Development of treatment media for advanced stormwater treatment at an industrial site

The individual stormwater discharge permit for Santa Susana Field Laboratory (SSFL) contains numeric thresholds and requires that the discharge locations on the site be monitored during all storms for compliance. Activities at the SSFL site are now limited to demolition, remediation, and restoration. Much of the site is open space. Due to severe site constraints at two of the compliance monitoring discharge locations (i.e., natural drainages located near the property boundary), ‘end-of-pipe’ stormwater controls are not feasible. Therefore a watershed-based stormwater management approach has been implemented, including the use of sediment and treatment controls that replicate natural processes and are distributed throughout the watersheds to capture the pollutants of concern for the project.

A review of the literature on filtration media and onsite monitoring data (including existing treatment system performance results and previous media pilot testing studies) indicated that several promising media exist for consistently treating the pollutants of interest to the required effluent concentrations. This paper is a brief summary of the research conducted by Pitt & Clark (2010) and is not a review of stormwater treatment media. Clark & Pitt (2012) presents a brief overview of the literature on stormwater treatment (including media) and selection for specific treatment objectives.

The costs of the media examined during these tests ranged from about $0.33 to $3.30/kg (at about 1 to 2 g/cm³, or about 1,000 to 2,000 kg/m³); with potential media costs being significant given the large volumes required for the systems based on early designs (estimated media volumes for the project have ranged from 4,000–10,000 cubic meters). Special laboratory and pilot-scale tests were therefore conducted to evaluate candidate materials under procedures that have proven successful during past media investigations for stormwater treatment effectiveness. The objectives for the media tests included identifying media mixture components and operational characteristics to provide stormwater treatment of constituents of concern to below numeric effluent limits under a wide range of site and rain conditions, extending operational life of the media considering chemical capacity and sediment clogging, increasing treatment flow rate, and reducing costs.

A series of controlled laboratory and pilot-scale batch and continuous column tests were used to investigate a number of different treatment media for a wide range of pollutants, using natural stormwater as a test material. Stormwater was used as the test solution as it contained a broad range of constituents and particle sizes, reflecting typical interferences and competition for treatment as would occur for filtering, sorption, and ion exchange treatment of stormwater. A few of the concentrations in the collected stormwater were increased to represent expected site conditions. Some of the pollutants targeted included: cadmium, copper, lead, mercury, zinc, gross alpha and gross beta radioactivity, oil and grease, and TCDD. Detailed test descriptions (along with results and statistical analyses) are provided by Pitt & Clark (2010). The testing methods were divided into four main activities and developed to represent expected stormwater conditions:

• Clogging, breakthrough, and removal tests. In these traditional downflow column tests, the media were subjected to intermittent stormwater flows over several months. The primary information from these tests included: treatment flow rates, pollutant removal, and clogging/maintenance requirements.

• Contact time and media depth tests. These tests determined the effect of contact time (as determined by the media depth and porosity and the treatment flow rate) on pollutant removal.

• Media capacity tests and kinetics tests. These traditional isotherm and kinetics batch tests were adapted to meet the range of conditions seen in stormwater biofilter treatment systems. The purpose of these tests was to determine the amount of contaminant that can be retained by the media, given a specific contact time. These tests, unlike some of the tests reported in the literature, are multi-component tests with stormwater as the base test water that was modified to better represent expected site conditions.

• Aerobic and anaerobic effects on contaminant retention in media. These tests examined long-term retention of captured pollutants by the media under varying porewater chemical conditions.

As noted previously, the developed media has been successfully used for a variety of stormwater treatment facilities at the SSFL site (and other Boeing facilities).

Clogging often is the cause of premature failure of stormwater infiltration/biofilter devices. While pre-treatment may substantially reduce the load of larger particulates, a portion of the suspended and colloidal particulates pass through pre-treatment devices. These smaller particulates contribute both to surface clogging and depth clogging of the treatment devices. The long-term continuous column test results are summarized in Figure 1, showing the relative flow rates and loading capacities for the different media combinations tested. The cumulative load to initial maintenance (kg/m²) values describe the period from initial construction and operation to the time when the flow rate first dropped to a pre-determined maintenance trigger point. Scraping the surface was performed in an attempt to restore the flows, but had little long-term benefits. In most cases, clogging resulted in flow rates that could not be restored after two or three surface scrapping intervals. The total load to clogging was about 1.5 to 2 times the load before the first maintenance (except for the sand where the improvement was about 6.5 times due to material trapped more on the surface compared to the other media).

Example media filtration performance plots for copper are shown on Figure 2, a set of traditional box and whisker plots showing the influent concentrations used during the tests, along with the observed effluent concentrations for each of the long-term full-depth columns. This figure also shows the site effluent limit as a dashed red horizontal line (14 μg/L). Obviously, it is desired that all effluent concentrations would be below this line. However, because the long-term full-depth tests were conducted to determine the ‘life’ of the media, some media-constituent concentrations exceeded breakthrough near the end of the tests. The goal was to conduct the tests with influent concentrations above the benchmark limits, but close to the site conditions. The test stormwater generally met these objectives, but was also adjusted where necessary. In some cases, the test stormwater was already well above the site stormwater quality (copper, for example) and it was not possible to reduce them for these tests. The radioactive constituents were also not adjusted due to safety considerations and other limitations.

