Stormwater Monitoring Using On-line UV-Vis Spectroscopy for Surrogate Measurements of Chemical Contaminants and Compared With Microbiological Data

doi:10.21203/rs.3.rs-761054/v1

Stormwater runoff contains a myriad of pollutants, including faecal microbes, and can pose a threat to urban water supplies, impacting both economic development and public health. Therefore, it is a necessity to implement a real-time hazard detection system that can collect a substantial amount of data, assisting water authorities to develop preventive strategies to ensure the control of hazards entering drinking water sources. An on-line UV-Vis spectrophotometer was applied in the field to collect real-time continuous data for various water quality parameters (nitrate, DOC, turbidity and total suspended solids) during three storm events in Mannum, Adelaide, Australia. This study demonstrated that the trends for on-line and comparative laboratory analysed samples were complimentary through the events. Nitrate and DOC showed a negative correlation with water level while turbidity and total suspended solids indicated a positive correlation with water level during the high rainfall intensity. The correlations among nitrate, DOC, turbidity, total suspended solids, and water level are the opposite during the low rainfall intensity. Nitrate, one of the main pollutants in stormwater, was investigated and used as a surrogate parameter for microbial detection. However, the microbiological data (E.coli) from captured storm events showed poor correlations to nitrate and other typical on-line parameters in this study, possibly explained by the nature of the stormwater catchment outside of rain events, where the sources of bacteria and nutrients may be physically separate until mixed during surface runoff as a result of rainfall. In addition, the poor correlations among the microbiological data and on-line parameters can be due to the different sources of bacteria and nutrients that end up into the stormwater drain.

Environmental Economics

stormwater

real-time monitoring

UV-Vis spectrophotometer

surrogate parameter

nitrate

E.coli

Access to a safe and reliable water supply is necessary for economic development and public health. Water-borne diseases, including cholera, amoebic dysentery, typhoid and paratyphoid fevers are commonly found and reported in the less-developed areas of the world (Bricker 2007, Petney &Taraschewski 2011, WHO 2006). Sewer is one of the main pollutants of water resources and sources of enteric diseases (Petney &Taraschewski 2011). With increasing populations, the availability of good quality freshwater in urban areas is becoming an urgent issue (Hamilton et al. 2005, Thayanukul et al. 2013). Additionally, expanding urbanisation has caused significant changes in local hydrological conditions, generated urban runoff, and led to the degradation of ecosystems (Göbel et al. 2007). Strong correlations were found between urbanisation, stormwater and wastewater discharges, and the transfer of pollutants and the quality of natural surface and ground waters (Göbel et al. 2007). Increase in population and water demand is making stormwater a more important resource. Thus, stormwater water quality monitoring will need to be developed as a component of integrated water resource management to minimise both environmental and public health risks.

In Australia, many policies have been developed to regulate wastewater discharges, but the same is not true for stormwater, and stormwater has remained as one of the main sources of pollution to urban freshwaters. Urban stormwater contains dissolved, colloidal and solid constituents in a heterogeneous mixture, which includes nutrients, heavy metals, organic and inorganic compounds (Bricker 2007, Gnecco et al. 2005). Additionally, urban stormwater runoff has been shown to contain large quantities of faecal microbes, such as Escherichia coli (Selvakumar &Borst 2006). Stormwater Quality Improvement Devices (SQIDs) have been applied to trap rubbish and pollutants that end up in stormwater drains, preventing large quantities of pollutants entering the stormwater drainage system (Hamilton et al. 2005). However, SQIDs are not effective in removing all contaminants, for instance, soluble nutrients, heavy metals, organics, suspended solids and faecal microbes can still flow from streets and gutters into creeks, streams and rivers, contaminating the surface water and posing a threat to our urban water supply (Eriksson et al. 2007, Greenway et al. 2002).

