The effects of flooding on a bank filtration site: lessons learned from the Breslau (Wrocław) manganese calamities of 1906

The first sign of an ongoing catastrophe was the sudden appearance of air in the pipeline network on the evening of 28 March 1906, which led to problems with the vacuum-driven siphon system. It took over 12 h of intensive efforts to remove the air. Apparently, the massive piston of infiltrating floodwater had entrapped the air in the dewatered aquifer and pushed it downward, from where it escaped, at least in part, through the wells. The water quality of Collector shaft 1, into which well groups I and III dewatered, was affected first. Iron concentrations suddenly rose to around 100 mg/l on 29 March (Figs. 3 and 4). Collector shaft 2, into which the wells of group II dewatered, followed 1 day later with values of up to 80 mg/l (Debusmann 1908). The iron removal system, although designed for maximum concentrations of around 20 mg/l, was surprisingly able to cope with these high concentrations. However, a few days later elevated concentrations of manganese suddenly appeared, on average around 30 mg/l, sometimes even more in individual wells (Figs. 3c and 4b). Manganese had not been taken into account while designing the Breslau (Polish: Wrocław) water supply and was hardly considered or analyzed at all in Germany at that time (von Raumer 1903). The treatment plant was not able to cope and manganese-containing water broke through the filters. The first indicators of this were a discoloration of the drinking water, followed by turbidity, a metallic taste and flocs of manganese oxides, which stained the laundry, rendering it unusable for the customers. This so-called Breslau manganese calamity led to the total breakdown of the brand-new water supply of one of the major cities of Germany. This, of course, caused a huge stir in the contemporary news but also a lot of scientific interest, reflected in a large number of publications (e.g., Woy 1906; Lührig 1907a, b, 1908; Lührig and Blasky 1907; Luedecke 1907; Beyschlag and Michael 1907; Debusmann 1908; Oettinger 1908; Klut 1911; Prinz 1919). The case was even reviewed in US literature (Weston 1910).

Full size image

The iron concentrations recorded after the event did not go back to the pre-flood levels of around 20 mg/l (Fig. 4a). Instead, they actually showed a rising trend. More than 1 year after the flood, the average iron concentrations (≈ 125 mg/l) were higher than the flood peak. Several iron peaks occurred after the March 1906 event. The one from September 1906 can be related to another flood that occurred at that time. The other two, visible in Fig. 4, were related to operational procedures, when pumping was shut down for a few days. During this time, groundwater levels probably recovered and the rising water washed out remainders of the pyrite oxidation products. It is interesting to note that there was a temporal delay between the peaks of iron and manganese (Fig. 4a, b). The iron peak occured on 30 March, the manganese peak on 04 April, thus 5 days later. This is a first indication that the two were not released at the same time and via the same process.

Flood water can bring high loads of organic matter. In 1906, the content of dissolved organic matter was measured as the consumption of the strong oxidant potassium permanganate (KMnO₄). Its values initially rose after the flood event, but then quickly decreased to values even below the initial groundwater concentrations (Fig. 5a). This shows that the infiltrating water did not bring a high load of organic matter with it.

Full size image

Similar to iron, the concentrations of sulfate rose dramatically after the flood event, from 0.5 mmol/l (≈ 48 mg/l) to around 4 mmol/l (≈ 385 mg/l) (Fig. 5b). More than 1 year after the flood, maximum concentrations of up to 12 mmol/l (≈ 1150 mg/l) sulfate still prevailed, with an average of around 6 mmol/l (≈ 575 mg/l) (Fig. 3c). On the other hand, the conservative tracer chloride showed almost no change (Fig. 5b).

Parallel to iron and manganese, the concentrations of calcium and magnesium rose significantly (Fig. 5c), also reflected in the jump in total dissolved solids (Fig. 5a). This indicates a dissolution of carbonate phases from the aquifer matrix. Bicarbonate concentrations, on the other hand, decreased significantly (Fig. 5b). This was most probably an effect of a low pH of the water. It must have been below pH 5.0, as measurements using the indicator rosolic acid showed. It probably was not lower than pH 4.3, as indicated by the low, but noticeable concentrations of bicarbonate (Lührig 1907a, b). Since these low pH values were observed after the buffering dissolution of carbonates, even lower pH must have prevailed further above.