Figure 2 indicates very large reductions for most media types tested for copper, likely due both to the removal ability of the media themselves, plus the relative ease of reducing higher pollutant concentrations, compared to reducing low pollutant concentrations. This figure shows that the best reductions were found for the granular activated carbon (GAC), the peat moss (PM), and the mixtures that contained large fractions of these materials. The rhyolite sand-surface modified zeolite-granular activated carbon mixture (R-SMZ-GAC) performed well with most of the effluent concentrations below the site effluent limit. Similar excellent performance is shown for the layered site sand-granular activated carbon-site zeolite mixture. The other mixtures and individual components provided significant removals, but were not as consistently below the effluent limit for copper as these two mixtures and the GAC alone.

Similar analyses for lead and dioxin highlighted the challenges for constituents whose influent values were low and only periodically exceeded the effluent limits. However, almost all of the effluent lead concentrations were below the effluent limit value (5.2 μg/L), and in fact, most of the effluent concentrations were below the very low detection limits for lead (1 μg/L). Therefore, all of the media types and combinations provided excellent effluent quality for lead, as expected due to the large fraction of lead that exists as the more easily removable particulate fraction. The dioxin test results (TEQ, the toxicity equivalency quantity, relative to 2,3,7,8-TCDD) also indicated good control for some of the combination media columns tested. However, few data are available due to the complexities and costs of the dioxin analyses. The detected effluent concentrations were at least an order of magnitude less than the observed influent concentrations, indicating consistently good removals to close to, or below, the extremely low site effluent limit for dioxin (2.8 × 10⁻⁸ μg/L).

Pitt & Clark (2010) present similar results for all of the tested media and constituents. The Rhyolite sand (R), surface modified zeolite (SMZ), and GAC media mixture met all current site discharge permit limits, except for copper and mercury during periods with unusually high influent concentrations (such as when the influent total copper concentration was greater than 100 μg/L and when the influent mercury concentration was greater than 1 μg/L, which are not expected to occur at the site). This media mixture also had statistically significant removals for all constituents measured, except for phosphorus and the very-low-concentration gross beta radioactivity values. The addition of peat (PM) to the mixture improved removal of certain constituents with relatively low influent concentrations and tended to retain good removals when conditions allowed for only short residence times, such as during periods of high flows.

The tested media mixtures performed more consistently under a broader range of conditions than individual components examined separately (details shown by Pitt & Clark 2010). The mixtures capitalize on the pollutant removal strengths of their components, while providing other attributes that may address weaknesses (such as the release of cations or anions in large concentrations during ion exchange). Table 1 summarizes some of the major performance characteristics for the mixed media column tests.

The media mixtures that are most robust (longest run times before clogging, with moderate flow rates and suitable contact times for pollutant removal) are the Rhyolite sand, SMZ, and GAC mixture (blended mixture) and the Rhyolite sand, SMZ, GAC-PM mixture (blended mixture). They had very similar performance attributes. The added peat provided some additional benefits for metal reductions at high flow rates. The GAC in these mixtures (when mixed with the other components) also provided better control for a number of other constituents, including nitrates.

The following are the most important conclusions found during these tests:

Removal of specific constituents:

• Particulate removals (90% or greater) were high for all media, even down to very small particle sizes (as small as 3 μm). Good removals were seen for pollutants that strongly associate with the particulates (such as for aluminium, iron, and lead for most of the media).

• Radionuclides, mercury and TCDD also had significant and large removals (75 to 90 + % reductions) by most of the media mixtures tested when detectable influent concentrations were seen.

• Nitrate removal occurred only in media columns containing GAC. Removals typically were approximately 90% for fresh media, with decreasing removals as nitrate capacity was (relatively rapidly) exhausted. Removal of nitrate by GAC resulted in the release of phosphate.

• Phosphorus and phosphate had statistically significant and low to moderate concentration reductions (about 30 to 70%) removals in the Rhyolite sand, the site sand, the site zeolite and the surface modified zeolite.

• The filtered forms of cadmium, thallium, and nickel had statistically significant and moderate to high removals (50 to 90%) by most media, while filtered lead and filtered zinc were poorly removed (0 to 15%) by all of the tested media and mixtures. Filtered copper removals were statistically significant, but highly variable (5 to 80% concentration reductions). Except for filtered lead, these removal efficiencies could be related to the likely form of the pollutant in the influent (complexed vs. dissolved), with the complexed form being more difficult to remove. Filtered lead removal likely was poor because the influent concentration was very low and near the detection level of the analytical method.

Contact time:

• Some constituents and some media required a certain contact time before retention, while others were more capable of pollutant retention more rapidly and at lower influent concentrations. For example, when the contact time was less than 10 minutes, the metal removals were much less than for the longer contact times. Also, greater contact with GAC resulted in slightly better nitrate removals, while the greater contact time resulted in greater losses of phosphate from the media. This type of trade-off between improved removal and increased leaching was seen for several media-constituent combinations.

