PREFACE

From building with sludge on the Marker Wadden to identifying microplastics with machine learning

This is the sixteenth edition of Water Matters, the knowledge magazine of specialist journal H₂O. This edition consists on a variety of topics, written by water professionals based on solid research.

The editorial board, consisting of experts from the sector, made a selection based on a clear relationship with daily practice in the water sector, which is the purpose of Water Matters. Research, results and findings form the basis for articles describing new knowledge, insights and technologies with a view to practical application. In this edition, Ruurd Noordhuis, Thijs van Kessel (both Deltares) and Joep de Leeuw (Wageningen University & Research (WUR)) provide insight into the lessons learned from five years of applied research into building with sludge and ecology on the Marker Wadden, a new group of islands in the Markermeer. You will also find articles on the new treatment method with powdered carbon dosing (PAH) and granular activated sludge (Nereda), phosphate removal with iron sand filters, monitoring with DNA metabarcoding, the ecological risks of psychopharmaceuticals and illegal drugs in European surface waters, safe reuse of wastewater and accurate identification of polymers and microplastics with machine learning. Water Matters, like the journal H2O, is an initiative of Royal Dutch Water Network (KNW), the independent knowledge network for and by Dutch water professionals. KNW members receive Water Matters twice a year free of charge as a supplement to their trade magazine H₂O. The publication of Water Matters is made possible by leading players in the Dutch water sector. These Founding Partners are Deltares, KWR Watercycle Research Institute, Royal HaskoningDHV and Stichting Toegepast Onderzoek Waterbeheer (STOWA). With the publication of Water Matters, the participating institutions aim to make new, applicable water knowledge accessible.

You can also read the Dutch version of Water Matters digitally on H₂O-online (https://www.h2owaternetwerk.nl/water-matters).

Would you like to respond? Let us know: redactie@h2o-media.nl

Monique Bekkenutte Publisher (H₂O Foundation)
Huib de Vriend Chairman editorial board of Water Matters

Marker Wadden: lessons from five years of applied research on building with mud and ecology

Between 2016 and 2022 the Marker Wadden were constructed with sand, clay and mud from Lake Markermeer, in the form of seven artificial islands. The goal was to create bird habitat and gradients between deep and shallow waters and dry zones. Moreover, Marker Wadden is intended to reduce the turbidity of Lake Markermeer by sheltering and the capture of mud, thus promoting the ecological development of the lake. The developments in the first five years were followed intensively. In this article we discuss ‘building with mud’ and the development of new nature.

The Marker Wadden are 1300 hectares in size and the distance from north to south is around 4 kilometres. The Wadden are a combination of seven islands with shallow water between them, in an environment that is 4 to 5 metres deep. The islands are built by filling compartments within raised sand ring dykes with soft mud. The sand and clay are extracted from the immediate environment of the islands in sand extraction pits and a mud trap dug for that purpose. On its extraction and transportation, the clay from these pits is mixed with water and pumped into the compartments at a low bulk density (approx. 1200 kg/m³).

Figure 1. Marker Wadden with mud compartments, regular compartments A - E, a mud trap and sand extraction pits (source: KIMA synthesis report. Together, A and B form the freely accessible main island. C, D1, D2 and D3/4 are four non-accessible nature islands. The two E-islands have not yet been completed here.

The programme KIMA [Marker Wadden knowledge and innovation programme], completed in 2022, followed the project’s development. Diverse partners collaborate within KIMA, including WUR, Deltares, NIOO, RWS, Arcadis, Witteveen+Bos and several independent researchers. The results are summarised in a synthesis report (De Rijk & Löffler 2022). In this article, we present a selection of the results in respect of the building with mud and ecology components.

Building with mud

They key question is how to reliably and affordably create a stable substrate for a nutrient-rich lake-marshland situation by using mud. To this end, particularly in the mud compartments (see illustration 1), the consolidation of the mud layer and the settlement of the subsoil were monitored for two years (2019-2021) (illustration 2). The development of the strength of the mud layer and crust formation in the top layer after drying out were also followed, and spatial gradients were mapped in the consolidation, settlement and sediment properties.
The consolidation in the first year is fast, much faster than the settlement of the subsoil. The consolidation in the second year is much slower, not much faster than the settlement of the subsoil.

Figure 2. Elevation of the thin mud compartments, eleven months after the second filling (5 January 2021). Height in m NAP. Markermeer winter level between NAP -0.2 and -0.4 m; summer level between NAP -0.3 and -0.1 m.

The density of the mud package increases with time, but also varies spatially. Close to the filling point there is more sand and the density is higher. After two years, the maximum bulk density of silty deposits is approximately 1600 kg/m³ (975 kg of dry matter/m³) and then stops increasing. Compared to the bulk density during filling of approx. 1200 kg/m³ (325 kg dry matter/m³), this means that the mud volume has decreased by a factor of three, so per 3 m³ of pumped watery mud, 1 m³ of thickened mud is left. This is less than was expected beforehand based on small-scale experiments.
Water level management has a major influence on the process of consolidation and settlement, as it steers the thickness and surface of the crust. Strength develops quickly in this crust and provides sufficient bearing capacity for vegetation development.

This has led to the following conclusions on building with mud:
– The unmixing of sediment fractions during filling (by varying speeds of settlement) causes a spatial variation in the level and composition of the bottom, which is favourable for ecological diversity.
– After settlement and the draining of water, a crust forms which quickly gains strength. The consolidation of the layer of mud below largely takes place in the first few years. The settlement of the subsoil continues longer.
– Water level management is an important steering factor for crust-formation, consolidation and settlement and therefor also for vegetation development.
– The process of consolidation and settlement is easy to predict if the amount of mud and the sediment properties are known precisely. Because these are rather variable from one place to the next, model predictions are uncertain and the desired final level can be the most reliably realised by local ‘adjustments’ by means of extra mud.

Influence of the Marker Wadden on the mud fluxes in Lake Markermeer

The capture of the mud serves a double purpose: 1) the mud can be used for the Marker Wadden (man-made or naturally) and 2) a reduction of the turbidity of Lake Markermeer.
The mud dynamics in Lake Markermeer are strongly governed by the wind and are very variable. A combination of measuring and modelling has therefore been used to determine the effect of the Marker Wadden.
Illustration 3 shows the change determined in the mud concentration in the water column (left) and the amount of mud at the bottom (right). In the shelter of the islands, the mud concentration decreases. There and in the pits, a few decimetres and a few centimetres of mud per year respectively are deposited. On the scale of the entire Markermeer, this means a small but permanent effect (approx. 1-2 mg/l decrease). The mud deposited in the mud trench and the pits can be used for maintenance, i.e. to compensate residual settlement.

Figure 3. The influence of the Marker Wadden on the mud dynamics in Lake Markermeer. Top: reduction of mud concentration in water column. Bottom: change of the amount of mud at the bottom.

Ecology

The key issue in terms of ecology was how a nutrient-rich marshland with corresponding shallows and bank areas can develop and be sustainably maintained as a functional part of the Lake Markermeer ecosystem.
Building with nutrient-rich holocene clay has directly ensured high biological production and fast germination and growth of vegetation. In the past years there was a pioneer situation, with a nutrient-rich reed marsh gradually developing on the islands. The first thing to arise on a large scale was pioneer vegetation, especially including marsh fleawort and red goosefoot on the shallow mudflats and land-water transitions (illustration 4a). In addition, cattail and reed began to grow, whether or not aided by the contribution of seeds or rhizomes, as well as protection against overeating by geese. In this way new habitats have developed with a mosaic of shallows, gradual land-water transitions, mudflats and higher sandy parts. In shallow water, submerged vegetation gradually grew, including pondweed and stoneworts, with a spatial variation in species composition and structure (Illustration 4b). In the water bottoms, communities of benthic fauna arose. The shallow zones in the shelter of the islands with a rich structure and benthic fauna quickly provided an attractive habitat for many species of fish, at least 14 of which used the area as very suitable spawning and breeding grounds (De Leeuw et al. 2021).

Figure 4: Vegetation and habitat development on and between the islands. Left: vegetation charting of the dry parts of the Marker Wadden in 2021. Light brown = bare muddy areas, light green = marsh fleawort, dark green = reed, orange = cattail, purple = other species (Klein Schaarsberg & Ivushkin 2021). Right: example of the presence of an aquatic plant species, the species Zannichellia in the trenches between the islands in 2021. From white to red in increasing density (Kers & Zielman 2022).

High densities of plankton, insects, benthic fauna and young fish are food for larger fish and local breeding birds. From the beginning, colonies of black-headed gulls and common terns settled on the sandy heights and dykes of the compartments, and numerous birds – particularly ducks and waders – found their way to the muddy compartments. In 2021, the total number of breeding bird pairs had increased to around 10,000, divided over almost 50 species. While in the first years the pioneer species of bare soils such as common terns, plovers and avocets were dominant (see the table), in recent years, marsh birds such as bearded reedlings, sedge warblers and water-rails are making their presence felt. Moreover, tens of thousands of birds make use of the islands while migrating or in the winter as foraging and sleeping areas. The birds use the Marker Wadden on a landscape scale, meaning that there is lots of exchange with surrounding areas such as the Oostvaardersplassen and Trintelzand, and that the areas strengthen each other.

