The study was conducted in 2009 and 2010 in 16 shallow man-made ponds located at the outskirts of the Brussels-Capital Region, Belgium (Figure 1). Historically, all ponds have been excavated along stream courses, and generally a flow-through hydrological regime is maintained. Thirteen ponds are situated in the Woluwe catchment, a small river largely fed by forested headwaters. Eleven ponds are situated within parks with broad grass lawns, while the remaining five have perimeters that are predominantly forested. Pond area ranges from 0.11 to 3.57 ha (average 1.10 ha), with depths spanning from 0.47 to 1.30 m (average 0.90 m).

As a result of past anthropogenic eutrophication largely sustained through internal loading, nutrient levels in ponds of the Brussels-Capital Region typically correspond to hypereutrophic conditions (OECD, 1982). To battle algal and cyanobacterial blooms and increase ecological value, all ponds had been biomanipulated within the past decade prior to the study. Biomanipulation in the region typically encompasses water drawdown with subsequent fish removal (Peretyatko et al., 2012). The short-term success of this management procedure has proven to be almost complete (De Backer et al., 2012), so that most ponds at least during the initial phase following biomanipulation reside in a clear-water state and develop SAV. In one pond, however (PchR), active biological interference was limited to stocking with juvenile Northern pike (Esox Lucius L.), without success. Three ponds (MlGr, MlKl, and Prm2) were emptied early 2009, refilled after the growth season and only sampled in the summer of the next year. At the time of sampling, at least two upstream ponds were presumably fishless (Beml, Dens; De Backer et al., 2012), while in several others large benthivorous fish and/or schools of zooplanktivores were observed (Leyb-a, Leyb-b, PchR, PRB1, PRB2, Prm2, Tenr, VKn2, and WPk1).

We evaluated two ecological aspects of the aquatic vegetation: (1) the community composition, and (2) total pond-scale cover of SAV. Firstly, macrophytes growing along the three sediment sampling transects (including five stops/transect) were monitored twice, once in July 2009 and again in July 2010. From a boat, a floating PVC frame (1 m²) was placed on top of the macrophyte assemblage followed by species cover estimation. Low-growing vegetation, as well as the floor of turbid ponds, was sampled using a rake. Vertical stratification of vegetation was taken into account, allowing the combined cover to exceed 100%.

Secondly, in both years waterbody-scale relative cover of submerged macrophytes was visually surveyed by traversing the pond in several directions, on three occasions: end of May, beginning of July and mid August. The maximal SAV extent during the growth season was used in the statistical analyses.

During two campaigns per year (end of May and at the time of vegetation community analysis in July), we measured a number of environmental variables (Table 1): water depth, Secchi depth, physical–chemical variables (temperature, pH, conductivity, and oxygen concentration) and nutrients (NO_x, sum of nitrate and nitrite; NH₄⁺, ammonium; SRP, soluble reactive phosphorus; TP, total phosphorus). When the Secchi disk was still detectable at the bottom, a correction was added in function of disk visibility (either 0.5, 0.3, or 0.1 m). Per pond, a water volume of 10 L was collected with a vertical tube sampler for measurements of physical–chemical variables (recorded on the field with a portable WTW Multi 340i multimeter) and nutrient analysis. Nutrient subsamples were frozen upon return and concentrations were analyzed using spectrophotometry (APHA, 1995).

Additionally, we integrated results of phytoplankton and zooplankton analyses cumulated in an internal database (Table 1). Phytoplankton and zooplankton samples had been collected end of May and beginning of July using a tube sampler and a water volume of 10 L. Two measures of phytoplankton abundance were extracted: pigment concentrations (Chl a + phaeophytin) and biovolume.

For zooplankton, we used identifications and length measurements of large Cladocera (mainly Daphnia; functional group as defined by Moss et al., 2003) to calculate both density (LCD, Large Cladocera Density) and mean length (LCL, Large Cladocera Length) per pond and growth season. Both measures of zooplankton grazing efficiency have been tested in Brussels peri-urban ponds, with LCL generally displaying the strongest relation with pond turbidity level (Peretyatko et al., 2009; De Backer et al., 2014). Also, LCL can be considered to be a surrogate for zooplanktivorous fish predation, because of down-regulation of cladoceran size (Brooks and Dodson, 1965).

