The lack of extreme water level fluctuations in managed, non-peat forming wetland ecosystems can result in decreased productivity through the loss of heterogeneity of these ecosystems. Stochastic disruption, such as a water level drawdown, can effectively reverse this effect and return the wetland to a more productive state, associated with higher biodiversity through new vegetation development. Yet, aside from the effect on vegetation dynamics, little is known about longer-term effects (30 years) of a water level drawdown, hereafter referred to as legacy effects, and how this may impact future water level drawdowns.

Here, we aim to unravel the legacy effects of a water level drawdown, stand alone and along a water level gradient, on seed bank properties and nutrient availability in a eutrophic clay wetland. To identify these, we studied the hydrologically managed nature reserve Oostvaardersplassen in the Netherlands. Here, one section was subjected to a multi-year water level drawdown and another section was kept inundated. We determined seed bank properties in both areas, spatially and along a soil elevation gradient (20 cm). Nutrient availability was measured by taking sediment samples along the water level gradient and through experimental manipulation of the water level in an indoor mesocosm experiment.

Germination was higher in locations with a water level drawdown history, especially at relatively high elevations. Additionally, the proportion of pioneer species in the seed bank was higher in the water level drawdown area. Overall, nutrient concentrations were higher compared to other systems. Nutrient availability was higher in the inundated area and did not respond to the water level gradient. We conclude that 30 years after an induced water level drawdown there is no depletion of nutrients, while we still observe a legacy effect in the number of viable seeds in the seed bank.

2.1 Study site

The study was conducted in Oostvaardersplassen in the Netherlands (coordinates: 52.456857, 5.355935). This eutrophic clay wetland of about 5600 ha consists of a 3600 ha marsh and a 2000 ha dryer border zone. This study took place in the marsh part. The marsh is characterized by large water bodies, reed vegetation and willow forests. Oostvaardersplassen is part of the polder Zuidelijk Flevoland, which is located in the former Zuiderzee estuary, a marine habitat (see van Leeuwen et al., 2021 for a detailed description). For water safety reasons the decision was made to separate the inland Zuiderzee from the North Sea through the construction of a dike, named the Afsluitdijk. After completion of the construction and within five years, the Zuiderzee transformed into a freshwater lake, IJsselmeer. In this freshwater lake, several polders were established to create land for agriculture; Zuidelijk Flevoland was reclaimed in 1968. Since Oostvaardersplassen is located in, what was then, the lowest part of the polder, it remained wet during the first years after reclamation and no actions were taken to develop this area into the industrial site as it was planned to be (Cornelissen et al., 2014).

The marine clay soil and its associated high nutrient concentrations (eutrophic) in combination with the unmanaged and wet conditions, led nature to develop quickly. This made the area into an important breeding and resting area for many wetland birds and therefore became a protected wetland nature reserve in 1974. In 1989 it became a protected area within the European Bird directive and under the Ramsar agreement. Additionally, it was qualified as a Natura 2000 area in 2009. Later on, the relatively high water levels at the end of winter, due to the height of the weir, in combination with high grazing pressure by moulting greylag geese (Anser anser) from May to July, resulted in the loss of reed cover (Phragmites australis) (Vulink and Van Eerden, 1998). This in turn resulted in decreasing bird numbers due to lower food and habitat availability (Beemster et al., 2010). To restore reed-dominated wetlands and to increase food and habitat availability for birds, a complete multi-year water level drawdown was induced in the western part of the marsh from 1987 till 1991 (Vulink and Van Eerden, 1998). The eastern part was hydrologically separated from the western part by a low dike and water levels and dynamics remained unchanged in this area. The implemented water level drawdown resulted in the development of c. 600 ha of reed-dominated vegetation in the western part, after which typical wetland birds, e.g., bearded reedling (Paranrus biarmicus), marsh harrier (Circus aeruginosus) and Eurasian bittern (Botaurus stellaris), increased in numbers (Beemster et al., 2012; Vulink and Van Eerden, 1998).