The islands also influence the ecology of Lake Markermeer as a whole. The most obvious effect is the addition of habitats and species that were underrepresented as a result of the unnatural depth distribution of the lake by the steep stone banks. Rich habitats such as marshland, mudflats, shallows with aquatic plant species and gradual land-water transitions come about. In these habitats, benthic fauna develops well, and in turn are a major source of nutrition for fish and birds.
The intended positive effect on the food web and the productivity by the improvement of the water quality around the Marker Wadden is less evident, but the work has only just been completed. In the shelter of the islands, new gradients of clear to turbid water arise, in combination with spatially varied patterns of erosion and sedimentation (illustrations 3 and 5). This is favourable for the functioning of the entire ecosystem of Lake Markermeer. Large numbers of fish can gather in the mud traps.

Figure 5. Shelter to the south of the Marker Wadden with wind from the north, 13 May 2019, with the Marker Wadden in the centre of the images. Sentinel satellite image, processed for suspended material (spm) and chlorophyl (chl). The images show reduced concentrations of suspended material (light colour) and chlorophyll (dark colour) (De Rijk & Löffler 2022).

Table 1. Numbers of breeding pairs of gulls, sterns and avocets on the Marker Wadden and the average number of young birds that have fledged per pair (including Dreef et al. 2021; for other sources see De Rijk & Löffler 2022).

What does this mean in practice?

The first five years of the Marker Wadden show that the higher sandy edges provide stability in the archipelago, and breeding grounds for birdlife. The dynamic wetlands with lake-marshland mudflats on nutrient-rich soil ensure high food production, thus forming an ideal foraging area for many birds and a breeding ground for fish. Through connections between the compartments and the surrounding water via trenches and wash-overs, dynamics of water and sediment flows arise between the shallows between the islands and Lake Markermeer.
The results to date also show that ecological development in the mud-built marshland is coming along well. Especially for birds, high natural values have been achieved, which are of national or even international significance in the light of Natura 2000. These developments also signify a major enrichment of the Markermeer ecosystem.
The fact that the vegetation is an early state of succession however, means that the current state of nature there is temporary. The high nature values are partly related to the scarce pioneer situations, which rare species can only make temporary use of. Upon further succession, new habitats arise, in turn offering opportunities for other species. Future developments are uncertain. It is clear, however, that dynamism is required for the sustainable presence of reed marsh and functional land-water interactions, either in the form of water motion (water-level fluctuations), or by specific management. The natural dynamics in Lake Markermeer are restricted by the use of a fixed target water level. This also limits the exchange of nutrients between the Marker Wadden and the lake. As long as the quays of the compartments have not yet been opened, in principle, the compartments can manage their own water level. By (permanently) opening compartments at a specific moment, the exchange of water, mud, nutrients, fish, etc. between Lake Markermeer and the Marker Wadden is possible, whereby the timing of the opening is important for the balance between the consolidation, erosion and sedimentation of mud and the influence of vegetation. Wash-overs in the edges (constructed lowerings) of the compartments form an intermediate solution.
The construction of the islands has also impacted the physical environment. In the sand extraction pits and the shelter of the islands, for instance (with gradients of turbid to clear water) new habitats have arisen for fish, birds and benthic fauna, for instance. However, part of the nutrients from the lake and swamp appear to accumulate in deep pits in the environment of the islands, which could possibly negatively affect the exchange between both sub-systems. This is one of the subjects for a follow-up study in KIMA-2.

Ruurd Noordhuis
(Deltares)
Thijs van Kessel
(Deltares)
Joep de Leeuw
(WUR - Wageningen Environmental Research)

Summary

After five years of research on the Marker Wadden there is now a first overview of the results. This article discusses two aspects: building with mud and the ecology. The construction of the islands with mud from Lake Markermeer took place as scheduled, although the settlement of the mud layer was stronger than expected. In the shelter of the islands and the sand extraction pits, the mud concentration decreases. This means a small but permanent reduction of the turbidity of the whole Markermeer.

The ecological development in the constructed swamps is well underway, which is especially apparent from the abundance of birdlife. In the sand extraction pits and the shelter of the islands, new habitats arise for fish, birds and benthic fauna. This is a major enrichment of the Markermeer ecosystem. The succession continues, meaning that some nature values will turn out to be temporary.

Sources

De Rijk S. & M. Löffler. 2022. SyntheserapportKIMA. De eerste vijf jaar onderzoekop Marker Wadden. Deltares,Utrecht. [KIMA Synthesis report. The first five years’ research on Marker Wadden. Deltares, Utrecht].

De Leeuw, J.J., J. Volwater, O. van Keeken, J. Elings & C. van Leeuwen, 2021. Paai- en opgroeigebieden voor vis in en rond Marker Wadden [Spawning and breeding areas for fish in and around Marker Wadden]. Wageningen Marine Research report C058/21

Dreef, C., J. van der Winden & Y.I. Verkuil. 2021. Broedvogels en pleisteraars op Marker Wadden 2020-2021 [Breeding birds and plasterers at Marker Wadden 2020-2021]. Report 2021-02, Camilla Dreef, Amsterdam.

Kers A.S. & J. Zielman 2022. Toelichting bij de Waterplantenkartering Marker Wadden 2021. [clarification for the charting of aquatic water species in the Marker Wadden 2021]. Rijkswaterstaat, CIV, Delft.

Klein Schaarsberg F.L.H. , and K. Ivushkin, 2021. Monitoring vegetatie ontwikkeling op land. Rapportage t.b.v project WN08 2019 Marker Wadden -Remote Sensing [Monitoring vegetation development on land. Report for project WN08 2019 Marker Wadden -Remote Sensing]. Witteveen + Bos, report 119777/21-017.059, Deventer.

^ Back to start

Removing micropollutants by a combination of powdered carbon and Nereda

The quality of surface water in the Netherlands must improve. One of the necessary measures is removing micropollutants (including medicine residues) from the effluent of waste water purification. To this end a new method has been tested on a demo scale: a combination of powdered carbon dosing and granular activated sludge (Nereda purification technology).

Micropollutants form a threat to water quality. For many of these substances, such as medicine residues, effluent from wastewater purifications is a major route to surface water. Together with STOWA [Foundation for Applied Water Research] and the Ministry of Infrastructure and Water Management (I&W), the water authorities are looking into possible measures to reduce the discharge of these substances into surface water in the IPMV [Innovation Programme for the Removal of Micropollutants at WWTPs]. In this programme, eleven guide substances have been determined (table 1). Nine of these are medicine residues. Continually monitoring the same guide substances in all studies makes it easy to compare the study results. The required removal efficiency is at least 70% for seven of the eleven guide substances.

Table 1. The eleven guide substances in the IPMV

One way to remove micropollutants is dosing powdered carbon into WWTPs, as was demonstrated in the WWTP at Papendrecht in the PACAS project [1]. Here, powdered carbon was dosed in a continuously activated sludge system.

There are also WWTPs that make use of the Nereda^®-purification technology. This is an upwardly flowing batch system with granular sludge. Is powdered carbon dosing also an effective removal technology in combination with Nereda? And does the plant otherwise keep doing what it has to do? To answer this question, in 2021 a study was started in the Nereda plant at the WWTP Simpelveld.
The results of this trial are described in detail in the STOWA report Pilot Powdered Carbon Dosing Nereda Simpelveld [2]. This articles summarises the test set-up and the primary conclusions.

Test set-up

The Simpelveld WWTP of the Limburg Water Authority (WBL) works with a Nereda plant, consisting of two parallel ‘lanes’. One lane purifies one third of the influent, the other two thirds.
In this test, powdered carbon (PAC, the same as in the PACAS-study) is dispensed in the smallest lane, while the other one functions as a reference. The effluents of the reference lane and of the PAC-lane are mixed in an levelling buffer and subsequently guided to the sand filters. For 14 months, the dosing level of powdered carbon was increased step by step from 5 to 10, 15 and 20 mg PAC/l influent. Subsequently, a brief test was carried out with (more sustainable) powdered carbon instead of regular coal, with 15 mg PAC/l.

The powdered carbon dosing took place at the end of the aeration phase, at least 30 minutes before the following phase commenced. The amount of powdered carbon to be added per batch was calculated based on the supply situation. In dry weather, the fixed dosing ratio of 5, 10, 15 or 20 mg/l influent was used. In the event of a rainy weather supply, these amounts were reduced to compensate for the rainwater (which did not, after all, contain any medicine residues).

Monitoring programme

Each dosing period, daily samples were taken (and frozen) of the influents and effluents of the reference lane and the PAC-lane. After checking the working of the WWTP and powdered carbon dosing, for the analysis of the guide substances, the 24-hour samples collected were combined to form 48-hour samples. The effluent sampling was started 24 hours later than the influent sampling. This was in order to take the hydraulic residence time in the reactor into account.
The samples were extensively monitored for: micropollutants (the guide substances), nutrients (P and N), other macro-parameters (COD, BOD, suspended material), sludge composition and substances of high concern such as PFAS. Based on these data, the removal efficiency and the effect of the dosing on the effluent quality, the sludge quality and the (calculated) extra sludge production were determined.
The removal efficiency is calculated by the ‘Provisional work instructions on sampling and the chemical analysis of medicine residues in WWTP wastewater’ of the Ministry of Infrastructure and Water Management, and STOWA. The calculated efficiency is hereby the average value of the purification efficiencies of the best seven guide substances removed in each water sample taken.

