The herbicide Glyphosate was detected in River Havel (Berlin, Germany) in concentrations between 0.1 and 2 µg/L (single maximum outlier: 5 µg/L). As the river indirectly acts as drinking water source for the city's 3.4 Mio inhabitants potential risks for drinking water production needed to be assessed. For this reason laboratory (sorption and degradation studies) and technical scale investigations (bank filtration and slow sand filter experiments) were carried out. Batch adsorption experiments with Glyphosate yielded a low KF of 1.89 (1/n = 0.48) for concentrations between 0.1 and 100 mg/L. Degradation experiments at 8 °C with oxygen limitation resulted in a decrease of Glyphosate concentrations in the liquid phase probably due to slow adsorption (half life: 30 days).During technical scale slow sand filter (SSF) experiments Glyphosate attenuation was 70-80% for constant inlet concentrations of 0.7, 3.5 and 11.6 µg/L, respectively. Relevant retardation of Glyphosate breakthrough was observed despite the low adsorption potential of the sandy filter substrate and the relatively high flow velocity. The VisualCXTFit model was applied with data from typical Berlin bank filtration sites to extrapolate the results to a realistic field setting and yielded sufficient attenuation within a few days of travel time. Experiments on an SSF planted with Phragmites australis and an unplanted SSF with mainly vertical flow conditions to which Glyphosate was continuously dosed showed that in the planted SSF Glyphosate retardation exceeds 54% compared to 14% retardation in the unplanted SSF. The results show that saturated subsurface passage has the potential to efficiently attenuate glyphosate, favorably with aerobic conditions, long travel times and the presence of planted riparian boundary buffer strips.

In the initial phase of the project "Organic Trace Substances Relevant for Drinking Water – Assessing their Elimination through Bank Filtration (TRACE)" the total herbicide glyphosate was classified as highly relevant for further investigations [Chorus & Wessel 2007]. Glyphosate is one of the most widely used and distributed herbicides in the world. Even though it has been on the market since 1974 its use increased with the expiry of the patent at the beginning of the 1990s, in the context of “soil conserving” agriculture (no ploughing) and with the introduction of glyphosate resistant, genetically manipulated cultures like corn, soy beans and cotton wool in 1997. To estimate the occurrence of glyphosate and its main metabolite AMPA in the surroundings of Berlin samples from 22 surface water sites were analysed within this study. In 5 samples the glyphosate concentration was above the European threshold for herbicides of 0.1 µg/L in drinking water. Up to 70 % of Berlin’s drinking water is produced via bank filtration and aquifer recharge characterized by comparatively low flow velocities (< 1 m/d), long contact times (3-6 months) and mainly anoxic redox conditions. To evaluate the potential of bank filtration to protect the drinking water from glyphosate contaminations an experimental study was conducted in the second phase of the TRACE project. Three enclosures at the UBA’s center for aquatic simulations were dosed with three different concentration levels (average concentration: 0.7, 3.5 and 11.6 µg/L) over a time period of 14 days. The effluent was sampled daily for 34 days. Glyphosate and AMPA were analysed applying the HPLC method according to the German Standard DIN 38407-22/2001. In parallel the applicability of the ELISA kit of the company Abraxis was tested without adequate results. The one-dimensional substance transport model VisualCXTFit was applied to obtain substance specific parameters of glyphosate and hydrodynamic parameters of the filter substrate from observed and measured breakthrough curves. The obtained results show that the breakthrough of glyphosate was retarded remarkably (retardation coefficient (R): 18.3 to 25) despite of the initially postulated low adsorption potential of the sandy filter substrate. Also a significant reduction, probably due to degradation was observed (1st order decay-rate (alpha): 0.069 to 0.092 d-1). In addition to the semi-technical scale enclosure experiments laboratory and lysemeter tests were carried out to investigate the processes responsible for glyphosate removal during subsurface passage. The laboratory experiments yielded a KF-value of 1.8998 mgLkg-1 and a Freundlich exponent of 0.4805, from which a retardation coefficient of 53.4 was calculated for a glyphosate concentration of 20 µg/L. Furthermore, delayed degradation under sub-oxic conditions could be observed. The lysemeter experiments ensured no glyphosate breakthrough in the effluent of a 2 m thick column of fine to medium sandy material within 7 months. The data obtained in this project prove that there is a potential of bank filtration to eliminate the herbicide glyphosate: Taking into account that glyphosate concentrations in surface water are highly variable a good protection of the drinking water source by bank filtration especially in respect to peak concentration is ensured. However, adsorption and degradation parameters obtained in the laboratory and semi-technical experiments vary significantly due to the difficulty to imitate natural conditions in the laboratory. Therefore the experimental study of the project TRACE emphasises the need to conduct semi-technical experiments in a near-natural environment to evaluate the risk of contamination.