The study area experiences seasonal variation in water level, but lacks long-term dynamics in water level that would be caused by extreme climatological periods. As the marsh is rainwater fed, natural water level dynamics occur with a high water level at the end of winter (March) and low levels at the end of summer (September;). The surplus of water in winter leaves the marsh via a weir. The average difference in water level between summer and winter is approximately 30 cm. During ‘dry’ summers the water level can drop 50 cm at the end of the growing season. Due to both the climate conditions in combination with the height of the weir, set as to pertain high water levels in the reed beds during late winter and spring, these naturally occurring ‘dry’ summers did not result in enough mudflat exposure throughout the area to allow extensive marsh recovery. At the time of sampling, both the water level drawdown and the non-water level drawdown area were characterized by a sharp border between vegetation and open water. The vegetation on the shores was similar in both areas and dominated by Phragmites australis, Salix spp. and, to a lesser extent, Convolvulus spp.. At drier sites, with greater proximity to the lake, Urtica dioica and Carduus spp. were present in higher abundances. The shores of the lake, that sometimes fall dry during dry summers, are colonized quickly by species among which Tephroseris palustris (also known as Senecio congestus), Epilobium hirsutum and Ranunculus sceleratus.

2.2 Experimental design

We examined the legacy effects of a water level drawdown, a water level gradient and water level fluctuations on seed bank germination and nutrient availability using field sampling and mesocosm experiment. The unique field situation consisting of areas with and without a water level drawdown history allows to explore legacy effects on seed bank properties (Part 1.1) and nutrient availability (Part 2.1). This approach focusses on the long-term effects of inducing a four-year water level drawdown, in this case 30 years after the event, by sampling 20 locations in each subarea that have been inundated since the last water level drawdown. In addition, soil samples have been taken in these two hydrologically distinct areas, along a water level gradient that is dictated by elevational differences of about 20 cm. With this approach, we used the elevational gradient to distinguish between higher locations, that would fall dry more often due to for example dry summers, and lower locations. The latter had not fallen dry for 30 years in case of the water level drawdown area and 50 years in case of the non–water level drawdown area. By taking soil samples on 7 (germination) or 5 (nutrient) locations along this water level gradient, we were able to research how changes in water level alter seed bank properties (Part 2.1) and nutrient availability (Part 2.2) on a smaller seasonal time scale. In addition to the above two sampling campaigns, a mesocosm experiment was conducted to study the effects of water level on germination (Part 3.1) and nutrient availability (Part 3.2) . With this approach it was possible to determine effects of a specified water level (inundated, saturated, dry) on an even smaller time scale of weeks/months and how such a response might be influenced by events in the past, in this case drawdown history. 

2.2.1 Part 1: Water level drawdown history

To investigate the legacy effects of a previously induced water level drawdown on the seed bank (part 1.1) and on nutrient availability (part 1.2), we compared seed bank properties (density, diversity, species composition) and sediment nutrient concentrations between an area with water level drawdown history and an area without. For the method on sediment nutrient concentrations we would like to refer to the section on water level gradient (2.2.2) for field sampling and lab protocols.

2.2.1.1 Seed bank properties (part 1.1)

We collected sediment samples from both areas in Oostvaardersplassen in June 2021, when both areas were still inundated. To cover the spatial heterogeneity of the area, 40 locations were sampled. 20 Sample points were located in the area that was continuously inundated for 50 years (non-water level drawdown history, n = 20) and 20 in the area that had undergone a water level drawdown from 1987 till 1991 and was subsequently inundated for 30 years (water level drawdown history, n = 20).

In June 2021, we took ten sediment cores of 23.8 cm² (diameter = 5.5 cm) to a depth of 10 cm and pooled the 0-5 cm and 5-10 cm depth in separate plastic bags at each location (Verhofstad et al., 2017). The bags were stored in the dark at 4°C for approximately one month to allow seed stratification, after which the sediment was sieved (mesh width: 150 µm) and the residue, containing the seeds, was spread across a tray (37×27 cm) containing sediment for propagation and germination (Lensli substrates; pH = ~5.3; electrical conductivity = ~0.5mS/cm). The trays were placed in a greenhouse with supplementary light from 6:00-22:00h so that light conditions on plant level corresponded with 250 μmol.m²/s. The temperature in the greenhouse was on average 21°C between 6:00-22:00 and 16°C between 22:00-6:00. The relative humidity (Rh) in the greenhouse was on average 60% (-5/+5%). To ensure optimal sediment moisture, the trays were watered at least once a week with rainwater. The germinating plants were then identified to species level and removed afterwards. This was done to minimize possible competition effects between seedlings. Unidentified plants were transferred from the trays to individual pots, providing the space for them to grow and/or flower until their identification could be determined. When germination stopped, the sediment was mixed to allow seeds deeper in the sediment to germinate. The trays were kept in the greenhouse until germination stopped again, which lasted up to 5 months.