Removal efficiency and other effects

The calculated removal efficiency of the medicine residues in the reference lane is an average of 35%. With a PAC dosage, the efficiency rises to 45% for 5 mg PAC/l and to 84% for 20 mg PAC/l (illustration 1). The removal efficiency at 10 mg/l PAC meets the minimum requirement of the Ministry of I&W of 70% exactly.
The sand filters realize an average of 8% extra removal of medicine residues. Further research will be needed to demonstrate whether this is also the case if the entire stream (so also in the reference lane) has been treated with powdered carbon.
The removal of PFAS and heavy metals does not increase as a result of dosing powdered carbon [2].

Figure 1. Removal efficiency of medicine residues in connection with different PAC-dosages, based on the 7 best of 11 guide substances. Number of observations: 5 mg PAC/l n = 3; 10 mg PAC/l n= 5; 15 mg PAC/l n = 6; 20 mg PAC/l n = 6.

The dosing of powdered carbon does not have a negative effect on the (biological) removal of nutrients: the sludge settling properties and the formation of the granular sludge are not influenced by the applied dosing amounts of powdered carbon.

The dosed powdered carbon is captured in the Nereda sludge grains and removed through the excess sludge. The resulting extra amounts of sludge are low. Based on a dose of 15 mg PAC/l, sludge production on the scale of WWTP Simpelveld will rise by approximately 19 tons ds/j, which is approximately 5%.

The possibility of powdered carbon being washed out with the WWTP-effluent has been examined by the so-called Schwarzgradbestimmung. This method was used in Germany and Switzerland [3, 4] and is now also prescribed by STOWA as the standard (publication is in preparation). This method (detection limit 1 mg/l) has not detected any leaching of carbon. At WWTP Simpelveld, the amount of carbon not captured in the sludge is probably largely removed in the sand filters.

The raw material Kaumera can be recovered from the Nereda sludge grains. The possibility of powdered carbon having an effect on this has therefore been looked into. The tests with excess sludge and reactor sludge showed that this is not the case: in both lanes, both types of sludge produced almost the same amounts of Kaumera (efficiency coefficient of the PAC-lane was 25.7%, while the reference lane was 27.5%).

Powdered carbon from pine wood

At the end of the trial period, testing of sustainable carbon produced from pine wood was carried out for one month (Acticarbone 2SW, Chemviron). The removal efficiency by a dosage of 15 mg/l does not significantly differ from the previously tested powdered carbon (Pulsorb WP 235): 74% and 76% respectively based on the seven best of eleven guide substances (illustration 2). The low density of the Acticarbone (180 kg/m³) made it less easy to dose with the pilot set-up.

Figure 2. Removal efficiency based on the 7 best of 11 guide substances when using sustainable PAC compared to fossil PAC with a dosage of 15 mg/l. Number of observations: with fossil coal n = 6; with sustainable carbon n = 4.

Promising technology

This study shows that micropollutants can be effectively removed in a Nereda-plant by powdered carbon dosing, with negligible negative side-effects. The removal efficiencies required by the IPMV are realised using a dosage of between 10 and 15 mg PAC/l influent. The required powdered carbon dosing can be set very accurately. The sustainable powdered carbon had a lower density than the fossil coal and was not as easy to dose. In the case of a full-scale application, the silo dosing plant must be suitable for coals with different properties.
An additional advantage is that in the same way as in this test, the plant can be used as a plug-and-play module, and is thereby easy to integrate into a modular building concept. As a result, the technology is also suitable for other Nereda-plants where specific medicine residue removal is necessary, the so-called hotspot WWTs [5].

Acknowledgments

The WBL thanks STOWA and the Ministry of Infrasctructure and Water Management for the financial support awarded to the pilot hereby described.

Sandra Malagón
(WBL)
Ad de Man
(WBL)
Wout Pannekoek
(WBL)
Herman Evenblij
(Royal HaskoningDHV)
Saskia Moll
(Royal HaskoningDHV)
Pascalle Deenekamp
(Royal HaskoningDHV)

Background picture:
Powdered carbon dosing unit (right) and powdered carbon stock (left) at WWTP Simpelveld

Summary

In the IPMV, together with STOWA and the Ministry of Infrastructure and Water Management, the water authorities are looking into possible measures to reduce the discharge of these substances into surface water. According to the study described here, the combination of powdered carbon dosing (PAC) and the Nereda purification technology is promising. The removal efficiency is a good deal higher than without the powdered carbon; with 10 mg/l PAC it precisely meets the minimum 70% requirement for seven of the eleven guide substances. The powdered carbon dosing does not have any significant negative side-effects on the (biological) removal of nutrients, the amount of sludge or the yield from Kaumera recovery.

Literature

1. STOWA, 2018; PACAS poederkooldosering in actiefslib voor verwijdering van microverontreinigingen [PACAS powdered carbon dosing in activated sludge for the removal of micropollutants]. Research on effectiveness and efficiency at WWTP Papendrecht (STOWA, 2018-02)

2. STOWA, 2023; Pilot Poederkooldosering Nereda Simpelveld (STOWA 2023-02) in voorbereiding [Poederkooldosering Nereda Simpelveld (STOWA 2023-02) pilot in preparation]

3. Platz, 2015; S. Platz; Charakterisierung, Abtrennung und Nachweis von Pulveraktivkohle in der Abwasserreinigung, PhD-thesis, Universität Stuttgart.

4. Metzger, 2011; S. Metzger, A. Rössler, H. Kapp; Optimierung der Pulveraktivkohleabtrennung durch Filtration als Grundlage zur Anlagendimensionierung; Abschlussbericht, Hochschule Biberach.

5. STOWA, 2017; Landelijke Hotspotanalyse Geneesmiddelen RWZI’s; STOWA 2017-42 [National Hotspot Analyse of Medicines at WWTPs; STOWA 2017-42]

^ Back to start

DNA metabarcoding provides a clear impression of the water quality

To improve the surface water quality and aquatic biodiversity, many restoration measures are performed. But do these sufficiently improve the ecology? And if not, what extra measures are required? These questions are hard to answer, given that specific monitoring of the effects is often lacking. Wageningen Environmental Research examines which new measuring techniques can be used to obtain more insight into the effects of restoration measures.

A specific monitoring programme to measure the effectiveness of restoration measures is often lacking, and existing monitoring networks generally fail to take measurements at the right locations and times. Scaling up the monitoring with the currently used technologies is, however, a labour-intensive and expensive affair. Consequently, adapted or new monitoring technologies are required to cost-effectively scale up sampling in a (sub)catchment. But with that, we’re not there yet. The next step is to convert the data into a meaningful diagnosis of the potential bottlenecks in the water system. This step is also necessary for conventional monitoring.

The aim of this study was to determine whether DNA metabarcoding of macroinvertebrates can generate lists of species that correspond to the species that have been found in the area (regional species pool). This would make it possible to use the method to diagnose ecological bottlenecks, as well as assess the effectiveness of the restoration measures carried out. The study was performed in the Run, a stream in the province of Brabant.

Why DNA metabarcoding of macroinvertebrates?

Macroinvertebrates in a stream, including amongst others dragonfly larvae, water beetles, leeches and water mites, reflect the state of the water. Each species has a preference for a certain environment, such as a specific range of flow velocity, nutrient richness and temperature. These are referred to as environmental and habitat preferences [1]. The environmental and habitat preferences of the species found in a certain waterbody can be used to calculate stress scores and bottlenecks. Hence, it is possible to determine whether restoration measures have led to an ecological water quality improvement based on the macroinvertebrates in the water.

However, identifying all the species in the traditional manner, using determination keys, is labour-intensive [2]. It requires extensive specialist knowledge of the morphological characteristics of the species concerned. Moreover, it is not possible to identify all the collected specimens to species level, given that the characteristics of young or non-sexually mature specimens are often insufficiently developed. This means that it is sometimes necessary to visit a location twice, in two different seasons.

Molecular technologies, particularly DNA Metabarcoding, may be able to circumvent these issues. A barcode (a characteristic piece of DNA-profile) [3] is available for more than 90% of the macroinvertebrate species used to determine the ecological status for the Water Framework Directive (WFD). In this study, two methods to collect DNA have been tested. The first is to isolate the DNA directly from the collected organisms (total-DNA). The second is to filter the DNA expelled by organisms in the water through cell residues, mucus and excretion. This DNA is referred to as ‘environmental DNA’ or in short eDNA. These technologies are not new in themselves, but are still being optimised.

The Run: degradation and restoration

The study area (the stream the Run in the province of Brabant) has greatly changed in the past: the flow-through marsh was changed into an oversized, straightened agricultural waterway. Furthermore, agricultural intensification led to a strong increase in pollution by fertilisers and pesticides used in the surrounding stream valley. These changes in and around the stream have led to significant deterioration of the water quality.

To improve the ecological water quality and make the stream valley ‘climate robust’ by 2021, water authority The Dommel commenced the restoration of more than three kilometres of the stream at nature reserve Grootgoor. Here, the Run has been made narrower and shallower, and meanders again through the landscape like it did in the past. A larger area surrounding the stream has been allocated to buffer heavy rainfall and to counteract drought.