Nitrogen species cannot have contributed to the changes in hydrochemistry since the concentrations of all species (nitrate, nitrite, ammonium) in river and groundwater both before and after the event were at the detection limit (Lührig 1907a, b; Oettinger 1908).

A major threat of the flooding of shallow wells by surface water is the influx of potentially harmful microorganisms. Oettinger (1908) cited a bacterial count of 10,000 colony forming units per cm³ (CFU/cm³) for water of the Oder River, while Lührig (1907a, b) gave a lower value of 560 CFU/cm³. Water from collector shaft 2 was daily monitored for total bacterial counts before and after the flood event (Fig. 6). The water had never been fully free of microorganism before the event, as the measurements in Fig. 6a from 1905 show.

Full size image

Bacterial counts after the March flood showed a surprisingly small and short-lived (few weeks) increase (Fig. 6b). Higher, although shorter peaks had already been recorded before (Fig. 6a). The pumped groundwater became finally bacteria-free in December 1906 (Oettinger 1908).

Figure 7a and b show that the groundwater temperatures in the alluvial aquifer usually ranged between 8.8 and 10.2°C over the course of a year, which are typical values for central Europe. The lowest groundwater temperatures occurred between April and August 1906, followed by the highest between August and February 1907. This pattern trails behind the seasonal air temperatures of the region, indicating a time lag of several months between air and groundwater temperature. The flood event of the March 1906 occurred during a cold spell with air temperatures around the freezing point and river water temperatures slightly above (Lührig 1907a, b, Fig. 7c). The infiltrating floodwater was significantly colder than the groundwater, which at that time was at around 2 °C, thus a difference of up to 7.5°C compared to groundwater (Fig. 7c).

Full size image

Several theories were proposed to explain the manganese–iron calamities. An early one, claiming an upconing of artesian groundwater from the underlying Neogene formations, due to the diminished water pressure in the overlying Quaternary aquifer, was quickly discarded since both the chloride concentrations and the temperature of the Neogene water did not fit the observations at all (Lührig 1907a, b). The initial argument was that the fluvial loam at the ground surface was assumed to be completely impermeable and the problematic water thus had to come from below. Salt and heat tracer experiments by Lührig (1907a, b) quickly disproved this assumption of an impermeable soil cover. Finally, the renewed rise of iron after the September 1906 flood clearly showed that the floods were the culprit (Fig. 4).

The elevated concentrations of iron and sulfate, combined with the loss of carbonate hardness (due to acid release), quickly convinced several investigators that pyrite oxidation must have played a decisive role (Lührig 1907a, b; Oettinger 1908; Debusmann 1908). Indeed, Luedecke (1907) had identified finely dispersed iron sulfide (FeS₂) in both the top soil and oxbow sediments. He also quotes the results of a well drilled in the Oder Valley, where a 15 cm layer, chiefly comprising iron sulfides was found. Sediment analyses confirming the quantity of pyrite in sediment or soils were not done. Elevated contents of total iron in soils, sediments, and surface waters within the Oder River valley were later confirmed by geochemical studies by Tamassi-Morawiec et al. (1998) and Konieczyńska (1998). These contents could be, to a large extent, atrtributed to iron sulfides. Formation of pyrite, often as framboids, in sediments under anoxic conditions was reported for geologically similar locations by Postma (1982). Several authors erroneously attributed the elevated manganese concentrations to the simultaneous oxidation of manganese sulfides (Luedecke 1907; Oettinger 1908). Manganese sulfides do occur in nature, but only in a narrow stability field, i.e., at strongly reducing redox potentials (Eh < 0.0 mV) and at a strongly basic pH (> 8; e.g., Hem 1972). These conditions are rarely met in aquifers of central Europe, including the Oder River valley aquifers.

