Discharge and fate of biocide residuals to ephemeral stormwater retention pond sediments

Retention ponds are important parts of stormwater management in urban areas. Since 2009, the German Federal Water Act requires to avoid, retain or infiltrate stormwater locally (WHG 2009 §55). Besides stormwater, ponds can also retain pollutants from the inflowing stormwater including diverse contaminants (Sébastian et al. 2015). Two types of ponds can be distinguished: first, treatment wetlands often constructed primarily to remove pollutants and improve water quality (Kadlec & Wallace 2008), and second, retention ponds primarily designed for water retention that are dry during phases of low precipitation and/or high evaporation. The type of pond and its water variability are crucial to understanding the fate of pollutants. For example, while treatment wetlands in agriculture have been found to control pesticide dissipation (Imfeld et al. 2021) and remove pesticides (Maillard et al. 2011), pollutants in a retention pond connected to an urban catchment revealed low retention of pesticides (Sébastian et al. 2015). At the latter pond, pesticides occurred mainly in low concentrations of 2–286 ng g⁻¹ in the sediment (Wiest et al. 2018). However, most studies in urban areas address pollutants in treatment wetlands or infiltration systems (Allinson et al. 2015; Tedoldi et al. 2016) and thus knowledge about pollutant retention in stormwater retention ponds is comparatively limited.

One example of contaminants in urban stormwater is biocides that act as active ingredients in coatings to prevent algae and fungi growth on façades. During wind-driven rain events, these biocides are partly washed off, enter the environment (Burkhardt et al. 2012) and may have detrimental effects on aquatic organisms (Musgrave et al. 2011) and soil microorganisms (Fernández-Calviño et al. 2021). Biocides also degrade into transformation products (TPs) that are often more mobile and persistent in the aquatic environment. TPs might have similar or even higher toxicity for various species compared to their parent compounds (Hensen et al. 2020). The focus of this study is on three commonly and complementary used biocides, namely diuron, octylisothiazolinone (OIT) and terbutryn, that act as herbicide, fungicide and algaecide, respectively. All of them can be toxic to environmental organisms even in low concentrations with predicted no-effect concentration (PNEC) values in the freshwater of 0.02 μg L⁻¹ for diuron, 0.013 μg L⁻¹ for OIT and 0.034 μg L⁻¹ for terbutryn (Paijens et al. 2019).

While various studies confirm biocide entry into urban stormwater (Burkhardt et al. 2012; Hensen et al. 2018; Wicke et al. 2021a) only a few studies so far focused on sediments in urban stormwater infrastructure as a target compartment of biocide pollution. Two studies sampled sediments in urban wetlands in Sweden (Flanagan et al. 2021) and in Australia (Allinson et al. 2015) and detected the biocide terbutryn only in one and in no sample, respectively. For retention ponds, there is little knowledge of the amount of biocides retained in pond sediment and we are not aware of a mass balance. Sébastian et al. (2015) analyzed samples of a detention pond for biocides used in film preservatives, such as terbutryn, irgarol, and isothiazolinone, but did not detect any of them in pond sediments.

Due to the utilization of terbutryn as an agricultural pesticide, several studies exist on its interaction with soil/sediment (Muir & Yarechewski 1982; Avidov et al. 1985). The sorption of terbutryn generally increases with pH and soil organic carbon (Johann et al. 2018). Degradation kinetics and formation of TPs in soil are described in Bollmann et al. (2017) pointing to a long half-life time of 231 days for terbutryn. Junginger et al. (2022) differentiated degradation pathways of terbutryn, namely biodegradation, photodegradation and abiotic hydrolysis, using compound-specific isotope analysis. Lower degradation rates of terbutryn were found with lower redox potentials (Muir & Yarechewski 1982) and less microbial activity (Avidov et al. 1985). As all biocides may have toxic impacts on the environment, it is important to understand their environmental fate including entry pathways and potential sinks such as ponds.