Collecting and analysing data

In May 2022, 28 measuring sites were sampled in the restored section of the Run, as well as upstream and downstream. Macroinvertebrates were collected by sampling 1.5m of the dominant habitat types present at each site using a standard macroinvertebrate net. For eDNA, a sample was collected at each site by filtering 0.5–0.75 litres of water (Silphium eDNA Dual Filter Capsule, pore size 0.8 µm).

In the laboratory, the organisms from the macroinvertebrate net sample were first sorted and counted using a macroinvertebrate quickscan method. Specifically, the macroinvertebrates were identified to genus (mayflies, stoneflies, caddisflies, water beetles), order (water mites, freshwater bristle worms), and family level (all others). Subsequently, the total DNA of the macroinvertebrates was extracted (using the PowerMax Soil kit by Qiagen). DNA metabarcoding was carried out on both the eDNA and the total DNA sample using the LeRay primer set [4]. For two sites, the DNA-metabarcoding of the eDNA sample failed due to the large amounts of filamentous algae, therefore, these sites were not considered in the comparison.

Calculating the bottlenecks

The bottlenecks or stress scores for the macroinvertebrates were calculated based on their environmental preferences [5]. For the quickscan samples, the environmental preferences of all species that could be present within the same WFD water type were combined, and calculations were made with log10(x+1)-transformed number of individuals. The eDNA and total-DNA data was scored based on the presence/absence at species-level.

Comparison of methods in identification of species

The presence of 160 macroinvertebrate species was established by the eDNA samples from the water (Figure 1). The results give a limited impression of the community present, because the eDNA was mainly derived from non-biting midges, flies and freshwater bristle worms. Many species that indicate flow (certain species of dragonflies, true bugs, water beetles and water mites) were missed as they expel little DNA in the water. On the other hand, the total-DNA (DNA in the macroinvertebrate samples) provides, with 230 different species, a detailed impression of the species richness in the Run. The majority of the species found were also classified at a higher taxonomic level (e.g. family) by the quickscan method. Moreover, the total-DNA results are comparable to the composition of the regional species pool (270 species, based on all water authority data available of the Run). A large number of taxa that are hard to identify morphologically, including freshwater bristle worms and non-biting midges, could be identified by species using total-DNA.

Figure 1. The number of macroinvertebrate species per group identified in the Run with only eDNA from water samples (blue), only total-DNA of the animals in the macroinvertebrate samples (orange), or with both methods (green).

Usability of the bottleneck analysis

Next, stress scores were calculated based on the species found at each site with each method. In this article we focus on the stress-score for flow. High stress levels indicate the (temporary) reduction or absence of flow, which is undesirable in streams. In the restored section of the Run, measures were taken to improve the flow conditions. One may expect to find more species with preference for high flow at these sites, and that the stress score would be lower than in the unrestored stream sections. Both the quickscan and the total-DNA method confirm this, but the total-DNA method provides a more positive assessment over a longer trajectory than the quickscan method (Figure 2). This was to be expected, as a larger number of species is identified with the total-DNA method. A more detailed identification of the quickscan samples to species level would most likely have a similar result. However, such a detailed identification requires a considerably greater time investment. The stress scores based on the eDNA method varied negatively from the quickscan method, as many species that indicate flow were not detected, which distorts the picture.

Figure 2. Stress scores for flow, calculated by data from the quickscan (left), total-DNA (middle) and eDNA (right) method. The restored section where the flow conditions have been improved is marked by a pink dotted line (roughly in the middle of each Figure).

Conclusions

This study shows that the eDNA sampled from the water provides limited information about the macroinvertebrate community in the Run, but that the total-DNA from the macroinvertebrate samples provides a very complete impression of the presence of different species. The information from total-DNA method can be used to make a diagnosis of the ecological bottlenecks in the stream. The large amount of information at species level that can be simply and quickly obtained for a large number of measuring points with this method is a major advantage compared to traditional identification based on morphological characteristics by a macroinvertebrate specialists. The total-DNA method makes it possible to determine the (sub)catchment-wide effectiveness of restoration measures with limited resources.

Acknowledgments

The study in the Run was realised in collaboration with Aquon and MicroLAN, and was supported by the Ministry of Agriculture, Nature and Food Quality, in connection with the Knowledge and Innovation Agenda (KIA); theme: Agriculture, Water, Food; mission C: Climate Proof National and Urban area).

Gea van der Lee
(Wageningen Environmental Research)
Marcel Polling
(Wageningen Environmental Research)
Ralf Verdonschot
(Wageningen Environmental Research)
Iris van der Laan
(Water Authority The Dommel)

Background picture:
The stream restoration project of the Run at Grootgoor (photo: Matthijs de Vos)

Summary

New monitoring technologies may be of great value to learn more from restoration measures that are caried out in future projects. This study examined the use of DNA-metabarcoding of macroinvertebrates in the Run, a stream that was recently restored. Only a limited part of the macroinvertebrates present could be detected by the eDNA from the stream water. On the other hand, the total-DNA of all animals in the macroinvertebrate samples provides a complete impression of the community. This completeness is an excellent basis to analyse the ecological bottlenecks in the stream. The total-DNA method provides a detailed impression of the sites where the lack of flow forms a bottleneck in the stream. The traditional approach (identification based on morphological characteristics) may provide the same picture, but requires a greater time investment. DNA-technologies thereby make it possible to identify the effectiveness of restoration measures at catchment scale with only a limited use of resources.

Literature

1. Verberk, W.C.E.P., Verdonschot, P.F.M., Van Haaren, T. & Van Maanen, B. (2012) Milieu-en habitatpreferenties van Nederlandse zoetwater-macrofauna [Environmental and habitat preferences of Dutch freshwater macroinvertebrates]. STOWA

2. Bijkerk. R (red) (2014) Handboek Hydrobiologie [Hydrobiology Manual]. STOWA.

3. Van der Hoorn, B. & Beentjes, K. (2020). Genetische biomonitoring van macrofauna [Genetic biomonitoring of macroinvertebrates] H2O online.

4. LeRay, M., Yang, J.Y., Meyer, C.P., Mills, S.C., Agudelo, … & Machida, R.J. (2013). A new versatile primer set targeting a short fragment of the mitochondrial COI region for metabarcoding metazoan diversity. Frontiers in zoology, 10(1), 1-14.

5. van der Lee G.H., Bakker, A.M., Verdonschot R.C.M., Verdonschot P.F.M. (2021). Doorwerking van lokaal beekherstel op de ecologische kwaliteit van het hele stroomgebied [Impact of local stream restoration on the ecological quality of the entire catchment area]. Water Quality Knowledge Impulse (KWIK) Memorandum.

^ Back to start

Assessment of chemical water quality for safe recycling

Climate change is causing longer dry periods and a growing demand for water. This makes the future availability of sufficient freshwater of good quality a huge challenge. A circular approach could provide a solution: the recycling of purified sewage water or industrial wastewater can (temporarily) compensate for shortages. How can we improve guarantees that water is safe for recycling?

At present, the ‘water balance’ in the Netherlands is unhealthy. 977 Mm3/year more groundwater is drained than is supplemented naturally. At the same time, wastewater treatment plants (WWTPs) and industrial wastewater treatment plants in the Netherlands discharge 2160 Mm3/year into the surface water [1]. Recycling this effluent could restrict desiccation and help meet the demand for water.
In principle, water can be recycled for many different uses. Examples are drinking water, agricultural irrigation, cooling, (industrial) process water, nature, recreation and supplementing groundwater.

Water recycling policy

However, recycling water also raises questions. Water may be polluted by biological and chemical components from previous use. This may make it unsuitable for recycling. However, it is striking to realise that we generally discharge purified wastewater into surface water. Downstream, this surface water is then used for various purposes, such as irrigation [2]. Actually, this means that purified waste water is already being ‘reused’. However, any risks that this reuse entails are currently not structurally identified, nor are they minimised. Irrigation by means of WWTP-effluent is subject to a European directive [3]. The directive provides for permit requirements, whereby risks must be identified and monitored. For other forms of water reuse, national and/or international policies or regulations are still lacking. For example infiltration in the soil, where monitoring the water quality could be important to preventing negative effects on the ecosystem. Quality requirements and the required legal framework for water reuse in the drinking water sector [4] have been considered in the research programme Water in the Circular Economy (WiCE).

Assessment of water quality for responsible reuse

Chemical safety plays an increasingly important role in reusing water from alternative sources, besides microbiological safety. A clear picture of the water quality is required to assess what purification efforts need to be made to make water suitable for responsible recycling.
However, assessing the chemical water quality is a special challenge due to the extremely high number of (often unknown) micropollutants that may be present in used water. Information about the origin, possible (mixture) toxicity and the fate (‘where do the substances end up in the environment’) of micropollutants is often lacking. The usual method for assessing chemical water quality often focuses on a limited number of substances (target substances), whether the monitoring of drinking water, surface water, purified WWTP-effluent or industrial wastewater is concerned. Given the large number of potential micropollutants, an incomplete picture may therefore be given. Presumably, this is too limited for a sound assessment of safety in the event of recycling. But there are more analysis methods available, namely target analysis, non-target screening and bioassays. A combination of these methods could increase confidence in the outcome of a water quality assessment (Figure 1).

Figure 1. Methods for water quality analysis may produce more results if they are applied in combination

Using innovative tools to discover relationships between substances and trends over time, and using new methods for toxicological determinations (for instance QSARs and machine learning), it is possible to prioritise smartly while still getting a grip on which micropollutants are the most important. Although the specific substance selected for targeted analysis are often carefully chosen, the drawback is that this approach fails to identify all the substances potentially present. In fact, the selected substances might not the most representative in every case for potential risks to people and the environment.

