Phosphorus Availability and Growth of Benthic Primary Producers in Littoral Lake Sediments: Are Di ﬀ erences Linked to Induced Bank Filtration?

: Submerged macrophytes and periphyton are benthic primary producers that play an important role for several ecosystem functions of lakes. Their growth often depends on the availability of phosphorus (P) in sediments and overlying water. This P availability is assumed to potentially be a ﬀ ected by induced bank ﬁltration (IBF), a cost-e ﬀ ective method for drinking water production. In this study, we tested whether littoral sediments sampled at sites with high and low inﬂuence of IBF in a temperate eutrophic lake used for bank ﬁltration since more than 100 years a ﬀ ects periphyton and macrophyte growth. Sediments di ﬀ ered in aerobic desorbed water-soluble phosphorus (P H2O ) and iron (Fe) content and the growth of macrophytes in sediments with a high impact of IBF was lower compared to sediments with low impact of IBF. We also found that P addition to the nutrient solution increased periphyton growth and that periphyton limited macrophyte growth. While these results point to a potential impact of IBF on P availability in sediments that can cascade to benthic primary producers, we could not prove mechanistic links between high rates of IBF and the lower macrophyte growth. Additional research to assure a sustainable application of this valuable drinking water production method is therefore needed. C.H. and S.H.; project M.G. and S.H.; C.H. and S.H.; S.H.; C.H. and S.H.; visualization, draft, M.G. and S.H.; writing—review and S.H.


Introduction
Benthic primary producers play an important role for biodiversity, productivity and several other ecosystem functions of lakes, but have been little studied so far compared to their planktonic counterparts [1,2]. Primary production of macrophytes and periphyton in lakes is often similar to or greater than that of phytoplankton and thus constitutes an important energy input to lake food webs [1], especially in shallow and/or small lakes [3,4]. Even in large lakes, the vast majority of species inhabit the shallow, nearshore littoral zone and most fish species exploit benthic resources [5].
Nutrient and light availability are the major bottom up controlling factors of primary production in lakes. Within stream and lake pelagic systems, nitrogen (N) and phosphorus (P) limitation are equivalent, however, subhabitat differences indicated that P limitation predominates in lake benthos [6]. P availability for benthic primary producers depends on P concentrations in both water and sediment, with sediments usually containing more P per volume than water. When P is limiting, both macrophytes and periphyton increase their biomass with increasing P concentrations. Above a certain threshold, however, periphyton attached to macrophytes attenuates too much light and macrophyte growth becomes limited by light [7,8]. This process has been shown to be a major mechanism of macrophyte decline during eutrophication of lakes [9].
The complex balance between P availability in sediments and overlying water can be affected by several factors such as redox/pH conditions, supply with organic matter and transport conditions [10], but also by groundwater influx [11]. Recently, induced bank filtration (IBF) has been suggested to potentially affect P availability and thus growth of benthic primary producers in lake littoral areas [12]. This cost-effective pre-treatment method has been applied since the 19th century and is expected to be increasingly used for drinking water production in the future [13,14]. A recent study showed that IBF in many cases promotes turbid states in shallow lakes [15].
On the one hand, IBF interrupts groundwater inflow to lakes in the area surrounding the well galleries [12]. Consequently, groundwater-induced P fluxes from sediments into water, which can enhance periphyton growth [11,16,17], would be interrupted. On the other hand, IBF might result in an increased flux of particulate P from water into sediments by clogging with fine sand, silt or clay particles and by particulate organic matter such as planktonic algae and detritus. This process has been found to promote microbial activity in sediments affected by IBF [18,19] and might thus increase P availability in sediments as compared to locations unaffected by IBF. But so far, changes in P availability and consequences for benthic primary producers and their interaction in lake littoral zones affected by IBF have not been studied.
In this study, we compared the characteristics of littoral sediments sampled at sites with high and low influence of IBF in a temperate eutrophic lake used for bank filtration since more than 100 years. Subsequently, the effects of the different sediments on periphyton and macrophyte growth with and without P addition were tested in laboratory studies. We hypothesized that 1) littoral lake sediments affected by bank filtration show a higher content of organic matter and consequently 2) have a higher P availability, 3) periphyton growth is stimulated by higher P availability and 4) hampers macrophyte growth when its shading impact is high. If true, bank filtration could affect the abundance and structure of different benthic primary producer groups in lake littoral areas.

