The impact of climate change on the morphology of a tidal freshwater wetland affected by tides, discharge, and wind

Tidal freshwater wetlands are threatened by climate change, especially by rising sea levels. Until now, research in these wetlands has focused mostly on determining historical and present‐day accretion rates without analysing the influence of climate change on future developments. We study a recently constructed freshwater wetland under influence of tides, wind, and riverine discharges and carry out a scenario analysis to evaluate the impact of climate change on morphodynamics. We use a numerical model that describes the hydrodynamics and morphology in the study area and includes the impact of vegetation and carry out transient scenario runs for the period 2015–2050 with gradually changing boundary conditions. We conclude that the simulated accretion rates are significantly lower than the rate of sea level rise, meaning that the wetland will gradually convert to open water. We also find that the morphological changes can largely be attributed to morphological stabilization of the constructed wetland and not to climate change. Wind plays an important role through resuspension and redistribution of fine sediment, and neglecting it would lead to a significant overestimation of accretion rates on the flats.

such as supply of sediment (Neubauer et al., 2002), impact of tides and wind (Orson, Simpson, & Good, 1990;Vandenbruwaene et al., 2011;Verschelling et al., 2017), and wetland parameters such as average water depth, wind fetch lengths, and distance to creeks (Delgado et al., 2013;Hupp & Bazemore, 1993;Temmerman et al., 2003). This paper focuses on the long-term effects of CC on the morphology of freshwater wetlands. More specifically, our objective is to understand and quantify the effects of CC on the morphology of a freshwater wetland affected by discharges, tides, and wind. To this end, we use a small microtidal flow-through wetland, located in the Biesbosch National Park in the Rhine-Meuse delta in the Netherlands, as a case study and compared net sediment deposition rates and sedimentation patterns of two distinct CC scenarios to present-day climate conditions. For this analysis, we deployed a numerical model that accounts for the interactions between hydrodynamics, morphology, and vegetation. We carried out transient scenario runs for the period 2015-2050, in which the boundary conditions changed on a yearly basis, using synthetic yearly time series of discharge, water level, and wind with similar statistical properties (auto and cross correlations) as the original measured time series. The simulation results were then post-processed to net sedimentation/erosion patterns over the years and bed levels and the evolution of average and flat levels over time. Finally, we tested the sensitivity of the accretion rates to an alternative vegetation scenario and to the impact of wind.

| STUDY AREA
The study area consists of a number of former polders in the Biesbosch National Park, a 9,000 ha tidal freshwater wetland in the lower Rhine and Meuse delta in the Netherlands (Figure 1). During 2007-2008, a channel network was constructed through the polders, and the embankments were opened at the northern and southern sides of the area, connecting the newly created wetland to the river Nieuwe Merwede and Gat van de Noorderklip, respectively. The objective of this opening was to lower upstream flood levels by enlarging local conveyance capacity of the river Rhine. Due to the location of the newly created wetland in the backwater of the North Sea (Kleinhans, Weerts, & Cohen, 2010), it is potentially threatened by SLR, making it a relevant study area. It has a limited number of connections to the surrounding water system, which facilitates the construction of sediment and water balances.
The flow-through wetland has a surface area of about 700 ha and has a dominant flow direction from north to south. The sandy material that was dug from the channels was used to create an artificial island in the centre of the area. The rest of the area has an average surface level of 0.3 m above the Dutch Ordnance Datum (NAP) and consists of mud flats still covered by the original layer of polder clay. The channels occupy around 25% of the surface area of the study area. Tides in the region are semidiurnal with a typical range between 0.2 and 0.4 m. Average depth on the tidal flats ranges from 0 to 0.5 m. Occasional exposure of the flats occurs at low tide or during easterly winds in the summer. Water levels in the area are strongly affected by Rhine discharge and tides. Storm surges from westerly winds at sea occasionally cause large setup in the wetland. Winds play an important role in shaping the morphodynamics of the system. Locally generated wind waves cause a substantial amount of resuspension due to the long fetch lengths across the inundated flats (Verschelling et al., 2017).
Suspended sediment concentrations (SSCs) at the inlet of the system typically vary from 10 to 40 mg/L during average flow conditions. During peak discharges, SSCs in the upstream Rhine River can reach up to 140 mg/L (Asselman, 2000). Since the de-embankment of the area in 2008, sedimentation has taken place in the central section of the channel system with values up to 14.3 mm/year, whereas channel sections close to the inlet and outlet have experienced significant erosion (van der Deijl, van der Perk, . The flats have remained relatively intact due to the erosion resistance of the layer of polder clay in combination with low flow velocities. Since 2008, most of the flats have been covered by a layer of mud 2 to 5 cm thick, with most of the aggradation occurring close to the channels (Verschelling et al., 2017). We refer to van der Deijl et al. (2017) and Verschelling et al. (2017) for more details on the morphological evolution in the study site. Vegetation in the study area is sparse. It consists mostly of shrublands and softwood riparian forests on the higher grounds, reed fields, pioneer herbs, and multiple types of grassland on the banks along the island and the flats and multiple species of mudwort and macrofytes on the flats (Van der Werf, 2016). Currently, vegetation is kept short by grazing (cows and geese) and by mowing. This effectively removes most of the vegetation from the mudflats and the island and also conserves the grassland.

