Restoration transforms coastal wetland ecosystem services over time

Abstract

Habitat restoration is increasing globally, aiming to recover biodiversity and ecosystem services. To assess the success and value of such restoration projects, we need to understand how and when they produce desired restoration outcomes. However, there is currently limited data‐driven understanding of how biodiversity and ecosystem services develop over extended timescales (e.g., decades) following restoration. While ecosystem services can be challenging to quantify, functional traits can be used as effective proxies that are more easily measured. Here, we employed a space‐for‐time approach at multiple restored sites (6–33 years old) in two estuaries in eastern Australia (Hunter River estuary and Maroochy River estuary) to explore temporal trajectories of functional traits in coastal wetlands. We found diverse, often nonlinear temporal trends for numerous functional traits in mangrove and saltmarsh habitats, including saltmarsh cover, saltmarsh burrows, mangrove height, and mangrove saplings, all of which directly link to ecosystem services. Further understanding of ecosystem service changes after restoration can inform ongoing monitoring and management, which may improve predictions of recovery timescales. These will help refine restoration goals and manage project expectations, thereby increasing future restoration success.

INTRODUCTION

In the second decade of the 21st century, the anthropogenic damage to global ecosystems and climate is widespread, and we are increasingly turning to ecosystem restoration to restore and protect biodiversity and ecosystem functions (Waltham et al., 2020). However, many restoration projects fail to quickly meet success criteria or anticipated restoration outcomes (e.g., Beheshti et al., 2022; Simenstad & Thom, 1996), and there is an urgent need to rapidly upscale restoration efforts to meet ambitious, but necessary, global targets (Joly, 2022; Waltham et al., 2020). To understand if restoration goals are being met or are on a promising trajectory, and thereby inform adaptive management of current and future projects, we need to monitor and assess restoration outcomes. Studies often assess restoration “success” by comparing the restored site to an equivalent natural or reference site/s (Bayraktarov et al., 2016). However, many studies only measure restoration outcomes at one or few time points, usually over short timescales (<10 years for terrestrial projects and <5 years for marine projects) (Atkinson et al., 2022; Danovaro et al., 2025). This means that we lack information on how different ecosystem characteristics and values develop over longer timescales (i.e., 20+ years; Mitsch & Wilson, 1996), or how long they might take to reach “maturity” or equivalence with natural sites (Ferraz et al., 2020). To address this research gap, Lanceman et al. (2025) conceptualized an approach for characterizing ecosystem service trajectories, outlining seven components that define ecosystem service development following restoration. This approach can help inform restoration planning, monitoring and assessment, and should be tested in the field. Here, we demonstrate how several of Lanceman et al.'s (2025) recommended trajectory components can be quantified on‐site, using coastal wetlands as a case study.

A major challenge for understanding restoration trajectories is the lack of long‐term restoration data. Given it may take decades to compile a comprehensive long‐term dataset for a single restoration site, using a space‐for‐time or chronosequence approach (collecting data at differently aged sites with similar environmental conditions) (e.g., Sutherland et al., 2016; Thura et al., 2022) can enable broadly applicable inferences to be made about long‐term restoration trajectories with a limited period of data collection (Morgan & Short, 2002). For example, Sutherland et al. (2016) used a space‐for‐time approach to determine nonlinear trends in recovering Canadian forest ecosystem services, and Atkinson et al. (2023) used this approach to explore temporal trends in restored Australian forest biodiversity indicators. Ideally, a space‐for‐time approach should be undertaken alongside monitoring of natural reference ecosystems, to provide a positive baseline level to assess restoration progress (Lanceman et al., 2025). Further knowledge of long‐term restoration trajectories has the potential to provide valuable insights for restoration practitioners about when and what to monitor, and to enable more accurate restoration predictions, thus increasing the likelihood of achieving project goals (Abelson et al., 2020).

A second challenge in quantifying ecosystem service trajectories is the difficulty of directly measuring most services, particularly at scale (Tomscha et al., 2021). However, functional traits can be effective proxies for ecosystem services (Carlucci et al., 2020; Miedema Brown & Anand, 2022). For example, leaf nitrogen content is consistently positively correlated with soil fertility (Miedema Brown & Anand, 2022). Functional traits are morphological and/or physiological characteristics of taxa that relate to organism functioning (Reich et al., 2003) and are closely associated with the provision of ecosystem services (Miedema Brown & Anand, 2022). Hundreds of plant functional traits have been measured in the literature (e.g., nearly 500 for Australian species; Falster et al., 2021), and many of these are correlated with ecosystem services. For example, Cebrian‐Piqueras et al. (2021) found that a combination of plant traits (e.g., relative quantity of fine roots) and community assemblies helped explain ecosystem service provisioning (e.g., belowground carbon storage) in a European saltmarsh. Functional traits have also been used for monitoring ecosystem services in restoration (thoroughly reviewed by Carlucci et al., 2020); however, this rarely translates to use by restoration practitioners (Merchant et al., 2022). Thus, there is a need for identification of suitable, easy‐to‐measure traits correlated with management‐relevant ecosystem services in specific ecosystems and environmental contexts.

Coastal wetlands, including mangroves, saltmarsh, and mudflats, are intertidal habitats that provide many important ecosystem services, including carbon sequestration, coastal protection, and fisheries production (Barbier et al., 2011; Friess et al., 2020; Marin‐Diaz et al., 2022). Globally, more than 50% of historical intertidal wetlands have been lost since European colonization (Li et al., 2018), with many more areas degraded or threatened. Coastal wetland restoration practices aim to counter these losses, with a common approach being tidal restoration (Hardy & Wu, 2020; Zhao et al., 2016). Tidal restoration, including managed realignment, often involves the breaching of artificial levees or dykes to reinstate intertidal flows into a bare or cleared area, followed by the recolonization of wetland plants from existing seeds in the estuary and/or soil seedbank (Glamore et al., 2021; Herbert et al., 2016; Sadat‐Noori et al., 2021). Tidal restoration is the basis of blue carbon accreditation methods (Lovelock et al., 2022) and is promising for future co‐benefit valuation policies (Hagger et al., 2022). Understanding tidal restoration trajectories would greatly improve prediction and economic valuation of these ecosystem services over time. Trajectories could also be used to help inform model reference systems and goals for restoration in locations where nearby reference sites are not available (Cadier et al., 2020). Identification of measurement parameters that provide insights into restoration progress would also improve monitoring consistency.