2.2.2 Part 2: Water level gradient

To determine how a water level gradient, induced through a gradient in soil elevation of around 20 cm, affects seed bank properties (density, diversity, species composition; Part 2.1) and nutrient availability (part 2.2), we collected sediment samples in the field. Sample collection occurred at seven locations (seed bank) and five locations (nutrient availability) along four transects perpendicular to the border of the reed vegetation. The indicated direction was chosen to cover differences in soil elevation, with locations on a relatively higher elevation falling dry more often due to small fluctuations in the water level and locations on a lower elevation falling dry less often.

2.2.2.1 Seed bank properties (part 2.1)

To assess how a water level gradient alters seed bank properties, we collected sediment samples in June 2021 along four transects, each consisting of seven sampling points (n = 28). The sampling points cover a gradient of soil elevation, where the locations indicated by a 1 are located at the highest elevation, and thus fall dry the most, while locations indicated by a higher number (2-7) are decreasing in soil elevation and thus fall dry less often or never. Each transect covered around 777.5 ± 418.7 meter. Two transects were located in the area without water level drawdown history and two in the area with water level drawdown history. The sampling and germination protocol was identical to the one described in section 2.2.1.

2.2.2.2 Nutrient availability (part 2.2)

To examine how a water level gradient affect nutrient availability, sediment samples were collected along four transects (different from the transects in 2.2.2.1) in November 2021. Each transect consists of five sampling points that were sampled in duplicate (n = 40). The sampling points cover a gradient of soil elevation, where the locations indicated by a 1 are located at the highest elevation, and thus fall dry the most, while locations indicated by a higher number (2-5) are decreasing in soil elevation and thus fall dry less often or never. Each transect covered around 237.5 ± 17.9 meter. Two transects were situated in the area without water level drawdown history and two in the area with water level drawdown history. At each sampling location, four sediment cores of 23.8 cm² (diameter = 5.5 cm) to a depth of 0-10 cm and 20-30 cm were collected for pore-water extraction and one sediment core of 23.8 cm² (diameter = 5.5 cm) to a depth of 0-10 and 20-30 cm was collected for sediment nutrient analyses. Soil elevation measurements were conducted with a dGPS (Topcon, HiPer SR). At each location, we took three measurements which were averaged.

Pore-water extraction was initiated in the lab on the same day as sediment collection and collected the next morning. Pore-water samples were extracted using vacuum syringes attached to rhizons (Rhizon SMS; Rhizosphere Research Products; Eijkelkamp Agrisearch Equipment, Giesbeek, The Netherlands). The pore-water was analyzed for pH, alkalinity (Metrohm, 877 Titrino plus), total inorganic carbon (TIC; infrared carbon Analyser, IRGA; ABB Analytical, Frankfurt, Germany) and nutrient concentrations.

Sediment samples were analyzed on water content, bulk density loss of ignition (LOI; proxy for organic matter content) and bioavailable phosphorus and NH₄⁺ and NO₃^-. The elaborated method can be found in the supplementary material S1. Nitrite (NO₂^-) concentrations were barely detectable and therefore left out of the analysis.