| Model setup and calibration
We used the depth-averaged version of the Delft3D modelling framework (Lesser, Roelvink, van Kester, & Stelling, 2004) to simulate water flow, sediment transport, and bed level changes in the study area. Here, we provide a short overview of the model setup and calibration, both of which are described in more detail by Verschelling et al. (2017).
Bed level changes are simulated based on differences in sediment transport rates of cohesive and noncohesive sediment fractions, calculated with the Krone and Ariathurai-Partheniades equations (Partheniades, 1965) (Verschelling et al., 2017). These datasets were provided by the Dutch National Water Authority (Rijkswaterstaat). Initial bed composition consisted of one uniformly mixed layer of two sediment fractions (one cohesive mud fraction and one noncohesive sand fraction) with 100% mud on the flats and 100% sand in the rest of the system (channels and island). The hydrodynamic and sediment transport models were calibrated and

| Scenario calculations
We used the so-called 2050GL and 2050WH CC scenarios     model using a 1D model of the Dutch river network, described in (Chbab, 2012).
Third, we constructed synthetic time series of 2-hourly discharges, water level, and wind speed (assuming dominant wind direction of WNW) with a duration of 1 year using an approach described in detail in Van den Boogaard, Uittenbogaard, and Mynett (2003). This approach consists of an iterative optimization algorithm that uses random seeds to construct synthetic time series that satisfy several a priori defined statistical properties. In this case, those properties are the marginal probability density distribution and autocovariance function of the routed signals at the locations of the Delft3D boundaries. For the water level, a distinction was made between a so-called carrier signal (setup) and a residual signal (tide).
Although the algorithm was originally developed for univariate distributions only, we manually checked for cross correlations between the discharge, water level, and wind speed. This is important because in our case, these variables are correlated due to the location of our  (2015) climate and the synthetic version of these series

| Vegetation
We modelled the hydraulic resistance due to vegetation using the   (2033). We constructed this vegetation cover as follows: first, we selected six representative vegetation types, and for each of these types, a key species was selected to represent that type (Table 3a) (Table 3a) to arrive at the fixed cover for the 2015-2050 scenarios (Figure 5b).

| RESULTS
For all scenarios, most of the changes in morphology occur in the channel system between −0.5 m NAP (Figure 6a) and 0.3 m NAP and below −3 m NAP (not shown in Figure 6): they become narrower and deeper. Compared with the 2050REF, the 2050WH and 2050GL scenarios both lead to enhanced sedimentation on the channel banks and an elevated bed level of the flats. Nevertheless, the difference between 2015 and 2050REF is much larger than the difference between 2050REF and 2050WH or 2050GL, implying that most of the morphological changes in the area can be attributed to morphological stabilization of the relatively recently reclaimed wetland and not as much to the impact of CC.

