Eutrophication has been identified as the primary cause of water quality deterioration in inland waters worldwide across decades (Schindler 1974; Smith et al. 2006). Eutrophication is caused by the overenrichment of nutrients in surface waters, such as nitrogen (N, Howarth and Marino 2006) and phosphorus (Carpenter 2008), resulting in increased primary productivity. The nutrients that originate from external sources in the catchment and from weathering of lake sediments stimulate autochthonous organic matter production (Smith et al. 1999), which may accumulate in lake sediments. This internal nutrient storage, mostly P due to denitrification driven N losses, may periodically be recycled in the water column (Søndergaard et al. 2013). A key symptom of eutrophication is the development of algal blooms (Kalff and Knoechel 1978) that may hamper provisioning of ecosystem services such as drinking water and recreation, resulting in economic losses as well as negative impacts on quality of human life (Grizzetti et al. 2016). A consequence of eutrophication-related organic matter deposition is oxygen depletion that may influence global warming potential by increasing methane emissions from lakes and impoundments (DelSontro et al. 2018; Beaulieu et al. 2019).

To counter environmental degradation, there is a need to control eutrophication and reduce nutrient loading. Apart from catchment-level nutrient abatement techniques such as wastewater treatment and control of fertilizer application, in-lake measures are becoming an effective tool for minimising algae nuisance (Lürling and Mucci 2020). In-lake restoration measures can generally be divided into two categories: symptom-oriented, i.e. not directly targeting nutrients but rather targeting the nuisance associated with water quality deterioration, and source-oriented mitigation, i.e. directly targeting nutrients. Examples of the former are aeration/mixing that improves hypolimnetic oxic conditions and hampers surface accumulations of algae (Beutel and Horne 1999; Visser et al. 2016), coagulation of algae cells (Liu et al. 2013), fish-stock reduction that improves clarity and sediment resuspension (Hosper and Meijer 1993). Examples of source-oriented mitigation are dredging of nutrient-enriched sediments (Zhang et al. 2010) or the application of phosphorus locking agents that precipitate and immobilize phosphorus (Lurling et al. 2016). Phosphorus locking agents that reduce P availability for organism growth provide a promising avenue for in-lake measures. A widely used P fixative is lanthanum-modified bentonite (LMB), sold under the Phoslock® trademark (Douglas 2002), which is designed to remove dissolved reactive phosphorus from the water column and to block P release from the sediment by forming an insoluble lanthanum-phosphate complex (LaPO4). LMB settled at the sediment can thus provide an active barrier for P fluxes from the sediment, promoting oligotrophication. In comparison with many other chemical locking agents, LMB can keep P locked under a wide range of environmental conditions, such as under high pH, unlike iron and aluminium based products (Mucci et al. 2018); under low pH, unlike calcium-bound P (Lin et al. 2015; Copetti et al. 2016); and under low redox conditions, unlike iron- and manganese-bound P) (Mucci et al. 2018). Binding/complexation with other oxyanions/humic substance is only a kinetic hindrance of LMB (Lurling et al. 2014). A laboratory study by Zamparas et al. (2012) showed that the P-adsorption capacity by LMB can be enhanced with increased temperature (from 5 to 35 °C) due to the enlargement of pore size and/or activation of the bentonite surface.

Lake heatwaves (periods of extremely warm lake surface temperatures) are reaching higher temperatures and are lasting longer under climate change (Woolway et al. 2021), which, through subsequent impacts on mineralization rates and internal nutrient loading, could potentially counteract the current efforts in lake restoration. Heatwaves enhance thermal stability, resulting in deep-water anoxia (Jankowski et al. 2006). As a consequence, under anoxic conditions iron- and manganese-bound P in lake sediments can be released and become bioavailable (Beutel et al. 2008). Rising water temperatures can also increase internal loading by enhancing carbon (C) mineralization rates (Gudasz et al. 2010) and liberating nutrients. Moreover, heatwaves could reinforce global warming regimes through greenhouse gas (GHG: carbon dioxide—CO2, methane—CH4, nitrous oxide—N2O) emissions from lake sediments (Bartosiewicz et al. 2016). Although studies on combined effects of nutrients and warming show strong interactive effects on GHG emission (Davidson et al. 2015, 2018; Aben et al. 2017), little is known about specific effects of heatwaves and restoration measures. Microbial processes like mineralization, nitrification and denitrification are all temperature dependent (Veraart et al. 2011; de Klein et al. 2017), while restoration measures may modulate these processes by reducing C, N and P availabilities needed for microbial activities (Redfield 1958). As a result, the nutrient cycling and GHG emission can be largely changed by the altered microbial processes due to restoration and climate change even without considering the role of primary producers.