The objective of this study was to explore how coastal wetland (saltmarsh and mangrove) ecosystem services develop over time following tidal restoration. Functional traits were used as proxies for ecosystem services, and a space‐for‐time approach was employed to represent long‐term change. Specifically, we aimed to (1) characterize temporal trends in saltmarsh and mangrove functional traits following restoration, and (2) test for differences in traits between natural and restored saltmarsh and mangrove communities. We hypothesized that there would be differences in saltmarsh and mangrove traits between natural and restored sites, and that traits would trend towards natural levels with increasing restoration age.

METHODS

Study sites

This study was conducted in April to June 2023 in intertidal wetlands in the Hunter River (NSW) and Maroochy River (QLD) estuaries, in eastern Australia (Figure 1). Both estuaries have a history of intertidal wetland reclamation and interacting stressors, including industrialization, dredging and pollution in the Hunter River estuary (Williams et al., 2000), and sugarcane farming practices in the Maroochy River estuary (Griggs, 2006, 2007). To offset some of these impacts, promote biodiversity, and prepare for sea level rise, tidal restoration is being undertaken, and there are areas (hereafter referred to as sites) in each estuary that have been tidally restored at different times, as well as natural reference sites (Figure 1, Appendix S1: Table S1). We sampled six restored sites in the Hunter estuary, tidally restored between 1990 and 2017. All sites contained saltmarsh and three contained mangroves. We also sampled three restored sites (tidally restored between ~2007 and 2018) in the Maroochy River estuary, all of which had mangroves and one of which had saltmarsh. Two remnant mangrove and saltmarsh wetlands (natural reference sites; Figure 1) were also sampled in each estuary. These sites were identified using historical maps, aerial imagery (Geoscience Australia, 2024; NSW Government Spatial Services, 2025; Queensland Government, 2025), and personal accounts (Moss, 1983; Waterhouse, 1981; Williams et al., 2000).

FIGURE 1.

Spatial distribution of natural and restored sites in the Hunter River and Maroochy River estuaries (with inset: locations in Australia). In the Hunter River estuary, natural reference sites are Ash Island (AI) and Smiths Island (SI), and restored sites of different ages are Cobbans Creek (CC—restored in 1990), Crabhole Flats (CF—1993), Tomago A (TA—2007), Hexham Swamp (HS—2011), Tomago B (TB—2012), and Area E (AE—2017). In the Maroochy River estuary, natural reference sites are Maroochy Wetland Sanctuary (MWS) and Vivigani Street Reserve (VSR), and restored sites of different ages are Coolum Creek Wetland (CCW—restored ~2004–2010), David Low Wetlands (DLW—2014), and Yandina Creek Wetlands (YCW—2018). Imagery source: Esri (2024).

Field sampling and processing

At each saltmarsh site, 12 (Hunter River) or eight (Maroochy River) replicate quadrats (1 × 1 m) were sampled. This difference was due to greater saltmarsh variability in the Hunter River estuary (D. Lanceman, personal observation) and is consistent with research that has shown that saltmarsh diversity increases with latitude (Friess et al., 2012; Saintilan, 2009b). Quadrat locations were selected a priori across tidal inundation levels using tidal planes data (Couriel et al., 2012; Martime Safety Queensland, 2022) and LiDAR data (NSW Government Spatial Services, 2014; State of Queensland [Department of Resources], 2014) in QGIS Version 3.26.2 (QGIS Development Team, 2023). LiDAR‐derived digital elevation models (DEMs) for estuaries were categorized into different tidal levels (e.g., mean sea level, mean high water springs) based on tidal planes data for the flow monitoring station closest to each site. Quadrats were randomly placed in saltmarsh communities at least 50 m apart from one another, within as many tidal levels as possible at each site. Although sampling was not stratified by elevation (elevation was included as a continuous variable in models), sampling across tidal levels helped ensure representative sampling across diverse tidal environments. At each quadrat, we measured several plant functional traits and ecosystem attributes (henceforth referred to collectively as “traits”). These included total saltmarsh cover, community composition (the relative cover of each species), and, in a 0.5 × 0.5 m subquadrat, the number of crab burrows (community traits; Appendix S1: Table S2). We also determined the stem density and reproduction (presence/absence of flowers and seeds) of each species present in the quadrat (species traits; Appendix S1: Table S2, Figure S1). In addition, for each species in each quadrat, we measured stem height and stem diameter of 10 stems (or all stems, for quadrats with <10 stems of a species—stem traits; Appendix S1: Table S2).

At each mangrove site, we sampled five quadrats (10 × 10 m), preselected across tidal levels and randomly placed, as above. In particularly dense areas with >200 trees and saplings per 100 m², 5 × 5 m quadrats were sampled instead. Where possible, mangrove quadrats were set at least 5 m from the edge of mangrove forests to avoid edge effects (Kanniah et al., 2021). At each quadrat, we measured tree density, number of saplings (>1 m tall, <3 cm diameter at breast height [DBH]), number of hollows, and canopy cover (community traits; Appendix S1: Table S3). In addition, for each species, we measured the height, DBH, and number of stems of five trees (or all trees when there were <5 trees—tree traits; Appendix S1: Table S3). We also harvested five healthy leaves per tree and photographed five additional random leaves per tree. Where there were no reachable leaves on mature trees in the quadrat, leaves were taken from saplings, with tree maturity noted. In the lab, healthy leaves were digitally scanned, weighed and dried (at 70–80°C for 48–72 h; Ball et al., 1988; Naidoo, 2010) to calculate leaf wet mass and leaf dry mass. Leaf photos and scanned healthy leaf images were visually analyzed to estimate percentage herbivory (after training with the ZAX Herbivory Trainer; Xirocostas et al., 2021) and leaf area (using a combination of manual area calculations and machine learning analysis; Guo et al., 2024; Schneider et al., 2012), respectively (leaf variables; Appendix S1: Table S3). Additionally, we counted the number of pneumatophores (mangrove roots, counted in a 50‐cm subquadrat), small seedlings (<30 cm) and large seedlings (30–100 cm) in two 1 × 1 m subquadrats within each quadrat (subquadrat variables; Appendix S1: Table S3). The height and thickest diameter of 10 pneumatophores per subquadrat were also measured (pneumatophore variables; Appendix S1: Table S3). All measured saltmarsh and mangrove traits were related to ecosystem functions or services (Figure 2).

