Experimental study of sediment washout from stormwater sumps

Suspended sediments in stormwater are generally considered as contaminants, since they can increase turbidity and facilitate the transport of pollutants to receiving water bodies, leading to many water quality, habitat, and aesthetic problems (Ashley et al. 2004). Catchbasin sumps have long been used as sediment traps in stormwater collection and conveyance systems (Lager et al. 1977). A number of studies have investigated the performance of sumps as stormwater quality control devices by evaluating their sediment and pollutant removal capacities (Lager et al. 1977; Aronson et al. 1983; Pitt 1985; Butler & Karunaratne 1995; Pitt & Field 1998; Faram et al. 2003; Andoh et al. 2007; Tang et al. 2016; Yang et al. 2018). The traditional catchbasin sumps rely on simple gravitational separation and have an overall satisfactory performance in capturing large debris, retaining floatables and minimizing deposition in drains.

Standard sumps are commonly used in stormwater collection systems. However, during less frequent storm events, if maximum treatment rates are exceeded, previously captured sediments can resuspend and discharge out of the devices. Most studies of proprietary settling devices have focused exclusively on the efficacy of initial pollutant removal. While it is a good indicator of the device's abilities in low-flow conditions, it does not provide sufficient information on how the system will work during heavy rain events (Andoh et al. 2007).

For sedimentation, settling velocities are an important parameter (Fan et al. 2003; Ferguson & Church 2004), whereas for resuspension, shear stresses are crucial. At low flow rates, the shear forces are too low to carry a significant amount of suspended solids, allowing them to settle, whereas during precipitation events, flow rates and shear forces are high enough to resuspend these solids and transport them downstream. Avila et al. (2008) found that the flow rate, particle size, overlying water layer thickness, and their interactions are significant factors that affect the scour of sediment in a conventional catchbasin sump, through a 2⁴-full factorial experimental design. The overlying water layer above the sediment has an important function in protecting the sediment from scour. The inlet geometry has a considerable effect on the sediment resuspension due to impinging impact, and an inlet pipe without any energy dissipating device can cause more washout than a wide rectangular one. Following experimental studies using heterogeneous sediment mixtures and different ranges of particle diameter about 45–4,750 μm (Avila & Pitt 2009; Avila et al. 2011) revealed two protective mechanisms: (1) the depth of the overlying water significantly reduces the scour potential, and (2) an armoring layer by large particles is formed, which provides protection for finer sediments below.

Extensive lab testing and field work were conducted in the St. Anthony Falls Laboratory at the University of Minnesota to understand the sediment retention and washout of standard sumps (Wilson et al. 2007; Saddoris 2010; Saddoris et al. 2010, Howard et al. 2011, 2012; McIntire et al. 2012). It was found that as discharge decreases or particle size increases, removal efficiency of the standard sump increases. For the same particle size and discharge, the removal efficiency increases with increasing the chamber size. Along with settling velocity, the sump diameter, sump depth, and flow rate are important in determining the removal efficiency (Howard et al. 2012). The washout behavior of the accumulated sediment can occur under the condition that the energy of the water flowing into the sump should exceed the power required by the resuspension processes (Howard et al. 2012). During washout testing, sediment at the bottom of the sump was reshaped by a large vortex, and under high incoming flows the vortex caused previously captured sediment to be washed out.

When evaluating the performance of catchbasin sumps and related hydrodynamic separators in stormwater systems, both pollutant-removal efficiency and pollutant-retention efficiency should be taken into account. Compared to vortex separators, traditional sumps are surprisingly prone to sediment washout (Andoh et al. 2007). As such, these catchbasins may represent a pollution source because of ineffective retention. Pitt (1985) reported that catchbasins can capture sediments up to approximately 60% of the sump volume. Studies suggest that increasing the frequency of maintenance can improve the performance of catchbasins, particularly in industrial or commercial areas (Mineart & Singh 1994). However, there are only very limited guidelines for the schedule of routine maintenance, which do not have a rigorous basis of physics (Erickson et al. 2010; Guo 2017). In addition, internal structures can be better designed to enhance the pollutant removal performance (Andoh et al. 2007; Pathapati & Sansalone 2009, Howard et al. 2011; Ma & Zhu 2014; Brar et al. 2016), once the dominant parameters of the retention process could be thoroughly understood, particularly the washout problem.