Most experimental studies on the efficacy of eutrophication control measures under climate change to date used continuous warming temperature scenarios (Cabrerizo et al. 2020), while few have investigated sudden and large temperature boosts (i.e. heatwaves) and the potential for post-heatwave recovery in lakes. Given its rather sudden and short-term characteristics, the heatwave impacts may be transient rather than long-lived. In our study, we conducted an exposure scenario where we exposed our systems to a heatwave, with prior- and post-heatwave monitoring of water quality, enabling us to study the potential for post-heatwave recovery in lakes. We investigated if the effectiveness of the well-established nutrient abatement technique Phoslock® (hereafter LMB) is impacted by an extreme heatwave, and how this affects potential lake GHG emissions. We measured sediment nutrient release and dissolved GHG concentrations at the sediment–water interface in a three-week sediment incubation experiment. We used unmodified bentonite as a control treatment to account for the potential effects of bentonite clay in the LMB treatment.

We tested three hypotheses: (1) Heatwaves will enhance sediment nutrient releases by increasing mineralization rates and decreasing oxygen concentrations; (2) LMB can reduce the potential heatwave-induced P-release by reducing P-availability in the sediments; and (3) GHG emissions in LMB treated systems will be mitigated due to a reduction in nutrient availabilities.

Materials and methods

Experimental design


On 21 June 2018, sediments were collected in 39 cores (60 cm in length and 6 cm in diameter) with overlying water column at a depth of 1.85 m from a eutrophic Dutch pond (52°02′20.6"N 5°38′51.4"E) using a UWITEC gravity core sampler. Pond water was sampled from three depths of the water column (0 m, 0.5 m, 1.5 m). Three of the 39 sediment cores were selected for analyses of P pools in the sediments, whereas the remaining 36 cores were designated for the experiment.

On these 3 cores, we carried out a P fractionation method following Cavalcante et al. (2018) for the top 10 cm sediment to determine loosely bound P (H2O-P), redox-sensitive P (BD-P), metal oxide-bound P (NaOH-P), Calcium-bound P (HCl-P) and residual P (residual-P). Each fraction of P consists of soluble reactive P (SRP) and non-reactive P (NRP) that represents the organic part. The sediment pools of mobile P, which can be released in anoxic conditions or by organic matter degradation and become bioavailable, was determined by the sum of the SRPs of the H2O-P and BD-P fractions and the NRP of the NaOH-P fraction (Cavalcante et al. 2018). The sum of other P forms represents the non-mobile sediment P pool, i.e. the difference between the sum of all P forms and the above-mentioned mobile P forms. In our treatments, we determined LMB doses based on the mobile sediment P pool.

Upon transportation into the laboratory, the undisturbed sediment cores were placed in a temperature-controlled water bath at 20 °C. 20 °C is regarded as a baseline temperature for the experiment, similar to the pond water temperature during core collection. Water temperature was continuously recorded during the course of the experiment. The cores were open to the air throughout the experiment and were kept dark most of the time to prevent algal growth. The cores were exposed to light during the sampling events of around one hour, when the light ranged between 0.01 and 104.0 µmol photons m−2 s−1. Sediments were acclimatized to the laboratory conditions for 5 days prior to the experiment.


LMB was obtained from Phoslock® Europe GmbH (Manchester, UK). To test the effectiveness of LMB under a heatwave scenario, we split the 36 sediment cores into three groups such that 12 cores were treated with LMB, 12 cores were treated with bentonite (Bent) and the other 12 cores were left untreated (Ctrl) (Fig. 1). We made a slurry of 380 mg LMB resuspended in water from each LMB treatment unit and added the slurry at the top of the core, targeting the amount of potential releasable P in the top 3.3 cm of the sediment to achieve an LMB:P ratio of 100:1 (see Table 1 for details on calculation of LMB dose). The same dose of bentonite was added to the ‘Bent’ treatment to control for potential physical capping of sediment in the LMB treatment.

Fig. 1
figure 1

Experimental design. ‘Ctrl’ represents control without adding P-binding materials; ‘Bent’ represents bentonite treatment; ‘LMB’ represents Phoslock® treatment. Shaded boxes represent sediment cores exposed to heatwaves

Table 1 Calculation of Phoslock® (LMB) dose

All of the 36 sediment cores were kept at 20 °C during the first week (Fig. S1). During the second week, a heatwave scenario was simulated by exposing half of the cores of all P control treatments (i.e. LMB, Bent and Ctrl, 18 cores in total) to a temperature of 30 °C, whilst half of the cores of all P control treatments (LMB, Bent and Ctrl) were kept at 20 °C, yielding 6 replicates in each treatment unit (Fig. 1). After one week at 30 °C, the heatwave cores were all returned to the baseline temperature of 20 °C. The heatwave scenario is similar to the recent summer conditions (2015–2020) in the Netherlands where the average temperature during heatwave that lasted weeks was around 30 oC (KNMI data, In the experimental heatwave scenarios, the heating and cooling of the water columns were realized within one day to simulate a sudden temperature boost scenario. As a result, we were able to monitor the dynamics of different water quality parameters before, during and after the heatwave in different treatment units.