FIGURE 2.

Relationships between measured saltmarsh and mangrove traits and select ecosystem functions and services. Saltmarsh traits were total cover, community composition, number of burrows, stem density, reproduction, stem height and stem diameter (Appendix S1: Table S2). Mangrove traits were tree density, number of saplings, number of hollows, canopy cover, tree height, diameter at breast height (DBH), number of stems, leaf area, leaf herbivory, leaf mass per area (LMA), leaf dry matter content (LDMC), pneumatophore (pneu.) density, small seedling density, large seedling density, pneumatophore height and pneumatophore diameter (Appendix S1: Table S3). References for saltmarsh relationships are (1) Chatanga et al. (2020), (2) Carlucci et al. (2020), (3) Xie et al. (2020), (4) Xie et al. (2022), (5) Schoutens et al. (2020), (6) Abbott et al. (2019), (7) Friess et al. (2012), (8) Fouse et al. (2022), (9) Baaij et al. (2021), (10) Raoult et al. (2018), (11) Rogers et al. (2012), and (12) van Proosdij et al. (2023). References for mangrove relationships are (a) Alongi (2008), (b) Faunce and Serafy (2006), (c) Laegdsgaard and Johnson (2001), (d) Hashim et al. (2013), (e) Friess et al. (2012), (f) McConville et al. (2013), (g) Miedema Brown and Anand (2022), (h) Zaman et al. (2023), (i) Thomsen et al. (2022), (j) Burrows (2003), (k) Carlucci et al. (2020), (l) Quadros et al. (2021), (m) Nagelkerken et al. (2008), (n) Norris et al. (2017), (o) Martuti et al. (2017), (p) Horstman et al. (2018), (q) Hill et al. (2024), (r) Zhang et al. (2020), (s) Gijsman et al. (2021), (t) Krauss et al. (2014), and (u) Lovelock et al. (2015).

Environmental attributes

The elevation of each quadrat was recorded using a real‐time kinematic GPS (RTK‐GPS). Where mangrove canopy height prevented GPS readings, these were taken nearby (within 20 m), at a similar elevation on the edge of the mangrove forest. Where RTK‐GPS readings could not be taken near the quadrat, handheld GPS readings were taken and the elevations for those sites were derived from the relevant LiDAR‐derived DEM (NSW Government Spatial Services, 2014; State of Queensland [Department of Resources], 2014). To check consistency of the RTK‐GPS and DEM‐derived elevation data, DEM data were derived for each RTK‐GPS data point and the elevations were plotted against one another. This revealed a 0.3 m difference between these datasets, so DEM values were offset accordingly to align with RTK‐GPS values.

Data analysis

All data were cleaned and analyzed in RStudio Version 2023.09.1 (Posit Software, 2023) with R Version 4.2.1 (The R Foundation for Statistical Computing, 2022). To test our hypotheses that saltmarsh and mangrove functional traits vary with (1) time since restoration and (2) restoration status (natural vs. restored), a generalized linear mixed model (GLMM) was fit for each trait and aim using glmmTMB (Brooks et al., 2017). For each trait in each habitat (saltmarsh and mangrove), we generated two models—one to test the effect of time since restoration, which included only the restored sites, and one to test for differences between natural and restored sites, regardless of restoration age.

Saltmarsh time since restoration models only included data from the six restoration sites in Hunter River estuary, because there was only one restored site in the Maroochy River estuary with saltmarsh. As such, temporal trends could not be calculated at that location. Mangrove time since restoration models only included data from one species (Avicennia marina, i.e. grey mangrove), as there was only one restored site with other species. The specific traits and model details (distribution types and univariate/multivariate statuses) are detailed below in Response variables (see Appendix S1: Tables S7 and S8), while the variables selected for each model are detailed below in Independent variables (see Appendix S1: Tables S4 and S5).

To test each hypothesis, equivalent null models were created without time since restoration or restoration status. Likelihood ratio tests were used to test for differences between the models using the “anova” function. Where differences were detected, and to investigate species‐specific trends in species‐level variables, post hoc analyses were run using “emmeans” and “emtrends” (emmeans package; Lenth, 2024) to calculate confidence intervals (CIs). We determined whether time since restoration effects were polynomial or linear by comparing Akaike Information Criterion (AIC) values for equivalent models. Trend shapes for polynomial predictor variables were investigated using “emmip” (emmeans package). Trends in saltmarsh community composition were visualized using a principal components analysis (PCA) plot (“prcomp” function). To account for the large number of variables measured in this study and reduce the likelihood of spurious results, a graphically sharpened false discovery rate p value adjustment (Benjamini & Hochberg, 2000; Verhoeven et al., 2005) was applied to each of the saltmarsh and mangrove datasets, following Pike (2010). This was applied in two steps: first with only interaction p values, and then for main effects p values (with interaction p values substituted for main effects p values where they were <0.05 in step 1). This method ensured that less than 5% of p values reported as <0.05 are false discoveries. All p values quoted in this paper have been adjusted accordingly, but raw p values are also included in the result tables for reference (Appendix S1: Tables S7 and S8). Additionally, results are reported in the language of evidence, rather than significance, as suggested by Muff et al. (2022) and Smith (2019), with CIs for all tests reported in Appendix S1: Tables S7 and S8.