2.2.3 Part 3: Water level fluctuations

Experimental setup

To unravel how water level influences germination (part 3.1) and nutrient availability (part 3.2), we performed a mesocosm experiment with different water levels on intact sediment cores from sites with and without water level drawdown history from Oostvaardersplassen. The different water levels reflect the different stages the system goes through during the first phase (drying) of a water level drawdown cycle: (1) Dry, the water level was 20 cm below sediment surface level (‘dry’ for brevity), (2) saturated, the water level was equal to the sediment surface level (‘saturated’ for brevity), and (3) wet, the water level was eight cm above sediment surface level (‘wet’ for brevity). The experiment ran for eight consecutive weeks in which each core experienced one of the water level treatments (inundated, saturated or dry) following Vonk et al. (2017). In November 2020, intact sediment cores were collected from Oostvaardersplassen at ten locations that were inundated. Half of these locations were situated in an area with a water level drawdown history (n = 5, water level = 13.8 +/- 3.9 cm), while the other half were situated in a continuously inundated area (n = 5, water level = 17 +/- 5.4 cm). At each location, four sediment cores with a diameter of 16 cm and a depth of 40 cm were collected by pressing a PVC-tube in the sediment and sealing it with a cap on the bottom. Three of the intact cores for each location were placed in a climate room for an acclimation period of six days, after which the experiment started. The cores were placed in the climate room with a temperature regime of 20°C from 6:00-22:00 and 15°C from 22:00-6:00. The average humidity in the climate chamber was 45% and the average light conditions at sediment level were 554 μmol.m²/s (LI-COR LI-250 photometer) with 16 hours light and 8 hours dark. The cores were placed using a randomized block design (n = 5), each block consisted of six sediment cores. The treatments were applied by drilling holes in the PVC-tube at the corresponding water level treatment height (-20 cm, 0 cm, +8 cm relative to the sediment height). To regulate the water level in the core, we placed the PVC-tube in a larger water-proof PVC-core (diameter = 20 cm, length = 50 cm). Water collected from the Oostvaardersplassen was used to initiate the treatments. During the experiment, water was replenished till treatment level with rainwater (pH = 5.18, alkalinity = 0.33 mEQ/L). The fourth core was used to determine sediment nutrient starting conditions by taking two sediment samples of 40 cm deep (23.8 cm²) after which it was split in two sections of 10 cm (0-10, 20-30). The two sediment samples from the sediment core were pooled per location and per depth and stored in the freezer at -20°C until further analyses. The same analysis protocol was used as in approach 2 (section 2.2.2.2).

2.2.3.1 Seed bank properties(part 3.1)

Through the use of intact soil cores in an experimental setup, we could identify possible environmental filters that would exert selection on the type of plants that were able to germinate during different phases of a water level drawdown cycle. During the 8-week experiment, the mesocosms were checked weekly for plant germination. Germinated plants were counted and identified to species level if possible. Plants were not removed during the experiment.

2.2.3.2 Nutrient availability (part 3.2)

The experimental setup allowed us to assess how a certain water level regime impacts nutrient availability in the system, in this case, we selected three water levels to mimic different phases of the water level drawdown cycle. By monitoring these changes it would be possible to identify possible nutrient depletion in the system upon repeated water level drawdown implementation. Nutrient concentrations were determined in both the pore-water and the sediment. To collect pore-water samples during the experiment, rhizons (Rhizon SMS; Rhizosphere Research Products; Eijkelkamp Agrisearch Equipment, Giesbeek, The Netherlands) were installed in the sediment core at a depth of 10 cm and a vacuum syringe could be attached to extract pore-water. This was done at the start of the experiment (day 0), and repeated five times on day 7, 14, 21, 35 and 56. Pore-water samples were analyzed in the same way as in approach 2. At the end of the experiment, sediment samples were taken from the sediment cores at two different depths (0-10 cm and 20-30 cm) following the same sampling strategy as at the start of the experiment. These samples were stored in the freezer at -20°C until further analyses, following the analysis protocol as described in approach 2 (section 2.2.2.2).

2.3 Statistical analyses

Data were analyzed in RStudio version 4.0.3 (R Core Team, 2023). For all hypotheses testing procedures the significance level was set at α = 0.05. All data are shown with their average ± standard deviation (sd).

