A new decentralised settling system based on the principle of lamella separation was developed for the treatment of road runoff. Two different laboratory test methods, the DIBt (Deutsches Institut für Bautechnik) procedure and our own approach, were applied in order to evaluate the efficiency of the system based on the separation of fine mineral particles and a mixture of mineral and organic particles, respectively. Overall efficiencies (88% after DIBt and 61% according to our own method) were comparable to results obtained for commercial systems. The lamella system was then applied in the field for 1 year to treat runoff from a road area of 420 m2. The amount of solids separated that was calculated from a mass balance (10.1 kg) was consistent with the amount of sediments measured (8.6 kg). However, the average separation efficiency was only 30% in the field study. This is related to the size and composition of the particles in runoff, which are not represented well by the material used for the test procedures. It is concluded that the test methods should be improved, and that more field studies are needed in order to obtain a better understanding of the settling behaviour of particles in road runoff.
In industrialised countries both combined sewer overflows and discharges from storm water sewers are a major source of aquatic pollutants (Sieker 2013). Since storm water runoff from traffic areas (roads, streets and parking lots) contributes significantly to overflows, several states of Germany now plan to require pre-treatment of polluted runoff before local discharge or infiltration. Besides semi-central facilities, decentralised systems treating runoff from traffic areas of 100–500 m2 in size will play an important role in this concept.
According to a recent review of European studies (Huber et al. 2015), the following median concentrations of heavy metals in runoff from motorways were found: 230 μg/L for zinc, 52 μg/L for copper, and 17 μg/L for lead, respectively. The data reveal that a considerable amount of these metals is bound to particulate matter. Similar findings were reported for North America (Kayhanian et al. 2012). Other investigations have shown that polycyclic aromatic hydrocarbons are also associated with particles to a large extent (Nielsen et al. 2015; Ozaki et al. 2015). Therefore mechanical treatment for the removal of particles from runoff can reduce significantly the load of pollutants. However, there is evidence that runoff from traffic areas contains a substantial fraction of fine particles with d < 63 μm (Li et al. 2005, 2006; Kim & Sansalone 2008). Thus the removal efficiencies obtained by applying mechanical processes might be limited.
The development of decentralised systems for runoff areas between 250 and some 1,000 m2 that can be integrated into street inlets started some 10–15 years ago (Fettig et al. 2007). They can be grouped into settling systems, filtration systems or a combination of both. At present 14 different systems are on the German market (Sommer et al. 2015); however, it is difficult to predict their performance under field conditions. In order to overcome this problem, test methods have been suggested that will simulate the separation of fine mineral particles (DIBt 2011) and both mineral and organic particles (Fettig et al. 2008). Recent work has indicated that particle separation efficiencies obtained in the field differ considerably from the test results (Barjenbruch et al. 2016). Therefore one specific goal of our work was the comparison of particle removal results obtained in laboratory tests and under field conditions.
In this study a new decentralised settling system based on the principle of lamella separation was developed. This principle has been applied in semi-central treatment of storm water, i.e. units for runoff areas of more than 5,000 m2, for a number of years (Clark et al. 2009; Hermann et al. 2010; Langeveld et al. 2012; Fuchs et al. 2014; Weiss 2014). The system was first tested under laboratory conditions and then applied in the field over a period of 1 year. The results are discussed with respect to the validity of the test methods. Hence the main objective of our study was to demonstrate whether lamella separation is a suitable method at the scale of decentralised systems with respect to both hydraulic properties and particle removal efficiency.
MATERIAL AND METHODS
The effective settling area of the system, i.e. the projected lamellae area seen from above, is 0.41 m2 and thus about 2.5 times larger than the cross-sectional area of the chamber. The minimal filling volume of the system is 133 L. Inflowing water causes an increase of the filling level, which determines the flow rate through the system. The bypass is activated when a filling volume of 198 L corresponding to a flow rate of 3 L/s is reached; i.e. the water flow in excess of 3 L/s will not pass through the lower part of the system. Thus turbulences close to the bottom sediment are limited and the sediment is prevented from being washed out at very high hydraulic loadings.