For all models, assumptions were tested with diagnostic plots (residuals vs. fitted values and normal QQ plots using “simulateResiduals” and “plotQQunif” in the package DHARMa; Hartig, 2022). Collinearity of predictors was tested using the “check_collinearity” function in the package performance (Lüdecke et al., 2021) on models without interaction terms.

Response variables

Saltmarsh traits modeled (dependent variables) were community composition, total cover, burrows, stem density, stem height, stem diameter, and reproduction (Appendix S1: Table S7).

Mangrove traits modeled were the number of trees, number of saplings, number of small seedlings, number of medium seedlings, number of hollows, canopy cover, number of stems, tree DBH, tree height, number of pneumatophores, pneumatophore height, pneumatophore diameter, leaf area, leaf herbivory, leaf dry matter content (LDMC), and leaf mass per area (Appendix S1: Table S8). LDMC was calculated as the ratio between leaf dry mass and wet mass, and leaf mass per area was calculated as the ratio between leaf dry mass and leaf area. We found that leaf dry mass and leaf wet mass were strongly correlated with leaf area (89% and 94% correlation); as a result, neither metric was analyzed.

Elevation was also modeled as a dependent variable for both saltmarsh and mangroves to test whether elevation increases with site age. Multivariate models were used where traits were measured for multiple species (including saltmarsh community composition), while univariate models were used for other community‐level traits. An appropriate regression family was chosen for each model based on each trait's distribution type (see Appendix S1: Tables S7 and S8).

Independent variables

Years since restoration or restoration status (i.e., restored vs. natural references sites) were included as predictors in the relevant saltmarsh models. Other factors included in models were elevation (continuous variable), estuary (i.e., Hunter or Maroochy), and for species‐level traits, species (Appendix S1: Table S4). To account for differences in community composition and inter‐ and intra‐species competition (Janikova et al., 2024) between quadrats, we also included the dominant species within the quadrat (i.e., Sporobolus, Sarcocornia, or other, which included bare ground) as a covariate. Species‐level models also included an interaction between species and species proportion, which was the cover of the measured species in relation to the total saltmarsh cover in each quadrat. Site and quadrat were included as random factors. Reduced‐rank correlation structures were included in species‐level models where different species' traits within a quadrat may be nonindependent. An offset was included for saltmarsh stem density to standardize sample effort while maintaining the mean–variance relationship, as subquadrat size varied among sites.

Mangrove models included the primary predictor time since restoration or restoration status, as well as estuary (and its interaction with restoration status), and random effects for site, quadrat, tree, and subquadrat (where relevant). We excluded elevation due to a strong (visual observation and strong evidence, p = 0.003) polynomial association between mangrove elevation and time since restoration (Appendix S1: Table S5), not unexpected given mangroves accrete sediments following restoration (Lanceman et al., 2025). Leaf models included the binary covariate tree/sapling where data included leaves from saplings, as saplings' leaves are morphologically and physiologically distinct from mature trees' leaves (Thomas & Winner, 2002), and models with multiple species included the covariate species and its interaction with restoration status. To account for differences in quadrat size during sampling, offsets were included for mangrove trees, saplings, and hollows.

Across models, where residual plots showed deviations from linearity, time since restoration and elevation were included as polynomial variables.

RESULTS

We observed several examples of trends in saltmarsh and mangrove traits, and therefore ecosystem services (Figure 3), with time since restoration. We also observed several examples of traits that differed between natural and restored sites. These included traits and services that exhibited distinct trend types (Figure 3: classified as lagged [a, b], neutral [c], asymptotic [d, f], and parabolic [e]) and times to natural equivalence (Figure 3: ~30+ years [a, b, e], ~15–25 years [d, f], and ~<5 years [c]) as per Lanceman et al. (2025).

FIGURE 3.

Stylized summary of key observed saltmarsh and mangrove trajectories following restoration (a: saltmarsh elevation (in Australian Height Datum [AHD]), b: saltmarsh burrows, c: saltmarsh cover, d: mangrove elevation (AHD), e: mangrove saplings, f: mangrove height), demonstrating predicted temporal trends (solid lines), direction of change (positive “+,” negative “−,” or neutral “o”) and variability (95% CIs; dark gray), compared to measured natural reference levels (horizontal dashed lines) and variability (95% CIs; light gray). Vertical dashed lines indicate estimated times when restored levels reach natural levels, and colored symbols represent related ecosystem services (a: sediment accretion, b: habitat provisioning, c: carbon sequestration, d: sediment accretion, e: flow attenuation, f: carbon sequestration). Graphics were created by author Dana Lanceman.

Saltmarsh

Species sampled

We surveyed 21 species in saltmarsh quadrats in the Hunter River estuary and 8 species in the Maroochy River estuary (Appendix S1: Table S6). Of these, only three species (Sporobolus virginicus, Sarcocornia quinqueflora, and Triglochin striata) were frequently observed across sites and thus used for species‐level morphological trait analyses.

Elevation

There was moderate evidence of a nonlinear increasing trend in saltmarsh surface elevation with time since restoration (p = 0.017, Figures 3 and 4, Appendix S1: Table S7). There was also moderate evidence that natural sites had, on average, higher elevations (by 0.3 m) than restored sites (p = 0.034, Appendix S1: Table S7, Figure 4).

FIGURE 4.

Elevation (in meters, Australian Height Datum [AHD]) of surveyed quadrats (a), and number of burrows per 0.5 × 0.5 m subquadrat (b) in restored saltmarsh sites of different ages (left plots), compared to elevations at natural sites (right plots; AI, Ash Island; MWS, Maroochy Wetland Sanctuary; SI, Smiths Island; VSR, Vivigani Street Reserve) in the Hunter River and Maroochy River estuaries, with model output mean restored and natural elevations (dashed lines) and 95% CIs (gray boxes), and raw data (black points). Note that only Hunter River estuary data were used for time since restoration analyses. Boxplots show 25th percentile (lower bound), median (middle line), and 75th percentile (upper bound) values, with lines extending to 1.5× the interquartile range on each side of the median line.