Combining analysis methods

In the chemical-analytic area, suspect screening and non-target screening (NTS) have really taken off. These technologies enable us to detect an extremely broad spectrum of substances. In just one analysis, suspect screening can identify many expected pollutants in water (due to be reused). Using non-target screening it is even possible to detect as yet unknown chemical substances in reusable water, and to provisionally identify these without reference material. Using data analysis, additional information can be obtained from the results, for instance about sub-structures, estimated concentrations or the formation of transformation products. For new unknown substances, knowledge about (sub)structures can be used to estimate possible toxicity by quantitative structure-activity relationships (QSARs) or recently developed machine learning models. Estimated concentrations can be tested against substance-specific safe concentrations or generic risk thresholds (threshold of toxicological concern; TTC). However, for more reliable testing against limit values, more accuracy is required regarding the concentration. This is not always easy to provide by means of suspect and NTS. A follow-up step could also be to confirm what substance is involved by means of target substance analyses, and to determine the precise concentrations.

Figure 2. Placing samples into the high resolution liquid chromatograph, for chemical analysis of a spectrum of known and unknown organic compounds as wide as possible

Biological test systems (bioassays, Figure 3) give insight in the combined effect of all substances present – whether or not detected chemically – on biological systems. A choice must be made as to which possible effects are looked into, as active substances may have different effects. Frequently, the bioassays chosen are those that are the most relevant for human health. There are also bioassays that are focussed on ecotoxicological parameters; these identify environmental risks. An increasing number of bioassays have so-called trigger values or signal values, which indicate when a response is so high that there might be a risk for people or the environment. Because harmful substances can nevertheless be present in the water in the case of a negative response in a bioassay, combining them with chemical analyses is important.

Figure 3. Bioassay in process

If chemical analyses or bioassays show non-acceptable exposure or risks for people and the environment, it is necessary to take measures. For some forms of water recycling, estimating risks for people and the environment is not simple. This is especially the case if the characteristics of substances indicate that they may also end up in food consumed by animals (cattle, wild animals, fish) or in crops. This broadens the issue from water quality to food safety. To ensure effective measures and set up routine monitoring, it is therefore necessary to understand the origin and the fate of the substances detected, including trends and conversions in the environment. Forensic methods are available for this purpose.

Water recycling: pilots and full scale applications

In recent years, multiple recycling pilots have been carried out. Below, three of these are described where WWTP-effluent and rainwater are reused as industrial water and for irrigation. In these pilots, diverse analysis methods were used to obtain insight into water quality, and we describe opportunities to obtain a yet more complete image.

1. Sub-irrigation with WWTP-effluent

Studies have been performed in Haaksbergen on irrigation with purified WWTP-effluent from an underground infiltration system. This pilot study examined the extent to which micropollutants in the soil are removed and can be washed into the groundwater. To this end, for several years, the groundwater was analysed for organic micropollutants at different depths and locations [5]. More than 100 substances were measured, including transformation products. Also, their combined (mixture) ecotoxicity was calculated. A more complete picture thus arose of the distribution of the substances when purified effluent is used for irrigation. The results show that the irrigation of the soil does not just increase water availability during drought, but also that a considerable part of the organic pollution remains in the soil. Of the 133 substances, 89 were found in the field. Bioassays could provide more information about the possible effects of these remaining substances.

2. WWTP-effluent and rainwater as process water

Dow (previously: Dow Chemicals) in Terneuzen uses considerable amounts of process water. For this purpose, among other things, further purified effluent from WWTP Terneuzen is used. From 2019 to 2021, a pilot was carried out to examine whether this can be supplemented by rainwater [6]. An important question was which substances are present in the various sources, how these substances behave in the process steps and whether they can form an obstacle for reuse as process water. Part of the study was also on the usability of ‘constructed wetlands’ for purification prior to reuse. Suspect analyses and NTS-analyses were especially used to screen more than 2,000 relevant compounds (including pharmaceutical products and pesticides). Some of these could not be identified in the samples measured with complete certainty, but nevertheless with sufficient certainty to give a general picture of the total burden. The suspect screening, for instance, showed the presence of the pharmaceutical metformin, as well as substances normally found in the effluent of WWTPs and not in the effluent of industrial wastewater treatment plants. Through NTS-analyses, the differences between influent and effluent samples were also identified. The location of specific substances in the Dow-process and of substances that are still unknown are could thus be seen. Insight into the suitability of the influent water as process water has been improved. It could also be important to examine the impact on the environment (in the ‘constructed wetlands’). Bioassays could provide more information about the potential (eco)toxicity of these compounds. This would enable an even more extensive assessment of the water quality.

3. WWTP-effluent or pre-purified surface water for high-quality recycling.

One example where, in addition to measurements of substances, bioassays have been used is the pilot study ‘Closing the Water Cycle in North-Holland’. This pilot aimed to examine whether WWTP-effluent or pre-purified surface water are suitable for high-quality recycling, for instance as irrigation water or even drinking water. Here, the effectiveness of various supplementary purification technologies for effluent of a WWTP plant were studied [7] by combining two techniques. The removal of eleven selected pharmaceuticals (also referred to as guide substances) was monitored by targeted analysis. The results were compared to outcomes of bioassays. The bioassays showed that when using advanced purification technologies, in most cases the levels of substances did indeed decrease. However, in some cases, hormone-disrupting activity, genotoxicity and oxidative stress increased. Consequently, additional measures are required for high-quality reuse. Using this combination of technologies (targeted analysis and bioassays), it is possible to determine the extent to which the guide substances are representative for the removal of a relevant toxicological effect. In this way, the pilot compares the quality of purified effluent to that of pre-purified surface water. The outcomes of the combination of technologies suggest that ozone treatment should be adjusted, in combination with other purification techniques, to enable high-quality recycling. The pilot therefore gives a more comprehensive picture of the improvement of water quality after purification than targeted analyses can give on their own.

Figure 4. Sampling groundwater from several depths (between groundwater level and twelve metres below ground level), on a plot in Haaksbergen. WWTP effluent has been infiltrating from drains here for several years. Monitoring well contains microfilters.

Conclusion

The described studies on the recycling of water require a combination of technologies to provide a broader picture of the number and the identity of (harmful) substances in water to be recycled. The pilot in North-Holland shows by additional research (bioassays) that the toxicity of these substances can also be determined, but this approach is not yet frequently used. We are therefore calling for a new framework for the assessment of the (chemical) water quality in the context of water recycling. We have shown that chemical and biological technologies are easy to apply. The technologies described here are available in various laboratories, both within and outside the Netherlands. However, to carry out the analyses, interpret the results and assess water quality requirements, specific expertise is necessary. It is therefore important to involve various experts in water quality assessment for water reuse. For a complete framework in which microbiological water quality is also assessed, experts in this area must also be brought in.

Summarising, in combination with knowledge of the origin and properties of the pollution, advanced chemical-analytical, biological and data-analysis technologies offer possibilities to properly assess water quality. This is necessary to consider possibilities for reuse and if necessary, to select risk-limiting measures.

An understanding is thereby required of the origin and the fate of the detected substances, including trends and conversions in the environment. Here too, different technologies supplement each other. Together, they give a more complete picture of the chemical water quality. This increases the reliability of the assessment of the water quality and reduces the likelihood of unexpected water quality issues.

Frederic Béen
(KWR Water Research Institute)
Milou Dingemans
(KWR Water Research Institute)
Nienke Koeman
(KWR Water Research Institute)
Stefan Kools
(KWR Water Research Institute)
Thomas ter Laak
(KWR Water Research Institute)

Background picture:
Sampling at sub-irrigation location Haaksbergen

Summary

Climate change causes longer dry periods and a growing demand for water. Recycling water, for instance for agricultural irrigation, cooling, (industrial) process water or nature (groundwater recharge) can help to meet the growing demand. The water for recycling can be rainwater collected, but can also come from sources such as WWTP-effluent or industrial waste water.

The contamination of water that is to be reused can lead to risks, depending on the intended use. A thorough assessment of the risks and the substantiation of the selection of risk-restricting measures, if any, are necessary. To this end, the usual analysis technologies to determine water quality produce information that is too limited. This article proposes using a combination of analysis technologies which together, give a broader and more reliable picture.

Literature

1. Pronk, G.J. Dooren, T.C.G.W. van Stofberg, S.F. Bartholomeus (2020). Waterhergebruik en de Zoetwatervoorziening [water reuse and freshwater provision] https://library.kwrwater.nl/publication/60884959/.

2. Jack E. Beard, Marc F.P. Bierkens, Ruud P. Bartholomeus. 2019 Following the Water: Characterising de facto Wastewater Reuse in Agriculture in the Netherlands. Sustainability | Free Full-Text | Following the Water: Characterising de facto Wastewater Reuse in Agriculture in the Netherlands (mdpi.com).

3. European Commission 2020. Minimumeisen voor hergebruik van water [minimum requirements for water reuse. https://eur-lex.europa.eu/legal-content/NL/TXT/PDF/?uri=CELEX:32020R0741&from=NL.

