may be used to decouple eustatic
from vertical land movements at shorter timescales and evaluate uncertainty. The magni-
tude of vertical land movement in the Cascais region (some 50 km north of Sado estuary)
has been assessed by [
52
], comparing the 1990–2019 relative sea level data, retrieved from
Cascais tide gauge, with 1993–2016 data on absolute sea level derived from satellite al-
timetry. He inferred vertical uplift rates from 0.08 to 0.28 mm/year. Different results have
been obtained by [
56
]. These authors computed vertical land motions along the Iberian
Atlantic coast using GPS time series with more than 8 years of observations. GPS-derived
estimates of the vertical land movement were further used to correct data on sea level
obtained at tide gauges. The latter was compared with satellite altimetry absolute sea level.
In Cascais region, tide gauge results suggest small subsidence (
−
0.40
±
0.08 mm/year) over
1997–2020. Another recent study [
57
] analyzed a global database of high-precision GPS data
acquired after 2002 to set up an interpolated field of vertical land movement rates, from
Remote Sens. 2022,14, 3400 9 of 25
which median rates at the location of tide gauges in the Permanent Service for Global Mean
Sea Level may be obtained ([
57
] provides methodological details). Vertical land movements
shown in this study for Cascais and Setúbal-Tróia tide gauges indicate low subsidence
at both stations, at rates of
−
0.766
±
0.238 and
−
0.766
±
0.320 mm/year, respectively.
As correctly pointed out by [
56
,
57
], the above estimates should be taken with caution, given
that different studies use different methodological approaches, contrasting lengths of the
data series, heterogeneity in the spatial distribution of GPS source stations, and possible
undocumented problems affecting both GPS and tide-gauge stations, and differences in
data analysis and correction.
The contrast between the magnitude of vertical land motion (in the order of 10
−1
mm/year)
affecting the study area recently and coeval multi-decadal changes in relative sea level (in
the order of 10
0
mm/year) indicates that eustatic components have largely dominated the
latter. The future increase in sea level rise rate (in the order of 10
1
mm/year by the end of
the 21st century) will enhance this contrast. It is reasonable to disregard the influence of
vertical land motions in relative sea-level patterns, and marsh inundation projected until
the end of the 21st century for Sado salt marshes, considering the objectives and timescale
of this study.
4.3. Sea-Level Rise Scenarios
This study considers four SLR scenarios: IPCC RCP4.5 and RCP8.5 [
58
], Mod.FC_2b [
52
],
and NOAA Extreme [
59
]. IPCC SLR scenarios were chosen to consider global projec-
tions, MOD.FC_2b is an empirical projection based on the Cascais (Portugal) tide gauge,
and NOAA Extreme is a high-end SLR projection, which can be useful for land manage-
ment measures.
SMRM requires an initial SLR rate and an acceleration value. These parameters are
published for MOD.FC_2b, but only decadal absolute mean sea-level values are available
for IPCC RCP4.5/RCP8.5 and NOAA Extreme. The provided values were plotted, and a
polynomial function was calculated. The initial SLR rate (2020) and the acceleration were
obtained through the first and second derivates of the functions, respectively (Table A2).
All the SLR projections were adjusted to the Portuguese datum to be concordant with other
parameters (Figure 4). A remark must be made on the NOAA Extreme projection since
the 2020 SLR rate inferred from the respective model gives a much higher rate than the
most recent observed SLR rate [
52
]. In 2100, such SLR would be only possible through an
exponential model, related to a disruptive process, that could respect the actual SLR rates
and the final sea level. However, the model and respective values of SLR and rates were
used as presented by the report’s authors [59].
Remote Sens. 2022, 14, x FOR PEER REVIEW 10 of 27
In addition to the previous scenarios, the IPCC RCP2.6 SLR projection was also
included to perform a sensitivity analysis, since it can be described as a scenario with an
almost linear behavior.
Many different scenarios can be incorporated in models of marsh response to SLR. It
is fundamental to select a wide range of scenarios, to obtain different results. The
previously presented scenarios include rises in sea level from 55 (IPCC RCP2.6) to 261 cm
(NOAA Extreme) until 2100, and intermediate values (111 cm—MOD.FC_2b). Other SLR
scenarios, e.g., [60], are within the presented set of values.
Figure 4. SLR evolution until 2100. SLR rates and acceleration, regarding the Portuguese local
datum, for the considered scenarios: IPCC RCP2.6/4.5/8.5, MOD.FC_2b, and NOAA Extreme.
SLAMM allows the user to introduce a specific SLR projection or to select an SLR
scenario from a list (mainly from IPCC). However, the included SLR projections are from
the outdated IPCC 4th Assessment Report [61].
