Stormwater treatment evaluation on a commercial site in Nambour, Queensland

Water sensitive urban design (WSUD), low impact design (LID) or Sustainable Urban Drainage systems (SUDs) and manufactured stormwater treatment devices have been incorporated into new developments for many years (Braune & Wood 1999; Taylor & Wong 2002). Though vegetated assets, including biofilters and wetlands, are popular solutions for large developments, many smaller, high intensity land-uses implement manufactured devices to conserve site area for higher value uses (Hipp et al. 2006; Sample et al. 2012). Due to the cost and complexities of laboratory and field testing, these manufactured stormwater quality improvement devices (SQIDs) have not been extensively studied and published (Cates et al. 2009). Through a collaborative effort between the authors’ respective organisations, a robust protocol was developed to evaluate the pollutant removal of a Class 1 SQID at a commercial site in Nambour, Queensland. This protocol was based on the Auckland Regional Council Proprietary Device Evaluation Protocol (ARC 2012) and the Stormwater BMP Database protocol (Geosyntec Consultants and Wright Water Engineers, Inc. 2009) and developed prior to the release of the Stormwater Australia protocol (Stormwater Australia 2014). The stormwater treatment train installed at this site included a SPEL Stormceptor^® (a BS EN 858-1 compliant Class 1 horizontally configured two-chamber SQID) equipped with a coalescer unit to promote settling. This device was specifically designed to target suspended solids and hydrocarbons, with nutrient capture possible but unknown. Testing was underway from March 2014 to April 2015 and recorded 59 events. Of these events, 18 qualified against the protocol. This paper presents the results of the field monitoring against the original protocol.

Testing has been under way for more than 12 months at a commercial premises at Nambour, about 101 kilometres north of Brisbane, Australia. The township is located in the sub-tropical hinterland of the Sunshine Coast. The site is commercially-zoned with a total area of 2,800 m² of which approximately 1,848 m² is roof area (66%), 924 m² is impervious driveway (33%) and the balance, 280 m² (1%) is landscaped area as shown in Figure 1. The stormwater treatment train includes an underground rainwater tank for roof water capture and reuse, gully pits and surface drains, and a Stormceptor^® Class 1 treatment device. Surface runoff from the site drains into the gully pits or surface drains and into the underground pipe network. Roofwater drains firstly to the underground rainwater tank, then overflows to the underground pipe network that discharges to the Class 1 device. Outflow from the SQID discharges to the underground drainage network discharging downstream into Petrie Creek, a sub-catchment of Maroochy River. This treatment train configuration is typical for a commercial development in southeast Queensland. The site is also representative of typical applications for this treatment device.

Water samples were collected from the inlet and outlet of the SQID as shown in Figure 2 during rainfall events and tested for total suspended solids (TSS), total nitrogen (TN) and total phosphorus (TP), ammonia N (NH₃), nitrogen Oxides (NOx), orthophosphate (PO₄) and particle size distribution (PSD). Dry weather sampling was also performed daily for one week between rainfall events to evaluate potential leaching. Qualifying rainfall events were triggered from rainfall events with more than 0.75 mm rainfall depth within a 30 minute period. A minimum antecedent dry period (ADP) of 72 hours was adopted in the protocol, based on previous research indicating that pollutant concentrations did not reach detectable levels until after this time (Liu 2011). Subsequent experience from similar field monitoring programs has refined the ADP to 6 hours following observations that rainfall volume and intensity can be more influential than ADP (Lucke & Nichols 2015). Therefore, events were defined by an absence of rainfall for 6 hours and influent pollutant concentrations above detection levels. Automated water samplers were programmed to collect composite samples in 200 mL discrete aliquots at 1,000 L catchment flow intervals. Considering the total volume required for laboratory analysis, a minimum of eight aliquots was initially considered a desirable number of aliquots to ensure a representative event mean concentration was achieved. Advice from the laboratories over time indicated that this volume could be reduced, however, where events were of sufficient volume, eight or more aliquots were targeted. The detailed sampling protocol is detailed in Table 1. The samples collected by University of the Sunshine Coast (USC) were transported to the USC (NATA registered) laboratory for processing and testing of the parameters.

Automatic water samplers were installed in the access chambers to the Class 1 device to collect water samples at the inlet and outlet. The first sampler inlet (a) location was immediately upstream of the inlet chamber in the pipe. The outflow collection point (b) captured water from the internal treated flow pipe carrying water that has passed through the second settling chamber and coalescer. The treatment flowpath is shown in Figure 2 by arrows. Design treatment flows for this installation was 20 L/s.