The oxidation of pyrite via oxygen proceeds in two steps (1a,b,c), of which the second (1b) often remains incomplete due to an insufficient supply of electron acceptors. If no ferrous iron is oxidized via reaction (1b), the molar ratio of sulfate to iron in water should be 2:1. Higher ratios indicate that some ferrous iron has been removed, either by reaction (1b) or secondary reactions such as cation sorption and redox reactions with other species, e.g., manganese (IV).combined (1c)

The observed sulfate concentrations after the flood were relatively stable around 4 mmol/l (≈ 385 mg/l), of which almost all had to come from reaction (1a), since the infiltrating river water brought only a few tens of milligrams of sulfate with it (Fig. 5b). Figure 8 shows that the molar sulfate to ferrous iron ratio in groundwater is close to 2:1, i.e., very close to the molar ratio of pyrite. This indicates that almost no oxidation of ferrous iron to ferric iron according to reaction 1b took place. The concentrations of dissolved oxygen in infiltrating flood water are far too small to explain the oxidation of two millimoles of pyrite per liter of water (= 240 mg FeS₂). As stated, other dissolved oxidants such as nitrate were absent. Therefore, only atmospheric oxygen remained. This shows that pyrite present originally in soil, subsoil and shallow aquifer matrix had been exposed to air due to pumping-induced dewatering, exacerbated by drought effects. The flood water had then washed out the resulting acidic iron sulfate solution and its byproducts (calcium, magnesium, manganese). The good correlations visible in Fig. 8 also suggest that the analytical techniques applied in 1906 were quite precise.

Full size image

The acid released during pyrite oxidation can be buffered by reactions with carbonates, leading to release of calcium and magnesium and an increase of pH (2).

Theoretically, the released acid can also lead to the dissolution of iron and manganese oxides according to reactions (3) and (4), but the solubilities at the observed pH range of 4–5 are generally small.

As Fig. 4b shows, the manganese peak occurred several days later than the iron peak. This rules out a simultaneous release with iron based on reaction (4). The mobilization of manganese is thus more likely to be facilitated by the reduction of solid manganese oxides by ferrous iron (reaction 5), a process already noted in previous studies (Postma and Appelo 2000).

Therefore, Fig. 8 also contains the initial iron concentrations (Fig. 8, white symbols), calculated from reconverting the manganese concentrations into ferrous iron, using the reverse of reaction (5). This brings the molar ratio even closer to 2:1. The correlation between iron and sulfate on one side and manganese on the other is quite poor. An interesting observation is that higher manganese concentrations tend to occur only at sulfate concentrations of 4.5 mmol/l and higher (or 2.25 mmol/l of iron).

Manganese could potentially also be mobilized by interactions of manganese oxides with organic matter. The fact that manganese concentrations in groundwater before the flood were negligible, shows, however, that this process was not at work in the aquifer itself. Figure 5 also shows that the—rather low—concentrations of dissolved organic matter (as KMnO₄ consumption) did not increase with the flood event, but rather decreased slightly. This pathway is thus unlikely.

The exact location of the oxidized pyrite deposits remained unclear at first. Luedecke (1907) had found iron sulfides (FeS₂) in both the topsoil and the aquifer. Therefore, Lührig (1907a, b) analyzed depth-specific groundwater samples (Fig. 9). He found that the uppermost two water samples were very acidic, the third slightly acidic and the deepest one around neutral. Mineralization and concentrations related to pyrite oxidation decreased with depth. He thus concluded that the source had to be close to the surface.

Full size image

Thermodynamic and kinetic reaction modeling with PHREEQC was done to quantify the hydrochemical processes discussed above and to reconstruct some parameters which could not be measured in 1906, e.g., pH. Modeling was undertaken in several steps, where the output of each step is taken as the input of the next, namely: (1) oxidation of pyrite in the unsaturated zone, thus with an exposure to atmospheric oxygen concentrations, (2) outwash of oxidation products by flood water (a high ratio of flood water to soil water was assumed (20% soil water, 80% flood water), since the soil water in the unsaturated zone is mostly present in the capillary-bound porosity, which is small in sandy sediments), (3) equilibration with calcite (buffering) after mixing (an equilibration already in the unsaturated zone was also tried but the resulting pH did not fit the observations), (4) kinetic reaction of ferrous iron with manganese oxides (taken the output from step 3 as input).