4.4. Accretion Rates
Sediment availability is fundamental for the survival of saltmarsh areas [62]; thus,
any models trying to predict how these areas will evolve need to include this parameter.
In this paper, sedimentation and accretion are used interchangeably to mean “the vertical
increase in surface elevation (in mm) relative to a specific layer of soil” [63].
Vertical accretion combines the deposition and erosion of sediments, the
accumulation of dead biomass on the marsh surface, and a certain amount of local
subsidence (also called settlement or autocompaction) [63]. Many factors can influence the
vertical accretion rate (e.g., sediment availability, presence of vegetation, tidal dynamics).
The advantage of using radiochronology (e.g., 137Cs and 210Pb dating) to derive
sedimentation rates is that the results already incorporate most of these factors [63].
In C. Tróia (N), accretion rates derived from the 210Pb dating method are 2.9 mm/year
in the high marsh, 3.0 mm/year in the low marsh, and 2.5 mm/year in the tidal flat. From
the 137Cs measurements and considering 1963 maximum atmospheric fallouts as the time
marker, the sedimentation rates are 2.9, 2.9, and 3.3 mm/year, respectively. Appendix B
presents the results from the 137Cs and 210Pb determination, and a detailed explanation of
the used methods. The obtained values represent the last century’s sedimentation pattern
and are suitable for integrating a model that will not exceed 2100 in its projections. We
assume that accretion rates will not vary until the end of the century. Regarding the
proximity of the two study sites and their similarities, the same accretion rates were also
assumed for C. Tróia (S) marsh. Furthermore, other studies performed nearby found
values of sedimentation rates ranging from 2.80 to 4.40 mm/year (depending on the
technique) in the marshes of Malha da Costa (38°25′58′′N, 8°49′31′′W) and Faralhão
(38°31′04′′N, 8°47′03′′W), both on the Sado estuary [64]. The similarity of the results
demonstrates how consistent sedimentation is in this region, implying that, in the absence
Figure 4.
SLR evolution until 2100. SLR rates and acceleration, regarding the Portuguese local datum,
for the considered scenarios: IPCC RCP2.6/4.5/8.5, MOD.FC_2b, and NOAA Extreme.
Remote Sens. 2022,14, 3400 10 of 25
In addition to the previous scenarios, the IPCC RCP2.6 SLR projection was also
included to perform a sensitivity analysis, since it can be described as a scenario with an
almost linear behavior.
Many different scenarios can be incorporated in models of marsh response to SLR. It is
fundamental to select a wide range of scenarios, to obtain different results. The previously
presented scenarios include rises in sea level from 55 (IPCC RCP2.6) to 261 cm (NOAA
Extreme) until 2100, and intermediate values (111 cm—MOD.FC_2b). Other SLR scenarios,
e.g., [60], are within the presented set of values.
SLAMM allows the user to introduce a specific SLR projection or to select an SLR
scenario from a list (mainly from IPCC). However, the included SLR projections are from
the outdated IPCC 4th Assessment Report [61].
4.4. Accretion Rates
Sediment availability is fundamental for the survival of saltmarsh areas [
62
]; thus,
any models trying to predict how these areas will evolve need to include this parameter.
In this paper, sedimentation and accretion are used interchangeably to mean “the vertical
increase in surface elevation (in mm) relative to a specific layer of soil” [63].
Vertical accretion combines the deposition and erosion of sediments, the accumulation
of dead biomass on the marsh surface, and a certain amount of local subsidence (also called
settlement or autocompaction) [
63
]. Many factors can influence the vertical accretion rate
(e.g., sediment availability, presence of vegetation, tidal dynamics). The advantage of using
radiochronology (e.g.,
137
Cs and
210
Pb dating) to derive sedimentation rates is that the
results already incorporate most of these factors [63].
In C. Tróia (N), accretion rates derived from the
210
Pb dating method are 2.9 mm/year
in the high marsh, 3.0 mm/year in the low marsh, and 2.5 mm/year in the tidal flat. From
the 137Cs measurements and considering 1963 maximum atmospheric fallouts as the time
marker, the sedimentation rates are 2.9, 2.9, and 3.3 mm/year, respectively. Appendix B
presents the results from the
137
Cs and
210
Pb determination, and a detailed explanation
of the used methods. The obtained values represent the last century’s sedimentation
pattern and are suitable for integrating a model that will not exceed 2100 in its projections.