To assess the fate and risks of terbutryn in pond sediments, it is necessary to quantify the amount of biocides leached from facades in the contributing catchment. Biocide emissions depend on temperature, sunlight irradiation, amount of biocide used, precipitation amount, and antecedent moisture, among others (Paijens et al. 2019). Usually, limited data are available and only models can fill this gap including various assumptions and uncertainties. For example, Coutu et al. (2012) modeled yearly biocide emissions and concentrations in receiving surface waters for the city of Lausanne based on experimental data of Burkhardt et al. (2011); they arrived at total emissions for Lausanne of 900 kg year⁻¹. Paijens et al. (2021) calculated the annual mass load discharged from Paris to the Seine River using Monte Carlo simulations based on measurement data and estimation of stormwater proportions. The annual terbutryn load discharged by wastewater treatment plants was 10–20 kg year⁻¹, whereas the load from combined sewer overflow was by a magnitude lower (0.5–0.7 kg year⁻¹). However, these authors did not provide information about the amount of biocide originally leached from individual facades that might be important for investigations in smaller urban areas. The model software COMLEAM calculates biocide wash-off from facades (Burkhardt et al. 2021) and was already applied to urban areas studies (Wicke et al. 2022). In this study, we use COMLEAM to determine the biocide emissions of a residential area.

Besides emissions, the transport of biocides and the arrival at selected sinks such as surface water, soil or groundwater are important. Wicke et al. (2022) investigated a residential area in Berlin and established a 1.5-year mass balance to show that over 90% of emitted biocides were diffusely lost around buildings. Thus, ponds might only receive a small part of emitted biocides of a connected area. Besides the amount of biocide flowing into a pond, it is important to understand biocide accumulation and degradation within a pond to determine the biocides’ fate over time and the possible biocide retention.

In this study, we aim to determine (a) typical biocide loads from a well-defined urban area and losses on the way to a retention pond and (b) a biocide mass balance for the pond and hence assess its remediation potential.

Our study area is a homogenous, clearly delimited 3-ha residential area of one-family dwellings built during the last 13 years. It represents a typical residential development area of Germany and other European countries. The connected retention pond has limited infiltration capacity. We conducted measurements of diuron, OIT, terbutryn and four TPs in urban stormwater entering the pond. In parallel, we analyzed terbutryn and its TPs in sediment samples over a period of 2 years. We then used the model COMLEAM and the urban rainfall–runoff model RoGeR_WB_Urban (Steinbrich et al. 2021) to calculate continuous terbutryn emissions and pond inflow. Finally, we combined all data to establish a 13-year mass balance of terbutryn in pond water and sediments.

Vegetation differs within the pond: reed grows at Pipe West, and grass at Pipe East. Clearing takes place once a year during winter. At Pipe West, continuous seepage creates saturated soil and small puddles which was already observed during pond construction. Only little infiltration losses can be expected due to the low water permeability values of the predominant loess soils. Their hydraulic conductivity values are in the range of 5.5 × 10⁻⁷–1.6 × 10⁻⁸ m s⁻¹ (Stadt Landau in der Pfalz 2004). The total retention volume of the stormwater retention pond is 625 m³ and groundwater is 8–10 m below ground (Stadt Landau in der Pfalz 2004). The average width and length of the pond are 58 and 21 m. The average depth from the terrain surface is 1.4 m. All weather data were available from a station about 0.6 km west of the pond (Agrarmeteorologie Rheinland-Pfalz 2022), except for wind direction data which were obtained from the Weinbiet weather station of the German Weather Service, located 17 km north.

First, stormwater was sampled at Pipe East and Pipe West during two events to investigate biocide occurrence. Following a two-step approach (Linke et al. 2021), sampling continued after concentrations were found to be above-defined test criteria, here the PNEC value. Then, the campaign was extended to test a possible connection between biocides in stormwater and pond sediments. For this purpose, four campaigns consisted of sediment sampling during dry conditions followed by water and sediment sampling shortly after a storm event, approximately 1 month later (Figure 1).