The composition of the river water and the native groundwater were taken from Luedecke (1907). Their pH were set to 5.5 and 5.0, respectively. Owing to the extensive interactions of river water and groundwater, both were of the calcium bicarbonate type and their initial concentrations quite similar. They can thus not be used as distinguishable end members for a mixing calculation. The model input, however, could be constrained by the analyses presented in Figs. 3, 4, and 5. For the flood event of March/April 1906, the model had to emulate an increase of sulfate from 0.5 mmol/l, as an initial concentration in groundwater, to 4.0 mmol/l afterward (Fig. 5b). Additionally, a pH below 5 was aspired, combined with a shift from bicarbonate as the predominant inorganic carbon species to CO₂. Through calcite buffering, the calcium concentration should rise from 1 to 3 mmol/l, as measured (Fig. 5b). The available calcite content was thus restricted to 2 mmol/l. For the sake of simplicity, the slight increase in magnesium was not considered. Potential buffering effects of aluminum phases were not considered, as aluminum concentrations were not measured and pH was likely not lower than pH 4.3.

As discussed above, the iron to sulfate ratios shown in Fig. 9 indicate that very little ferrous iron released from pyrite oxidation is oxidized and precipitated as iron oxides. Therefore, reaction 1b was disabled in the PHREEQC solution species.

Table 1 shows selected results obtained from the thermodynamical modeling. For the flood event of March/April 1906, a good fit was found when oxidizing 8.7 mmol/l (1044 mg/l) of pyrite. Only the resulting electrical conductivity did not fit the observed TDS of around 700 mg/l well (Fig. 5a), maybe because some nonreactive constituents had been ignored in the model, e.g., potassium, ammonia, and organic matter. For the average sulfate concentrations of 6 mmol/l, observed in May–June 1907 (Lührig 1907a, b, 1908, Fig. 3c), a pyrite consumption of 13.5 mmol/l (1620 mg/l) was modeled. This model had the closest fit to the field data in general. For the extreme case of 12 mmol/l sulfate (Lührig 1907a, b, 1908, Fig. 3c), 28.8 mmol/l of pyrite (3456 mg/l) had to be used. The first two scenarios are probably the most representative of the general processes at play. The oxidation of pyrite in the unsaturated zone of the drawdown cone via atmospheric oxygen led to a very acidic soil solution (pH ≈ 2). The resulting mixing with flood water diluted the acidity (pH ≈ 3) and calcite buffering finally brought it up to pH ≈ 5, close to what the acidity indicators used at the time suggested.

The reaction of manganese oxides with ferrous iron (reaction 5) is not implemented in the PHREEQC thermodynamic database. Instead, a kinetic rate law modified from Postma and Appelo (2000), adapted for the mineral Hausmannite (Mn₃O₄) was used (reaction 6). The lag time between the iron and manganese peaks visible in Fig. 4 of 5 days was used as total reaction time (time step = 1 hour). The Fe²⁺/Fe³⁺ pair was recoupled in the PHREEQC solution species to allow the consumption of ferrous iron according to reaction 5.where

The kinetic model predicts a maximum release of 0.51 mmol/l (28.0 mg/l) manganese after 5 days. This is very close to the observed peak value of 28.4 mg/l on 04 April 1906. This maximum concentration corresponds to a consumption of 1.02 mmol/l of ferrous iron (57 mg/l). The shape of the manganese release curve is practically linear. Unfortunately, the number of measured manganese concentrations is very small, there are just four values for the rising part of the peak (Fig. 4b). Additionally, the first value is from 2 days after the flood and only days and no hours are given for the analyses. Furthermore, the data shown in Fig. 4a and b come from different sampling points. Therefore, no comparison of the kinetic release curve to the actual breakthrough curve is merited here. It is sufficient that the kinetic model is able to emulate the maximum concentration and the steep, almost linear rise of the manganese breakthrough curve fairly well.

After discarding the initially discussed hypothesis of ascending groundwater from the underlying Tertiary (Neogene) formation being the source of the calamities, all contemporary authors studying the Breslau (Polish: Wrocław) calamities identified pyrite oxidation close to the surface as the fundamental reaction. They attributed the oxidation of the pyrite to its aeration, caused by excessive drawdowns, amplified by the previous drought. Only Lührig (1907a, b) and Debusmann (1908), however, identified the reduction of manganese oxides by dissolved ferrous iron as the underlying process for the release of manganese.