4. Krajenbrink, H.J.Handgraaf, S.Koeman-Stein, N.E.Cirkel, D.G.Stofberg, S.F. 2022. Juridisch kader aanvulling watersysteem met industrieel restwater [legal framework for the recharging of the water system with industrial wastewater. https://library.kwrwater.nl/publication/69265309/.

5. Narain-Ford, D. M., A. P. van Wezel, R. Helmus, S. C. Dekker and R. P. Bartholomeus (2022). "Soil self-cleaning capacity: Removal of organic compounds during sub-surface irrigation with sewage effluent." Water Research 226: 119303.

6. Ioanna Gkoutzamani, L. Wyseure, T Steenbakker, N. van Belzen, A. de las Heras Garcia, O. Schepers; Wetlands & hybrid desalination at Dow Terneuzen Technical report of pilot study April ’19 – August ’21, 2021-10, 2021.

7. Bertelkamp, C., Dingemans, M.M.L. Roest, K. Hornstra, L.M. Hofman-Caris, C.H.M. Reus, A.A. (2020) TKI Sluiten watercyclus Noord-Holland https://library.kwrwater.nl/publication/61261355/

^ Back to start

Removal of phosphate by iron-coated sand filters: what works and what doesn’t?

One of the causes of the poor water quality in the Netherlands is the high levels of phosphate and nitrates. A major part of the phosphorus is due to agricultural crop fertilisation. Iron-coated sand (ICS) can bind phosphate, thus preventing it from running into rivers and lakes. What requirements must a good iron-coated sand filter meet?

Part of the phosphate from phosphorus fertilisation is washed from the soil into rivers and lakes, where high phosphate concentrations lead to eutrophication. This is subsequently the cause of, among other things, algal blooms. To improve the ecological quality of the surface water it is important to reduce the diffuse phosphorus coming from agriculture. One way to achieve this is to implement mitigation measures such as iron-coated sand filters. Iron-coated sand filters are placed around drainage pipes can thus prevent phosphate from running into drainage water [1, 2].

The ability of iron-coated sand filters to bind phosphate is influenced by adsorption kinetics and redox-conditions. More specifically: what is the effect of the residence time and flow rate on the adsorption capacity of the ICS? And what happens if the filter is permanently under water and the conditions are reducing.

Figure 1: Schematic of iron-coated sand filters around drainage pipes to remove phosphate. The most important problems that are expected to influence the efficiency of phosphate adsorption are shown in a) a redox gradient and in b) short retention times (kinetic effect) caused by heavy rainfall or pumps.

The experiment

The fieldwork for this study was performed in two fields in the Flower Bulb Region, an area with intensive agriculture on sandy soils. In 2018, iron-coated sand was applied around the drains here as an experiment. After that, the filter material was under groundwater level for long periods. The effect of iron reduction on phosphate retention was studied in two trial fields and with microcosm experiments in the laboratory. Furthermore, column studies were performed in the laboratory to determine the adsorption kinetics [3, 4].

Because drainage pipes are under groundwater level for a great part of the time, the oxygen supply is limited. In combination with the presence of quickly degradable organic material, a reduced environment can arise as micro-organisms use nitrate, manganese, iron (hydr)oxides, sulphate and/or CO2 as terminal electron acceptors (anaerobe respiration) (Figure 1a).

Results

We saw that three-year old iron-coated sand filters still remove phosphate efficiently, despite the iron and manganese reducing conditions. We also tested the effect of the deterioration of the conditions on the stability of the iron-coated sand and the adsorption of phosphate. The laboratory conditions were comparable with those in the field. For iron-coated sand with a molar P/Fe-ratio of 0.013 and exposure times of 45 days, the phosphate remained bound to the iron-coated sand, even after the partial dissolution of iron and manganese under weakly, moderately and strongly reducing conditions. Almost no changes in the phosphate adsorption capacity or the mineral structure of the iron-coated sand were observed. Under strongly reducing conditions, iron and manganese were released and low percentages of iron sulphides were formed in the outer layer of the iron-coated sand, but without any phosphate being released.

The flow speed in the filter influences the retention time and the kinetics of phosphate adsorption in the iron-coated sand filters (Figure 1b). The adsorption of phosphate to iron-coated sand takes place in two stages: fast adsorption on the surface of the grains, followed by slow kinetic adsorption in the micropores in the iron coating.

We calibrated and validated a reactive transport model for the relationship between transport by flow (advection) and adsorption in order to describe the kinetics of phosphate adsorption in iron-coated sand. The parameters are based on long-term column experiments with two flow speeds, and successfully describe adsorption and desorption processes, even in stop-flow conditions. It is thus possible, for instance, to model the effluent phosphate concentrations with varying flow speeds and stop-flow. Any desorption is also calculated. [4]. (Our model code is freely available in GitHub and can be implemented in coding programme R for the design of new filters or the verification of existing ones). The lifespan of the filter is determined by the time required for the diffusion of phosphate into the inner layers of the iron coating. Only approximately 5% of the adsorption capacity is fast an takes place outside of the iron-coated sand grains. The remaining 95% of the retention capacity is deeper, inside the iron coating. Reaching equilibrium in the iron coating by diffusion is a slow process that takes around 300 days.

Table 1. Properties of iron-coated sand filters

Implications for filter design

We recommend that iron-coated sand filters placed under ground water level are designed with a low molar P/Fe-ratio [3]. Under the studied conditions, we expect that the phosphate release is not relevant within the lifespan of the drainage pipes, approx. 20 years. The formation of iron sulphides is reversible: if the material is exposed to (water with dissolved) oxygen, the reduced iron is oxidised again to form iron oxides. The sensitivity of iron to redox conditions can result in restrictions in the use of iron-coated sand or iron sludge under strongly reducing conditions. Using iron-coated sand may not be the best option if the sulphide concentration in the drainage water is high or if there are signs of sulphate reduction.

Under strongly reducing conditions, manganese and iron oxide partially dissolve in iron-coated sand. Manganese generally dissolves before iron, so manganese concentrations can be measured as an early warning for reduced conditions [4].

To make use of the total phosphate retention capacity, it is important to give the phosphate time to diffuse into the iron-coating . We recommend the use of iron-coated sand filters in slow-flow systems with a flow of under 10 cm/hour. Fitting iron-coated sand filters around the drains instead of in units at the end of one or more drains [6, 7] has the advantage that lower flowrates and a lower P/Fe-ratio are possible. More filtering material is needed , but as iron-coated sand is a residual product of drinking water production it can be available in large amounts. The biggest disadvantage of placing the filter under the ground is that it is not easily accessible for removal or maintenance purposes.

Figure 2. Placing of drainage pipes and iron-coated sand (Photo: Victoria Barcala)

Slow flow though velocities or resting periods allows the phosphate diffuses from the outer to the inner layers of the coating. This can be achieved in a practical sense by having two parallel filters and using them alternately. This is easy to realise in decentralised systems such as those for greenhouse wastewater treatmenr. Slow adsorption is also a good phosphate removal method in natural systems. Between rainy periods, these systems have natural rest periods in which the phosphate can diffuse in the coating.

Another attractive application of iron-coated sand filters is in the outflow of water retention basins, or in combination with constructed wetland. Iron-coated sand filters function by adsorption and not by biological processes. This is why they also work well in the winter, when the removal efficiency by constructed wetland may be lower.

Victoria Barcala
(TAUW)
Arjon Buijert
(Arcadis)
Thijs Lieverse
(Arcadis)
Leonard Osté
(Deltares)
Joachim Rozemeijer
(Deltares)
Stefan Jansen
(Deltares)

Background picture:
One of te test locations: drainage with iron-coated sand filter in the underground; ditch with ecologically healthy water (Photo: Victoria Barcala)

Summary

The Netherlands has an urgent need to reduce the burdening of surface water with phosphate. One of the methods is the use of iron-coated sand filters around drainage pipes. The iron-coated sand adsorbs the phosphate. The main issue is how iron-coated sand filters can be designed and optimised, taking adsorption kinetics and reducing conditions into account. To research this field and laboratory experiments were carried out . The iron-coated sand filters do not release any phosphorus under weakly and moderately reducing conditions, even after the partial dissolution in iron and manganese. Strongly reducing conditions can reduce the effectiveness of the phosphate adsorption, especially through the formation of (iron) sulphides. The pump speed or the natural flow velocity influence the phosphate retention in the iron-coated sand filters. The optimal functioning of an iron-coated sand filter requires rest periods that extend the diffusion time of phosphate inside the granular coating.