Water samples were taken as grab samples at both pipes using 1-L brown water bottles previously washed with deionized water. During six events in the years 2019–2021, water was collected in a stainless steel container and then filled into the bottles. For the third and fourth events, an additional sample was collected, about 2 h after the first, to determine intra-event variability. For these events, the mean biocide concentration was calculated. Water samples were kept cool until preparation started, usually within 2 days after sampling. Duplicate samples were taken during the first three events.

Sediment samples were collected by digging a 0.3 × 0.3 m pitch of approximately 0.1 m depth. Vegetation was removed and the sediment was homogenized in a large container. At selected locations, sediment was sampled up to 0.3 m depth in 0.1 m steps. All samples were filled into plastic bags, immediately cooled, and prepared for analysis directly after sampling or frozen if this was not possible. To avoid resampling of disturbed sediments, samples were taken about 0.5 m away from the previous locations. Yet, this was not possible for samples directly at the pipes. A reference sample was taken outside the pond during each sampling campaign. In total, 116 sediment samples were collected (Tables S2–S4, Supplementary material). Figure S2, Supplementary material shows a map of sampling locations.

After the first sampling campaign, soil moisture and redox potential were measured by portable instruments across the pond to assess influencing factors of biocide concentrations in pond sediments (Table S5, Supplementary material). Volumetric soil moisture was determined by an ML3 Theta Probe (Delta-T Devices, Cambridge, UK) connected to the readout device Moisture Meter HH2 (Umwelt-Geräte-Technik GmbH, Münchberg, Germany) and redox potential using a redox probe (Paleo terra, Amsterdam, the Netherlands).

Water samples of event 3 were analyzed for major ions to determine possible groundwater contributions to pond inflow. Concentrations of chloride, nitrate, sulfate, sodium, potassium, magnesium and calcium were determined by ion chromatography (Dionex ICS-1100, Thermo Fisher Scientific Inc., USA).

Grain size, organic matter (OM) and total carbon content were determined for sediment samples taken during campaign 1. The grain size was determined with a Malvern Mastersizer following the procedure outlined by Abdulkarim et al. (2021). The content of OM and carbon were determined using loss on ignition, drying samples at 105 °C and combusting them subsequently at 550 and 900 °C (Heiri et al. 2001).

Stormwater samples were analyzed for three biocides (diuron, terbutryn and OIT) and four TPs (diuron-desmethyl, terbuthylazine-2-hydroxy, terbutryn-desethyl and terbumeton). Table S6, Supplementary material provides an overview of the analyzed substances. Pesticide standards were purchased from Neochema GmbH (Bodenheim, Germany) except diuron-d6 (HPC standards GmbH, Borsdorf, Germany) and OIT (LGC Standards, Teddington, Middlesex, UK). Analysis was performed as shown by Linke et al. (2021). Briefly, samples were filtered, prepared using solid phase extraction and analysed by HPLC-MS/MS (Agilent Technologies, Inc.; 1200 Infinity LC system and 6430 Triple Quad; Waldbronn, Germany). Acetonitrile (LC-MS grade, VWR International GmbH, Darmstadt, Germany) was used for stock solution preparation and as an organic mobile phase in chromatography. Limits of detection (LOD), limits of quantification (LOQ) and recovery rates were assessed by procedures outlined in Linke et al. (2021); Table S7, Supplementary material gives the results.

The preparation of sediments was conducted using an extraction protocol similar to Droz et al. (2021). Sediments were sieved to 2 mm. 5 g of sample was filled up to 50% of water content and put into a centrifuge tube (50 mL, Fisher Scientific, Darmstadt, Germany). To each sample, 3 mL of a mixture of pentane:dichlormethane (3:1 v/v) was added (Carl Roth GmbH + Co. KG, Karlsruhe, Germany). Samples were vortexed for 5 s, put into an ultrasound bath for 5 min, vortexed for 1 min, and centrifuged at 5,000 RPM (Thermo Scientific Heraeus Megafuge 16R). The supernatant was sampled and put into a new glass vial. The previous steps were repeated three times until 9 mL of supernatant was obtained. This supernatant was completely vaporized under nitrogen flow and resuspended in 1 mL of acetonitrile. 75 mg of magnesium sulfate anhydrous and 12.5 mg of PSA Silica were added (Sigma Aldrich, Taufkirchen, Germany). Samples were vortexed for 30 s, centrifuged for 5 min and the supernatant was transferred into a vial. Measurements of two biocides (terbutryn and OIT) and three TPs (terbuthylazine-2-hydroxy, terbutryn-desethyl and terbumeton) were conducted via HPLC-MS/MS using the same method as for the water samples. LOD, LOQ and recovery rates are given in Table S8, Supplementary material.