Water and sediment measurements

We sampled a suite of water quality variables including dissolved oxygen concentrations (O2), pH, conductivity, dissolved inorganic nutrients and dissolved GHG at 10 cm above the sediment surface of each of the experimental units. We sampled SRP, NO3-N + NO2-N and NH4-N, and carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) in total eight times along the course of the experiment (2–3 times/week). Metals, i.e. lanthanum (La), calcium (Ca), iron (Fe), aluminium (Al) and manganese (Mn) were sampled three times during the experiment (1 time/week). We used handheld sensors to instantaneously measure O2 (HQ40d Portable probe, Hach, Colorado, US), pH and conductivity (WTW Multi 350i, Geotech Environmental Equipment Inc., Colorado, US). After filtration of water samples over prewashed GF/F filters (Whatman, Maidstone, U.K.) we analysed the filtrate for SRP, NO3-N + NO2-N and NH4-N, using an QuAAtro39 Auto-Analyzer (SEAL Analytical Ltd., Southampton, U.K.). In addition, we measured concentrations of La, Ca, Fe, Al and Mn in the filtrate by inductively coupled plasma optical emission spectrometry (iCAP6500 Duo, ThermoFisher, U.K.).

The concentrations of GHGs in the water column were determined by a headspace equilibration technique described in Magen et al. (2014) and Halbedel (2015). Briefly, water samples were introduced in a syringe, where air was introduced and the dissolved GHGs were equilibrated with the headspace by shaking the syringe vigorously for two minutes. Afterwards, the equilibrated headspace gas was collected and analysed for GHG concentrations in a Gas Chromatography (TRACE™ 1300 GC, ThermoFisher, U.K.) machine equipped with a Flame Ionization Detector (FID) and an Electron Capture Detector (ECD). Calculations for the water column concentrations were based on Halbedel (2015), in which we used parameters from Weiss (1974) for CO2, Yamamoto et al. (1976) for CH4, and Weiss and Price (1980) for N2O. Note that the CO2 concentrations in our system were far beyond atmospheric CO2 equilibrium (≈ 412 ppm), therefore the limits of the headspace method to analyse the partial CO2 concentration in water discussed in Koschorreck et al. (2021) is not relevant in our case.

To gain a close understanding of the anoxic conditions in the sediments in addition to the O2 concentrations measured in the overlying waters, 12 cores (2 cores randomly chosen from each of the six treatment groups) were equipped with redox probes. Redox potential dynamics were continuously recorded (time interval of 10 min) at 12 different depths above and/or below the sediment water interface (max = 14 cm depth below the sediment surface).

After the experiment was completed, the P-fractions in 24 out of 36 cores (the ones without a redox probe, i.e. 4 cores from each treatment) were determined using the same method described for the three initial cores above. Besides the mobile P, their non-mobile P (i.e. the sum of the NRPs of the H2O-P, BD-P fractions, the SRP from NaOH-P fraction and HCl-P and residual-P fractions) was also determined.

Statistical analysis

The effects of the experimental treatments on the dynamics of different variables including O2, pH, SRP, NH4-N, NO3-N + NO2-N, CO2, CH4, and N2O were analysed by a linear mixed-effect model (LME; Lindstrom and Bates, 1988), with each of the variables as the univariate response variable. The P treatments (Ctrl, Bent, or LMB) and temperatures were included as fixed effects. Day in the experiment was taken as an additional fixed effect to evaluate the changes in response variables through time. Differences among subjects (cores) were taken as a random effect in the LME model. The LME model corresponds to:

$${y}_{ik}={\beta }_{0}+\sum_{m=2}^{3}{\beta }_{1mk}\times {I[P]}_{imk}+{\beta }_{2k}\times {Temp}_{ik}+{\beta }_{3k}\times {Time}_{ik}+{b}_{0k}+{\varepsilon }_{ik}$$

where \(i=\mathrm{1,2},\dots ,288,\) index \(k=1, 2, \dots ,36\) corresponds to the 36 cores, \(\mathrm{m}=\mathrm{2,3}\) corresponds to the P control measures. \(\beta\) are the fixed-effects coefficients. \({I[P]}_{im}\) is the dummy variable representing the level \(\mathrm{m}\) of the P treatments. \({Temp}_{ik}\) stands for temperature, and \({Time}_{ik}\) stands for time. \({b}_{0k}\) is the random effect for level \(\mathrm{k}\) of cores, and \({\varepsilon }_{ik}\) is the observation error for observation \(i\). The random effect has the prior distribution,