We assume that accretion rates will not vary until the end of the century. Regarding the
proximity of the two study sites and their similarities, the same accretion rates were also
assumed for C. Tróia (S) marsh. Furthermore, other studies performed nearby found values
of sedimentation rates ranging from 2.80 to 4.40 mm/year (depending on the technique)
in the marshes of Malha da Costa (38
◦
25
0
58
00
N, 8
◦
49
0
31
00
W) and Faralhão (38
◦
31
0
04
00
N,
8
◦
47
0
03
00
W), both on the Sado estuary [
64
]. The similarity of the results demonstrates
how consistent sedimentation is in this region, implying that, in the absence of new
measurements, the obtained accretion rates in one marsh can be applied when modeling
others located nearby.
Both SMRM and SLAMM consider one accretion rate for each environment, which is
a simplification, since there are many aspects not considered: accretion rates will usually
decrease with the increase in height, accretion rates tend to be higher near the tidal creeks,
and others. Furthermore, changes in SLR could potentially affect sediment availability in
the entire estuary. Investigating this aspect is, in itself, an entirely separate work. It might
be possible to identify different accretion rates in areas with very high sedimentation rates
over time. SMRM considers the possibility of working under nonlinear accretion rates. This
is done by introducing a positive or negative acceleration value that affects the initial rates.
5. Application of SMRM to the Test Areas and Comparison with SLAMM
This section presents the results of SMRM application to both sectors of C. Tróia (the
output datasets are available in [
65
]). They are similar in the two test areas, although the
southern sector is less resilient to SLR. In addition, the obtained results are compared with
SLAMM simulations for C. Tróia (N). The comparison is presented according to each SLR
Remote Sens. 2022,14, 3400 11 of 25
scenario’s severity (IPCC RCP4.5/8.5, MOD.FC_2b, and NOAA Extreme). In line with each
SLR projection, the obtained results are quite different (Table 4).
Table 4.
Area and HM/LM ratio values for 2020, 2050, and 2100 for C. Tróia (N) and C. Tróia (S) salt
marshes, considering IPCC RCP4.5/RCP8.5, MOD.FC_2b, and NOAA Extreme SLR scenarios using
SMRM and SLAMM.
Scenario
SMRM SLAMM
2020 2050 2100 2020 2050 2100
Area
(ha) HM/LM Area
(ha) HM/LM Area
(ha) HM/LM Area
(ha) HM/LM Area
(ha) HM/LM Area
(ha) HM/LM
C. Tróia (N)
IPCC RCP4.5
4.48 2.9
4.49 2.3 4.33 1.0
4.31 2.8
4.30 2.2 3.99 0.8
RCP8.5
4.45 2.1 3.79 0.4 4.27 2.0 3.63 0.3
MOD.FC_2b
4.36 1.4 1.89 0.7 4.17 1.3 1.67 0.5
NOAA
Extreme 3.42 0.3 1.25 1.2 3.25 0.2 1.03 0.8
C. Tróia (S)
IPCC RCP4.5
10.29 1.4
10.12
1.2 9.10 0.8
10.31 1.4
9.93 1.2 8.60 0.7
RCP8.5
9.95 1.2 6.65 0.4 9.74 1.1 6.37 0.4
MOD.FC_2b
9.44 1.0 4.39 1.1 9.22 0.9 4.01 0.8
NOAA
Extreme 5.98 0.3 5.40 1.5 5.71 0.3 3.81 1.0
The results are very similar for both models. However, some differences should be
noted. The initial area is manually defined based on field and satellite imagery observation.
Nevertheless, each model adjusts the initial area according to the local tidal levels (SMRM)
or other relevant parameters (slope, salt elevation, great diurnal tidal range—SLAMM).
For this reason, in 2020, the area is not precisely the same for both models. Thus, in the
C. Tróia (N) salt marsh, the initial area is overestimated by 3% with SMRM and 1% with
SLAMM (Figure 5A,B). In the case of C. Tróia (S), the real limits coincide with the local tidal
levels (Figure 5C,D).
Remote Sens. 2022, 14, x FOR PEER REVIEW 12 of 27
Figure 5. Modeled morphological map for 2020 for C. Tróia (N) and (S), using SMRM (A,C) and
SLAMM (B,D). Coordinate system: ETRS1989 (PT06). Basemap: 2014–2015 DGT Orthophotos.