Burrows

There was moderate evidence that burrows increased nonlinearly over time since restoration (p = 0.017, Appendix S1: Table S7), with burrow numbers remaining low (mean < 1) for ~25 years post‐restoration, before increasing with high variability (e.g., at 33 years, predicted mean number of burrows was 13 (95% CI: 1–126); Figure 4). There was no evidence of a difference in the number of burrows between natural and restored sites (p = 0.342, Appendix S1: Table S7).

Community composition

There was weak evidence of an effect of time since restoration on saltmarsh plant community composition (p = 0.078, Appendix S1: Table S7), and very strong evidence of a difference in community composition between natural and restored sites (p = 0.0002, Figure 5, Appendix S1: Table S7). Natural sites had similar community compositions to one another, whereas restored sites were highly variable, with older sites appearing more like natural sites than younger sites in the PCA plot (Figure 5). Drivers of differences in community composition were the proportion of Sporobolus cover and the relative proportion of other species' cover (Figure 5).

FIGURE 5.

Principal components analysis (PCA) plot showing community composition (relative cover of the six most common species—Sporobolus virginicus, Sarcocornia quinqueflora, Triglochin striata, Aster subulatus, Phragmites australis, and Bolboschoenus caldwellii) surveyed in restored saltmarsh quadrats of different ages (years since restoration [YSR]) and natural saltmarsh quadrats (gray) in the Hunter estuary. Arrow lengths are proportional to the relative importance of each species in explaining differences in community composition and directions relate to which axis (PC1 and PC2) they are contributing to.

Total cover

Saltmarsh cover was consistent with time since restoration for Sarcocornia and Sporobolus‐dominated quadrats (80% of quadrats), but declined by 50%–80% over time in other species‐dominated quadrats (very strong evidence of interaction effect; p = 0.0008, Appendix S1: Table S7). There was no evidence of a difference in mean saltmarsh cover between natural and restored sites (p = 0.553, Appendix S1: Table S7).

Stem traits

There was no evidence of an effect of time since restoration or differences between restored and reference sites on species‐specific stem densities (p = 0.264 and 0.625, respectively), stem heights (p = 0.264 and 0.183, respectively) or stem diameters (p = 188 and 0.512, respectively; Appendix S1: Table S7).

Reproduction

We found no evidence of effects of time since restoration on the likelihood of reproduction (flowering or seeding) of different saltmarsh species (p = 0.183, Appendix S1: Table S7). We also found no evidence of effects of restoration (i.e., restored vs. natural sites) on saltmarsh reproduction (p = 0.264, Appendix S1: Table S7).

Mangroves

Species sampled

We sampled one mangrove species (Av. marina) in the Hunter estuary and four species (Av. marina, Aegiceras corniculatum, Bruguiera gymnorhiza, and Rhyzophora stylosa) in the Maroochy estuary (Appendix S1: Table S6). R. stylosa was only sampled at natural sites, so could not be included in analyses. Br. gymnorhiza and Ae. corniculatum were only observed at one restored site, so were only included in restoration status analyses.

Elevation

There was moderate evidence of a nonlinear increasing trend in surface elevation with time since restoration (p = 0.029, Figure 6, Appendix S1: Table S8), with predicted average elevations at the 33‐year‐old mangrove site 0.54 m greater than at the 5‐year‐old site. There was weak evidence of a difference in elevation between restored and natural sites (p = 0.071, Figure 6, Appendix S1: Table S8).

FIGURE 6.

Elevation (in meters, Australian Height Datum [AHD]) (a) of, and number of saplings (b) in surveyed quadrats in restored mangrove sites of different ages (left plots), compared to elevations at natural sites (right plots; AI, Ash Island; MWS, Maroochy Wetland Sanctuary; SI, Smiths Island; VSR, Vivigani Street Reserve) in the Hunter River and Maroochy River estuaries, with model output mean restored and natural burrow numbers (dashed lines) and 95% CIs (gray boxes), and raw data (black points). Boxplots show 25th percentile (lower bound), median (middle line), and 75th percentile (upper bound) values, with lines extending to 1.5× the interquartile range on each side of the median line.

Number of trees

There was no evidence of an association between time since restoration and the total number of trees (p = 0.255), or of a difference between the number of trees in natural and restored sites (p = 0.310, Appendix S1: Table S8).

Number of saplings and seedlings

There was moderate evidence of a polynomial association between time since restoration and the number of mangrove saplings (p = 0.0229, Figure 6, Appendix S1: Table S8). Younger sites (<16 years old) tended to have greater, and more variable, numbers of saplings than older restoration sites (>30 years old), with the greatest numbers at moderate aged sites (9–16 years old; Figure 6). There was weak evidence of a difference in sapling numbers between natural and restored sites (p = 0.060, Figure 6, Appendix S1: Table S8).

There was no evidence of an effect of time since restoration on, or difference between natural and restored sites in, the abundance of small seedlings (<30 cm; p = 0.310 and 0.253, respectively) or medium seedlings (30–100 cm; p = 0.574 and 0.611, Appendix S1: Table S8).

Number of hollows

We observed no evidence of effects of time since restoration on the number of hollows in mangrove sites (p = 0.573, Appendix S1: Table S8). However, there were substantially more hollows at natural sites (moderate evidence; p = 0.029; average of 12 hollows per 100 m²) than at restored sites (average of 2 per 100 m²; Appendix S1: Table S8).

Canopy cover

There was no evidence of an association between the time since restoration and canopy cover (p = 0.304), or of a difference in canopy cover between natural and restored sites (p = 0.154, Appendix S1: Table S8). However, the 95% CI for canopy cover in the Maroochy estuary was higher for natural (59%–79%) than restored (35%–53%) sites (Appendix S1: Table S8).

Number of stems

There was no evidence of an association between time since restoration and the number of grey mangrove stems per tree (p = 0.611), or of a difference in the average number of stems between natural and restored sites (p = 0.253, Appendix S1: Table S8) across species and estuaries.

Diameter at breast height

There was only weak evidence of a trend in mangrove tree DBH over time since restoration (p = 0.060, Appendix S1: Table S8). However, there was moderate evidence that average mangrove tree diameters were greater at natural than at restored sites across species and estuaries (p = 0.029, Appendix S1: Table S8). For example, average grey mangrove diameters in the Maroochy estuary were 6 cm (120%) thicker at natural sites than at restored sites (Appendix S1: Table S8).