To further improve potential predictive power of zooplankton on SAV structure, we combined information on density and body size in a metric predicting the filtering rate of the total population of large Cladocera (LCFR, Large Cladocera Filtering Rate; Reynolds, 2006). Calculating individual filtering rate for 20 specimens of known size (incorporating temperature data; Burns, 1969) and multiplying this rate with LCD provided an indication of the grazing potential, expressed as the total volume of water filtered per day by large zooplankters present in 1 L of pond water (Reynolds, 2006).

For environmental, phytoplankton, and zooplankton variables, the warm-period average (May and July) for each pond was used in statistical analyses.

Effect of fully or partially herbivorous avifauna species on macrophytes was tested using records obtained from a regional database compiled with year-round input from various sources, including citizen science (Leefmilieu Brussel¹). These data are reliable since a large fraction is validated by experts on the basis of attached images or originates from experienced bird-watchers. Coordinates were used to assign bird observations to particular ponds. A 50 m buffer extended the zone of interest to surrounding green areas potentially containing resting, roosting or grazing waterfowl. If a clear obstacle was present between the observation and a particular pond, or if distance to a nearby pond was shorter, the count was not retained. Only waterfowl species that potentially include macrophytes in their diet were selected (Wood et al., 2012).

Since the type and magnitude of waterfowl herbivory differs between macrophyte growth stages, avifauna observations were partitioned over three time frames per year, roughly corresponding to meteorological seasons and the pattern of bird migration. This allowed separating effects of belowground herbivory in autumn and winter from primarily direct grazing in the sprouting phase and the main growth period. During vegetation senescence and throughout the major part of autumn and winter, conjointly termed the cold period (WF_Cold; September 1st – February 28th), herbivory would mainly impact the macrophyte propagule bank (Perrow et al., 1997; Hidding et al., 2010b). On the other hand, the spring period or recruitment phase (WF_Spring; March 1st – May 15th) is known to be a delicate window of opportunity for macrophyte establishment (Marklund et al., 2002). Finally, during the warm period following initial SAV expansion (WF_Warm; May 16th – August 31st, including the peak of summer), herbivory likely affects both standing crop biomass and propagule production (Hidding et al., 2010b; Gayet et al., 2011; Marco-Méndez et al., 2015). In total, bird observations spanned September 1st 2008 till August 31st 2010.

Bird counts were converted to a measure of effective herbivorous biomass, by multiplying them with species-specific average biomass and percentage plant material (including seeds) in diet, provided by Wood et al. (2012). Because of supposedly substantial amounts of multiplicates and false negatives in the observation database, we used the maximum herbivorous biomass per period and pond as a proxy for the influence of a species in that particular period, and summed the species maxima to obtain the total herbivorous biomass of the avifauna community. Finally, total herbivorous biomass was divided by pond size. The resulting biomass density (kg/ha) represents a potential peak in waterfowl-induced stress (Wood et al., 2012). Features of the aquatic vegetation in July were related to waterfowl herbivorous biomass density in the cold and spring periods immediately prior to, and the warm period during, the summer of macrophyte survey.

Summer values of abiotic and biotic variables are given in Supplementary Material. Conductivity was high (>800 μS/cm) in Beml, PRB1, Prm2, and WPk1, and low (<500 μS/cm) in ponds supplied by forest streams (VKn1, VKn2, Wtml). A few ponds experienced hypoxic conditions, due to high lemnid or floating-leaved plant cover. Most ponds were characterized by low acidity values (average pH of 8.1 ± 0.5 SD). DIN varied widely, with an average of 0.226 ± 0.294 SD mg N/L. Total phosphorus status of the ponds also fluctuated strongly, but was overall high (average of 0.370 ± 0.374 SD mg P/L). Mean water depth was 0.9 ± 0.2 SD m, while Secchi depth was 1.2 ± 0.5 SD m on average – suggesting light penetration generally permitted SAV colonization.

Transparency levels ranged from very clear (minimum phytoplankton pigment concentration and biovolume were 3.1 μg/L and 0.2 mm³/L, respectively) to extremely turbid in ponds experiencing phytoplankton blooms (maximum values were 257.0 μg/L Chl a + phaeophytin and 36.6 mm³/L biovolume).