On account of the wide use of sumps in urban drainage systems but limited data for washout, a detailed investigation is required for a better understanding of the mechanism and quantifying scour processes. This study investigated the sediment washout from a model sump for different particle sizes, sediment depths, and flow rates. The objectives are to quantify the washout amount of previously deposited sediment under different conditions, provide insight into the mechanism, and guide the maintenance program development and/or optimization.

The supply flow rate was measured by a magnetic flow meter (KROHNE OPTIFLUX 2300F), with a relative error of approximately 0.2%. A camera (Sony FDR-AX60) was used to record the flow at 50 frames per second, of which the video could be used to perform image measurements. A handheld three-dimensional scanner (Shinning 3D Einscan-Pro 2X 2020) was used to digitally rebuild the eroded bed, of which the scan accuracy was about 0.05 mm. Removable filter bags were fitted at the outlet during the washout testing, and the captured solids were dried by a drying chamber (DHG-9240A Heating and Drying Oven) and weighed by a high precision electronic scale (QUINTIX35-1CN). The differences between measured temperatures at the beginning, middle, and end of each test were below 0.5 °C, thus negligible.

To investigate the washout process and mechanisms, sands of homogeneous particle size distribution (PSD) were used. Three size groups of quartz sand (specific gravity = 2.65) were tested in this study. A sieve analysis was conducted to determine the characteristic diameters of the sediment. Sand A corresponded to a sediment material, of which the median diameter d_s was 0.6 mm, and the uniformity coefficient (d₆₀/d₁₀) was about 2.3. The median diameter of Sand B was about 1.2 mm, with a uniformity coefficient about 2. Sand C had a median diameter of 2.5 mm and a uniformity coefficient about 1.5. As the discharge and sediment height were expected to be significant parameters in determining the flushing, the test cases were designed as listed in Table 1.

The procedure for the experiments was as follows: (1) sediments of known particle sizes (Sand A, B or C) were loaded with wet sump to a desired height and the flow was started at a preset flow rate; (2) washout solids at the pipe exit were collected using filter bags every minute for at least 40 min; and (3) the amount was determined after drying. The quantitative analysis of scour was based on the amount of washout solids.

The sediment deposit in the sump would change the effective volume/depth of the chamber, leading to changes in the velocity distribution, particularly the magnitude of shear stress at the bed surface. As a result, the solids could be entrained, and the surface of the sediment deposit became distorted due to the scour (Figure 2). During such a process, the resuspension at the downstream side of the sump (below the outlet) was significant and the erosion increased continuously at the early stage of flow. The deposition at the upstream side of the sump (below the inlet) was also eroded, but the scour depth was smaller, compared to that of downstream side. Entrainment of particles was intense at the beginning, as shown in the video capture at t = 2 min, and the change of scour hole becomes indistinguishable after 30 minutes.

If there was no overlying water initially above the deposited sediment, the jet flow would plunge into the bed, inducing extra disturbances during the flow build-up process. As a result, the erosion at the very beginning was much larger. Nonetheless, the percentage of the first five-minute washout mass was about 30%, which was nearly identical to that of the case with initial overlying water layer above the deposited sediment, as calculated from the data in Figure 3. After a thorough analysis, it is believed that the influence of the absence of initially overlying water was negligible.

The growth rate of the total mass of washed out solids continuously decreased, as shown in Figure 3. The quantitative analysis of deposit change by image measurements shows that after a short time period, e.g., 16 min for the case with Q₀ = 8 L/s and y_s = 4 cm, changes of scour depths below the inlet and outlet were found to be marginal. In general, the scour hole rapidly develops at the initial stage and then approaches a static equilibrium condition. This is because the acting shear stress on the sediment surface decreases with temporal development of the erosion (Avila et al. 2011, Ota et al. 2017), until equilibrium is reached.

The expected results for the washout experiments are that, for cases with given particle size and deposition depth, there exists a threshold value for the incoming flow rate to enable washout, and vice versa. It can be seen from Figure 5(b), when the initial depth of bed surface was larger than 8 cm, the maximum washout mass flow rate is only about 33 g/min. Thus, at Q₀ = 8 L/s, the threshold value of the initial depth of bed surface y_s is around 10 cm. From the temporal perspective, when y_s >= 10 cm, the relative difference of the washout mass flow rates over the testing process was less than 15%.