Microbial fate and transport in aquatic environments can be affected by various physical, chemical, and biological factors, which include season, temperature, nutrient availability, adsorption/desorption processes, hydrologic processes, predation and others (Ferguson et al. 2003, McCarthy et al. 2007, Selvakumar &Borst 2006). The fate and transport of these contaminants, especially those microbiological, have received high attention recently due to public health concern. Studies have shown an increased concentration of microorganisms during storm flows, indicating a relationship between flow magnitude and microorganism transport (Davis et al. 1977, Olivieri 1977). Other studies presented a significant relationship between microbes and the incidence of rainfall and rainfall intensity (Davies &Bavor 2000, Haydon &Deletic 2006, Kelsey et al. 2004). Davis et al. (1977) presented the correlations between microorganisms and both discharge and suspended solids for stormwater runoff. Duncan (1999) reported that faecal coliforms were strongly correlated with some stormwater quality parameters, such as total phosphorus and turbidity. Kelsey et al. (2004) and Mallin et al. (2009) evaluated the relationships between land use and faecal coliform bacterial pollution. However, most of these studies have been conducted in streams, rivers and/or estuaries (Ferguson et al. 2003, Kelsey et al. 2004, Mallin et al. 2009, Selvakumar &Borst 2006). Also, analyses are mainly performed on single grab samples, which tend to have a relatively low resolution to identify the inter-event trends (Mallin et al. 2009, McCarthy et al. 2012, McCarthy et al. 2007, Selvakumar &Borst 2006). Therefore, an alternative monitoring technique that can capture real-time data, such as nutrients, dissolved organic carbon, turbidity and total suspended solids, for stormwater will provide a better understanding of variations in microbe concentrations related to different water quality parameters. This information will be useful for decision making to perform rapid corrective actions regarding water management and public health.

On-line monitoring systems have been identified as useful tools with the introduction of UV spectroscopic techniques for wastewater and stormwater quality monitoring, such as total suspended solids and turbidity (Brito et al. 2014, Gruber et al. 2006, Torres &Bertrand-Krajewski 2008). Additionally, the application of UV spectroscopic techniques for qualitative and quantitative analyses of organic pollutants has been extensively investigated, although it may be limited for the survey of non-absorbing compounds, i.e. saturated hydrocarbons, carbohydrates, and almost all mineral species except oxyanions such as nitrite and nitrate (El Khorassani et al. 1998). Mrkva (1975) suggested that the UV absorbance at a certain wavelength is proportional to the content of dissolved organic matter present in the surface waters and in some types of effluent with predominant organic compounds of aromatic character. This is based on a basic interaction between UV light and unsaturated ionic or molecular structures (chromophores) (Thomas et al. 1999). For instance, the absorbance value at 254 nm has long been used for the estimation of non-specific parameters, i.e. COD and BOD₅, in water and wastewater (Briggs &Melbourne 1968, Brito et al. 2014, Lepot et al. 2016, Torres &Bertrand-Krajewski 2008), whilst the presence of nitrate was found to increase absorbance intensity for wavelengths around 210 nm etc (Causse et al. 2017, van den Broeke 2007).

South Australia depends heavily on River Murray water as a source water supply. The 60 km long Mannum-Adelaide Pipeline (MAPL) is one of the major pipelines supplying raw water from Mannum (a country town along the River Murray) to Adelaide, a city of over 1 million residents. Stormwater from the Mannum township enters the River Murray close to the pipeline intake with consequent potential water quality risks. In addition, there are other potential hazards, such as agricultural activities from upstream of the intake. It is, therefore, a necessity to implement a real-time hazard detection system and use this system to collect a substantial amount of data, helping the water authorities to develop preventive strategies to ensure the control of hazards potentially entering the drinking water sources (Chow et al. 2009).

This paper describes a stormwater monitoring study using an on-line UV-Vis spectrophotometer in the Mannum township near the Mannum-Adelaide Pipeline (MAPL). A key component of this study was an investigation to develop a technique to identify the chemical and microbiological signatures of water samples using on-line spectral data with the aims of determining the relationships among various parameters, rainfall events and E. coli to assess stormwater impact on water quality. The knowledge gained will be used to evaluate the feasibility of implementing future early warning systems to develop the hazard analysis and critical control point (HACCP) approach for water quality incidents (including stormwater runoff) as part of a decision support system for water supply operations.

2.1 On-line monitoring instrumentation and sampling system

An online UV-Vis spectrophotometer, s::can spectro::lyser, with optical path length of 0.5 cm was installed at a stormwater drain at Mannum, South Australia and equipped with an automatic carousel sampler (24 samples) with a flow proportional campsite sampling (FPCS) system. This spectrophotometer was designed based upon the principle of a photodiode array (PDA) spectrophotometer, which has no moving parts, and has the advantage of reagent free operation. Full spectrum absorbance (200–720 nm) was measured and calculated equivalents, such as nitrate (NO₃⁻), dissolved organic carbon (DOC), turbidity and total suspended solids (TSS) were determined.