Studied Lake System and Impact of Bank Filtration
Lake Müggelsee ( Figure 1) is a flow-through lake located in south-eastern Berlin, Germany. Its surface area is 7.3 km 2 and average depth is 4.9 m with a maximum depth of 7.9 m [20].
In 1905, the first groundwater wells were installed north of Lake Müggelsee, and later additional wells were installed west and south of the lake [20] (Figure 1). As groundwater is being pumped and the groundwater level drops below the lake level, lake water infiltrates through the sandy sediments of the shallow littoral lake area and continues through the subsurface until reaching the well. North of the lake, where most of the water is being pumped (Figure 2A), the groundwater level drawdown is larger than five meters ( Figure 1). Groundwater abstraction rates of galleries C and D at the north-eastern shore were about twice as high as compared to galleries E and F at the south-western shore between 2008-2017 ( Figure 2A).
After very high nutrient loadings into Lake Müggelsee during the 20th century peaking in the 1970s and 1980s, nutrient loads started dropping steeply in the beginning of the 1990s and continued reducing after year 2000 [21]. In parallel with the high nutrient loading, almost all macrophytes disappeared from Lake Müggelsee in the 1970s. After the nutrient reduction started, however, no significant recovery of macrophyte abundance and species diversity was seen for about 20 years despite increasing water transparency, especially during spring [22]. Light attenuation by high periphyton biomass was significantly contributing to this delay [23]. The macrophyte population was dominated by Stuckenia pectinata (formerly known as Potamogeton pectinatus), a species known for survival under turbid conditions in shallow littoral areas of highly eutrophic lakes [22]. Only after around 2006, macrophyte maximum colonization depth and biomass started increasing [22] due to decreasing periphyton biomass (unpublished data). Periphyton biomass, however, is still high (unpublished data). Since 2011/2012, the lake experienced a strong invasion of quagga mussels (Dreissena rostriformis bugensis) which significantly increased water transparency all year round facilitating the expansion of macrophytes other than S. pectinata particularly in deeper littoral areas between 2 and 4 m [24]. The average concentrations for total phosphorus was 66 µg/L, total organic carbon was 6.8 mg/L, dry matter was 3.9 mg/L, dissolved iron was 0.01 mg/L and total iron was 0.08 mg/L for the period March 2016 to March 2017 in Lake Müggelsee (data from Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany).

Sampling of Sediments and Plant Tubers
We chose six sediment sampling sites along the north-eastern (N) shore, influenced by water abstraction through IBF (Figure 1), and six sites along the south-western (S) shore, where water abstraction is much lower than along the N shore ( Figure 2A and Table S1). For a pre-analysis, carried out to generally characterize the sediment and to determine best practice for P analyses (see below), we took one sediment sample each from sites 5 and 10 at 1.5 m depth using a tube sampler attached to a push rod in February 2017. For the main analysis and growth experiment we took two samples at every site (1 to 12) at 1.5 m depth in March 2017.
For the growth experiment, we collected tubers from S. pectinata in 50-70 cm water depth at the N shore in March 2017.

Description
In a pre-analysis, we analyzed two sediment layers (0-2 cm and 2-6 cm), while in the main analysis we pooled the top 5 cm of the two sediment cores taken at each site and used this material in the continued analysis and growth experiment. Sediment samples were very sandy and some (mostly the samples from the S shore) had a 2-4 mm layer of fluffy brown organic material on top (Table S2). This upper layer and any mussels were removed in the main analyses and the growth experiment. We determined sediment dry weight by drying 10 g of fresh sediment sample at 105 • C for 24 h and weighing them after cooling down in a desiccator. Subsequently, loss on ignition (450 • C for 5 h) and the grainsize distribution were determined by using the following sieve sizes: 0.063; 0.125; 0.18; 0.25; 0.63; 1.25; 2.0; 2.8 mm. Statistical comparison between sediments of the N and S shores was done using the values for D10 and D50, which are the diameters where 10% and 50% of the material's mass has a diameter smaller than the respective diameter.