The DIBt (Deutsches Institut für Bautechnik) laboratory test procedure (DIBt 2011) is based on the loading of a system with fine mineral particles (Millisil W4, Quarzwerke Gruppe, Germany). Characteristic diameters of the material by weight are: d10 = 7 μm, d50 = 63 μm and d90 = 160 μm, respectively; and its density is 2.7 g/cm3. Three flow rates of 0.1 L/s (for 480 min), 0.24 L/s (for 200 min) and 1 L/s (for 48 min), respectively, are applied in the first part of the test. Here, altogether 20 kg of solids are fed into the system, resulting in total suspended solids (TSS) concentrations in the inflow of 1.2–3.5 g/L. This part of the test will simulate the solids load of 1 year. Finally a maximum water flow of 4 L/s is applied for 15 min without the addition of solids in order to estimate remobilisation of the sediment. For each flow rate at least 10 time-proportional effluent samples of 1 L are taken in order to obtain a representative composite sample. Individual separation efficiencies for each flow rate are determined from the ratio of the measured solids load in the effluent and the calculated solids load in the influent, respectively. By taking the results of the remobilisation test into account, the overall separation efficiency is calculated (DIBt 2011).
For our own test method (called HS-OWL test method) that was developed independent of the DIBt approach, a mixture of 42% of fine mineral particles and 58% of dried and sieved potting soil is used. The mineral particles have the characteristic diameters d10 = 36 μm, d50 = 120 μm and d90 = 500 μm, while the soil particles' diameters are <800 μm. Since the latter also contain inorganic compounds, the test material has a total organic content of 30%. The flow rates used (0.2 L/s, 0.5 L/s, 2 L/s for loading and 5 L/s for remobilisation, respectively), are higher than proposed by DIBt. Meanwhile, the total solids load (153 g) is lower because influent concentrations of 150 mg/L are applied, and the test period (28 min) is considerably shorter. About 25 time-proportional effluent samples of 100 mL are taken for each flow rate. As with the DIBt test, the overall separation efficiency is calculated (Fettig et al. 2008). In both tests, the particles are quantified as TSS according the German Standard DIN 38409 H2.
In addition to the lamella plate settler, two commercial decentralised settling systems, the Centrifoel Separator (Roval Umwelt Technologien, Germany) and the Aco System (ACO Tiefbau, Germany) that are designed to treat runoff from areas of 400–500 m2 like our system, were tested in the laboratory in order to compare their performance with the lamella plate settler. The Centrifoel Separator has three settling compartments which are flowed through successively, while the Aco System is a sludge trap with a deflecting plate and other devices that minimise turbulences in the water flow. Details are given by Sommer et al. (2015).
For the field study the lamella system was connected to an inlet that received runoff from a road area of 420 m2 frequented by about 12,000 vehicles daily. The inflow had to pass through a sieve first which removed leaves and other coarse material. The chamber of the system was equipped with a device for continuous flow measurement that calculates the flow rate from the storage depth recorded in intervals of 10 s based on a corresponding calibration curve (UFO-Ex, W.A.S. GmbH, Germany). Flow-proportional samples were taken from the half-open container at the inflow pipe and from the outflow pipe (TP 5, MAXX Mess- und Probenahmetechnik GmbH, Germany). A sample of 50 mL was collected after each 50 L of throughput, from which weekly composite samples were prepared. The system was cleaned twice (after 6 and 12 months), and the sediments were quantified.
Rainfall intensity data for the period of the field study are given in the Supplementary Material (available with the online version of this paper). For different events, rainfall data were compared with the flow through the system, resulting in event-related runoff coefficients between 0.6 and 1.0. As documented in the Supplementary Material, the runoff coefficient was 0.76 on average. A time series of the recorded storage depths in the system for the whole period as an indirect result of runoff variation is shown in Figure 6. The corresponding statistical evaluation of the hydraulic loading of the system is given in Figure 7.