Here, for the first time, the data sets by the different authors were combined and modern interpretation techniques such as hydrochemical modeling were applied. This allowed us not only to test the old hypotheses but also to quantify the reactions involved. The hydrochemical hypotheses proposed by the early investigators of the 1906 calamities were confirmed by this study to be correct in principle. The molar correlations between iron and sulfate confirmed the central role of pyrite oxidation by oxygen (Fig. 8). The PHREEQC model was generally able to recreate the hydrochemical interplay between the hydrochemical species and mineral phases. Its utility, however, went beyond identifying the main reactions. It also allowed a quantitative assessment, something the original researchers could not do. This included the quantification of the masses of pyrite involved (Table 1). It also confirmed and quantified the role of the carbonates in buffering the released acidity and yielded the pH values, which could only be measured in a rather crude way in 1906. Finally, the model was also able to confirm and emulate the kinetic release of manganese by the reduction of manganese oxides by ferrous iron. By adapting the input concentrations of, e.g., pyrite and calcite, the model could be used for other RBF sites as well. It could also easily be expanded for further hydrochemical scenarios, e.g., when not only dissolved oxygen but also nitrate is available as oxidant.

Figure 10 summarizes the conceptual model of the events of the Breslau (Polish: Wrocław) 1906 calamities. Figure 10a shows the initial situation with high groundwater levels, reducing conditions in soil and aquifer, which lead to the formation of pyrite. Figure 10b shows the drawdown caused by pumping, aggravated by drought conditions and the resulting exposure of pyrite to atmospheric oxygen, followed by the inundation of the wellfield and the outwash of the pyrite oxidation products (Fig. 10c). Figure 10d shows the remediation chosen for the system (see below), the installation of infiltration ponds close to the wells, which maintain high groundwater levels and prevent pyrite oxidation.

Full size image

One of the greatest fears for RBF operators is the breakthrough of microorganisms into their well water, since surface waters are basically open recipients for any fecal contamination in their catchment. Therefore, a flood infiltrating huge volumes of surface water from above is a nightmare scenario, since the distance from the ground surface to the top of the well screen is commonly small, usually much smaller than the passage from the riverbed to the well. Surprisingly, the 1906 calamities show a rather mild and short-lived microbial contamination.

The reason for this positive development is probably twofold. Since the screens were installed close to the bottom of the aquifer, the infiltrating floodwater had to pass through several meters of soil and sediment, likely leading to a removal of microorganisms by sorption and decay. On the other hand, the infiltrating water was, as indicated above, likely quite acidic, a condition many freshwater bacteria do not tolerate well. This was already confirmed by both Lührig (1907a, b) and Oettinger (1908) who experimentally exposed water from the Oder River, having high counts of bacteria, to (acidic) iron sulfate solutions and recorded a significant reduction in numbers over time.

The 1906 flooding event provides an interesting large-scale natural heat tracer experiment. Figure 7b shows that the flood caused a sudden drop of groundwater temperatures of around 1°C. From there, however, temperatures rose again quickly and after around 10 days, the original temperatures had been re-established. Furthermore, the negative peak occurred around a week later than the flood and several days after the dramatic changes in hydrochemistry. The phenomena nicely show the delayed transport of the temperature signal and the fast attenuation of the input of cold water into the system.

The 1906 data compare favorably with other heat tracer experiments, both natural and artificial. For, example, Watson et al. (2018) studied the temperature effects of a large natural flood event on the temperature of adjacent groundwater and also found a return to the initial temperature distribution within a few days.

Immediate measures taken after the catastrophe included disconnecting the wells of the most affected group III. Over a period of 40 days, these wells were pumped to waste at a rate of 20,000 m³/d and the water disposed to the river Oder, after some lime treatment. Lührig (1907a, b) estimated that with an average concentration of 14.4 mg/l Mn, a total of 11.5 tons of manganese were removed this way. Pumping, however, did not decrease the concentrations, as had been hoped. More than 1 year after the flood, manganese concentrations were still at the same high level (Fig. 3c). On April 5, 1906, artificial sand bed filtration of river water was reactivated to compensate for the lacking volume of drinking water; the old plant had luckily not been dismantled yet.