Phosphorus Availability
In order to compare the content of different P fractions [25], potentially relevant for macrophyte growth, and for diffusion into the overlying water affecting periphyton growth [8], we analyzed water-soluble P (P H20, aerobic, two replicates), reductive-soluble P (P reductive , anoxic, two replicates), acid-soluble P (P HCl , two replicates) and total P (TP, five replicates). 10 g from each sediment sample were put into a centrifuge tube, mixed with 20 mL extraction solution (H 2 O, reductive solution (10 g Na 2 S 2 O 4 and 4.6 g NaHCO 3 mixed with 500 mL distilled water) or acid solution (0.5 N HCl)). Total P was determined using 0.025 g grinded sample with 2mL H 2 SO 4 and 2mL H 2 O 2 filled to 50 mL with distilled water. The samples of sequential P extraction were shaken for 2 h with an overhead shaker. After centrifuging for 5 min (10142 g), the supernatant was decanted and subsequently filtered to 0.45 µm (Minisart ® NML Syringe Filter, Sartorius, Göttingen, Germany). The desorption solution was diluted when needed, digested with K 2 S 2 O 8 in the digestion tube (autoclave) and the P concentration was determined photometrically. This procedure was repeated two more times for P H2O and P reductive , three more times for P HCl and no more for TP: 20 mL extraction solution, 2 h shaking with overhead shaker, 5 min centrifuge, filtered, digestion and photometric measurement.
Wünscher [26] used water soluble P as a measure for plant available P. We took a similar approach based on the results from the pre-analysis ( Figure S1) and chose P H2O to measure P availability for macrophytes and periphyton. P HCl and P reductive give results that are irrelevant for macrophyte and especially periphyton available P, but TP was analyzed to see if a difference in P H2O could be explained by a difference in TP. During shaking, the friction of the sand could increase the P binding at particle surfaces, so that even after a very high number of repeated measuring processes the P released from the sediment would not decrease, we therefore chose to perform the process when analyzing P H2O .
10 g from each sediment sample were put into a centrifuge tube, mixed with 20 mL H 2 O (aerobic desorption) and shaken for 2 h with an overhead shaker. After centrifuging for 5 min, the supernatant was decanted and subsequently filtered to 0.45 µm (Minisart ® NML Syringe Filter, Sartorius, Göttingen, Germany). The desorption solution was diluted when needed, digested with K 2 S 2 O 8 in the digestion tube (autoclave) and the P H20 concentration was determined photometrically.

Growth Experiment
Three different types of sediment were put in small (≈ 2 dl) glass flasks: Pure sand as a control (C), sediment collected from the N shore of Lake Müggelsee and sediment collected from the S shore. Two tubers from S. pectinata (sampled from the N shore of Lake Müggelsee) were planted in each of the sediments, along with an artificial strip (transparent polypropylene; General Binding Corporation, Chicago, Illinois) serving as a substrate for periphyton growth sticking out ≈15 cm above the sediment surface ( Figure 3). Before the tubers were planted their length and width were measured ( Figure S2). With the assumption that the tubers have the shape of a prolate spheroid the volume was calculated using 4/3π*length*width 2 . There was no significant difference in tuber volume between the treatments (Kruskal-Wallis rank sum test). The small glass flasks with sediment and tubers were put into 2 L bottles (Ikea Korken) filled with a nutrient solution after Körner and Nicklisch [27] that emulates the ionic composition of Lake Müggelsee, but without P. We ran five treatments (six replicates each): Controls, N and S sediment, and N and S sediment but with P addition to the nutrient solution in the form of monopotassium phosphate (KH 2 PO 4 ) to reach the concentration 0.1 mM (~320 µg P/L) (Figure 3). The combination of pure sand and P addition was omitted as it did not contain any periphytic algae inoculum.
The growth experiment was performed in a climate chamber at 20 • C (+/-0.5 • C) under a 12 h light/dark regime for 21 days. This time-length was chosen to give sufficient time for the periphyton to develop but was held short enough to prevent nutrient depletion for the macrophytes that can initially use the nutrients stored in the tubers. Before starting the experiment, the light intensity was measured (TriOS GmbH, RAMSES-ASC-VIS) and 30 spots with similar light conditions (92-112 mW/m 2 /nm, median = 102 mW/m 2 /nm) were chosen for the placement of all bottles. The bottles were placed randomly, and their positions were changed randomly twice a week.
At the end of the experiment all plants were harvested, separated into roots and shoots, measured, dried in 60 • C and weighed to determine final biomass. Before comparison and analysis the macrophyte biomass was divided with the tuber volume to account for the fact that tuber size determines macrophyte growth [28]. We used the entire macrophyte biomass for comparing treatment effects, however, we had to estimate the root weight of the control based on the correlation between shoot and root weight in the other treatments, because it was impossible to entirely remove the pure sand from the roots. Periphyton was scraped from the artificial substrates using a brush and by flushing nutrient solution. Aliquots of the suspension were filtrated onto pre-weighed filters (Whatman glass microfiber filters GF/F, diameter 25 mm). Filters were dried at 60 • C to weight constancy and final periphyton biomass determined from the difference between initial and final filter weight.