Considering IPCC RCP4.5 (SLR of 65 cm until 2100), SMRM projects that C. Tróia (N)
will maintain its area until 2050 and a loss of only 3% is expected until 2100 (4.33 ha)
(Figure 6A). The HM/LM ratio will decrease to 2.3 and to 1.0 by 2050 and 2100,
respectively, which is clear in the morphological maps (Figure 7A,C). C. Tróia (S) will lose
2% (10.12 ha) and 12% (9.10 ha) of its area (Figure 8A) and the HM/LM ratio will be 1.2
and 0.8 in 2050 and 2100 (Figure 9A,C). The losses are greater with SLAMM: in C. Tróia
(N), there will be no losses in the area until 2050 and a loss of 7% (Figure 6B) (HM/LM =
2.2/0.8 for 2050/2100—Figure 7B,D) is expected until 2100. In C. Tróia (S), the area will be
reduced by 4% and 17% until 2050 and 2100 (Figure 8B) (HM/LM = 1.2/0.7 for 2050/2100—
Figure 9B,D).
With IPCC RCP8.5 (SLR of 86 cm until 2011), SMRM projects 1% (4.45 ha) and 15%
(3.79 ha) losses until 2050 and 2100 for C. Tróia (N) (Figure 6C). The HM/LM ratio will
decrease to 2.1 and 0.4 for these reference dates, respectively (Figure 7E,G). In C. Tróia (S),
the area will be reduced by 3 and 35% (Figure 8C) and the HM/LM ratio will be 1.2 and
0.4 in 2050 and 2100 (Figure 9E,G). With SLAMM, C. Tróia (N) area will decrease by 1%
and 16% (Figure 6D) (HM/LM = 2.0/0.3 for 2050/2100—Figure 7F,H) and C. Tróia (S) will
decrease by 6% and 38% until 2050 and 2100 (Figure 8D), respectively (HM/LM = 1.1/0.4
for 2050/2100—Figure 9F,H).
Figure 5.
Modeled morphological map for 2020 for C. Tróia (N) and (S), using SMRM (
A
,
C
) and
SLAMM (B,D). Coordinate system: ETRS1989 (PT06). Basemap: 2014–2015 DGT Orthophotos.
Remote Sens. 2022,14, 3400 12 of 25
Considering IPCC RCP4.5 (SLR of 65 cm until 2100), SMRM projects that C. Tróia
(N) will maintain its area until 2050 and a loss of only 3% is expected until 2100 (4.33 ha)
(Figure 6A). The HM/LM ratio will decrease to 2.3 and to 1.0 by 2050 and 2100, respectively,
which is clear in the morphological maps (Figure 7A,C). C. Tróia (S) will lose 2% (10.12 ha)
and 12% (9.10 ha) of its area (Figure 8A) and the HM/LM ratio will be 1.2 and 0.8 in 2050 and
2100 (Figure 9A,C). The losses are greater with SLAMM: in C. Tróia (N), there will be no
losses in the area until 2050 and a loss of 7% (Figure 6B) (HM/LM = 2.2/0.8 for 2050/2100—
Figure 7B,D) is expected until 2100. In C. Tróia (S), the area will be reduced by 4% and 17%
until 2050 and 2100 (Figure 8B) (HM/LM = 1.2/0.7 for 2050/2100—Figure 9B,D).
Remote Sens. 2022, 14, x FOR PEER REVIEW 13 of 27
Figure 6. C. Tróia (N) decadal area evolution until 2100, using SMRM and SLAMM, respectively,
considering IPCC RCP4.5 (A,B), IPCC RCP8.5 (C,D), MOD.FC_2b (E,F), and NOAA Extreme (G,H).
Figure 6.
C. Tróia (N) decadal area evolution until 2100, using SMRM and SLAMM, respectively,
considering IPCC RCP4.5 (
A
,
B
), IPCC RCP8.5 (
C
,
D
), MOD.FC_2b (
E
,
F
), and NOAA Extreme (
G
,
H
).
Remote Sens. 2022,14, 3400 13 of 25
Remote Sens. 2022, 14, x FOR PEER REVIEW 14 of 27
Figure 7. C. Tróia (N) salt marsh morphological maps for 2050 and 2100, using SMRM and SLAMM.
Results are presented for the considered SLR scenarios: IPCC RCP4.5 (A–D), IPCC RCP8.5 (E–H),
MOD.FC_2b (I–L), and NOAA Extreme (M–P).
Figure 7.
C. Tróia (N) salt marsh morphological maps for 2050 and 2100, using SMRM and SLAMM.
Results are presented for the considered SLR scenarios: IPCC RCP4.5 (
A
–
D
), IPCC RCP8.5 (
E
–
H
),
MOD.FC_2b (I–L), and NOAA Extreme (M–P).
Remote Sens. 2022, 14, x FOR PEER REVIEW 15 of 27
Figure 8. C. Tróia (S) decadal area evolution until 2100, using SMRM and SLAMM, respectively,
considering IPCC RCP4.5 (A,B), IPCC RCP8.5 (C,D), MOD.FC_2b (E,F), and NOAA Extreme (G,H).