We chose terbutryn as a sample biocide to continuously model biocide inflow to the pond from the start of pond construction until the end of sampling (01.01.2008–04.11.2021). We used a rainfall–runoff model (RoGeR_WB_Urban) to calculate the amount of stormwater runoff entering the pond and a leaching model (COMLEAM) for the terbutryn load leached from the area connected to the pond. Both models have individually been applied and successfully tested in various urban areas. Our study combined both models to obtain daily terbutryn loads in stormwater entering the pond. Precipitation and temperature data were obtained from the weather station (Agrarmeteorologie Rheinland-Pfalz 2022) and potential evapotranspiration was calculated according to Haude (1954).

RoGeR_WB_Urban is a 1D rainfall–runoff model adapted to urban areas (Steinbrich et al. 2021). We used a 10 min timestep and a 1 m × 1 m resolution to adequately simulate runoff generation and summed up to arrive at daily values. COMLEAM is a model used to simulate the leaching of substances from buildings (Burkhardt et al. 2021). It first calculates wind-driven rain and corresponding facade runoff using data on precipitation, wind speed and wind direction. Thereafter, hourly biocide emission is quantified based on stored emission functions. In our study area, no information about initial terbutryn concentration on individual facades was available. We, therefore, estimated the facade area painted with terbutryn to be 50% of the total facade area excluding windows (3,649 m² in Catchment East and 6,016 m² in Catchment West, respectively). We chose initial terbutryn concentration and emission data from encapsulated biocides and arrived at an initial terbutryn concentration of 2,250 mg m⁻² which was within the expected range of 200–3,000 ppm in dry films of coatings (Burkhardt et al. 2012). The age of the buildings was set to the date of their first appearance in aerial images. At both inlet pipes, we coupled COMLEAM terbutryn emissions with stormwater inflow to the pond calculated by RoGeR_WB_Urban. At both inlet pipes, the obtained terbutryn concentrations were compared to samples collected during six measured events using a Nash–Sutcliff efficiency (NSE) index (McCuen et al. 2006). We optimized NSE to determine a loss factor for the area, i.e. how much terbutryn was lost in the contributing residential area before entering the pond.

Characterization of surface types next to houses helps to understand losses directly at the base of facades. We thus calculated the percentages of partly sealed (i.e. permeable pavements) and unsealed (vegetated soil) areas in a buffer of 1 m around houses based on available GIS data.

We neglected infiltration losses to groundwater due to the low infiltration rates of prevalent soils. In our mass balance approach, inflowing stormwater was assumed to fill up the pond, partly evaporate and spill over into the sewage system. Terbutryn in stormwater was assumed to remain in standing water, infiltrate into the sediment or leave the pond via the outlet. Degradation of terbutryn took place by photodegradation in standing water and by biodegradation in the sediment. Our mass balance approach consists of two storages, namely standing pond water and pond sediment. Both contain water and terbutryn concentrations in daily time steps. Details of individual model calculations are provided in section A of the Supplementary Material.

Measured biocide concentrations compared with other studies, for example, with measurements in urban stormwater infiltration systems for terbutryn, between 3 and 1,800 ng L⁻¹ with median values ranging from 30 to 100 ng L⁻¹ (Hillenbrand et al. 2016). Generally, concentrations at Pipe East were higher than at Pipe West. These differences could partly be explained by more recent buildings in Catchment East or by different biocide amounts and types used in paints and renders. We observed high variability of concentrations between events at both pipes for all substances. This variability was within the expected range since biocide emissions depend on various factors, such as moisture, temperature, remaining biocide concentration in the paint and/or render and on precipitation intensity (Paijens et al. 2019). Most substances showed the highest concentrations during the third event with the highest precipitation (Table S10, Supplementary material). As we did not take flow-proportional samples, our measurements are not necessarily representative of entire events. But for all three biocides, they exceeded PNEC values for surface water at least during one event. Without further dilution, this corresponds to a possible risk for ecosystems and calls for further investigations (Linke et al. 2021).