The full reconstruction of the waterworks and the stabilization of groundwater quality took about 15 years (Kowal 1993). To provide sufficient water quantities and quality, it was decided to develop a system of filtration ponds, located about 50 m from the siphon wells, fed by surface water taken from the Ohle (Polish: Oława) River (Lührig 1908; Kirchner 1931a, b). This approach was tested first with a full-scale prototype running for several months. After the successful proof of concept, that is, the water becoming chemically and bacteriologically safe, more filtration ponds were constructed. In spite of droughts in 1922–1923, water production increased up to 77,000 and 90,000 m³/d in 1925 and 1936, respectively. In 1922–1928, water production gradually decreased due to the precipitation of iron oxides within the collectors and siphon pipes, blocking up to 40% of the diameter of the latter. In 1929, a zone of sanitary protection around the well field was established, while in 1933 a system of new rapid filters was installed (Kowal 1993).

For many years, the infiltration ponds were kept free of vegetation as much as possible, especially of aquatic plants. Since the 1990 s, the Municipal Water and Sewage Company of Wrocław changed its attitude to vegetation, especially in the infiltration ponds which is now perceived as a significant factor in phytoremediation processes.

Other major flood events affecting Wrocław occured in 1975, 1985, 1997, 2010, and 2024 (Łoniewski 1998; Ligenza et al. 2021), of which the most catastrophic was the flood of July 1997 (Dębicki 1998; Przewłocki et al. 1998). The entire RBF site was flooded, as well as several water and sewage pumping stations and other facilities. The waterworks were out of order for more than 2 weeks and drinking water quality suffered for an extended period of time. The full restoration after the 1997 flood took about 6 months. Taking into account the experience of the 1906 calamities, prognostic modeling to predict the chemical composition of filtrate water was performed. Calculations showed that no excessive concentrations of iron and manganese were to be expected, which was confirmed afterward by hydrochemical analyses. Similarly, no increased concentrations of nitrogen species were observed. Liming, however, was used as a countermeasure against possible mobilization of heavy metals (Łomotowski 1998).

Interestingly, a case very similar to the 1906 events was observed about 90 years later in a wellfield at Racibórz (German: Ratibor), a city 140 km away from Wrocław (Fig. 1a), upstream the Oder River (Miotlinski et al. 2012). There, an extensive groundwater extraction from the Pleistocene aquifer had caused a significant decrease of groundwater levels, of up to 20 m. The situation was alleviated to some degree after 1994, when extraction decreased and some rainy years occurred. In July 1997, heavy rainfalls of 540 mm over the course of a few days caused a massive flooding of the Oder River valley. Parts of the floodwater infiltrated into the subsurface, leading to an increase of groundwater levels. Similar to the 1906 case, concentrations of sulfate started to rise from around 50–100 mg/l to up to 300 mg/l. Iron and manganese also showed significant increases, from 1–2 mg/l to up to 16 mg/l and from 0.2 mg/l to up to 0.9 mg/l, respectively. In some wells, the concentrations of bicarbonate decreased from around 300 mg/l to around 100 mg/l and the pH dropped by 0.5 pH units, although always remaining above pH 6.5. Again, the oxidation of pyrite and the dissolution of manganese oxides was invoked to explain these changes. The effects persisted for several years, Miotlinski et al. (2012) show data up to the year 2005. There are, however, some differences to the 1906 case. In Racibórz, the wells themselves were not flooded directly. This explains, why the high concentrations described above occurred significantly later than the 1997 flood. Sulfate started to rise as early as 1998 but the other parameters followed later on, in 1999 and 2000. Unlike the Breslau (Polish: Wrocław) case, the flood actually increased the chloride concentrations from around 25 mg/l to 40 mg/l. Owing to agricultural activities in the region, pyrite oxidation by nitrate must have played at least some role (e.g., Kölle 1985). The Racibórz example clearly shows that such events may recur, especially after drought–flood cycles, which may become more common in Central Europe due to climate change.