In zooplankton communities, Daphnia was the dominant large cladoceran genus. A number of ponds contained marginal numbers of Eurycercus and Simocephalus. Individuals were very small in some ponds (minimum LCL of 0.20 mm in Leyb-b in 2010), or completely absent (LCD = 0 n/L in Leyb-a in 2010). On multiple occasions (especially in fishless ponds), LCL and/or LCD were high, resulting in potential clearance of the complete basins’ volume each day (LCFR > 1 L). LCFR exceeded 3,800 mL/L/day in Sbsk in 2010. Generally, zooplankton parameters varied widely, indicating a broad range of potential filtering capacities (see Supplementary Material). LCFR was very low in turbid ponds.

In total, 515 waterfowl entries were extracted from the dataset (around 6,300 individuals belonging to 19 herbivorous or omnivorous taxa; see Supplementary Material). These included 16 species of Anatidae (ducks, geese, swans, sheldgeese, and Egyptian goose) as well as three Rallidae species (rails). Five common waterfowl species dominated the avifauna in terms of computed total biomass, two of which are non-native to Belgium. In decreasing order of importance, these were Cygnus olor Gmelin (Mute swan; 34% of herbivory-related biomass), Branta canadensis L. (Canada goose, non-native; 21%), Alopochen aegyptiacus L. (Egyptian goose, non-native; 16%), Anas platyrhynchos L. (Mallard; 14%) and Fulica atra L. (Eurasian coot; 9%).

The presence of waterfowl was unevenly distributed, both spatially and temporarily (Figure 2). Cold period peaks (WF_Cold) matched with Beml and Sbsk (616 and 644 kg/ha), while spring values (WF_Spring) were generally low. Peaks observed during the warm season (WF_Warm) were even more prominent, mainly in Beml and Dens (703 and 723 kg/ha, respectively). Within these ponds’ perimeter, extensive flocks of larger Anatidae species aggregated in summer. With a surface area of 0.43 ha, Beml endured the company of 60 A. aegyptiacus and 17 C. olor in the warm season of 2009. In Dens (surface area 0.33 ha), 70 B. canadensis and three Chloephaga picta Gmelin (Magellan goose, non-native) were noticed simultaneously. Average biomass densities for the six consecutive periods were 84 ± 16 SD, 37 ± 52 SD, 110 ± 237 SD, 164 ± 189 SD, 32 ± 49 SD, and 65 ± 97 SD kg/ha.

A total of 11,015 propagules from eight submerged macrophyte species germinated during the study period. Most (95.07%) belonged to three species of Charophyceae: oospores of Chara vulgaris L., C. globularis Thuillier and Nitella mucronata (Braun) Miquel, respectively, accounted for 65.60, 23.09, and 6.38% of all propagules. Angiosperm submerged macrophytes were represented mainly by Zannichellia palustris L. (3.07%) and Potamogeton pusillus L. (1.71%), accompanied by P. pectinatus L. (0.11%), Callitriche obtusangula Le Gall ex Hegelm. (0.04%) and P. crispus L. (0.01%).

Large differences in presence and densities of submerged species were observed between ponds. Top and sub layers contained on average 290.5 ± 388.2 SD and 107.7 ± 153.2 SD oospores/L, respectively. Mean angiosperm propagule densities were 15.8 ± 23.4 SD n/L in the top sediment layer and 7.9 ± 12.9 SD n/L in the underlying layer. Top and sub layers were correlated in terms of density (Charophyceae: r_s = 0.86, p < 0.0001; Angiospermae: r_s = 0.90, p < 0.0001) and on average showed relatively high Chao-Sørensen similarity (0.66 ± 0.45 SD). Overall, Shannon diversity within the propagule pool of both sediment layers was low, as a result of charophycean dominance and low species richness (H’ top: 0.31 ± 0.35 SD; H’ sub: 0.35 ± 0.36 SD).

A number of ponds were characterized by relatively diverse and/or dense submerged macrophyte propagule banks (Dens, Leyb-a, Leyb-b, MlGr, VKn1 and Wtml, for instance); others displayed poor recruitment potential. In one turbid pond, no propagules sprouted from the collected sediment (PchR). Five ponds lacked angiosperm propagules (PchR, PRB1, PRB2, Prm2, and VKn2), while germinating oospores were entirely absent from only one site (PchR).