RESULTS AND DISCUSSION
The results for all three systems are summarised in Table 1. The average removal efficiencies obtained are 65% (HS-OWL) and 74% (DIBt), respectively. According to the HS-OWL method the lamella system is comparable to the commercial systems, whereas it is clearly better based on the DIBt procedure. However, the Centrifoel system shows a less favourable performance in the DIBt test, whereas opposite results are obtained for the other systems. Thus there seem to be shortcomings with the test procedures, in particular with the DIBt test method, which might be partly caused by the results of the remobilisation test.
|Test method||Lamella system||ACO SSA||Centrifoel|
|Test method||Lamella system||ACO SSA||Centrifoel|
After both phases the sediment was removed, dried and weighed. From about 230 m3 of runoff passing through the system during 1 year, altogether 8.6 kg of solids with an average organic content of 35 ± 9.8% were separated. When calculated from flow-weighted TSS concentrations the mass of solids that was separated should have been 10.1 kg. In view of the conditions of sampling, these data are consistent. Moreover, no deposits which could have affected the flow were detected on the lamella plates.
The average TSS removal efficiency for the whole period was 30% and thus similar to the removal efficiency for zinc, whereas the other metal compounds were removed to a lesser extent (Table 2). However, TSS removal in the field study was clearly lower than in the laboratory tests. A possible reason is the fraction of fine particles in road runoff. During 5 weeks TSSfine concentrations (d < 63 μm) were determined in addition to the total TSS. In influent samples TSSfine amounted to 50% of total TSS. While the average separation efficiency during this period was 34% for TSS, the result was only 24% for TSSfine. These findings indicate that the efficiency of a settling system for the removal of fine particles might be limited. In order to find a conclusive explanation, the hydraulic conditions are evaluated in more detail first.
|Parameter||Phase I||Phase II||Whole period|
|Parameter||Phase I||Phase II||Whole period|
For the design of lamella settlers in semi-central systems, Weiss (2014) and Fuchs et al. (2014) recommend a maximum surface loading of 4 m3/(m2 h), equivalent to a storm intensity of 15 L/(s ha). In our study 78% of the total water flow passed through the system under these conditions. If we assume a surface loading of 5 m3/(m2 h) as a design value, the corresponding storm intensity will be about 18 L/(s ha), based on a runoff area of 420 m2 and an average runoff coefficient of 0.75. Fuchs et al. (2014) derived a maximum surface loading of 5.6 m3/(m2 h) and a mean surface loading of 1.9 m3/(m2 h) from their data. With a mean inflow concentration of 134 mg/L they obtained 49% removal of TSS. In other recent studies with settling systems for storm water treatment, particle removal has also been only moderate. For a hydrodynamic separator operated at surface loadings of 16–40 m3/(m2 h), Lee et al. (2014) found TSS removal efficiencies between 9% and 43%. Langeveld et al. (2012) investigated a semi-central lamella separator designed for a surface loading of 1 m3/(m2 h) and obtained these results: 34% removal of TSS, 23% removal of zinc, 21% removal of copper and 36% removal of lead. Thus it is concluded that the operating conditions of our system were not much different from those of semi-central systems. Taking into account that the removal efficiency for TSS might be improvable by up to 6% using a nozzle distributor for the inflow into the lamella plate pack (Salem et al. 2011), the efficiencies are also in the same range.