• Longer retention times can be achieved through deeper media beds or slower flow rates and larger surface areas. The column tests confirmed generally the results of the prior laboratory studies that showed that good removals could be achieved with relatively slow to moderate flow rates (5 to 60 meters/day) and moderate contact times of the water with the media (10 to 40 minutes).

Physical clogging and maintenance:

• Clogging by sediments generally occurred before chemical retention capacity was exceeded for most media mixtures. Highly effective pre-treatment (such as suitable sedimentation) is therefore critical to reduce the sediment load. This will result in longer useful lives of the devices and more economic use of the chemical capacity of the media.

• Maintenance by scraping the surface layers of most media was only partially effective at restoring the loading rate and for only short durations. Removal of the surface clogged layers from the sand, in contrast, did restore much more of the flow capacity. In all cases, these benefits were only temporary and after about 2 or 3 maintenance intervals, they ceased in being effective. It is expected that plants in a biofilter, with underlying media mixtures, will provide the longest run times before clogging, as they assist in the vertical migration of flows and captured material deeper into the media.

Trade-offs from ion-exchange reactions in the media:

• Both anion and cation exchanges occur in media filters. For the media tested here, phosphorus, chlorides, potassium, and sodium were found to be commonly released constituents. Shifts in pH were also found for some media, indicating changes in the H⁺ and OH⁻ ion concentrations with treatment. Use of mixtures where one component releases potential pollutants and another component captures the released ions provides the best overall pollutant removal performance. Sorption was also found to be a significant removal mechanism for many of the media types.

Development of anaerobic/micro-anaerobic zones and releases of previously captured material:

• Anaerobic conditions may develop in (poorly draining) filters during long dry periods, with a concurrent potential for release of some constituents (generally more of a problem for organics and nutrients than for metals). Anaerobic conditions lead to losses of previously captured contaminants and can increase the degradation of some of the media. During these tests, the release of nutrients was more severe during anaerobic conditions for the media having higher organic content or where removal was poor (the sands). In contrast, for most metals, retention was very good under all conditions for both anaerobic and aerobic storage conditions. Copper losses were found to be more common than for other metals. It is important that the design of the treatment systems minimize the potential for the formation of anaerobic conditions and poorly draining media.

Design summary:

• Fine grained media (such as the sands tested during these tests) clogged quickly and had poor flow rates, while large-grained media had high flow rates with very short residence times, generally resulting in poorer effluent quality. Therefore, the final treatment media composition should contain particles in the range of fine sands and silts to slow down the flow and increase the water's contact time with the larger-sized treatment media such as the GAC and zeolites. An alternative would be to use an effluent flow control on the biofilters (such as the SmartDrain^TM as described by Sileshi et al. 2018).

• The addition of GAC to a media mixture generally improved removal performance, especially for nitrates. However, the drawback to this GAC was the release of phosphorus and potassium. This can be minimized by the use of another component, such as peat, that has some affinity for these pollutants. Peat, when added in quantities of around 10% by volume, provided additional removal, especially for metals at shorter contact times, such as would be expected during periods of high flows.

• The GAC was the most important component in these mixtures, while the addition of zeolites was also needed. The specific choice of media mixture would be dependent on costs and specific ion exchange issues. Sand is critical to moderate the flow rates and to increase the contact times with the coarser media, unless other flow controls were used in the filter designs. Coarse sand (or other coarse granular material) also provides physical support decreasing collapse and compression of the media mixture under high flow conditions. The Rhyolite sand added some removal benefits compared to the silica sand. As noted, a small amount of peat added to the mixture increases metal removals during high flow rates. Therefore, the best mixture for removal of critical pollutants to levels that met the effluent discharge limits was the well-mixed combination of Rhyolite sand (30%), surface modified zeolite (30%), GAC (30%), and approximately 10% peat (percentages by volume). To minimize the leaching of constituents from the GAC, its concentration could be reduced, but then nitrate removals would be limited.

• The larger site biofilters incorporated native vegetation. Several cm of top soil was therefore added to the top of the media mixtures to help support the plant growth. It was not possible to locate an organic top soil that did not leach dioxins, as an example of unintended consequences of media mixture selections. Additional GAC was therefore added to compensate for this additional potential contaminant release. Discharges of phosphorus was not regulated by the site NPDES permit, but that is also a well-known problem associated with top soils added to biofilter media mixtures.

Funding for this project was provided by The Boeing Company and is gratefully acknowledged. Additional important research direction and assistance was provided by Geosyntec Consultants. This research was conducted as part of the Surface Water Expert Panel for the Santa Susana Field Laboratory (members include: Robert Gearheart, Humboldt State; Jon Jones, Wright Water Engineers; Michael Josselyn, WRA; Robert Pitt, University of Alabama; and Michael Stenstrom, University of California Los Angeles). In addition, many graduate students at Penn State – Harrisburg and at the University of Alabama participated on this project and their assistance was critical.