The results show that when Pe > 1.5, the maximum ESSC was about 109 mg/L, and the average ESSC was very low, generally less than 9 mg/L. However, the ESSC can increase rapidly when Pe < 1, particularly for cases with large F_p values. A fitted washout function proposed by Howard et al. (2012) was shown to accommodate scaling for standard sumps, but only one sediment gradation (US Silica F110) was tested. The influences of particle size and inflow rate are evident in Figure 8, although the dimensionless parameters collapse the data into generalized performance behavior. It is worth mentioning that when the height of sediment layer is equal to outlet pipe invert, i.e., y_s = 0 and Pe = 0, it is almost impossible to develop a generalized washout function. The dispersion of observed data was also reported in Saddoris (2010), where the data set for a Downstream Defender was fitted by separate curves. Nonetheless, the overall trend of the ESSC with regard to or is valid, and when > 2, the ESSC would drop down to an acceptable level.

Froude number scaling is applicable for sump washout studies as the flow is controlled by gravitational forces. For a prototype sump with a diameter 1.2 m, the length ratio L_r = 3.16 and the discharge ratio Q_r = 17.7. For the discharges of 4–8 L/s in the model sump, the prototype flowrates will be 70.8–141.6 L/s. The scaling for settling and scour of solids in sumps was given in Guo (2017), and scour similarity is often difficult to achieve simultaneously with Froude number scaling. Thus, it is recommended that scaling laws should be tested with combined model and prototype studies. In this study, the results should only be used as a diagnostic technique for sediment washout from sumps.

The presentation and discussion of experimental results show that as the sump depth decreases because of sediment capture, increased washout will occur, leading to decreased removal efficiencies. As a result, the primary maintenance activity required with stormwater sumps is cleaning of deposited sediments to minimize resuspension and washout. Cleanout frequency is one of the main concerns of stormwater management. In this study, a group of dimensionless parameters were found as dominant, which can be used to determine the need for action.

In studies of Pitt (1985) and Mineart & Singh (1994), the cleanout was associated with the basin volume occupied by sediments. In contrast, the depth of deposit surface y_s should be used instead (or ⁠), according to the results of this study. When the washout processes are well understood and threshold values are set in specific cases, the cleanout frequency can be reasonably predicted. However, many variables affect the accumulation of sediments, including watershed hydrology, stormwater runoff, suspended solids particle size distributions and sump dimension. Therefore, accurate prediction of net sediment accumulations in a specific device in a specific drainage basin is challenging.

Inflow variability over time was ignored in the present study, but is important to consider as it would be representative of typical conditions. In traditional methods, the dynamic inflow can be considered as quasi-steady, provided that the change is not too rapid. However, for washout of deposited sediment in a stormwater sump, the differences also lie in the initial bed form. In addition, the influence of the drop height between the inlet and outlet was not examined in this study. It is anticipated that the washout function will be profoundly affected by the drop height. It is important to note that armoring of particles may influence sediment washout under actual operating conditions; however, the effects of armoring were not tested. Thus, it is recommended that these parameters should be taken into account in future studies.

This study presents an analysis of sediment washout in a stormwater sump model based on experimental measurements under a variety of operating conditions. The washout mass decreased exponentially along with time as the predeposited sediments were continually resuspended and flushed out. The sediment bed morphology would evolve during the washout process and approach a final form, with the largest local scour depth being at the downstream sides of the sump (below the outlet), due to the combined effect of vertically oriented vortex motions and an erodible bed. The scour hole volume (or the total mass of washout particles) increased with the increasing of discharge and the reduction in particle size.

A number of parameters that were expected to affect sediment washout were taken into account in the dimensional analysis, through which dimensionless effluent concentrations were correlated with other non-dimensional parameters. It is believed that there is no universal washout function for such a complicated phenomenon. Nonetheless, a dimensionless parameter, ⁠, was found very useful to assess the washout mass flux and predict maintenance schedules with sufficient knowledge on sediment removal.

This work was partially funded by the Natural Science Foundation of Ningbo (202003N4134), the Natural Science Foundation of Zhejiang Province (LY21E090003), the Key Research and Development Program of Zhejiang Province (2020C03082), and the Fundamental Research Funds for the Provincial Universities of Zhejiang (SJLZ2021004). The authors are grateful to Dr Jiachun Liu and Dr Chunling Wang for their help in the experimental program, and the insightful comments from Dr Shengtao Du and Dr Yu Qian are also highly appreciated.

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

The authors declare there is no conflict.