The samplers were controlled by a CR800 datalogger and wireless modem to facilitate remote access and monitoring. A Waterlog H-3401 0.2 mm tipping bucket rain gauge was mounted on an adjacent structure and connected to the CR800 datalogger to provide the rainfall trigger for sampling. Flow through the SQID was monitored by a Magflux^® flow meter. Due to the constraints of retrofitting monitoring equipment after installation, the treated outflow from the Class 1 device was measured at the outlet. An ultrasonic flow meter was also installed in the bypass pipe.

After 14 months of monitoring, 59 rainfall events above the trigger criteria were recorded. However, only 18 events qualified under the testing protocol (Lucke & Nichols 2015). In particular, low influent concentrations contributed to many events being excluded as non-compliant. Table 2 presents the water quality data observed at the inlet and outlet. Being specifically designed to target suspended solids and hydrocarbons, the results confirm effective reduction of TSS. However, TN removal was minimal and TP reductions were significantly influenced by low influent concentrations.

The treatment device at Nambour is a ‘wet sump’ device in that it retains water in its chambers between storm events. Samples were collected on days after rainfall events on 14/3/15 and 7/4/15. As the auto-samplers require flow to collect samples, the dry weather samples were collected manually as grab-samples. The results of water quality analyses on the samples is presented in Table 3.

Comparison of the observed influent concentrations from the catchment against the relevant local guidelines (Mackay Regional Council 2008; Melbourne Water 2010; Water by Design 2010) indicate that the site is producing significantly lower pollutant levels. Key parameters are presented in Table 4. Initially this was thought to be related to the influence of the large percentage of roof area and rainwater tank in the overall treatment train. However, recent research has concluded that the Nambour results may be indicative of generally lower pollutant concentrations across all urban catchments at a lot scale (Lucke et al. 2018).

This has implications for designing stormwater treatment measures to achieve policy water quality objectives of 80% TSS, 60% TP and 45% TN load reductions (State of Queensland 2013). Where pollutant concentrations in runoff are closer to the background, or irreducible, concentrations than previously understood, the load may not exist for removal by traditional measures. Alternately, it will require much larger treatment measures than previously anticipated to achieve the design load reductions. Of note from the Nambour results, even at the low influent concentrations observed, the TSS reduction metrics were at or very near the TSS objective.

Evaluation of the data demonstrates ERs of 83%, 11% and 23% for TSS, TP and TN respectively. Of interest, the 3 metrics for TSS are similar but the TN and TP metrics have confidence intervals above ±20%. Normality tests confirmed that the TSS and TN datasets were log-normally distributed, however, parametric testing (Student's t test) confirmed only TSS influent and effluent were significantly different at the 90% confidence interval. Non-parametric tests (Wilcoxon-Mann-Whitney Rank sign test) on both TN and TP datasets could not confirm significant difference. The datasets that are considered statistically to be significantly different, are shown in bold in Table 5.

The results in Table 3 indicate that there is a small change from inlet to outlet in dry weather conditions for TSS and TN. This increase should, however, be considered in the context that the analytical variability on replicates has been observed to be up to 2.8 mg/L for TSS, 0.3 mg/L for TN, and 0.02 mg/L for TP. For the low influent concentrations observed this could mean variability in ER and CRE results of ±254% TSS and ±130% TN. Also, it should be noted that the TSS results are mostly below the typical LOD of 5 mg/L and were only tested to a 1 mg/L LOD due to the low TSS observed on initial samples. Also, evaluating the effluent concentrations over time does not show a conclusive trend that the nutrients are released with storage time. The authors recommend further testing over a longer period to definitively identify any nutrient loss.

USC, DEC and SPEL have worked together to adapt international protocols to suit local and regional conditions on a commercial site at Nambour, Southeast Queensland, Australia. After monitoring from March 2014 to April 2015, results for 18 protocol-compliant events indicate removal efficiency (ER) of 83% TSS, 11% TP and 23% TN for the Stormceptor^® Class 1. Based on the statistical analyses of the dataset it confirms that the TSS performance is statistically significantly different at a 95% confidence level. As anticipated from the technology's design to target suspended sediment and hydrocarbons, the SQID performance for nutrient removal was variable and not statistically significant. Dry weather testing of the device has shown minimal variation in the pollutant concentrations, and variation that is within the observed analytical variability. Accounting for this variability, there does not appear to be conclusive evidence that the wet sump exports pollutants over time, and the authors recommend further research before reaching definitive outcomes. Multiple events recorded during the monitoring period had influent concentrations below detection levels, and generally were much lower than anticipated from local guidelines. Recent publications suggest these guidelines may require revision.

The authors acknowledge the contributions to the field monitoring and laboratory analysis by staff and students at the USC, especially Terry Lucke, Peter Nichols, Michael Nielsen, Ronald Kleijn and Brian Pearson.