Height

There was moderate evidence that grey mangrove tree height increased nonlinearly with time since restoration (p = 0.040, Appendix S1: Table S8), with tree heights increasing till approximately 15 years following restoration, to approximately 6 m (Figure 7). There was no evidence of a difference between mean tree height in restored and natural sites (p = 0.253, Figure 7, Appendix S1: Table S8).

FIGURE 7.

Height of grey mangrove trees in surveyed quadrats in restored sites of different ages (left plot), compared to natural sites (right plot; AI, Ash Island; MWS, Maroochy Wetland Sanctuary; SI, Smiths Island; VSR, Vivigani Street Reserve) in the Hunter and Maroochy River estuaries, with model output mean restored and natural tree heights (dashed lines) 95% CIs (gray boxes), and raw data (black points). Boxplots show 25th percentile (lower bound), median (middle line), and 75th percentile (upper bound) values, with lines extending to 1.5× the interquartile range on each side of the median line.

Pneumatophores

There was no evidence of an effect of time since restoration on, or difference between restored and natural sites in, pneumatophore density (p = 0.343 and 0.568, respectively) or diameter (p = 0.253 and 0.573). There was only weak evidence of an effect of time since restoration on pneumatophore height (p = 0.098), but pneumatophores were substantially taller (by 6 cm or 50%) at natural sites than at restored sites (moderate evidence; p = 0.029, Appendix S1: Table S8).

Leaf traits

There was no evidence of effects of time since restoration on LDMC (p = 0.563), leaf mass per area (p = 0.343), or leaf area (p = 0.226, Appendix S1: Table S8). There was moderate evidence of an increasing trend in leaf herbivory over time since restoration, with mean herbivory three times greater 33 years after restoration compared to 5 years after restoration, after controlling for differences between estuaries (p = 0.040, Appendix S1: Table S8). LDMC was slightly higher at natural sites than at restored sites (moderate evidence; p = 0.040), while there were no differences between restored and natural sites in LMA (p = 0.200), leaf area (p = 0.267, Appendix S1: Table S8), or herbivory (p = 0.310, Appendix S1: Table S8).

DISCUSSION

Tidal wetland restoration projects are increasingly being conducted to promote valued ecosystem services (Bugnot et al., 2024; Hagger et al., 2024; Howie et al., 2024). However, the development of these ecosystem services over time towards targeted reference levels (or not) is largely unknown. Here, we have empirically demonstrated an approach for determining temporal trajectories of ecosystem traits, and likely related ecosystem services, following restoration. We observed examples of traits that developed over a range of timescales and recovery trajectories. Functional traits that relate to habitat structure and composition demonstrated the clearest changes over time since restoration, while individual‐level morphology traits were generally highly variable and trajectory patterns less clear. Our chronosequence approach, looking at trends over time using differently aged sites, allowed for a much deeper understanding of development of ecosystem services than a simple natural versus restored site comparison. As one of few studies to empirically assess how key diverse components of intertidal wetlands develop over time following tidal restoration, this work can provide insights into likely recovery pathways of other restored wetlands, with further applications across ecosystem types.

We observed saltmarsh trends only in measurements taken at the community or habitat structure level, suggesting that for saltmarsh, ecosystem‐level traits may be more responsive or sensitive to restoration changes over our study period (6–33 years after restoration) than individual plant morphology traits. Importantly, commonly and easily measured saltmarsh components (e.g., saltmarsh cover, community composition, and number of burrows) exhibited distinctive patterns, suggesting that existing, relatively rapid monitoring efforts may be sufficient to identify several key aspects of restoration development. Saltmarsh cover was remarkably consistent across time in Sporobolus and Sarcocornia‐dominated quadrats (Figure 3), suggesting that saltmarsh communities can regain maximum cover quickly (in <6 years) following tidal restoration, as found in previous studies focusing on early saltmarsh development in eastern Australia (Glamore et al., 2021; Laegdsgaard, 2006; Lanceman et al., 2022). Given that we also found no clear trends in saltmarsh morphology traits (density, diameter, or height) over time, this suggests that ecosystem services related to saltmarsh aboveground structure (e.g., sediment stabilization, erosion prevention, coastal protection; Carlucci et al., 2020; Schoutens et al., 2020, 2022) may recover within 6 years of restoration. While similarly rapid aboveground recovery has been observed outside Australia (e.g., Craft et al., 1999; Rowland et al., 2024), highly disturbed, slower growing marshes, such as those in the western United States, may take much longer (e.g., 20+ years, Beheshti et al., 2023) to recover, or not trend towards reference levels (Zedler & Callaway, 2002). This highlights that recovery timelines can vary vastly with environmental context, climate variability, site history, and/or restoration approach (Wasson et al., 2025; Zedler & Callaway, 2002).

Unlike cover, saltmarsh community composition exhibited strong differences between natural and restored sites, with restored sites (younger sites in particular) having greater species diversity than natural sites, which were generally Sporobolus‐dominated (Figure 5). This reflects known Australian saltmarsh successional patterns, where Sarcocornia and other herbaceous species are often initial colonizers, gradually replaced by Sporobolus over time (Adam, 1990; Laegdsgaard, 2008; Zedler et al., 1995). Given that even our oldest restoration sites (30–33 years since restoration) appeared to have more diverse community composition than natural sites (Figure 5), it is likely that this successional process may take longer than 30 years. Observed shifts in community composition, and therefore habitat structure and expected function, have implications for numerous ecosystem services, including habitat provisioning and fisheries production (Raoult et al., 2018; Saintilan & Mazumder, 2010), carbon sequestration (Arp et al., 1993; Ehleringer & Cerling, 2002), erosion prevention (Stoorvogel et al., 2025), and nutrient cycling (Ooi et al., 2022; Sousa et al., 2010). It is possible that in eastern Australia, younger restored sites, that have greater species diversity, may support greater invertebrate and microorganism biodiversity than natural sites by providing a greater diversity of resources and habitat niches (Palmer et al., 2016; Saintilan, 2009a). This link between resource availability and infaunal and epifaunal composition has been demonstrated in a Californian saltmarsh restoration study, which found that invertebrate community composition and food web dynamics shifted over time following restoration, due to changes in resource availability (Nordström et al., 2014).