The submerged macrophyte community in July was composed of nine species in total (Figures 3–5). A number of ponds completely lacked SAV during one (Beml and Dens in 2009) or both monitoring periods (PchR and PRB2). Others were sparsely colonized with submerged species (≤5% cover; Beml, Dens, Prm2 and VKn1, all in 2010). At macrophyte-rich sites, P. pectinatus, P. pusillus and C. globularis were particularly common, often (co-)dominating the aquatic vegetation. Other submerged species sporadically reaching high cover levels were Ceratophyllum demersum L., C. vulgaris, Elodea nuttallii (Planch.) H. St. John and Z. palustris. VKn2 was completely filled with C. demersum throughout the study period. Submerged species richness was relatively high in MlGr (five species in 2010), MlKl (five species in 2010), Tenr (six species in both years), and Wtml (four species in 2009 and seven in 2010).

Floating-leaved macrophytes [Nuphar lutea (L.) Sm. and Nymphaea alba L.] were locally abundant in Beml, WPk1, and Wtml, along the fringes or in smaller patches. During summer, two ponds (Sbsk and VKn1) became gradually covered by lemnids [including Lemna trisulca L., Lemna minor L., and Spirodela polyrhiza (L.) Schleid.], while WPk1 developed extensive detached algal mats [composed of filamentous algae and Enteromorpha intestinalis (L.) Nees]. Finally, Riccia fluitans L., Sagittaria sagittifolia L., and Veronica anagallis-aquatica L. were recorded within vegetation transects of a single pond each.

Macrophyte species exhibited variable relationships with propagule densities and herbivory by avifauna (average of spring, warm, and cold periods), although in this context the influence of both predictor variables on summer vegetation abundance seemed rather weak (Figures 3, 4). Maximum species-specific macrophyte cover values concurred with intermediate propagule densities for C. globularis, C. vulgaris, P. pectinatus, P. pusillus, and Z. palustris. Nevertheless, for three common angiosperm species (P. pectinatus, P. pusillus, and Z. palustris), maximum cover values were attained at sites with relatively rich propagule banks compared to the pattern observed for charophycean algae. Additionally, C. demersum, E. nuttallii, P. pectinatus, and P. pusillus were also able to develop medium or high summer cover along some vegetation transects despite absence of propagules in sediment. C. obtusangula was only found within the propagule bank.

For common macrophytes, maximal waterfowl densities – as retrieved for ponds Beml and Dens,– were systematically linked with (near) vegetation absence in summer (Figures 3, 4). In other ponds, the relationship between potential waterfowl herbivory and macrophyte occurrence revealed a fairly chaotic pattern. Some ponds developed high macrophyte cover unhindered by abundant waterfowl residing in or traveling through the area, while in others, species were unable to establish even at low bird biomass densities.

In terms of species incidence, the predictive power of propagule bank and of vegetation in the first year differed among macrophytes (Figure 5). C. obtusangula, C. vulgaris, and Z. palustris showed comprehensive unrealized potential, since few viable seeds and oospores managed a transition toward standing populations (Figure 5, left). Other species (C. globularis, P. pectinatus, and P. pusillus) expressed their recruiting abilities in the majority of transects. On the other hand, for all species presence within the macrophyte community was not exclusively explained by propagule availability in spring.

Within the vegetation, continuity over consecutive years was high for C. demersum, E. nuttallii and two common Potamogeton species (especially P. pusillus), but less for charophycean macroalgae and Z. palustris (Figure 5, right).

In 2009, two ponds were turbid (Turbid), three were clear but lacked abundant SAV (Clear L), while eight resided in a macrophyte-dominated state (Clear H; see Supplementary Material). In addition, in the first year three newly biomanipulated ponds were completely drawn down for the duration of the growth season. The ensuing year, the three re-inundated ponds showed high water transparency, yet only two were extensively colonized by submerged macrophytes (MlGr and MlKl). In 2010, three ponds were turbid, four had an intermediate ecological state (Clear L) and nine were abundantly vegetated (Clear H). In between years, three ponds switched categories: Leyb-b (from Clear H to Turbid), VKn1 (from Clear H to Clear L) and WPk1 (from Clear L to Clear H).

Compared to macrophyte-dominated waterbodies, turbid ponds featured significantly lower propagule densities in the ecologically accessible top sediment layer (Angiospermae: p = 0.015; Charophyceae: p = 0.022; Figure 6, left and middle; Table 2). In clear, vegetation-poor ponds the same layer was characterized by intermediate angiosperm propagule densities, but significantly richer oospore reservoirs in comparison to turbid ponds (p = 0.031). In the deeper layer, the situation was similar for Charophyceae (Clear H vs. Turbid: p = 0.018; Clear L vs. Turbid: p = 0.007), while for Angiospermae there was a significant difference between Clear H and Clear L ponds (p < 0.001), with Turbid ponds positioned in-between. Chao-Sørensen similarity between top or sub layer contents and vegetation was highly variable in Clear H and Turbid ponds, although the former group had a tendency for low overlap between compartments (Figure 6, right and Table 2). Similarity was extremely low in Clear L ponds and significantly weaker compared to Turbid ponds.