$${b}_{0k}\sim N\left(0, {\sigma }_{b}^{2}\right),$$

And the error term has the distribution,

$${\varepsilon }_{ik}\sim N\left(0, {\sigma }^{2}\right).$$

We analysed three phases in the experiment separately. First, we analysed the period before the heatwave (days 1–7), where we only evaluated the effects of the P treatments and time. Second, we analysed the period during the heatwave (days 8–14), where we evaluated the effect of the different temperature treatments (heatwave vs non-heatwave exposed treatments) in addition to the effects of P treatments and time. Finally, we analysed the post heatwave period (days 15–21), where we evaluated the effects of P treatments and the recovery from the heatwave over time. The normality and heteroscedasticity of the residuals were tested by Shapiro Wilk test (Ghasemi and Zahediasl 2012) and Breusch Pagan test (Waldman 1983), respectively. For the data that did not pass the test for normality, we applied a data transformation method (i.e. square root transformation for O2, pH, NO2 and NO3, and logarithm transformation for CH4). To account for heteroscedasticity, we deployed a weighted linear mixed-effect model (Zuur et al. 2009).

Determination of differences between the sediment P forms after the experiment were performed through two-way analysis of variance (ANOVA) with heatwave and LMB/Bent treatments as factors. In addition, we carried out two-way ANOVA analyses to determine the differences between metal concentrations in the heatwave phase as well as in the post-heatwave phase. All statistical analyses in this study were performed in R language (R Core Team 2019). We used the packages lubridate (Grolemond and Wickman 2011), ggplot2 (Villanueva and Chen 2019), nlme (Pinheiro et al. 2019) and vegan (Oksanen et al. 2019).


Pre-treatment conditions

At the moment of field-sampling, water transparency measured with a Secchi disc was 0.54 m, pH was 7.52 ± 0.0, temperature was 19.5 ± 0.0 °C and O2 concentration was 3.8 ± 0.1 mg/L and no apparent water stratification was observed. The soluble reactive phosphorus (SRP) concentration was 0.14 ± 0.02 mg P/L, the ammonium (NH4-N) concentration was 0.56 ± 0.13 mg N/L, and the sum of nitrate and nitrite (NO3-N + NO2-N) was 0.04 ± 0.08 mg N/L. The total dissolved carbon (TDC) averaged 49.1 ± 3.1 mg/L, with an inorganic dissolved carbon (IDC) concentration of 36.9 ± 0.4 mg/L and a dissolved organic carbon (DOC) concentration of 12.2 ± 3.0 mg/L.

The sediment layers in the cores were at least 15 cm deep. The sediment was very fluffy with an average water content of 90.7 ± 1.5%, a dry weight (DW) density of 0.10 ± 0.02 g/mL, and an average organic matter (OM) amount of 0.022 ± 0.001 g/mL wet sediment.

Nutrients and metals dynamics and responses

In all treatments the initial SRP concentration in the water columns started at a similar level of 0.71 mg P/L (SE =  ± 0.06) and changed significantly throughout the experimental periods (Fig. 2a). After LMB application the amount of SRP in the water column of our experimental units dropped by 83% within two days, whereas in the non-LMB groups (i.e., Ctrl and Bent treatments) a continuous increase in the SRP concentrations was observed along the entire course of experiment, up to 1.24 ± 0.10 mg P/L. No significant difference between Ctrl and Bent treatments were detected, with SRP concentrations increasing at a similar rate over time in both groups regardless of heatwave treatment (effect of increasing rate = 0.027 mg P/L/day, F1, 162 = 117.5, p < 0.001). In contrast, in LMB-exposed groups, upon exposure to the heatwave, SRP concentrations were significantly increased compared to the non-heatwave exposed treatments (estimate of difference = 49%, F1, 58 = 14.20, p < 0.001), with effects lasting to the end of the experiment. The heatwave-induced elevation of SRP concentrations in LMB exposed treatments ranged between 0.05 and 0.08 mg P/L.

Fig. 2
figure 2

a Dynamics of soluble reactive phosphorus (SRP) in the water column. Error bars indicate the standard error. The pink shading indicates exposure to different temperature regimes (20 °C in solid lines and 30 °C in dashed lines); before and after the pink shading shows all P-treatments (Ctrl, Bent and LMB) at 20 °C. b Phosphorus (P) fractions in sediments (µg P/g DW). The labels on the x-axis illustrate treatment units: ‘Ctrl’ for no P treatments (i.e., no LMB/bentonite), ‘Bent’ for bentonite and ‘LMB’ for Phoslock®; ‘20 °C’ for non-heatwave exposed groups and 30 °C for heatwave exposed groups

After completion of the experiment, the sediment P fractionation analyses showed different P pools among different P-binding treatments, with Ctrl treatments of 430.3 ± 81.6 µg P/g DW, Bent treatments of 494.6 ± 198.1 µg P/g g DW and LMB treatments of 590.9 ± 177.9 µg P/g DW (Fig. 2b). According to the two-way ANOVA test of different P pools, heatwave exposure resulted in a significantly lower redox sensitive P pool (BD-P) than the non-heatwave treatments (estimate of difference = 7%, F1, 113 = 12.2, p < 0.001). Such heatwave-induced decrease in BD-P was not observed in the LMB addition groups. Other sediment P forms showed no significant heatwave effects, but LMB addition led to significantly higher contents of metal oxide-bound P (NaOH-P, estimate of difference = 86%, F2, 114 = 7.9, p < 0.001) and calcium-bound P (HCl-TP, estimate of difference = 42%, F2, 114 = 7.9, p < 0.001).