The SMRM with the MOD.FC_2b SLR scenario (111 cm until 2100) produces major
changes until the end of the century. In C. Tróia (N), the area will be reduced by 3% (4.36
ha) and 58% (1.89 ha) (Figure 6E) and the HM/LM ratio will be 1.4 and 0.7 until 2050 and
2100 (Figure 7I,K). For C. Tróia (S), the losses will be almost identical: the area will be
reduced by 8% (9.44 ha) and 57% (4.39 ha) in 2050 and 2100 (Figure 8E). However, the
HM/LM ratio will be 1.0 and 1.1, respectively (Figure 9I,K). SLAMM produces area losses
of 3% and 61% for C. Tróia (N) (Figure 6F) (HM/LM = 1.3/0.5 for 2050/2100—Figure 7J,L)
and of 11% and 61% for C. Tróia (S) for the reference dates (Figure 8F) (HM/LM = 0.9/0.8
for 2050/2100—Figure 9J,L).
Finally, considering SMRM with NOAA Extreme (261 cm of SLR until 2100), the area
will be reduced by 24% (3.42 ha) and 72% (1.24 ha) in 2050 and 2100 in C. Tróia (N) (Figure
6G). The HM/LM will be 0.3 and 1.2, showing a recovery by the end of the century (Figure
7M,O). In C. Tróia (S), the losses in the area will be almost the same in 2050 and 2100: 42%
Figure 8.
C. Tróia (S) decadal area evolution until 2100, using SMRM and SLAMM, respectively,
considering IPCC RCP4.5 (
A
,
B
), IPCC RCP8.5 (
C
,
D
), MOD.FC_2b (
E
,
F
), and NOAA Extreme (
G
,
H
).
Remote Sens. 2022,14, 3400 14 of 25
Remote Sens. 2022, 14, x FOR PEER REVIEW 16 of 27
(5.98 ha) and 48% (5.40 ha) (Figure 8G). The HM/LM ratio decreases to 0.3 in 2050 and
increases to 1.5 in 2100 (Figure 9M,O). SLAMM produces area losses of 25% and 76% in
C. Tróia (N) (Figure 6H) (HM/LM = 0.2/0.8 for 2050/2100—Figure 7N,P) and 45% and 63%
in C. Tróia (S) until 2050 and 2100 (Figure 8H) (HM/LM = 0.3/1.0 for 2050/2100—Figure
9N,P).
In general terms, the results obtained with the two considered models are similar for
both study areas.
Figure 9. C. Tróia (S) salt marsh morphological maps for 2050 and 2100, using SMRM and SLAMM.
Results are presented for the considered SLR scenarios: IPCC RCP4.5 (A–D), IPCC RCP8.5 (E–H),
MOD.FC_2b (I–L), and NOAA Extreme (M–P).
Figure 9.
C. Tróia (S) salt marsh morphological maps for 2050 and 2100, using SMRM and SLAMM.
Results are presented for the considered SLR scenarios: IPCC RCP4.5 (
A
–
D
), IPCC RCP8.5 (
E
–
H
),
MOD.FC_2b (I–L), and NOAA Extreme (M–P).
With IPCC RCP8.5 (SLR of 86 cm until 2011), SMRM projects 1% (4.45 ha) and 15%
(3.79 ha) losses until 2050 and 2100 for C. Tróia (N) (Figure 6C). The HM/LM ratio will
decrease to 2.1 and 0.4 for these reference dates, respectively (Figure 7E,G). In C. Tróia (S),
the area will be reduced by 3 and 35% (Figure 8C) and the HM/LM ratio will be 1.2 and
0.4 in 2050 and 2100 (Figure 9E,G). With SLAMM, C. Tróia (N) area will decrease by 1%
and 16% (Figure 6D) (HM/LM = 2.0/0.3 for 2050/2100—Figure 7F,H) and C. Tróia (S) will
decrease by 6% and 38% until 2050 and 2100 (Figure 8D), respectively (
HM/LM = 1.1/0.4
for 2050/2100—Figure 9F,H).