The first two axes of the RDA explained 37.8% of variance in the species data and 58.1% of variance in the species-environment relationship (see Supplementary Material). In the PCA (Figure 7), Clear H ponds were directed toward the clear end of the spectrum (corresponding to higher SD/D values) and clustered in two subgroups, the first being species-rich and dominated by P. pectinatus and P. pusillus (associated with higher values of T_Angio and pH), the second related to high C. globularis, C. demersum and lemnid cover. With a few exceptions, Clear L sites grouped together with Turbid ponds, sharing high values of bare ground. Typical vegetation attributes (i.e., low abundance of macrophytes) of these two ecological categories were linked with high phytoplankton biovolume (Biovol), conductivity (Cond), and waterfowl herbivory in the warm period (WF_Warm).

Forward selection retained five significant contributors to macrophyte community composition (conditional effects; Table 3), with decreasing effect size: phytoplankton biovolume (Biovol; λ = 0.10; p = 0.004), angiosperm top-layer propagule density (T_Angio; λ = 0.09; p = 0.019), waterfowl herbivorous biomass density in the warm period (WF_Warm; λ = 0.08; p = 0.024), pH (λ = 0.07; p = 0.031) and conductivity (Cond; λ = 0.06; p = 0.026). Nutrients and cladoceran filtering rate (LCFR) behaved neutral with respect to macrophyte assemblages.

We obtained a shortlist of four candidate variables displaying significant singular relationships with SAV cover: phytoplankton biovolume (Biovol), pH, angiosperm propagule density in top layer (T_Angio) and waterfowl herbivory in the warm period (WF_Warm). The maximal summer extent of submerged macrophytes in ponds showed a positive trend with pH (CI range: unidirectionally positive) and a negative relationship with phytoplankton biovolume and waterfowl herbivory in the warm period (CI range: unidirectionally negative; Table 4 and Figure 8). The evidence for an effect of density of angiosperm propagules in the upper sediment layer is less convincing (i.e., CI range bidirectional relative to zero value; Table 4).

Apart from conductivity and pH, none of the abiotic variables included in the analyses were associated with characteristics of macrophyte composition and abundance. Acidity values were lower in ponds dominated by SAV, especially coinciding with high abundance of tall Potamogeton pectinatus (in agreement with Capers et al., 2010) and P. pusillus. Nevertheless, pH reached high levels in turbid ponds as well and correlated significantly with oxygen concentrations, indicating decreasing acidity should be interpreted as a by-product of carbon consumption in the process of primary production.

Conductivity was generally relatively low in macrophyte-dominated ponds and high in clear, vegetationless ponds, though not consistently. Possibly, conductivity values were a reflection of a combined effect of photosynthetic activity (Pedersen et al., 2013) and the source of water supply to the waterbody. Although higher values potentially indicate reception of contaminated runoff (Hatt et al., 2004), conductivity remained equivalent to freshwater conditions, and most likely did not profoundly influence macrophyte species filtering or growth. The only sign of a potential structuring effect of accumulated (de-icing or other) salts within the study area was the co-occurrence of high conductivity in WPk1 and proliferation of Enteromorpha intestinalis, a euryhaline, yet predominantly brackish-water macroalgal species (Martins et al., 1999; Gallego et al., 2015).

Nutrients presumably were not relevant since phosphorus is the primary limiting element in many freshwater ecosystems (Correll, 1998), yet readily available in the strongly eutrophic ponds within the study area (Peretyatko et al., 2012; Teissier et al., 2012). The elapsed time since TSB (more specifically, water drawdown combined with fish removal) did not affect macrophyte dynamics. Søndergaard et al. (2008) pointed out that successful biomanipulation outcomes in shallow lakes tend to last between 2 and 6 years, followed by gradual deterioration after recovery of fish. The ponds included in our study were mostly biomanipulated within 4 years prior to the analyses, and therefore did not show much relationship with the timing of biomanipulation. With only a partial fish reduction that was executed eight years before the onset of the study, the macrophyte-dominated Tenr pond seemed to be exceptionally resilient to degradation, most likely because of a balanced fish community including a strong piscivorous guild (Leefmilieu Brussel, pers. comm.).