We analysed NH4-N and NO3-N + NO2-N as indicators of N dynamics in the water columns (Fig. 3). In all treatments NH4-N had a slight increase in the first week from 3.54 ± 0.25 to 4.58 ± 0.29 mg N/L, and then decreased to approximately 0.55 ± 0.34 mg N/L by the end of the experiment. The NH4-N concentrations for cores subjected to the heatwave were lower (estimate of difference = 27%, F1, 249 = 117.5, p < 0.001) than those not subjected to a heatwave. The NO3-N + NO2-N concentrations started to increase from the second week. The NO3-N + NO2-N concentrations for treatments subjected to the heatwave were higher compared to the non-heatwave groups in the second week, with a decline after the heatwave leading to a lower concentration relative to the non-heatwave groups at the end of the experiment (estimate of difference = 8%, F1, 174 = 18.88, p < 0.001). No effects of LMB/Bent additions on NH4-N or NO3-N + NO2-N were detected.

Fig. 3
figure 3

Dynamics of a ammonium (NH4-N) concentration, b dissolved oxidized nitrogen (DON = NO3-N + NO2-N) concentration in the water column. Error bars indicate the standard error. The pink box indicates exposure to different temperature regimes (20 °C in solid lines and 30 °C in dashed lines); before and after the pink box indicates all P-treatments (Ctrl, Bent and LMB) at 20 °C

Fig. 4
figure 4

Dynamics of dissolved oxygen concentration (O2) in the water column. Error bars indicate the standard error. The pink shading indicates exposure to different temperature regimes (20 °C in solid lines and 30 °C in dash lines); before and after the pink shading indicates all P-treatments (Ctrl, Bent and LMB) at 20 °C

Dissolved La concentrations in the LMB-exposed groups were much higher in the pre-heatwave and during the heatwave phases (estimate of difference = 91%, F2, 33 = 35.1, p < 0.001), and decreased through time, with an end concentration of 1.50 ± 0.30 µg La/L, which is below the Dutch La standard for surface water (= 10.1 µg La/L). A decreasing trend of La was also observed in the treatments without LMB additions (Fig. S2a), without significant differences between LMB/Bent/Ctrl treatments in the post-heatwave phase. Dissolved Fe concentrations decreased through time (Fig. S2b), with heatwave exposure leading to higher Fe concentrations during and after heatwave phases (estimate of difference = 103%, F1, 68 = 9.9, p = 0.003), whereas no effects from LMB/Bent treatments were detected. Of other metals including Ca, Al, Mn and S we detected no effects from heatwave or LMB/Bent treatments (Fig. S2c). Both Ca and S concentrations decreased over time with a high correlation between them (Pearson correlation coefficient = 0.84, t = 19.3, df = 105, p < 0.001, Fig. S3).

Oxygen dynamics

In all treatments the oxygen concentrations significantly increased from 0.42 ± 0.16 to 3.01 ± 0.43 mg/L with time (Fig. 4, F1, 239 = 98.6, p < 0.001), with no effects from LMB/Bent additions detected. This was reflected by the redox dynamics (Fig. S4) which indicated reducing conditions in the water column of the cores in the starting phase, with the systems becoming more redox potential-positive towards the end of our experiment, irrespective of the treatments. In contrast to observations in the water column, redox potentials in the sediments were descending along the experiment period.

Our linear mixed effect model (LME), however, revealed a temporal reduction of oxygen concentrations during the heatwave in the heatwave treatments (estimate of difference = 16%, F1, 34 = 4.9, p = 0.03), with a maximum deviation by 0.87 mg/L at day 11. This heatwave effect disappeared after the heatwave. By the end of our experiment, the oxygen concentrations were still unsaturated (saturation concentration at 20 °C = 9.03 mg/L).

Dissolved greenhouse gases

We measured concentrations of dissolved greenhouse gases CO2, CH4 and N2O, as a proxy of their emission potential. Among the three gases, CO2 concentrations were the highest, followed by CH4 and N2O, with respectively a factor 10 and 100 lower levels. All three gas concentrations in the water column of the cores changed significantly over time, with significant increase during the heatwave phase (Fig. 5a for CO2, estimate of difference = 15%, F1, 243 = 27.9, p < 0.001; Fig. 5b for CH4, estimate of difference = 21%, F1, 243 = 20.4, p < 0.001; Fig. 5c for N2O, estimate of difference = 536%, F1, 243 = 116.6, p < 0.001), equalling an increased CO2-equivalent by 106%. No effects from LMB or Bent treatments on the greenhouse gas dynamics were detected.