We detected terbutryn in 85 out of 116 sediment samples, 54% of these samples showed a concentration below 1 ng g⁻¹ (Figure 2). At Pipe East, only 32 out of 57 samples had concentrations above LOD with a maximum concentration of 2.7 ng g⁻¹ in samples collected during campaign 4. At Pipe West, 53 of 59 samples had concentrations above LOD and were taken during different campaigns, 6 had concentrations above 10 ng g⁻¹ (Figure S3, Supplementary material). Flanagan et al. (2021) found higher terbutryn concentration in urban stormwater pond sediments in Sweden (100 ng g⁻¹), although terbutryn was only quantified in one out of 32 samples. The reason for the low concentrations of terbutryn found in our sediment samples could be a lower input. We found the highest concentrations close to the outlet of Pipe West, both before (maximum 26 ng g⁻¹) and after (maximum 25 ng g⁻¹) storm events during various campaigns (Figure 3). Two effects could be responsible for these findings. First, sorption might be higher directly at the pipe since all inflow enters here and terbutryn loads of small events might not reach distant locations. Second, we observed saturated soil during all sampling campaigns and measured low redox potentials (−90 mV) during campaigns 3 and 4, which might have retarded the degradation of terbutryn as observed in experimental studies by Muir & Yarechewski (1982).

Terbutryn-desethyl was the only TP detected in sediment samples and was found in low concentrations (<1 ng g⁻¹) close to both pipes. In the first and second campaigns, about half of the measured samples contained terbutryn-desethyl (63 and 55%), but all samples during the third and fourth campaigns were devoid of this TP. More was found at Pipe West, sometimes even in samples that did not contain the parent compound terbutryn. We explain this observation by terbutryn degradation due to photolysis in standing water (Hensen et al. 2020) or by terbutryn-desethyl formed on and leached from facades (Bollmann et al. 2016). In general, differences in detected TPs during different campaigns might be due to different pre-conditions such as soil moisture and temperature, and due to varying concentrations in inflowing stormwater (Figure 2). Also, Bollmann et al. (2017) frequently detected this TP below recently painted facades in Denmark and showed in a laboratory experiment an accumulation of terbutryn TPs over time. This we could not detect in our study, as our inflow concentration was continuously low. Moreover, we did not detect terbuthylazine-2-hydroxy and terbumeton in sediment, although both are reported to form on facades (Bollmann et al. 2016). We attribute this finding to higher LODs and lower concentrations in stormwater.

Diuron and diuron-desmethyl were not detected at all. OIT was detected in 10 out of 116 (8%) sediment samples, mostly directed at the pipes, both before and after rain events. Two samples at Pipe West contained OIT and eight at Pipe East. This corresponded to collected stormwater samples that contained more OIT at Pipe East. Concentrations of OIT in sediments were all below 1 ng g⁻¹. The half-life time of OIT in the soil is 9 days (Bollmann et al. 2017). This degradation might explain the absence of OIT in most sediment samples. TPs of OIT could not be assessed due to missing laboratory standards.

To check the plausibility of these diffuse biocide losses, we characterized surface types adjacent to the facades. In our GIS analysis, approximately one-third consisted of vegetated soil, i.e. was unsealed, while the remaining part consisted of permeable pavements. See Figure S7, Supplementary material for a picture of typical houses and their surroundings. Also, residential streets that diverted stormwater runoff to the pond were covered by permeable pavements. On these pavements, water and biocide losses may be expected. They depend on surface detentions and infiltration characteristics which are highly variable and change over time (Schaffitel et al. 2020).