| DISCUSSION
The current mean accretion rate on the intertidal flats in our study area is about 6 mm/year, which is just sufficient to keep up with the current rate of SLR . However, our results show that this rate will quickly drop to about 1 mm y −1 for the period 2015-2050 for the two CC scenarios. These rates are substantially lower than the local rate of SLR (2050GL: 8 mm/year and 2050WH: 10 mm/year), which implies that this marsh will likely gradually convert to open water over decades.
Because the bed level changes of the 2050REF scenario are much larger than the difference between 2050REF and 2050WH or 2050GL scenarios, most of the morphological changes in the relatively recently constructed Kleine Noordwaard wetland over the period 2015-2050 can be attributed to morphological stabilization and not to the impact of CC. CC does however increase the rate of these changes, especially in the western branch of the channel system: these tend to fill up more rapidly under CC. We speculate that this process will likely continue beyond 2050, up to a point where it starts to limit the water discharge into the system and with it, the influx of sediment.
The current sedimentation rates in the study area fall within the range of rates reported for other sites along the lower Rhine branches, such as a de-embanked polder with the rate of 1-2 mm/year (Bleuten, Borren, Kleinveld, Oomes, & Timmermann, 2009) and the floodplains along the river Waal with rates between 0.2 and 11.6 mm/year (Middelkoop, 1997). These values are low compared with accretion rates found for freshwater marshes in the USA which vary between 1 and 27 mm/year (Delgado et al., 2013;Mitsch et al., 2014;Orson et al., 1990). Sediment starvation often plays a large role in reduced sedimentation rates (Neubauer et al., 2002). This is also the case for the river Rhine, where upstream river regulation works have led to a significant drop in SSCs over the last decades (Snippen et al., 2005;Vollmer & Goelz, 2006). Current SSC at the inlet of our study area averages at about 15 mg/L, which is very low compared with other TFWs (van der Deijl et al., 2017).
Our study is one of the first where the impact of CC on the sediment budget of a microtidal freshwater wetland is quantified. Coastal wetlands have been studied more extensively and often gain elevation at speeds similar to SLR due to ecogeomorphic feedback loops that cause increased deposition of both mineral sediment and organic material as the water depth increases (French, 2006;Kirwan & Megonigal, 2013). However, these feedback loops only occur until a certain flooding threshold, beyond which the vegetation dies off and the feedbacks are stopped, causing wetlands to drown (Kirwan & Megonigal, 2013). For coastal wetlands, this threshold can be reached in case of very high rates of local SLR, in some cases causing wetland submergence (Cahoon, Reed, & Day, 1995;Kirwan et al., 2010  Our research suggests that it is important to ensure that abiotic conditions after de-embankment promote vegetation growth and succession in order to prevent drowning. This study also underlines the need to distinguish between the different components contributing to morphological changes: sedimentation patterns and rates are governed by a balance between boundary conditions (affected by external factors such as CC and anthropogenic modifications such as dams and gates) and internal conditions (such as the channel system and bed levels). Exciting future research directions would be to further assess the role of internal drivers such as wetland shape, channel configuration, and inlet sizes on the long-term accretion rates and patterns in TFWs.

| CONCLUSIONS
We carried out a scenario analysis to gain insight in the impact of CC on the morphology of a recently constructed TFW affected by a combination of tides, winds, and riverine discharges. The main conclusions are as follows: • The scenario study shows that the simulated accretion rates over the scenario period are significantly lower than the rate of SLR for the two CC scenarios. This means that the study area will gradually convert to open water.
• Nevertheless, CC leads to enhanced sedimentation on the channel banks and an elevated bed level of the flats.
• Most of the morphological changes that take place over the scenario period  can be attributed to morphological stabilization and not to CC.
◯ Present-day abiotic conditions do not lead to vegetation succession and ecogeomorphic feedbacks that may promote increased accretion rates. Instead, the considerable water depth and inundation frequency lead to vegetation die-off and corresponding increase in wind shear.
• Neglecting the impact of wind leads to a significant overestimation of accretion rates.