In our study, we also found that belowground habitat provisioning changed over time. Saltmarsh burrows exhibited a lag period before increasing in abundance (Figure 4), suggesting there is likely a minimum time required (e.g., 30 years) for crabs and other benthic invertebrates to establish in previously tidally disconnected saltmarshes, which may have altered soils (Bai et al., 2013) and be some distance from existing saltmarsh invertebrate communities. Note, this lag was not observed in our single restored saltmarsh site in the Maroochy River estuary (Figure 4); this is likely due to its immediate proximity to remnant saltmarsh habitat. Several Hunter River estuary restoration sites are also located near remnant intertidal habitats, but there are potential dispersal barriers like roads or levees separating remnant and restored patches. It may be important to deliberately create connectivity to facilitate infaunal and epifaunal dispersal. Lower crab numbers in young restored sites compared to reference sites have also been observed in previous studies (Alletson et al., 2006; Rozas et al., 2005), potentially due to altered sediment structure, moisture, chemical composition, and pH in reclaimed coastal wetlands prior to restoration (Bai et al., 2013). Altered soil structure, as expected in the Hunter River estuary due to its industrial history (Williams et al., 2000), may therefore explain the observed lagged burrow recovery in the Hunter River estuary. Slow crab recovery suggests that related ecosystem services, including decomposition, nutrient cycling (e.g., An et al., 2021; Saintilan, 2009a), and fisheries production, may also develop over several decades following restoration.

Mangroves also exhibited substantial shifts in habitat structure over time following restoration. This was most notable in the number of saplings (Figures 3 and 6b) observed at 9‐ to 16‐year‐old sites, which was substantially (10–35 times) greater than older (30–33 years old) restored sites and natural sites. Given that restored sites also had taller pneumatophores than natural sites, contributing to additional roughness, these substantial differences in stem density have likely implications for ecosystem service provisioning, with approximately 10‐ to 20‐year‐old restored sites likely more effective at flow attenuation and sediment accretion (Friess et al., 2020; Hashim et al., 2013) than older/natural sites. Similar sapling numbers in older restored sites and natural sites indicates that despite starting with highly elevated sapling numbers, restored sites can thin out and trend towards more open natural mangrove forest structure without management intervention (similarly observed in Alongi, 2008; Maza et al., 2021; Salmo et al., 2013).

As expected, mangrove trees grew in height and diameter over time since restoration (Figure 7, Appendix S1: Table S8), and were thinner in restored compared to natural sites. Tree heights were equivalent to natural sites approximately 15 years after restoration; however, diameters lagged substantially, with differences still present 33 years following restoration. This is consistent with Osland et al. (2020), who found that mature mangrove tree equivalence took at least 55 years. Slow tree diameter growth suggests that aboveground carbon storage develops over multi‐decadal timescales (e.g., 50 years, as observed in a Chinese study; Luo et al., 2010), similar to belowground carbon storage trajectories (e.g., 45–80 years; Osland et al., 2020). Indeed, measured differences in tree diameters, and therefore carbon storage, were likely underestimates of true differences, as we only sampled a small proportion of mangrove trees, and very old, thick (e.g., the thickest trees had 20–50+ cm DBH) trees would contribute disproportionately to carbon storage (Dontis et al., 2020; Jones et al., 2020; Zaman et al., 2023). This slow recovery of large, mature trees is also reflected in the substantially greater (6 times more) numbers of hollows observed in natural sites compared to restored sites. Lagged recovery of mangrove hollows suggests that terrestrial animal habitat provisioning (e.g., for bats; McConville et al., 2013) takes more than 30 years to recover following restoration, highlighting the importance of conserving old‐growth mangrove forests.

We observed threefold increases in Av. marina leaf herbivory over time since restoration. This suggests that older sites may provide greater terrestrial invertebrate habitat than younger restored sites, contributing to associated food web dynamics and nutrient cycling (Burrows, 2003). Lower rates of herbivory observed in younger restored mangrove forests suggest that herbivory is likely not a major barrier to mangrove reestablishment. Previous studies have conversely observed higher rates of insect herbivory in disturbed mangrove sites (Maldonado‐López et al., 2019) and younger leaves (Tong et al., 2006). Further research is needed to determine whether our herbivory observations are typical for eastern Australian mangrove forests. Our observed differences in mangrove LDMC between natural and restored sites were minor (2%), and therefore not necessarily ecologically relevant. However, the moderate evidence of this difference (Appendix S1: Table S8) suggests that natural sites had a slightly more conservative resource acquisition strategy, favoring longer lasting leaves. Higher LDMC within a species has been associated with increased competition (Janikova et al., 2024), and in our case also likely reflects a slight shift from fast growth to longevity strategies in younger versus established mangrove forests (Chelli et al., 2022; Ji et al., 2021; Liu et al., 2019).

Elevation exhibited clear changes over time in both saltmarsh and mangrove habitats, with older restored sites having higher surface elevations, more like natural sites (Figures 3, 4, and 6a). These elevation changes indicate that the restored habitats are effectively performing sediment accretion (Craft et al., 2003; Rogers et al., 2012), as expected following successful tidal restoration, which provides a sediment source (via tidal waters) and trapping mechanism (via growing vegetation) (Henderson & Glamore, 2024; Iram et al., 2022; Rayner et al., 2021). Rapid sedimentation following restoration likely also contributes to water filtration through turbidity reduction (Johnston, 1991; van Proosdij et al., 2023). While elevation was positively correlated with time since restoration, it also varied spatially across sites and is a known factor influencing and influenced by mangrove and saltmarsh zonation and traits (Bass et al., 2022; Kelleway et al., 2016; Saintilan, 2009a). Given that it is challenging to separate these two aspects of elevation variability (time‐driven and spatial variability), we note that both including (as in saltmarsh models, where there was a weaker correlation between elevation and time since restoration) and excluding elevation (as in mangrove models, where there was a strong correlation) may impact interpretation of trends. Including elevation may have masked or altered some time since restoration trends in saltmarsh models, while excluding elevation may have prevented some elevation effects not related to time since restoration from being controlled for.