Part 1: Water level drawdown history

Part 1.1 Seed bank properties

To determine the effect of water level drawdown history (Yes or No) on mean Shannon-Wiener diversity, mean species richness, and mean germination densities (log transformed), we used mixed linear models from the GlmmTMB package (Mollie et al., 2017), using location ID as a random effect. Differences in the total sum of germinated individuals between the water level drawdown and non-water level drawdown area were tested using a Chi-Square test. Shannon-Wiener diversity was calculated using the ‘vegan package’ (Oksanen et al., 2022). To assess the effect of water level drawdown history on species composition a permanova analysis with a Bray-Curtis dissimilarity index was used, in combination with non-metric multidimensional scaling (NMDS) (vegan package: Oksanen et al., 2022).

Part 1.2 Nutrient availability

To determine the effect of water level drawdown history and sampling depth (independent variables) on the nutrient availability (dependent variables) along the transect survey (method section 2.2.2.2), we used mixed linear models from the GlmmTMB package (Mollie et al., 2017). The model was performed for both the sediment- and the pore-water nutrient concentrations. Location ID was used as a random effect to correct for the duplicate measurements. Tukey-adjusted comparisons were done using “emmeans” (Russell, 2022). Normality and heterogeneity of the residuals of the models were assessed using histograms, and transformed if necessary.

Additionally, we used the nutrient starting concentrations from the experimental water level experiment (part 3) to determine differences in nutrient concentrations due to the water level drawdown history. To determine the effect of water level drawdown history (independent variable) on nutrient availability (dependent variables), we used mixed linear models from the GlmmTMB package (Mollie et al., 2017). Starting nutrient concentrations (day 0; field conditions) were used as the dependent variable. Field location ID was used as a random effect to correct for samples taken at the same location.

Part 2: Water level gradient

Part 2.1 Seed bank properties

To determine the best fit of the relation between germination and distance to the reed border, we compared the AIC of linear, parabolic, hyperbolic and exponential decay functions. An ΔAIC ≥ 2 was used to differentiate models ( ‘stats’ package (R Core Team, 2023). To assess the effect of water level drawdown history and location along soil elevation gradient on species composition, a permanova analysis with a Bray-Curtis dissimilarity index was used in combination with non-metric multidimensional scaling (NMDS) (Oksanen et al., 2022).

To determine differences in Shannon-Wiener diversity, species richness and germination densities (dependent variables) along the transect survey (location within transect as independent variable), we used mixed linear models from the GlmmTMB package with location ID as a random effect (Mollie et al., 2017). Species richness was fitted with a Poisson distribution. This approach was done separately for the water level drawdown and the non-water level drawdown area. Tukey-adjusted comparisons were done using “emmeans” (Russell, 2022). Shannon-Wiener diversity was calculated using the ‘vegan package’ (Oksanen et al., 2022). Differences in the sum of germinated individuals per location along the water level gradient were tested using a Chi-Square test.

Part 2.2 Nutrient availability

To test for differences in nutrient availability along the elevational gradient of current water level fluctuations in the transect survey, we performed Spearman correlations. The Spearman correlations were done between nutrient concentration as the dependent variable and elevation in meters NAP as the independent variable.

Part 3: Water level fluctuations

Part 3.1: Seed bank properties

Due to the low germination rate, no statistical analysis were performed on seed bank properties in relation to any of the water level treatments.

Part 3.2: Nutrient availability

To determine the effect of water level treatment (independent variable) on nutrient availability (dependent variables), we used mixed linear models from the GlmmTMB package (Mollie et al., 2017). Nutrient concentrations from the end of the experiment (day 56) were used as dependent variable. Nutrient starting concentrations were used as a covariate into the model and the blocking factor was used as a random effect. Additionally, nutrient concentrations were tested for changes over time during the eight-week experiment using mixed linear models from the GlmmTMB package (Mollie et al., 2017). Nutrient concentrations were used as the dependent variable, the blocking factor was used as a covariate in the model and date was used as the independent variable. To test for differences among the independent variables, Tukey-adjusted comparisons were done using “emmeans” for all models (Russell, 2022). All models were fitted with a Gaussian-error distribution. Normality and heterogeneity of the residuals of the models were assessed using histograms, and were transformed if necessary.

For more details we would like to refer to Figure 1 in the related manuscript.