Statistical Analysis
Welch's t-test (when comparing data with unequal variability), Wilcoxon rank-sum test (when analyzing non-normal data) and Kruskal-Wallis test (multiple comparisons) were used for testing for significant differences between treatments. Correlations (Pearson product-moment correlation and Spearman's rank correlation) between macrophyte biomass and periphyton biomass were calculated using all treatments and macrophyte biomass above sediment.
We used multiple linear regression (MLR) with periphyton biomass as the dependent variable and P addition, P H2O concentration in sediment, TP concentration in sediment, LOI, grainsize distribution (D10, D50) and lake side as independent variables. By using the Akaike information criterion and stepwise regression we found the most parsimonious model. We did the same with macrophyte biomass as the dependent variable and periphyton and the above-mentioned variables as independent variables. We also built a model where periphyton was the only independent variable explaining macrophyte growth. When necessary, data was log-transformed to meet the assumptions of normality.
We used the software R, version 3.5.0 for all statistical analysis and specifically the package ggplot2 [29] for plotting.

Sediment Characteristics
In all the sediment samples, more than 99% of the non-organic material was sand. D10 values were not significantly different between samples from the N and the S shore ( Figure 2B). The same holds for average values of D50, which were 0.389 ± 0.021 and 0.415 ± 0.019 mm, respectively. No significant difference in LOI and TP content was found between the samples of the N and S shore ( Figure 2C, D).
The P H2O content was significantly higher in the samples from the N shore than in the samples from the S shore (p = 0.03, Wilcoxon rank-sum test, Figure 2E), while the opposite was found for the Fe content (p = 0.015, Wilcoxon rank-sum test, Figure 2F).
No significant difference in the Al, Ca, Mg, Mn and S content was detected between the samples of the N and S shore ( Figure S3). The Pb content was below detection limit (<0.01 mg/g) in all samples.

Growth Experiment
Periphyton biomass was significantly higher in both treatments with lake sediments as compared to the control (white boxes in Figure 4A, p < 0.01, 95 % confidence intervals: [0.16, 1.86] (N vs. C) and p < 0.01, 95 % confidence interval: [0.28, 1.04] (S vs. C)). Periphyton biomass was significantly higher in treatments with P additions to the water (grey boxes) as compared to treatments without P additions (Figures 3 and 4A, p < 0.001, 95 % confidence interval: [1.35, 2.88], Wilcoxon rank-sum test), while periphyton biomass was not higher in treatments with sediment of the N shore containing more P H2O (white boxes in Figure 4A). P H2O contents and periphyton biomass were not significantly correlated. The linear regression model that best explained periphyton biomass contained both P addition to solution and content of P H2O in sediment as independent variables, both positively contributing to periphyton biomass (adjusted R 2 = 0.63, p < 0.001, Table S3).
Macrophyte biomass was not significantly different from controls in both treatments with sediments from S and N shore ( Figure 4B). Macrophyte biomass was significantly lower when the tubers were grown in sediment samples from the N shore (p < 0.05, 95 % confidence interval: [−0.32, −0.03], Wilcoxon rank-sum test) as compared to when the tubers were grown in sediment samples from the S shore ( Figure 4B). P additions resulted in significantly lower macrophyte biomass (p < 0.05, confidence interval: [−0.208, −0.001], Wilcoxon rank-sum test). However, this effect is not significant when analyzing the treatments N and S individually (white and grey boxes, Figure 4B).