As expected, high phytoplankton biovolume was the principal driver explaining submerged macrophyte community composition and pond-scale cover. This is in agreement with relative importance of turbidity in shallow lakes in the Netherlands (Van den Berg et al., 2003; van de Haterd and Ter Heerdt, 2007), and emphasizes the mutually exclusive nature of phytoplankton and submerged macrophyte dominance (Scheffer et al., 1993). In the absence of top-down control of phytoplankton, submerged macrophytes are readily out-competed and will disappear (Scheffer et al., 1993). The results stress the need for continuing management efforts to reduce external and internal loading affecting ponds in the region (Søndergaard et al., 2003; Jeppesen et al., 2005; Hilt et al., 2006).

Although zooplankton plays a paramount role in the control of phytoplankton (Jeppesen et al., 1997, 2005), the estimated filtering rate of large Cladocera (LCFR) did not directly relate to any SAV characteristic. Likely this results from the fact that large Daphnia in a number of fishless ponds (Beml and Dens) maintained high water clarity, while SAV did not successfully develop. Similarly, the filtering capacity of zooplankton in vegetated ponds can easily be misjudged if large Cladocera exhibit diel vertical (e.g., to and from Chara meadows) or horizontal migration (Burks et al., 2002), or if the contribution of macrophyte-associated cladocerans to total phytoplankton clearance potential is significant (Balayla and Moss, 2004).

Propagule banks in the peri-urban ponds were rarely species-rich and displayed low taxon evenness, because of a dominance of charophycean oospores. Oospores are ubiquitous in many freshwater sediments (de Winton et al., 2000; Bonis and Grillas, 2002; Kalin and Smith, 2007), and high densities might be required for establishment and colonization (Van den Berg et al., 2001). Angiosperm propagules were far less abundant, although any direct comparison with oospore densities would be biased given the small size of oospores. Correcting for propagule dimensions could enable to offset variations in storage capacity, providing a method to evaluate the relative contribution of each species to the propagule bank potential.

Since propagule densities in top- and sub-layers of the sediment were correlated, we focused on the upper 5 cm, supposed to be in close contact with ecological processes in the water column (Dugdale et al., 2001; Bonis and Grillas, 2002; Spencer and Ksander, 2002). However, the extent to which bioturbation by burrowing unionid bivalves, foraging benthivorous fish or waterfowl could mix layers and homogenize strata should not be underestimated (McCall et al., 1995; Vaughn and Hakenkamp, 2001; Huser et al., 2016).

Several elements indicated that the composition and density of the propagule bank plays a significant, albeit modest role in the structuring of SAV in nutrient-enriched ponds. Overall, propagules mainly influenced the assemblage of the macrophyte community (with wealthy stocks of angiosperm propagules accompanying development of small-leaved Potamogeton communities), rather than secured SAV dominance.

At the level of individual submerged macrophyte species, the relationship between recruitment reserves and cover within the aquatic vegetation could be summarized as ambiguous. The status of the propagule bank matched maximal vegetation abundance of the most common Angiospermae (P. pectinatus, P. pusillus, and Zannichellia palustris) slightly better than was the case for Charophyceae (indicating some positive influence of high propagule density on vegetation cover), but for each species there were clear dissimilarities depending on the pond. Propagules of Chara vulgaris and Z. palustris poorly contributed to standing vegetation. Recruitment failure of certain macrophyte species suggests that environmental conditions in the waterbody are not beneficial for germination or growth (the impact of which is mitigated by dormancy of the propagules; Bewley, 1997; Sederias and Colman, 2007; Hay et al., 2008; Nonogaki, 2014).

Local occurrence of submerged macrophytes was not exclusively predicted by presence of propagules, likely because most are capable of vegetative, either non-specialized (e.g., Ceratophyllum demersum and Elodea nuttallii) or rhizomatous (e.g., P. pectinatus and Z. palustris), spread and survival (Van Vierssen, 1982; Barrat-Segretain, 1996; Boedeltje et al., 2004; Vári, 2013; Vári and Tóth, 2017). For Charophyceae, establishment without prior availability of oospores would have been less plausible, suggesting that prevailing germination conditions did not always break dormancy (Sederias and Colman, 2007).