Literature

1. Chardon, W.J., Groenenberg, J. E., Temminghoff, E. J. M., & Koopmans, G. F. (2012). Use of Reactive Materials to Bind Phosphorus. Journal of Environmental Quality, 41(3), 636–646. https://doi.org/10.2134/jeq2011.0055

2. Groenenberg, J. E., Chardon, W. J., & Koopmans, G. F. (2013). Reducing Phosphorus Loading of Surface Water Using Iron-Coated Sand. Journal of Environmental Quality, 42(1), 250–259. https://doi.org/10.2134/jeq2012.0344

3. Barcala, V., Jansen, S., Gerritse, J., Mangold, S., Voegelin, A., & Behrends, T. (2022). Phosphorus adsorption on iron-coated sand under reducing conditions. Journal of Environmental Quality, 52(1), 74–87. https://doi.org/10.1002/jeq2.20432

4. Barcala, V., Zech, A., Osté, L., & Behrends, T. (2023). Transport-limited kinetics of phosphate retention on iron-coated sand and practical implications. In Journal of Contaminant Hydrology (Vol. 255). https://doi.org/10.1016/j.jconhyd.2023.104160

5. Chardon, Wim J., Groenenberg, J. E., Vink, J. P. M., Voegelin, A., & Koopmans, G. F. (2021). Use of iron-coated sand for removing soluble phosphorus from drainage water. In Science of The Total Environment. Science of the Total Environment. https://doi.org/10.1016/j.scitotenv.2021.152738 Lambert, N., Van Aken, P., Van den Broeck, R., & Dewil, R. (2020). Adsorption of phosphate on iron-coated sand granules as a robust end-of-pipe purification strategy in the horticulture sector. Chemosphere, 267, 129276. https://doi.org/10.1016/j.chemosphere.2020.129276

6. Vandermoere, S., Ralaizafisoloarivony, N. A., Van Ranst, E., & De Neve, S. (2018). Reducing phosphorus (P) losses from drained agricultural fields with iron-coated sand (- glauconite) filters. Water Research, 141, 329–339. https://doi.org/10.1016/j.watres.2018.05.022

^ Back to start

Occurrence, Hazard, and Risk of Psychopharmaceuticals and Illicit Drugs in European Surface Waters

Globally, the use of psychopharmaceuticals has increased significantly. These drugs are used to treat mental disorders and diseases, along with other conditions related to the nervous system, such as in the case of painkillers and anaesthetics. Psychoactive substances without medical application, such as illicit drugs, also fall into this category from an environmental point of view. Little is known about their occurrence in water, ecotoxicological effects, and ecological risks.

Psychopharmaceuticals alter the neurochemistry of the human brain. Due to the similarities between the nervous systems of humans and other animals, these drugs can also act on the nervous systems of non-target species, which can disrupt the ecosystem. For example, antidepressants can affect the predatory behaviour of fish, reducing their ability to hunt effectively.
Sewage treatment plants are major emission points for these compounds into the environment. Therefore, wastewater has become a much researched medium to reveal trends in the use of illicit drugs among the population. Currently, up to 60-70 per cent of consumed drugs, illicit drugs, and their transformation products are not removed by wastewater treatment plants, and thus increasingly end up in surface water.

Little is known about Psychopharmaceuticals

The number of studies on occurrence, hazards, and risks of psychopharmaceuticals is relatively limited compared to other causes of ecological changes, such as habitat loss or climate change. Moreover, the focus is often on common compounds such as carbamazepine, paracetamol, and fluoxetine; and not on the newer or more widely used compounds such as escitalopram. It is therefore important to conduct more research to determine the contribution of these compounds to the current degradation of the aquatic environment.
The aim of this study was to inventorise and evaluate data on the occurrence, hazards, and risks of psychopharmaceuticals and illicit drugs in European surface waters. This was done taking into account uncertainties and knowledge gaps. Risks were estimated by comparing concentrations in the aquatic environment (occurrence) and concentrations at which effects occur (hazard, or ecotoxicity). Then, for each substance, these risks were related to the drug type.

Methods

Based on pharmacological classification and the Dutch opium law, 702 psychopharmaceuticals were identified for this study. Ecotoxicity data (so-called chronic 'no observed effect' concentrations) and occurrence data of psychopharmaceuticals were searched for in eight databases (UBA Pharmaceuticals in the Environment, NORMAN Network, Naiades, EU WATERBASE, Water Quality Portal, RIWA database, EPA ECOTOX Knowledgebase, and UBA ETOX database). Finally, we collected data from the top 50 most prescribed drugs in the Netherlands (2015-2020), if they were missing from the aforementioned databases. The occurrence concentrations and the hazard concentrations were then used to calculate risk quotients (RQs):

In addition, we performed a data confidence test on each compound, assessing the frequency of detections and in how many countries a compound is monitored. We also used the European Commission’s Technical guidance 27 (TG 27) as a reliable way to score the ecotoxicity data. The scores were combined and ranged from very low to very high.

Few Reliable Data

Occurrence data were not available for more than half of the compounds (359 out of 702) and ecotoxicity data were missing for the vast majority of compounds (597 out of 702, see Figure 1). If data were available, confidence was often low because the lion's share of ecotoxicity studies were conducted on only a few organism species.
Both mean occurrence concentration and occurrence data confidence showed a positive correlation with prescription. However, no significant correlation was found between how often the compound was measured and how much it was prescribed. That is to say, the more a substance is prescribed, the higher the concentration in the environment, and the more reliable the measurement data, but not the more measurement data there were. In addition, there also appeared to be no relationship between how often a drug was prescribed and how much toxicity data was available. So even on frequently prescribed drugs, very little was regularly found to be known about environmental effects.

Figure 1. Pie charts showing the percentage of the 702 compounds that provided data

Environmental risks of psychopharmaceuticals and illicit drugs

Only for 87 of the 702 (12%) compounds were sufficient data on occurrence and hazard available to calculate environmental risks based on chronic NOEC values (Figure 2 shows a selection). Data above an RQ of 1 (red line) indicate a risk. The graph shows that a significant proportion of the psychopharmaceuticals about which much is known pose an ecological risk in surface waters. The five riskiest agents were risperidone, carbamazepine, paracetamol, cocaine and ibuprofen. For these agents, at least 10 per cent of the risk quotients were above 1. However, data confidence is often low. Only for carbamazepine, paracetamol, fluoxetine and ibuprofen was the reliability high or very high. This makes carbamazepine the only drug in all analyses with a median risk of more than 1 and very high reliability , and so we can confirm that it poses an ecological risk. However, with more data, we could confirm the risk of more drugs.

Figure 2. Risk quotients of ten selected psychopharmaceuticals based on surface water concentrations and chronic NOEC (ecotoxicity). The error bars represent the upper and lower quartiles, while the box represents the middle upper and middle lower quartiles. The central line indicates the median (L = low, M = medium, H = high, VH = very high).

We used TG 27 (European Commission) criteria for scoring the reliability of ecotoxicity data. No agent met all the criteria, and only 11 compounds met half of the criteria. This means that an ecosystem-wide risk assessment was not possible, mainly due to a lack of ecotoxicity data.
The most prescribed drugs also often appear to be the best studied; for example, the four drugs carbamazepine, paracetamol, ibuprofen and fluoxetine together accounted for 23 per cent of the ecotoxicity data and 28 per cent of the measurement data. When comparing Dutch prescription data in defined daily doses (DDDs) with the calculated risks, the compounds posing the highest risk also tend to be the most commonly used and prescribed. Nevertheless, Pearson correlations did not indicate that DDDs correlated with risks. How often the substance was measured correlated positively with higher perceived risk, as did the amount of chronic NOEC data. The reliability of the ecotoxicity dataset also correlated with risk. Therefore, the better a substance was studied, the higher its risk was estimated. This is worrying because it suggests that many poorly studied substances may carry hidden risks. More research on the presence and dangers of psychopharmaceuticals is therefore needed to clarify whether these hidden risks are indeed present.

Data were missing from a number of commonly prescribed drugs. For example, Risperidone (number 37 in the top 50 in NL) showed the highest median risk of all compounds, but had a very low data score making the calculated risk unreliable. For betahistine (number 7 in the top 50), the risk could not be calculated due to a lack of both measurement data and ecotoxicity data. For Tramadol, many measurement data existed, but only one ecotoxicity study was found. Even for the most prescribed psychopharmaceutical, paroxetine, the risk could only be estimated with medium reliability due to too little ecotoxicity data.
Thus, to better assess the ecological risk of psychopharmaceuticals, more data are needed, especially on ecotoxicity. Therefore, we argue that more research is needed on the occurrence of highly toxic and commonly prescribed drugs in surface water, but also and especially on the ecotoxicity of commonly prescribed and commonly found compounds. This requires better cooperation between environmental chemists, ecotoxicologists and local water authorities to assess water quality and evaluate measures to improve water quality.

Conclusions

Our analysis yields two main conclusions:
- Risks of many psychopharmaceuticals remain unknown due to a great lack of measurement data in the aquatic environment and an even greater lack of ecotoxicity data.
- The most commonly prescribed compounds in the Netherlands were the most studied and most common. However, not enough data were available for all commonly prescribed psychopharmaceuticals. This means that the risk of underestimating the environmental risks of psychopharmaceuticals is real, as ecological risks to aquatic organisms have been identified for many psychopharmaceuticals about which much data are available.

Charlie J. E. Davey
(Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam)
Michiel H. S. Kraak
(Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam)
Antonia Praetorius
(Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam)
Annemarie P. van Wezel
(Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam)
Thomas L. ter Laak
(Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam / KWR Water Research Institute, the Netherlands)

Summary

Little is known about psychopharmaceuticals and illicit drugs in surface water, especially data on ecological risks are lacking. Of 702 substances studied, measurement data were missing for more than half, and ecotoxicity data were missing for 85 per cent. However, a lot is known about some commonly prescribed agents. The four drugs carbamazepine, paracetamol, ibuprofen and fluoxetine together were found to account for 28 per cent of the monitoring data in surface water and for 23 per cent of the ecotoxicity data, with ecological risks to (aquatic) organisms found to be high. This leads to a pessimistic outlook for the risks of the many hundreds of substances about which little or nothing is known. It is high time for systematic research on presence and risks of psychopharmaceuticals in the aquatic environment.