Fig. 5
figure 5

Dynamics of a dissolved carbon dioxide (CO2), b dissolved methane (CH4) and c dissolved nitrous oxide (N2O). Error bars indicate the standard error. The pink shading indicates exposure to different temperature regimes (20 °C in solid lines and 30 °C in dashed lines); before and after the pink shading indicates all P-treatments (Ctrl, Bent and LMB) at 20 °C

CO2 concentrations decreased from 7.3 ± 0.3 × 103 to 6.4 ± 0.3 × 103 ppm in the before-heatwave phase, whereas during the heatwave phase in the heatwave-exposed treatments the CO2 concentrations rose up to 8.4 ± 0.5 × 103 ppm. In the post-heatwave phase, CO2 concentrations in the heatwave groups dropped to the same levels as the non-heatwave treatments, resulting in an average of 6.0 ± 0.5 × 103 ppm by the end of the experiment.

CH4 concentrations dropped from 391.98 ± 123.98 ppm to 62.70 ± 45.74 ppm in the pre-heatwave phase. During the heatwave phase, the heatwave-exposed cores were observed with higher CH4 concentrations (mean = 47.38 ± 3.21 ppm) than the non-heatwave treatments (mean = 26.69 ± 1.87 ppm). However, in the post-heatwave phase (days 15 – 21), the CH4 concentrations in the non-heatwave treatments surpassed that in the heatwave-exposed groups and reached a high average level of the same magnitude as that measured in the beginning of the experiment (mean = 174.25 ± 57.39 ppm), whereas the CH4 concentrations in the heatwave-exposed cores had a relatively low-end concentration (mean = 23.40 ± 5.84 ppm).

N2O concentrations stayed at a rather low level (mean = 0.15 ± 0.02 ppm) in the before-heatwave phase. During the heatwave phase, N2O emissions started to increase in all treatments, with N2O concentrations in heatwave-exposed cores increasing at a much higher rate than in non-heatwave groups. In the period after the heatwave (days 15–21) the non-heatwave groups showed a relatively stable N2O concentration (mean = 15.16 ± 2.52 ppm) until the end of the experiment. N2O concentrations in the heatwave groups increased to 72.84 ± 26.53 ppm during the heatwave, but in the post-heatwave phase, the concentrations dropped to 38.10 ± 19.14 ppm, which is, however, a still higher level than the observed N2O emission in the non-heatwave groups.


Using a full-factorial design, we investigated the combined effects of phosphorus control by lanthanum-modified bentonite (LMB) and exposure to an extreme heatwave event on the biogeochemistry at the sediment–water interface and the resulting fluxes between sediment and water. For shallow waters, one of the most abundant water types (Verpoorter et al. 2014), such fluxes between water and sediment play an important role in the phosphorus, nitrogen and carbon cycling. Despite the observation that LMB was able to reduce phosphorus levels in the water column up to 91% in comparison to the control groups by the end of our experiment, LMB effectiveness was hampered upon exposure to a heatwave, resulting in increasing P concentrations with 11%, persisting until the end of the experiment. There was no significant effect of LMB addition on nitrogen dynamics. Under low oxygen conditions (O2 < 4.8 mg/L in all treatments), nitrification was stimulated by increased oxygen and temperature, resulting in an accumulation of nitrate + nitrite and nitrous oxide. In addition, our results suggest that GHG dynamics were impacted upon heatwave exposure, but LMB did not affect this pattern.

The impact of heatwaves on phosphorus and nitrogen dynamics: eutrophic vs oligotrophic sediments

The continuous P-release in controls, leading to concentrations as high as 1.2 mg P/L towards the end of our experiment, could be explained by the large pool of bioavailable phosphorus in sediment (Fig. 2b). This pool of bioavailable phosphorus consists of phosphorus in the pore water, redox-sensitive P (BD-P), and organic P which can be mobilized under anoxic conditions (Cavalcante et al. 2018). LMB strongly reduced phosphorus concentrations in the water column and kept those persistently low until the end of our experiment, which indicates that LMB both stripped the water column P and hampered sediment P release, as was expected from its well-documented performance (Copetti et al. 2016). The bentonite-only treatment did not have any impact on P dynamics relative to our controls, which further underpins that the LMB effect was a result of the P inactivation and not caused by depositing a thin clay layer on top of the sediment. Some studies reported a P abatement capacity of unmodified bentonite (Zamparas et al. 2012), but the bentonite used in our study had no P-binding capacity (Mucci et al. 2018) and the layer was evidently too thin to act as a passive barrier (Kim et al. 2007).