Importantly, turbid ponds contained less angiosperm and charophycean propagules in the top sediment layer compared to macrophyte-dominated, clear ponds. Turbid ponds in the Brussels-Capital Region possess high densities of zooplanktivorous and benthivorous fish, resulting in persistent phytoplankton dominance over multiple years and a need for biomanipulation to force a shift. Following a crash of macrophyte domination and winter resetting of autotrophic production, moderate recruitment in spring might start depleting the propagule bank. In case dispersal input does not compensate for the loss, opportunities for submerged macrophyte establishment finally will be lost, leaving the ruined propagule bank as a critical bottleneck for ecosystem recovery. On the contrary, in the shallow Lake Terra Nova (Netherlands), van de Haterd and Ter Heerdt (2007) did not identify propagule limitation after a long period of turbidity. Here, (within-lake) dispersal or less intense phytoplankton blooming could have allowed sparse macrophyte reproduction under turbid conditions. Alternatively, Lake Terra Nova might experience less intense organic matter accumulation on top of propagule-rich sediment when compared to the sapropelium-rich turbid ponds in our study.

Interestingly, propagule densities in clear but sparsely vegetated ponds typically seemed sufficient, though not fully translated as a result of waterfowl herbivory (Beml and Dens in 2009), lemnid cover (VKn1 in 2010), or some other factor affecting SAV. This results in significantly lower similarity between sediment content and vegetation composition when compared to turbid ponds, where submerged macrophytes remain underrepresented in both compartments.

In terms of overall ecological status, the emerging trend seems to be that: (1) for clear, macrophyte-dominated ponds the propagule potential is realized; (2) clear ponds with low SAV cover are unable to utilize their potential; and (3) in turbid ponds recruitment opportunities are inadequate. This is different from patterns observed in interconnected fishing ponds in France, where Arthaud et al. (2012) found no relationship between density of propagules and Chl a concentrations in the water, supposedly because propagule banks in the French pond system are regularly replenished – either through dispersal or during recurrent clear-water phases. If propagule banks are known to be impoverished, reproductive success and natural dispersal functioning could be assessed before proceeding to active assistance of macrophyte establishment (Smart et al., 1998; Hilt et al., 2006; Van Onsem et al., 2018).

During the warm period of the growth season, peaks of plant-consuming waterfowl coincided with negligible SAV abundance, even in the presence of high propagule densities and in the absence of fish or turbidity stress. This confirms the need to incorporate macrophyte herbivory in aquatic research (Bakker et al., 2016; van Altena et al., 2016; Wood et al., 2017). The ranking of waterbird herbivory as second most important biotic predictor following phytoplankton-induced turbidity is consistent with van de Haterd and Ter Heerdt (2007).

It is difficult to situate the avifauna biomass densities obtained for Brussels ponds with respect to other regions, because of the use of peak densities instead of the periodic averages provided by Wood et al. (2012). Nonetheless, maximum densities found in our study seem considerable at times, potentially provoking loss of macrophytes on short notice. The swarming of swans, geese, shelducks, and ducks becomes especially problematic when population sizes overstretch the carrying capacity of the waterbody. In urbanized areas, the pond perimeter can offer a copious supply of staple or alternative food in the form of park lawns and thrown-away bread, thereby attracting and potentially bonding birds to a particular location. Near small, shallow ponds, such an aggregation of opportunistic herbivores lowers the likelihood of large-scale and sustainable SAV establishment.

Nevertheless, the impact of waterfowl seemed to be restricted to direct grazing or disturbance in summer. Although autumn foraging on seeds and tubers or spring grazing on freshly emerged shoots could have had a time-lagged effect (Marklund et al., 2002; Hidding et al., 2010b), high waterfowl biomass density during the spring or the cold season did not weigh on macrophyte performance at the height of the growth season. With an average depth of 0.90 m, the studied ponds are rarely sufficiently deep to prevent herbivory on belowground or freshly emerging biomass. Presumably, waterfowl occurring in the area of ponds was less inclined to forage directly on the water during colder periods. Limited disturbance by waterfowl during the spring establishment phase is in accordance with coot effects found in British broads (Perrow et al., 1997) and a Turkish lake (Sandsten et al., 2005).

Overall, the results illustrate that ecological pond management in densely populated regions should comprise some form of waterfowl control, especially during summer. This can be achieved by prohibiting bird feeding, in combination with regulation of resident, invasive alien waterfowl (specifically A. aegyptiacus and B. canadensis).