Discussion
Our results show that the investigated sediments from different locations in the shallow sandy littoral of Lake Müggelsee did not differ in their content of organic matter, even though sediments of the N shore were more strongly affected by bank filtration. This process was expected to lead to an accumulation of organic material in sediment interstices. However, significantly higher values of water-soluble P H2O in littoral sediments of the N shore were found as compared to sediments of S shore, explainable by either their lower Fe contents and/or an accumulation P H2O due to a high biological turnover of organic material under the influence of IBF. Growth experiments revealed that additional P stimulated the growth of periphyton and that periphyton hampered macrophyte growth. Differences in P availability and growth of benthic primary producers in littoral lake sediments were thus found between samples taken from shores with low and high impacts of pumping wells. A causal link to bank filtration, however, could not be drawn and further mechanistic studies are required to explain the differences and to clarify the impact of IBF on benthic primary producers.

Differences in Sediment Characteristics
Contrary to our expectations, sediments of the N and S shore did not differ in their particle size distribution or organic matter content. The intensive pumping of the wells at the N shore of Lake Müggelsee and the fact that the uppermost sediment layer is very efficient in retaining particles with sizes down to 0.45 µm during bank filtration [30] did thus not significantly influence the organic matter content, probably due to wave action clearing the pores. Organic matter content in littoral sediments affected by IBF of nearby Lake Tegel were found to be 1.5 ± 0.5 % of sediment dry weight in the upper 5 cm [31], which is about twice the content in this study.
An accumulation of fine organic material in coarse sandy sediments was probably prevented by a combination of sediment resuspension by wind (main wind direction in Berlin is from the southwest for most of the year [20]) and a high turnover of organic carbon by microbes and meiofauna of the interstices similar to what Hoffmann and Gunkel [19,30] studied in Lake Tegel. Results of studies on accumulation of organic matter during infiltrating conditions can often differ (compare for example [19,30] with [32]) and deserve further research. In the sediment and the subsurface below an infiltration pond, organic carbon content was found to vary by two orders of magnitude [33]. Also, during operation a clogging layer with high organic carbon content was formed and after around four months it fully hindered infiltration through the pond bottom. This led to unsaturated conditions below the pond and increased mineralization of sedimentary particulate organic carbon [33]. Such conditions, however, seem unlikely at wind exposed bank filtration sites in Lake Müggelsee.
Our results showed a significantly higher content in P H2O in sediment samples of the N shore, which was not explained by differences in TP content. One potential explanation is the higher Fe content in sediments of the S shore. P bound to Fe will not desorb when using an aerobic extractant such as water [34] unless it is loosely adsorbed [25]. P could also be bound by Mn, Al, Ca or S, but no difference in measured content was found between the N and S shores. Alternatively, or in addition, microbial mineralization of organic material accumulating in sediments during bank filtration could be responsible for the difference. This process has been described as being fast [18,19] and might result in a loading of the upper sediment layers with P. A similar accumulation of P in littoral areas has been described for invasive dreissenid mussels, which filter pelagic water and excrete P-rich faeces in benthic areas [35]. Dreissenid mussels, however, can filtrate the entire lake volume during one day [24,36], while bank filtration in Lake Müggelsee takes about a year for pumping the lake volume. More detailed biogeochemical analyses of the upper layers of sediments affected by bank filtration accompanied by experimental manipulations of the bank filtration intensity are needed to provide evidence for a causal relationship between bank filtration, P availability and binding forms in littoral lake sediments.