Literature

1. Bisesi, J.H., Bridges, W., Klaine, S.J., 2014. Effects of the antidepressant venlafaxine on fish brain serotonin and predation behavior. Aquat. Toxicol. https://doi.org/10.1016/j.aquatox.2013.12.033

2. aus der Beek, T., Weber, F.A., Bergmann, A., Hickmann, S., Ebert, I., Hein, A., Küster, A., 2016. Pharmaceuticals in the environment-Global occurrences and perspectives. Environ. Toxicol. Chem. https://doi.org/10.1002/etc.3339

3. European Commission, 2018. Technical Guidance For Deriving Environmental Quality Standards (Guidance Document No. 27). Eur. Community Rep. 11-12 June, 210p.

4. Huizer, M., ter Laak, T.L., de Voogt, P., van Wezel, A.P., 2021. Wastewater-based epidemiology for illicit drugs: A critical review on global data. Water Res. 207, 117789. https://doi.org/10.1016/j.watres.2021.117789

5. Davey, C.J.E., Kraak, M.H.S., Praetorius, A., Thomas, L., Wezel, A.P. Van, 2022. Occurrence , hazard , and risk of psychopharmaceuticals and illicit drugs in European surface waters. Water Res. 222, 118878. https://doi.org/10.1016/j.watres.2022.118878

^ Back to start

Accurately identifying polymers and microplastics by machine learning

Microplastics form a growing problem for the environment. There are different types, making identification a challenge. Infrared spectroscopy in combination with machine learning can enable accurate identification.

Microplastics, including rubber particles, form a growing problem for the environment because they spread everywhere and break down slowly. It is therefore important to identify these small plastic particles to determine the scope of the problem and find effective solutions. This is also endorsed by the European Union and laid down in the drinking water directive (EU) 2020/2184).
There are many different types of microplastics, varying in size, shape and material and identifying them can therefore be a challenge. Infrared spectroscopy in combination with machine learning can help by discovering patterns in the spectra and enabling accurate identification, even if the particles have been affected by the environment and are therefore hard to identify by other methods.

Introduction

Research into microplastics has been under development for years, but the data and insights are not yet at the level of solutes. The cause is in the nature of particles which, in contrast with molecules, can be considered unique. As with snowflakes, one particle is not the other. Chemically identical particles can vary in terms of size and shape. They may also be partially degraded (i.e. oxidised on the surface). In this way, in actual fact a mixture of the original material and the degraded material arises, in varying proportions, making identification considerably more difficult.
In the case of a solute, we are used to making a comparison with the same substance in a database/library. If the data correspond, we can be certain that we have identified the substance. In the case of particles this is not as easy, even if it is still possible to compare them to data in the database. As soon as a particle varies too much from the data in the library, it can in some cases not be classified, or is wrongly classified. Filling a database with all potential states of a particle is not possible, because the number is infinite. These data are nevertheless necessary for proper identification. Ways of solving this problem must therefore be sought.
To analyse microplastics, infrared spectroscopy is often used. Each particle is irradiated, for instance, by an infrared laser. This produces an infrared spectrum that is characteristic for this particle. The spectrum is then compared to spectra in a database. If the variations of the spectra are too large, identification is no longer possible. To increase the number of spectra in the library and improve identification, a two-step approach has been chosen. In the first instance, newly generated spectra were added to the database. The spectra of real particles were then compared with the spectra in the new database by machine learning.

New spectra

New spectra were created in the following manner. From the existing database, in this case 210 spectra of different particles, two random spectra were chosen from particles with a range of different characteristics, but from the same type of polymer (e.g. polyethylene). From this, a new spectrum was made from a linear combination. The individual spectra are given a random weighing factor between 0 and 1. However, the sum of both weighing factors is equal to 1. In this way, the two original spectra can generate a similar replicate sample. This step can be repeated with other spectra until sufficient spectra have been created.
A random number generator determines which spectra and which weighing factors are used. These new spectra were copied into a new database, and this database was subsequently used to classify the 210 original spectra of real particles. To this end, two different machine learning models were tested. The first model was an ensemble sub-KNN (subspace k-nearest neighbours) and the second was CNN (convolutional neural network). The sub-KNN model uses the one dimensional numeric signals to predict the type of microplastics, which an ensemble machine learning algorithm based on the calculation of distance between training data points.
In the second model, the numerical data of the spectrum are converted into visual data. In this case, the numbers are shown using a polar coordinate system and saved as figures. The figures that arise are compared with each other. CNNs are good at recognising patterns in an illustration, such as lines, textures, circles or colours. If there are big similarities between two illustrations, the corresponding particles are also comparable.

Complexity

Besides the two different working methods, CNN and sub-KNN are also distinguishable in terms of their complexity. CNN, known from image recognition, is a so-called deep learning algorithm, which requires a large amount of computing power and a considerable amount of data to function properly. Sub-KNN on the other hand requires less computing power and can also produce good results with small datasets.
To assess the methods, accuracy, precision and ‘recall’ are important, as is the amount of effort that is required, such as the size of the dataset and the computing time and power required. The relationship between correctly identified polymers compared to the total number of attempts at identification is the accuracy. Precision is the ratio between the number of times that a model has correctly identified a polymer and the total number of times that the model has attempted to make a prediction for this specific type of polymer. Therefore, if the model has identified polyethylene, for instance, 100 times and 80 of these were correct, the precision is 80%. Recall is the number of times that the model has actually made a correct identification, divided by the total number of this polymer. This gives an idea of how well the model performs in correctly identifying the polymers. Finding a good balance between accuracy, precision and recall is essential.

Results

Illustration 1 shows that the maximal values for accuracy, precision and recall are all >0.94 if the models are offered 100 generated spectra per type of polymer as reference spectra. This means that each of these models can accurately identify polymers with 6% incorrect predictions at the most. The best model with an accuracy score of 0.995 (i.e. sub-KNN, trained on 40 generated spectra) has just one wrongly classified spectrum; this model is the most suitable to classify polymers based on the infrared spectra. Although the model architecture of sub-KNN is simpler than that of the CNN, the former performs better than the latter in terms of all performance indicators.

Illustration 1: Performances of the two models. All models were trained on N generated spectra per class (N = 10, 20, 30, 40, 50, 75 or 100) and tested on 210 original spectra. A reference value was calculated based on the Sub KNN-model, which was trained and tested on the original spectra. The performances of the model are shown as (a) accuracy, (b) precision and (c) recall.

Moreover, illustration 1 shows that the number of training spectra (the x-axis) have a considerable influence on the model performances. The performances of all models increase according to the increase of the number of spectra in the library. No more than a slight fall of precision can be seen for sub-KNN. In other words, optimal model performances can be jeopardised if models are trained with few spectra. Sub-KNN already achieves the best performances with 40 spectra, while even with 100 spectra, CNN performs considerably worse.

Illustration 2: Training times for the two models used.

A benchmark simulation (performance test) carried out with sub-KNN was also trained and cross-validated (sample validation) based on the original dataset with just 210 spectra in total. In other words: studies are examining the point from which the use of artificially generated spectra leads to better results than just using the original dataset.
Illustration 1 shows the benchmark as a reference line. In general, even the use of 50 generated spectra or more per polymer class can lead to a performance that is better than the benchmark for all models Particularly sub-KNN (the most suitable model) needs only 20 spectra or more to perform better than the benchmark. This shows that the generation of spectra offers a solution in the event that too few real spectra are available.
Illustration 2 shows the training time as a function of the size of the dataset. Sub-KNN needs 9.5 s to train models, while the second CNN needs 5 min. Although the longest period (25 min) is still relatively short, it is important to take future applications into account, whereby the model is used to classify thousands of polymers in an online learning mode, and where the model must be repeatedly trained with added spectra of new polymer types. Where training time is concerned, sub-KNN is also the most suitable approach.
The methods proposed here to identify particles by means of machine learning are currently being integrated into a programme that can be used by everyone without a large amount of prior knowledge of machine learning. A user interface is made to this end.

Conclusions

The identification of microplastics in environmental samples is a challenge due to the differences that occur after their environmental degradation. As a result, microplastics demonstrate strongly varying properties compared to new plastics. Machine learning models can aid in accurately identifying and counting microplastics in samples. The study shows that it is not per se necessary to use the most advanced machine learning models. The results show that the simpler model is better, both in terms of its performance and its computing time. This stresses the importance of testing multiple models before a final choice is made for a specific application.

Patrick S. Bäuerlein
(KWR)
Xin Tian
(KWR)
Frederic Beén
(KWR)
Yiqun Sun
(KWR)
Peter van Thienen
(KWR)

Summary

Data and insights derived from studies on microplastics are not yet as accurate as those on solutes. This is due to the nature of the particles which, in contrast with molecules, can be considered unique. Particles can vary in size and shape and may be partly degraded. This makes it difficult to achieve correct identification with infrared spectroscopy, for instance.

To tackle this problem, machine learning models have been tested and used to identify microplastics. It became clear that even a relatively simple model is able to accurately identify microplastics. This makes the use of this technology suitable for a broader public and not just for specialists. Making and training a model requires knowledge, but once this has taken place the models can be used by anyone with knowledge of computers. The approach presented here can also be applied to other sorts of spectral data, for instance ultraviolet, Raman, FTIR and mass spectra.

^ Back to start