Our mass balance approach yielded a maximum terbutryn concentration in pond sediments of 4.2 ng g⁻¹ with mean concentrations of 1.9 ng g⁻¹. The modeled value is close to the mean measured sediment concentration of 2.6 ng g⁻¹ (Figure S6, Supplementary material). Terbutryn loads arriving in the pond were between 0.1 and 0.9 g per ha and year. Wicke et al. (2022) calculated a 3-year diuron emission of 305 g from which only 3.4 g arrived in the sewer system in an area of 1.85 ha. This corresponds to ten times higher diffuse biocide loss but to a similar load of 0.6 g per ha and year compared to the present study.

For our mass balance, we excluded groundwater inflow to the pipes by measuring the ions at samples taken at event 3. Ion concentrations were typical for stormwater and did not suggest any groundwater inflow to pipes (Table S9, Supplementary material).

Our mass balance approach could further be used to separate annual terbutryn amounts for inflow, storage, outflows and processes (Figure 5). We calculated the mass balance for 13 years since the start of construction and included the increasing number of houses. As a result, 11% of leached terbutryn reached the pond. We calculated this reducing factor by comparing measured and modeled terbutryn concentration during six events (see section 3.2.1). Incoming terbutryn partly reached the overflow (6% of leached terbutryn) or remained in the pond to finally infiltrate into the sediment (5% of leached terbutryn). Thereby, terbutryn is degraded by photodegradation in standing water and by biodegradation in the sediment. Photodegradation in pond water amounted to 98 mg per year, 0.9%, and biodegradation in pond sediments to 360 mg per year, 3.5%. During several years, the amount of degraded terbutryn in sediments was higher than influxes due to additions from previous years. Years with constructions in the contributing catchment were periods with the highest terbutryn inflow.

In a 3-ha residential area of single-family dwellings, measured biocide concentrations in stormwater had high variability but were within the expected range. We showed that a combination of a water balance model and a leaching model can be useful to assess biocide fate in such an environment. A comparison of measured and modeled data suggested diffuse losses of 89% of the leached biocide terbutryn on its way to a stormwater retention pond. There, overall low terbutryn concentrations in pond sediments and a 13-year mass balance suggested that only approximately 50% of the inflowing terbutryn was retained and degraded, while the remaining 50% spilled over into the sewage system. Our results are site-specific, and include various assumptions and uncertainties, but suggest that biocide retention in ponds is limited and environmental entry pathways are diverse in urban settings. Thus, end-of-pipe measures for urban biocide mitigation have limited efficiency. To prevent biocide emissions, measures should concentrate on the source.

We would like to acknowledge Lennart Lisy for taking samples in the spring 2020 and the team of sedimentology in Freiburg especially Alicia Sullivan for the sediment analysis. We thank Michael Burkhardt and his team for providing support with the software COMLEAM, Andreas Steinbrich for finding and helping to prepare the GIS data for modeling and Tobias Junginger for helpful discussions. Stadt Landau and David Elsässer kindly provided GIS data and aerial images. Further GIS data were obtained via OpenStreetMap and 3D Rheinland-Pfalz. A student course from the University of Landau recorded the GIS data regarding surface types in more detail. Different soil parameters were provided by Bundesanstalt für Geowissenschaften und Rohstoffe and the digital elevation model by Landesamt für Vermessung und Geobasisinformation Rheinland-Pfalz.

J.L., F.P. and F.L. conceptualized the study; F.L., J.L. and M.B. performed sampling and data visualization. F.L., L.S., O.O., F.P. prepared samples and analysed data. H.L. and F.L prepared the model. F.L. wrote the original draft. All authors wrote, reviewed, edited and approved the final version of the manuscript.

This research was funded by the EU within the European Regional Development Fund (ERDF), to support measure INTERREG V in the Upper Rhine as part of the NAVEBGO project 5.3 (sustainable reduction of biocide inputs to groundwater in the Upper Rhine region). We acknowledge support by the Open Access Publication Fund of the University of Freiburg.

Data analysis and visualization were done in R (Version 4.2.0) using R Studio (Version 2022.02.3).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.