The Aquisafe project aims at mitigation of diffuse pollution from agricultural sources to protect surface water resources. The first project phase (2007-2009) focused on the review of available information and preliminary tests regarding (i) most relevant contaminants, (ii) system-analytical tools to assess sources and pathways of diffuse agricultural pollution, (iii) the potential of mitigation zones, such as wetlands or riparian buffers, to reduce diffuse agricultural pollution of surface waters and (iv) experimental setups to simulate mitigation zones under controlled conditions. The present report deals with (iv) and evaluates the suitability of the technical scale experimental site at the UBA in Berlin, Marienfelde for simulating processes that impact the fate and transformation of nutrients in wetlands / riparian zones. A 3-month pilot investigation (Sep. to mid Nov. 2007) was conducted in order to assess the impact of vegetation on nitrate (NO3-) removal in slow-sand filters (SSFs) and identifying possible interference of glyphosate with N and C cycling processes in these systems. SSFs are engineered bio-reactors that can mitigate the transfer of a wide range of pollutants including nutrients and organic contaminants to water bodies. Two vertical-flow experimental SSFs (average area: 60 and 68 m2, depth: 0.8 and 1.2 m, respectively) at the UBA facilities in Berlin were used in this study: one unplanted and the other vegetated with Phragmites australis. The SSFs received water amended with nitrate (NO3-) and phosphate (PO4 -) without and with glyphosate (added for 2 weeks). Mineral N concentration at the mixing cell, SSF surface, 40 cm depth and at the SSF outlet was measured at least twice per week to calculate N removal rates. Physical water properties (pH, redox potential, temperature) and greenhouse gas emission (CO2, CH4 and N2O) were also monitored to gain insights into controlling processes. Results showed that N removal rates were several-fold higher in the vegetated than in the non-vegetated SSFs averaging 663 mg N m-2 d-1 (57 % of input) and 114 mg N m-2 d-1 (14 % of input), respectively. In both systems, most of the N removal occurred in the top 40 cm of the SSFs. Marked temporal variation in N removal rates was also detected with rates in general 3 times higher in late summer compared to mid/late autumn. In the latter period, a net release of N was observed in the non-vegetated SSF. The seasonal variation in N removal could be related to a lack of vegetation growth and thus plant N uptake, and may also reflect of the sensitivity of denitrification to climatic factors as suggested by strong (r2 > 0.77) linear relationships between weekly N removal rates and SSF water temperature. A clear impact of glyphosate addition on nitrate concentrations could not be observed. Denitrification, the process most responsible for the removal of nitrogen from waters and soils seems to be unaffected by the addition of glyphosate under the conditions in the experiment. The impact of glyphosate, if any, was probably much smaller compared to the strong influence of temperature on N dynamics in the SSFs. Difficulty of maintaining a constant concentration of glyphosate during dosing may have also contributed to this outcome. Nitrous oxide emission accounted for < 3 % of the total N removed was always lower in the vegetated (< 0.1 - 0.3 mg N2O-N m-2 d-1) than in the non-vegetated SSF (0.2 - 3.8 mg N2O-N m-2 d-1). Conversely, CH4 emission was always higher in the vegetated (range: +0.4 to +49.5 mg CH4-C m-2 d-1) than in the non-vegetated SSF (range: -2.1 to +1.32 mg CH4-C d-1). These results, in connection with much lower oxidation reduction potential readings in the vegetated filter, suggest that the reduction of N2O to N2 was important in the SSF systems and that N2 was the dominant N gas produced. Thus, N2 production must be quantified in order to establish N mass balance of SSF systems. The results show that technical-scale experiments can realistically simulate mitigation systems, while having control over contaminant loading, flow conditions and monitoring. Important lessons learnt for future applications are the following (i) Denitrifying conditions can be established in both SSF of the experimental site by adjusting to low flow conditions (0.23 m³/h) and dosing nitrate. (ii) Dosing of trace contaminants (in this case glyphosate) needs to be improved, but will remain difficult for the large amounts of water involved. The results underline the importance of measurements in the mixing cell. (iii) Since seasonal effects play an important role in mitigation zone performance, any experiments need to be done in parallel, rather than in succession to be able to compare the results.