Unlike most restoration studies that only compare natural versus restored sites (of mixed, or random ages), our chronosequence approach, where we tested for temporal differences across sites of varying ages, had several advantages. First, we were able to detect numerous restoration impacts that would not have been detected, due to variability, in natural versus restored comparisons. These include saltmarsh burrows and mangrove saplings, which both exhibited substantial changes over the 30‐year study period, and likely influence ecosystem service provisioning. Second, we were able to estimate when different traits, and therefore ecosystem services, may be provided by restoration sites. In our study, traits evolved over four broad timescales (Figure 3, Appendix S1: Table S9), namely, (1) short‐term (<5 years), where we observed similarities between natural and restored sites of all ages, suggesting these traits developed over a shorter timescale than the age of our youngest site (e.g., Sporobolus‐dominated saltmarsh cover); (2) medium‐term development (5–30 years), where we observed trait changes at the temporal scale of our study, and similarities between older restored and natural sites (e.g., mangrove saplings); (3) long‐term development (>30 years), where we observed distinct differences between natural and restored sites (regardless of changes with time since restoration), suggesting these traits develop over longer timescales than assessed in our study (e.g., mangrove hollows); and (4) unknown development, where traits exhibited high variability among and within sites, indicating an unknown trend and timescale of restoration development (e.g., saltmarsh stem density). We were also able to determine trend shapes of trait or ecosystem service provisioning over time (i.e., trait provisioning may increase or decrease, or speed up or slow down, at different times following restoration). These include a range of trend types proposed in the literature (Bullock et al., 2011; Lanceman et al., 2025), including asymptotic (e.g., mangrove height), parabolic (e.g., mangrove saplings), and lagged (e.g., saltmarsh burrows) trajectories (Figure 3). Understanding relevant timescales and trend shapes for different traits and ecosystem services could facilitate the development of targeted restoration monitoring regimes with appropriate frequency, duration, and time points of monitoring.

Quantifying time periods required for provisioning of different ecosystem services is also essential for valuing restoration projects via environmental economic accounting frameworks (Hagger et al., 2022; King et al., 2024; Lovelock et al., 2022; Yang et al., 2025). To date, most frameworks have concentrated on a narrow range of ecosystem services (e.g., blue carbon; Glamore et al., 2023; Lovelock et al., 2022), and often utilize a presence or absence (i.e., binary) approach or assume services are provided immediately and at a constant rate following restoration (Carnell et al., 2024; Clean Energy Regulator, 2022). Our approach allows much more accuracy about the relative levels of ecosystem service provisioning at different times following restoration. This level of temporal detail is essential for ensuring that expectations are met regarding restoration outcomes, improving funders' and other key stakeholders' confidence in restoration success (Abelson et al., 2020).

The trait trends observed, and ecosystem service implications explained in this study can be used directly to inform restoration planning, prediction, monitoring, and assessment of intertidal wetland habitats in eastern Australia. Our observed trends are likely representative of wetland development in other eastern Australian estuaries, because both studied estuaries are wave‐dominated barrier estuaries, which make up 50% of the water area of NSW estuaries (Roy et al., 2001), and geomorphologically similar estuaries generally have similar ecologies (Dunlop et al., 2025; Henderson & Glamore, 2024; Mahoney & Bishop, 2018). Given that there is currently little information available about temporal development of functional traits or ecosystem services in intertidal wetland restoration globally (but see the modeling approach in Yang et al., 2025), our findings could also be an interesting comparison point for wetland restoration research around the world. The approach showcased in this study could be applied to other habitat types and restoration techniques as an initial step in quantifying how restoration progresses in different contexts. As more restoration is conducted, future research should include more time points, locations, climatic conditions (e.g., restoration during La Niña and El Niño events), and traits.

Using the approach outlined here to characterize restoration trajectories for a range of ecosystem services across coastal habitat types, restoration managers will be able to predict the timing of restoration outcomes, effectively plan monitoring regimes at appropriate time points, and meaningfully assess success against an expected development trajectory. Therefore, we believe that both the wetland‐specific trends outlined here, and the broader approach developed, will greatly assist in accelerating restoration efforts and success. This is much needed as we reach the midpoint of the UN Decade of Ecosystem Restoration.

AUTHOR CONTRIBUTIONS

Dana Lanceman: Conceptualization; data curation; formal analysis; investigation; methodology; project administration; visualization; writing—original draft preparation; writing—review and editing. William Glamore: Conceptualization; funding acquisition; supervision; writing—review and editing. Eve Slavich: Formal analysis; methodology; writing—review and editing. Mariana Mayer‐Pinto: Conceptualization; methodology; supervision; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Appendix S1.

Lanceman, Dana , Glamore William, Slavich Eve, and Mayer‐Pinto Mariana. 2026. “Restoration Transforms Coastal Wetland Ecosystem Services over Time.” Ecological Applications 36(6): e70292. 10.1002/eap.70292

DATA AVAILABILITY STATEMENT

Data and code (Lanceman, 2026) are available on Zenodo at https://doi.org/10.5281/zenodo.20016363. Additionally, data from the digital elevation model (NSW Government Spatial Services, 2014; State of Queensland [Department of Resources], 2014) were utilized; metadata are available at https://qldspatial.information.qld.gov.au/catalogue/custom/viewMetadataDetails.page?uuid=%7BA03104FF-B849-4748-9B29-4D615AB692B5%7D and the data can be downloaded from the ELVIS portal at https://elevation.fsdf.org.au/?maxx=153.151905&maxy=‐26.423364&minx=152.538297&miny=‐26.99476&metadata=A03104FF‐B849‐4748‐9B29‐4D615AB692B5.