The SMRM with the MOD.FC_2b SLR scenario (111 cm until 2100) produces major
changes until the end of the century. In C. Tróia (N), the area will be reduced by 3% (4.36 ha)
Remote Sens. 2022,14, 3400 15 of 25
and 58% (1.89 ha) (Figure 6E) and the HM/LM ratio will be 1.4 and 0.7 until 2050 and 2100
(Figure 7I,K). For C. Tróia (S), the losses will be almost identical: the area will be reduced
by 8% (9.44 ha) and 57% (4.39 ha) in 2050 and 2100 (Figure 8E). However, the HM/LM
ratio will be 1.0 and 1.1, respectively (Figure 9I,K). SLAMM produces area losses of 3%
and 61% for C. Tróia (N) (Figure 6F) (HM/LM = 1.3/0.5 for 2050/2100—Figure 7J,L) and
of 11% and 61% for C. Tróia (S) for the reference dates (Figure 8F) (HM/LM = 0.9/0.8 for
2050/2100—Figure 9J,L).
Finally, considering SMRM with NOAA Extreme (261 cm of SLR until 2100), the area
will be reduced by 24% (3.42 ha) and 72% (1.24 ha) in 2050 and 2100 in C. Tróia (N)
(Figure 6G). The HM/LM will be 0.3 and 1.2, showing a recovery by the end of the century
(Figure 7M,O). In C. Tróia (S), the losses in the area will be almost the same in 2050 and
2100: 42% (5.98 ha) and 48% (5.40 ha) (Figure 8G). The HM/LM ratio decreases to 0.3 in
2050 and increases to 1.5 in 2100 (Figure 9M,O). SLAMM produces area losses of 25%
and 76% in C. Tróia (N) (Figure 6H) (HM/LM = 0.2/0.8 for 2050/2100—Figure 7N,P)
and 45% and 63% in C. Tróia (S) until 2050 and 2100 (Figure 8H) (HM/LM = 0.3/1.0 for
2050/2100—Figure 9N,P).
In general terms, the results obtained with the two considered models are similar for
both study areas.
6. Sensitivity Analysis
6.1. Methodology
A sensitivity analysis was performed on the C. Tróia (N) salt marsh to evaluate the
influence of each parameter on the result of SMRM, establishing three risk levels: low,
medium, and high (Table 5). For each level, each parameter (DTM, SLR, and accretion
rates) was combined in all the available possibilities. In total, 36 different combinations
were considered in this exercise and the results were plotted in boxplots (Figure 10).
Table 5. Risk levels for the sensitivity analysis performed with SMRM to the C. Tróia (N) marsh.
Risk DTM SLR Acc. Rates
IPCC RCP2.6
Low SLR −RMSE IPCC RCP4.5 5.46/5.84 mm/year
(tidal flat/marsh)
Medium - MOD.FC_2b 2.73/2.92 mm/year
(tidal flat/marsh)
High SLR + RMSE NOAA Extreme 0 mm/year
Each risk level includes different values for each evaluated parameter (DTM, SLR,
and Accretion Rates). The three risk levels of the DTM were formed through the inunda-
tion heights, adding or subtracting the RMSE to the SLR [
53
]. Four SLR scenarios were
considered: the conservative one (low risk—IPCC RCP4.6), the intermediate (medium
risk—MOD.FC_2b), and the high-end (high risk—NOAA Extreme). In addition, IPCC
RCP2.6 was also considered since it is a scenario with an almost linear behavior. Lastly,
the determined accretion rates were used for the medium-risk scenario. For the low risk,
accretion rates were doubled (5.8 mm/year for high and low marsh; 5.5 mm/year for tidal
flat) and eliminated (0 mm/year) for the high-risk scenario. The evaluated results were the
change in the global area as well as the inundation area (which is the new area colonized
by marsh that is currently terrestrial area) until 2100.
Results yielded by SMRM are contained in a GeoTIFF file. Therefore, it is possible
to determine the total addressed area, partial high and low marsh areas in the simulated
time horizon, and the inundation area (which is the area now colonized by marsh vegeta-
tion). Therefore, both the changes in total and inundation areas in 2100 were used in the
sensitivity analysis.
Remote Sens. 2022,14, 3400 16 of 25
Remote Sens. 2022, 14, x FOR PEER REVIEW 17 of 27
6. Sensitivity Analysis
6.1. Methodology
A sensitivity analysis was performed on the C. Tróia (N) salt marsh to evaluate the
influence of each parameter on the result of SMRM, establishing three risk levels: low,
medium, and high (Table 5). For each level, each parameter (DTM, SLR, and accretion
rates) was combined in all the available possibilities. In total, 36 different combinations
were considered in this exercise and the results were plotted in boxplots (Figure 10).
Table 5. Risk levels for the sensitivity analysis performed with SMRM to the C. Tróia (N) marsh.