When looking at removal efficiencies the hydraulic conditions are interconnected with the size of the particles to be separated. As stated before, our grab samples showed a fraction of 50% of fine particles (d < 63 μm). Fuchs et al. (2014) found even larger fractions of 70–90% of TSSfine, while Selbig et al. (2016) measured a fraction of about 70% (mean value) in runoff from a parking lot and referred to data from the US Nationwide Urban Runoff Program where a generalised particle size distribution with a TSSfine fraction of even 88% is reported. On the other hand, Charters et al. (2015) observed a mean value of 39% for the fraction of TSSfine in urban runoff (TSS concentration of 158 mg/L), and Herr & Sansalone (2015) reported a value of 85 μm for the mean particle diameter (d50). However, in both studies a large variation of particle size distributions measured for different events was observed: the diameter d50 varied between 12 μm and 103 μm (Charters et al. 2015) and its standard deviation was ±76 μm (Herr & Sansalone 2015). Thus there seems to be some evidence that the fraction of fine particles in road runoff can be quite high during certain periods.
The removal of particles from road runoff by sedimentation can furthermore be affected by the following circumstances:
A low density of the solids. For road dust, a particle density of 2.2 g/cm3 was found in an orienting analysis in the field study. However, organic matter can take up water which will result in a reduction of the density (negative effect) but on the other hand in a swelling of the particles and thus in a positive size effect on sedimentation.
The shape of the particles in combination with a certain roughness of their surface. Particles with irregular shapes are likely to settle more slowly. However, in this respect there is still a lack of knowledge, so the effect of shape cannot be assessed at this stage.
The first flush effect. According to Kayhanian et al. (2012) the so-called concentration first-flush has been observed in numerous studies; i.e. the concentrations of solids and dissolved pollutants in road runoff tend to be higher during the first minutes after rain has started because of drag effects on the road surface. Although Sun et al. (2015) state that a first flush in their study area was not common based on strict definitions, the data of Bach et al. (2010) and Barjenbruch et al. (2016) support the first flush effect. Since the flow rates and thus the surface loadings of the system are often quite high during these periods, the removal efficiencies will then be particularly low.
An uneven flow distribution. It is likely that flow velocities through different channels of the plate pack are not constant. In that case even large particles can pass through the system due to high local flow rates. Since measurement of the flow distribution was not possible, this effect cannot be assessed in detail. Further work will include flow modeling, which might give more insight into the hydraulic conditions inside the system.
With respect to the test methods it can be stated that the flow rates seem to represent the range encountered in the field: the DIBt method corresponds to surface loadings between 0.9 and 8.8 m3/(m2 h), while the HS-OWL test even covers the range between 1.8 and 18 m3/(m2 h) and thus more than 90% of the water flow rates through the system (Figure 7).
However, it is strongly suggested to modify the test methods in such a way that the test material better represents the particles in runoff regarding size and composition. The DIBt test could be adjusted by using a second material with lower density besides Millisil and by reducing the influent concentration to realistic values. The HS-OWL method could be modified by using a larger fraction of both mineral and organic fine solids. Furthermore the first flush effect could be accounted for by conducting the high flow rate test with a higher influent concentration. The goal should be to obtain a better predictability of the systems' performance in the field based on laboratory test results.
The decentralised lamella system developed has shown good hydraulic properties. The bypass that limits the flow rate through the system to 3 L/s was activated only seldom during the field study.
The laboratory test results for particle removal were similar or better than those of commercial systems of the same size according to both test methods applied. However, the TSS removal efficiency in the field was considerably lower.
Among possible explanations for this discrepancy the hydraulic conditions may play a certain role, but it is likely that parameters such as size, shape and density of the particles are more important. In addition the first flush effect or an uneven flow distribution inside the plate pack can reduce the TSS removal efficiency. More studies particularly under field conditions are needed to clarify this puzzle.
The laboratory test methods need to be adjusted in order to better represent the practical circumstances and to provide more realistic results. This applies first of all to the type and concentrations of the test material. Meanwhile the hydraulic conditions seem to be similar to those encountered in the field.
The project was funded by the Ministry for Climate Protection, Environment, Agriculture, Conservation and Consumer Protection of the State of North Rhine-Westphalia, and by Urban Drainage Höxter GmbH.