When we exposed the LMB-treated sediments to a one-week heatwave, phosphorus concentrations increased by 0.08 ± 0.03 mg P/L (the overlying water volume = 863 ± 21 mL) in the water column at the end of our three-week experiment (Fig. 2a), equalling an 11% increase compared to the non-heatwave exposed group. In an earlier study, however, higher P-adsorption capacity by LMB was observed under increased temperature due to the enlargement of pore size and/or activation of the bentonite surface (Zamparas et al. 2012). It is noteworthy that Zamparas et al. (2012) used a rather simple environment (3 h of experimental time, heavy mixing of the slurry with solution), whereas the static sediment in our experiment is a much more complex matrix with biogeochemical processes playing an important role in determining the LMB capacity. Earlier model simulations indicate that concentrations of 0.05 mg P/L, similar to the increase in SRP concentrations observed at the end of this experiment, could cause systems to shift from transparent to turbid states (Janse et al. 2008). However, this has not been observed in whole lakes that have undergone LMB additions; for instance the LMB-treated shallow Lake Barensee in Germany did not develop any blooms during heatwaves (Epe et al. 2017).

The discrepancy between our lab study and the field observations might be caused by a larger dilution effect in lakes compared to the small water volumes in the cores, by not all binding sites of LMB being rapidly available, by a possible under-dosing in our experimental treatments, or erratic sediment–water transport through ebullitive processes in our cores. The P inactivation by LMB is a kinetic process, which means that it takes a certain time before all binding sites are occupied (Dithmer et al. 2016), especially in the presence of high DOC (Lurling et al. 2014) as was the case in our system. However, despite the notion that binding sites were potentially not all occupied, LMB was only partially able to counteract the temperature-enhanced P release in the week after the heatwave exposure, an observation that has been confirmed by an in-situ months-long enclosure experiment (Zhan et al. in prep). Furthermore, we may have potentially under-dosed LMB our experimental treatments, as we did not include the BD-NRP fraction which may also contain organic P (Jan et al. 2015) in our dose estimates. The microbial breakdown of organic matter results in release of organic phosphate to pore waters and can be considered one of the most important source of phosphate (Föllmi 1996). As organic matter breakdown is expected to be enhanced by increased temperature (Gudasz et al. 2010), part of the liberated P is taken up by the decomposing bacteria, and part will enter the overlying water column. In addition, the ebullition that transports sediment-P might be enhanced under heatwave conditions (Aben et al. 2017). Therefore, inclusion of the BD-NRP and targeting also deeper layers of the sediment in dose estimations are highly recommended for lake managers applying LMB as a eutrophication control measure, especially to counteract the heatwave impacts.

Our results showed that the rates in the N-cycle processes were significantly changed upon exposure to a heatwave, irrespective of treatment. At the start of our experiment, the sediment incubations were anoxic, with NH4-N as the dominant nitrogen form accumulating in the overlying water. As dissolved oxygen concentrations rose over time, nitrification became a more dominant process, decreasing ammonium accumulation and leading to increased oxidized nitrogen concentrations (Fig. 3). Rysgaard et al. (1994) demonstrated that nitrification is stimulated by increasing O2 in the O2 range of 0–9.6 mg/L, which was the case in our experiment. Upon exposure to a heatwave (30 °C), the oxidation of NH4-N was accelerated, indicating a positive response of nitrification to increasing temperature. Thamdrup and Fleischer (1998) demonstrated that the optimum temperature for nitrification in warm temperate sediment was near 40 °C. The decline of oxidized nitrogen concentrations (NO3-N + NO2-N) in the heatwave groups during the post-heatwave phase could be associated with the depletion of ammonium. Nitrification can operate at low ammonium concentration but at low rates (Dodds and Jones 1987). Moreover, denitrification is expected to be stimulated with increasing temperature (Veraart et al. 2011). Though we did not measure denitrification directly, denitrification was plausibly leading to an increased consumption of NO3-N and NO2-N, therefore decreasing N-release into the overlying water. The LMB treatment showed no effect on N-dynamics, which supports our conclusion that LMB acted as a chemical binding compound of DRP rather than a physical barrier for nutrient release from the sediments. A recent study by Zeller and Alperin (2021), however, showed that LMB can act as NH4-N source with an increase in NH4-N concentration by 10–275%. This difference between our results and Zeller and Alperin (2021) might be attributed to the fact that their LMB dose was more than double of ours (0.029 compared to 0.013 g LMB/cm2 sediment) and the overlying water volume was much smaller in their experiment (150 mL compared to 863 mL), resulting in higher LMB concentrations.

Heatwave effects on greenhouse gas emissions

In our experimental system, regardless of the treatments, concentrations of greenhouse gases (GHG) initially decreased (CO2 and CH4), or were close to zero (N2O). Earlier studies on freshwater ecosystems demonstrate increased GHG emissions under warming scenarios (Bartosiewicz et al. 2016; Aben et al. 2017; Bergen et al. 2019). In our experiment, exposure to a heatwave did lead to increased concentrations of CO2 and N2O. For CH4, however, after an initial short-lived increase during the heatwave itself, the heatwave-exposed treatments had lower CH4 concentrations relative to the non-heatwave exposed group towards the end of the experiments (Fig. 5). A potential explanation for this might be that on the longer term, substrate limitation for methanogens (Duc et al. 2010) may play a stronger role in the heatwave exposed cultures, potentially due to the short-lived increase in methane production during the heatwave. Note that our sediment incubation was conducted under dark condition, different patterns might emerge in real lakes in the presence of primary producers, which points to the importance of validating our results with field observations.