Risk DTM SLR Acc. Rates
IPCC RCP2.6
Low SLR − RMSE IPCC RCP4.5 5.46/5.84 mm/year
(tidal flat/marsh)
Medium - MOD.FC_2b
2.73/2.92 mm/year
(tidal flat/marsh)
High SLR + RMSE NOAA Extreme 0 mm/year
Figure 10. Boxplots of the sensitivity analysis performed on SMRM. The analysis includes three
levels of risk for each evaluated parameter (DTM, SLR, and accretion rates—Acc. Rates). The
analyzed results are the marsh area and the inundation area in 2100.
Each risk level includes different values for each evaluated parameter (DTM, SLR,
and Accretion Rates). The three risk levels of the DTM were formed through the
inundation heights, adding or subtracting the RMSE to the SLR [53]. Four SLR scenarios
were considered: the conservative one (low risk—IPCC RCP4.6), the intermediate
(medium risk—MOD.FC_2b), and the high-end (high risk—NOAA Extreme). In addition,
IPCC RCP2.6 was also considered since it is a scenario with an almost linear behavior.
Lastly, the determined accretion rates were used for the medium-risk scenario. For the
low risk, accretion rates were doubled (5.8 mm/year for high and low marsh; 5.5 mm/year
for tidal flat) and eliminated (0 mm/year) for the high-risk scenario. The evaluated results
were the change in the global area as well as the inundation area (which is the new area
colonized by marsh that is currently terrestrial area) until 2100.
Results yielded by SMRM are contained in a GeoTIFF file. Therefore, it is possible to
determine the total addressed area, partial high and low marsh areas in the simulated time
horizon, and the inundation area (which is the area now colonized by marsh vegetation).
Figure 10.
Boxplots of the sensitivity analysis performed on SMRM. The analysis includes three levels
of risk for each evaluated parameter (DTM, SLR, and accretion rates—Acc. Rates). The analyzed
results are the marsh area and the inundation area in 2100.
6.2. Results
The sensitivity analysis results show that SLR is the most critical parameter to consider
in the evolution of the marsh areas, because each SLR scenario produces a different set
of results. For instance, when the NOAA Extreme scenario is considered, C. Tróia (N)
marsh will lose between 70% and 80% of its original area until the end of the 21st century,
no matter which risk level is selected for the remaining parameters. An exception is made
for IPCC RCP2.6 and RCP4.5, which may produce similar results. Accretion rates are also
a fundamental input for the model (as other authors have reported [
66
,
67
]). Double rates
(low risk) may produce lower variations in the global area than the medium risk level.
However, the most significant impacts are observed when no accretion rates are considered
(high risk). Thus, it is possible to infer that accretion rates are an essential parameter to
consider, even if an increase in the value does not produce proportionally different results.
The RMSE of the DTM seems to have an impact on the global area that is proportional to
the size of the error.
Regarding the inundation area, SLR is also the most relevant parameter, although only
NOAA Extreme (high risk) produces a very significant change in the result. As in the global
area evolution, the RMSE of the DTM also impacts the inundation area proportionally
to the error’s size. Lastly, accretion rates do not influence the inundation area, which is
only controlled by the inundation height, which is different according to the chosen SLR
scenario and RMSE of the DTM.
7. Discussion
The discussion of this work is based on three topics: the advantages and drawbacks of
applying SMRM to marsh areas; the projection of the evolution of the test areas; and the
possibility of extrapolating the results to other areas.
Using a reduced-complexity model allows for obtaining projections of marsh evolution
considering only the fundamental data. This case includes tidal levels, SLR, accretion rates,
and the local morphology. Using as few parameters as possible without compromising
the reliability of the results is essential to study many areas where there are not sufficient
available data or to perform quick preliminary projections. Furthermore, these advantages
Remote Sens. 2022,14, 3400 17 of 25
are crucial to undertake the correct land management of these areas (e.g., to define which
areas should remain free of occupation).
In order to evaluate the reliability of the produced results, a comparison with SLAMM
was performed. Besides the SMRM inputs, SLAMM also requires at least the following
additional parameters: slopes, wetland categories map (spatial parameters), great diurnal
tidal range, and salt elevation (site-specific parameters). SLAMM also allows the user to
choose among predefined SLR scenarios, which are outdated. The alternative is to input
the mean sea-level position for each year until the end of the century. The results obtained
with SMRM are quite similar to those obtained with SLAMM regarding the global area,
HM/LM, and marsh distribution. For the initial year, both models recalculate the marsh
area, which is the basis for the simulations performed until the end of the century. SLAMM
seems more accurate in representing the initial salt marsh area, although the differences are
not significant. Differences between models are always related to the maximum inundation
line—the behavior of seaward limits is always similar in both models.