The city of Poznań (Fig. 1a) is supplied with water from two RBF sites. Both facilities are set within the unconfined sandy-gravely aquifer of the Warta River valley, which is a tributary of the Oder River (Chomicki et al. 2008; Górski et al. 2021; Matusiak et al. 2021). Flooding of the Warta River luckily never caused flooding of these RBF systems, nevertheless, some changes in physiochemical and microbiological parameters of filtration water caused by seasonal flooding have been detected (Kołaska et al. 2018; Matusiak et al. 2021).

Compared to the flood event investigated by Ascott et al. (2016) for an RBF site on the Thames River, England, the 1906 flood in Breslau (Polish: Wrocław) shows some remarkable differences. In the English case, the infiltrating floodwater induced an increase in dissolved organic carbon but a sharp decrease in electrical conductivity (EC) in the well water. The latter indicates that the dilute floodwater did not induce secondary reactions, e.g., pyrite oxidation, which could have added to TDS and EC. This is insofar surprising as the groundwater before the flood was generally suboxic to reducing (< 1 mg/l dissolved oxygen). The concentrations of dissolved oxygen markedly increased with the flooding (≈ 4 mg/l). An oxidant would thus have been available, but the sediment apparently contained no reactive minerals. There are, however, also some similarities. The high microbial load of the River Thames water at the time of the flood (Escherichia Coli >5000 cfu/100 ml) was strongly attenuated to around 4 cfu/100 ml in the well water. In addition, the microbial contamination disappeared after a few days. The comparison shows that the effects of an RBF flooding depend a lot on the site-specific geochemical settings, i.e., the reactive inventory of the aquifer, which, if present, can induce secondary reactions.

Between 1945 and the early 1950 s, the waterworks had to recover from the damages of the Second World War. Water production slowly but steadily, increased. Between 1953–1958, due to growing demand, water was again abstracted directly from the Oława (German: Ohle) River. Several problems regarding surface water quality were encountered, including high concentrations of phenols and the influence of wastewater. In the 1960 s, because of constantly growing water demand, the treatment facility “Na Grobli” (German: Weidendamm) was enlarged and upgraded. More wells, including radial collector wells, were installed in new well groups IV–IX (Fig. 1b). In 1982, construction of a new facility called “Mokry Dwór” near a small village of the same name, consisting of a pumping station and a treatment plant, was completed (Fig. 1b). This facility, with a planned water production capacity as high as 200,000 m³/d relied on two catchments, i.e., the Nysa Kłodzka (German: Glatzer Neiße) River and the Oława (German: Ohle) River, connected by a channel (Nalberczyński and Izbiańska 1998).

Meanwhile, water quality in the catchments of the Oława (German: Ohle) and Nysa Kłodzka (German: Glatzer Neiße) Rivers deteriorated owing to intensive agricultural and industrial activities, as well as insufficient municipal sewage treatment (Roszak 1987, 1988, 1989; Wasilewski and Szymańska 1989; Dubicki at al. 1998; Rejman 1999; Kolanek et al. 2007). In addition, several pollution point-sources were located in the vicinity of the RBF system. Since the end of the 1990 s and beginning of the 2000 s the situation started to improve, e.g., by improved water treatment (Urbanowska and Kabsch-Korbutowicz 2016). The issue of water quality at the RBF site became urgent again after the disastrous flood of 1997, which proved the high vulnerability of the system (Kryza et al. 1989; Łomotowski 1998; Łoniewski 1998).

After 1990, the waterworks became property of the city as the Miejskie Przedsiębiorstwo Wodociągów i Kanalizacji (MPWiK, English: Municipal Water and Sewage Company). Water use started to be measured, which together with the reduction of water losses, caused a drastic decrease of water demand, especially between 1990 and 2000 (Kowal 1993; Zięba 1998; Głowacka and Hotloś 2012; Hotloś et al. 2012). In 2010, domestic water production dropped to about 61,000 m³/d. The water sources started to change as well, i.e., between the 1980 s and 1999, about 60–64% of total abstraction came from surface water, the rest from the RBF. Since 2000, the share of surface water decreased to 50% (Hotloś et al. 2012). At present, water from the Oława (German: Ohle) River is connected by a channel with the Nysa Kłodzka (German: Glatzer Neiße) River and infiltrated via 59 ponds, having a total area of 48.4 hectares (Łoniewski 1998), finally extracted by more than 570 wells.