Previous studies on shallow aquatic systems demonstrated that greenhouse gas emissions are higher in eutrophic systems than in more oligotrophic systems (Davidson et al. 2015; Peacock et al. 2019). We therefore hypothesized that LMB treatments, by reducing nutrient availability, would inhibit bacteria growth and subsequently reduce GHG productions. In disagreement with our hypothesis, our results showed that addition of LMB, although effective in blocking sediment P-release, did not affect the GHG emissions from the sediments. Previous studies (Dithmer et al. 2016) demonstrated that LMB needs time (up to months) to bind all available P in sediments. Another potential cause may be that our doses applied theoretically can only target top 3.3 cm sediment, which is about 20–25% of the sediment column, leaving ample space for bacterial activity and methanogenesis in deeper layers. Under warmer conditions the sediment layers available for bacterial activity and methanogenesis may be located deeper, because anoxia is expected to be stronger (Jankowski et al. 2006), associated with enhanced methanogenesis (Schulz et al. 1997). Nonetheless, our results indicate that LMB might not be an effective method in controlling GHG emissions, at least in the short term. For reduction of carbon associated GHG emissions (CH4 and CO2) measures that directly target the reduction of organic matter inputs into sediments, such as an improvement of water treatment in the catchment (Jones et al. 2016) and control of bank soil erosion (Rickson 2014), may be more favourable.

N2O is typically derived during nitrification of NH4-N under oxic conditions and from the coupled nitrification–denitrification reactions under suboxic conditions, which explains the increase in N2O coinciding with the NH4-N decrease. Both nitrification and denitrification are strongly temperature-dependent (Veraart et al. 2011; de Klein et al. 2017). Heatwave treatments in our experiment had higher N2O concentrations, coinciding with a drop in O2 concentrations, which further indicates that freshwater N2O emissions can be strongly temperature dependent and can be boosted under climate change (Parton et al. 2001; Veraart et al. 2011). The decline in N2O concentrations in the post-heatwave phase in heatwave-treated groups could be because NH4-N for nitrification became limiting, while denitrification in deeper sediment layers increased in efficiency, reducing the N2O:N2 ratio in the final product (van de Leemput et al. 2011). Conventional biological denitrification requires low oxygen concentration less than 0.2 mg/L (Seitzinger et al. 2006). Even when the water column was oxygenated, these concentrations still occurred in anoxic microsites in the sediments of our experimental cores (Fig. S4). In addition, aerobic denitrification has also been observed in freshwater sediments (Trevors and Starodub 1987; Rysgaard et al. 1994; Lv et al. 2017) as well as in coastal sediments (Marchant et al. 2017). Moreover, the pathway of dissimilatory nitrate reduction to ammonium (DNRA) could also contribute to part of the production of N2O (Sun et al. 2016). A 15 N tracing technique (Müller et al. 2014) is needed to determine which pathway is mainly responsible for the production of N2O.

Conclusions and recommendations

Our sediment incubation experiment indicates that the rates of biogeochemical processes can be significantly accelerated upon heatwave exposure, resulting in a change in fluxes of nutrient and greenhouse gases between sediment and water column. The current efforts in eutrophication control will face more challenges under future climate scenario with more frequent and intense extreme events as predicted by the IPCC.

The effectiveness of widely established eutrophication control measure LMB was, at least temporarily, impaired upon exposure to an extreme heatwave, with an increase in concentration of 0.08 ± 0.03 mg P/L with an overlying water volume of 863 ± 21 mL, equalling 50% increase relative to non-heatwave treatments. Although the effect of the heatwave on P-release of LMB treated sediments persisted until the end of the experiment, long term studies should address whether the P-concentrations eventually return to lower pre-heatwave levels. In addition, further research is needed to explore whether increased LMB dosage can mitigate the negative impacts of a heatwave. Nonetheless, our study does suggest that our current abatement efforts may be hampered under climate change, which calls for consideration of more climate-robust measures, such as through revisiting dose–response relationships in the development of rehabilitation plans.

Exposure to a heatwave resulted in higher dissolved GHG concentrations with an increased CO2-equivalent by 106%, showing potential for increased emissions relative to non-heatwave exposed treatments. Our experiment showed that LMB addition did not lead to lower GHG concentrations, which implies that inhibiting microbial GHG production by creating a P-limiting environment through LMB is ineffective. Thus, alternative strategies directly targeting reduction in organic load such as sludge removal or erosion control should be explored to effectively mitigate greenhouse gas emission.