The sensitivity analysis shows that the parameter that most significantly controls the
evolution of marsh areas with SMRM is the SLR—the result is clearly different according
to the chosen scenario. It is the most powerful parameter regarding both the evolution of
the marsh area and the inundation area. Conservative and intermediate scenarios produce
comparable results, but the high-end scenario can project an almost complete inundation
of these areas. The other two parameters are relevant in different ways. When the RMSE of
the DTM is considered, a proportional change in the evolution of the marsh is identified
(i.e., the greater the error, the larger the expected inundation, e.g., [
53
,
68
,
69
]). Accretion
rates are crucial for estimating the marsh area in the future, but they are not relevant to
determining the inundation area, since it only depends on the DTM and SLR. On the one
hand, if the accretion rates double, it seems that the results are hardly affected. On the other
hand, considering no accretion rates, a major difference in the results can be produced.
Many studies have been published regarding the evolution of wetlands, trying to
find a global tendency of evolution, even if each wetland can have its behavior in the
future [
31
]. Recently, marsh areas have kept pace with SLR [
70
], with small area losses [
71
].
However, there are different results (from different studies) on the future of salt marshes.
Some authors project that in the future, these areas will survive with conservative SLR
scenarios [
72
] or even if high-end scenarios are considered [
73
], but other studies always
project a significant loss in area in these environments [
74
]. The results of SMRM indicate
that the accommodation space landwards will be fundamental to the survival of wetlands,
as other authors previously suggested [
75
]. Even so, it is reasonable to assume that the
stability of marsh areas should be guaranteed in the first half of the century, mainly when
conservative or intermediate SLR scenarios are considered. However, the HM/LM ratio
reduction seems inevitable in all projections for both study areas, mainly in the second
half of the century. This reduction occurs because, when sea level rises, the tidal limits
that define each environment’s existence also increase in height. If accretion rates are not
high enough to compensate for the SLR rates, areas currently colonized by high-salt marsh
vegetation will suffer longer periods of annual flooding. Consequently, these areas will be
colonized by vegetation more tolerant to inundation than high marsh vegetation. If sea level
continues to rise, the annual inundation periods in these areas will also continue to increase.
They may reach values that will be no longer compatible with low marsh vegetation.
As a result, these areas will not keep any vegetation type, and will become non-vegetated
areas (tidal flats). The consequences of these losses are reflected in the diminishing of
the ecosystem’s services, such as physical protection or carbon sequestration [
76
]. In fact,
the loss of marsh areas due to SLR could imply an emission of CO
2
to the atmosphere
equivalent to the increase in global greenhouse emissions [
77
]. That is why the preservation
of marshes is seen as a natural climate solution to climate changes [78,79].
In this work, four theoretical phases in the evolution of salt marsh areas until the end of
the 21st century were identified (Figure 11): (I) stability, (II) significant loss of maturity, (III)
significant loss of area and, in cases, (IV) a partial area recovery. High-end SLR scenarios
Remote Sens. 2022,14, 3400 18 of 25
tend to generate more evolution phases than conservative projections (this is always visible
in Figures 7and 8). The behavior is similar in all cases, which is different in the timing of
each evolution phase occurrence.
Remote Sens. 2022, 14, x FOR PEER REVIEW 20 of 27
Figure 11. Schematic representation of the four identified phases associated with salt marsh area
evolution in an SLR context. (I) Stability; (II) loss of maturity; (III) loss of area; (IV) recovery.
The above-described sequence of evolution indicates that many salt marsh areas will
probably disappear in the future. However, there are landward areas adjacent to the
present salt marsh limits that, in turn, will be flooded frequently enough to establish new
high and low marsh areas. The increase in the annual inundation periods may also
promote the deposition of sediments transported in suspension by tides, further
facilitating the colonization of these supratidal areas by marsh vegetation. A consequence
of these dynamics would be the formation of new salt marsh areas that may minimize
(although not compensate) the projected losses in the remaining marsh or even lead to a
partial recovery of the global area.
These assumptions are valid if there is no direct anthropogenic destruction of marsh
areas until the end of the century and if there is a constant flux of sedimentation. However,
this is not necessarily true, because these environments are increasingly threatened by
anthropogenic factors and sediment deficits [80]. In addition, it is also assumed that the
increase in storm intensity in the future [81,82] will not increase the impact on these areas,
which is likely true for the tested marshes, since they occur in a very sheltered area.
8. Conclusions
The SMRM is a reduced complexity model that aims to compromise between
complex models and simplistic approaches to quantify the evolution of salt marsh areas
in a climate-change context. SMRM does not consider as many parameters as other more
complex numerical and rule-based models and the number of required inputs to perform
a simulation greatly simplifies the process. In </