Differences in Growth of Benthic Primary Producers
Increased growth rates of periphyton after P addition to the water indicated a general growth limitation of periphyton by P in treatments without P addition. Final periphyton biomass was lower than values found in field experiments in River Spree flowing through Lake Müggelsee [37], but the periphyton biomass in treatments with P addition was higher than values reached on the same substrate in lab experiments simulating groundwater influxes to lakes [8], while periphyton biomass in treatments without P addition was similar or slightly lower. Differences in the availability of P H2O in the sediments of the N and S shore still did not translate into different periphyton biomass in the experiment. This might have been caused by growth of benthic algae in and on the sediment that lowered P fluxes into the water and thus P availability for algae growth on plastic strips. Hoffmann and Gunkel [19] measured primary production in the top 2 cm in the sediments of the nearby Lake Tegel and found significant production by diatoms. Since interstitial algae were found to be abundant down to 6 cm sediment depth [31], the production is most certainly ongoing down to those depths as well. In addition, phytoplankton may have taken up part of the P released from sediments, but benthic algae covering the sediment surface should limit this process under sufficient light supply [38]. However, the most parsimonious model that explained periphyton growth did contain P H2O as an explanatory variable, along with P addition. This indicates that diffusive fluxes of P H2O from sediments into the water were still relevant for periphyton growth in our experiment. In lakes with a higher organic content in littoral sediments such as Lake Tegel [31], a higher P availability in interstices can be expected, potentially increasing the impact of diffusive P fluxes on periphyton growth. In the field, higher periphyton biomass has been shown to occur in an oligo-mesotrophic lake at locations with groundwater influx [11]. Whether diffusive P fluxes from sandy littoral sediments are also relevant for periphyton biomass in eutrophic lakes remains to be proven.
Macrophyte growth was lower in the treatment using sediment from the N shore, where IBF pumping rates were higher. Furthermore, macrophyte growth was explained by periphyton growth and lake side in our most parsimonious model. While these findings generally confirm earlier findings on the negative effect of periphyton shading on macrophytes [8,23,37,39], a strong causal link between the additional P availability in the N shore sediments and the lower macrophyte development via light attenuation by periphyton cannot be proven with our data. Shading effects by periphyton in the treatments without P addition to the water were low. In those treatments, the periphyton biomass reduced light availability by up to 19 % using the formula provided in Köhler et al. [37]. Macrophytes are also facilitated by additional P availability in sediments, and this effect may have compensated for the additional shading effects by periphyton [7]. One could imagine that increased filtration would give a higher availability of dissolved inorganic nitrogen forms for plant growth. But that would lead to the opposite results from the ones we obtained in this study. Also, in the control treatment, where no nutrients at all were added, neither in the sediment nor in the nutrient solution, the macrophytes grew as much as in treatments with sediments collected from Lake Müggelsee. Therefore, there is no reason to believe, that the S or N treatments were more nitrogen limited than the control and therefore the role of nitrogen in the sediments, at least in our experimental setup, should have been negligible. Other factors potentially explaining the lower macrophyte growth on sediments of the N shore include exudates of dormant cyanobacteria that have been shown to negatively affect macrophyte seedling growth [40] pesticides or other organic micropollutants [41,42]. A high efficiency in removal of such substances during bank filtration has been shown recently [43] and consequently, an enrichment in sediments affected by bank filtration may occur and affect growth of benthic primary producers, but this topic was beyond the scope of this study.

Conclusions
We could not confirm our first hypothesis that (1) littoral sediments in Lake Müggelsee affected by bank filtration showed a higher content of organic matter. But we could confirm that (2) those sediments had a higher P availability, (3) additional P stimulated periphyton growth and that (4) higher periphyton biomass hampered macrophyte growth. We conclude that significantly higher P H2O contents and lower macrophyte growth can occur in littoral sandy lake sediments affected by bank filtration as compared to unaffected sites, but we cannot provide final evidence for a causal relationship between higher bank filtration rates, higher P H2O content and lower macrophyte growth. However, our data suggest the potential for significant effects of induced bank filtration on sediments cascading to benthic primary producers that should receive more attention in future research to assure a sustainable application of this valuable drinking water production method.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4441/11/5/1111/s1, Supplementary figures and tables.docx: Figure S1: Parameters of littoral sediment taken from two locations in Lake Müggelsee, Figure S2: Stuckenia pectinata tuber length and volume for each treatment, Figure S3: Parameters of littoral sediment taken from north-eastern (N) and south-western (S) shores, Table S1: Sample site coordinates, Table S2: Description of sediment samples, Table S3: Most parsimonious linear model explaining periphyton biomass, Table S4: Most parsimonious linear model explaining macrophyte biomass and linear model with periphyton biomass as the sole dependent variable explaining macrophyte biomass. Experimental data.xlsx: All data from sediment analysis and growth experiment.