The spectrophotometer was set up to record the full spectrum every 5 minutes, along with corresponding parameter values (calculated equivalents). During rain events, the level sensor was used to measure water level in the stormwater drain at 5-minute intervals, the auto-sampler was triggered by the flow condition, the collection of water samples was initiated when the water level was above a threshold of 25 mm by, the FPCS system, which is designed to collect individual samples during an event based upon the flow. The samples were collected based on representative distribution across the flow profile of the stormwater event, obtained from the storm drain water level sensor. Samples were collected more frequently at larger fluctuations of water level changes or during rapid changes of water level, and less frequently at smaller fluctuation. Samples from the three storm events, (1) July 29th 2010 (18 samples), (2) August 19th 2010 (13 samples) and (3) November 25th 2010 (24 samples), were captured and analysed. The collected water samples were kept at less than 4°C whilst being transported to the laboratory and filtered within 24 hours for microbiological analysis. The first event (Event 1) and the second event (Event 2) were conducted during the wet season (July–September) in South Australia (supported by rainfall data in 2010 provided by the Bureau of Meteorology). The last event (Event 3) was carried out to capture the stormwater quality after a period of the dry season to study the impact of seasonal change. The three events reported in this study were thus carefully selected to obtain the maximum amount of relevant information.

2.2 Laboratory analytical methods

2.2.1 Chemical

To validate the accuracy of the calculated equivalent results from the online UV-Vis instrument, 18, 13 and 24 grab water samples collected by the autosampler from Event 1, 2 and 3, respectively, at corresponding times, were analysed in the laboratory for NO₃⁻, DOC and turbidity using standard laboratory methods (APHA 2005).

2.2.2 Microbiological

According to the results of water quality parameters from the autosampler, about 9 to 10 samples with differences were selected for Escherichia coli measurement. E. coli numbers were estimated using Colilert-18 (Defined Substrate Technology, IDEXX Laboratories Pty. Ltd., Sydney) as previously described (Adcock and Saint, 1997, AS 4276.21–2005).

2.3 Statistical analyses

Statistical analyses were performed using Microsoft Excel and SPSS Statistics 24 (IBM). Differences between the on-line and laboratory measurement for various water quality parameters were tested using Passing–Bablok regression. Pearson Correlation was applied to evaluate any correlations among water quality parameters, water level and E.coli. Both correlation factor (r) and probability (p) values were used to determine significance.

3.1 Validation of online stormwater quality monitoring

Several water quality parameters (NO₃⁻, DOC, turbidity and TSS) were determined by the on-line spectrophotometer during the three rainfall events using partial least square (Shi et al. 2020). The grab samples were also collected by the autosampler from the three events at corresponding times and analysed in the laboratory (Event 1–18 samples, Event 2–13 samples and Event 3–24 samples). NO₃⁻, DOC and turbidity were measured in the laboratory with the exception of TSS. In the case of TSS, a volume of 300–500 mL of water sample is required for analysis and the auto-sampler was not able to provide sufficient volume for all the analyses. The TSS calculated equivalent value was used as an alternative. The other calculated equivalent results from the online spectrophotometer were compared with the laboratory results. These parameters have been validated previously for surface water samples, in good agreement with results obtained from laboratory analyses (Chow et al. 2007, Chow et al. 2008), but have not been validated for stormwater samples. Part of this study was to validate these parameters for stormwater samples. The calculated equivalent parameters were determined using the built-in algorithms provided by an acquisition software, ana::pro, with a global calibration.

In this study, the numerical values (concentrations) of stormwater quality parameters were obtained from both the online spectrophotometer and laboratory measurements. Passing–Bablok regression was used to compare the two methods. The results from Table 1 showed that the slope values (b values) from NO₃⁻ data were close to 1 and most of the intercept values (a values) were close to 0, indicating the online and laboratory measurements were comparable within the investigated concentration range. For turbidity results, the intercept values were negative and far from 1, while the slope values were close to 1 except Event 2, indicating there is a systematic difference (the online spectrophotometer overestimated the results) but no proportional difference between the two methods. However, neither the intercept values or slope values for DOC were close to 1, indicating the systematic and proportional difference between the two methods. The difference could be mainly caused by using global calibration in this study. The project team did not conduct local calibration during the monitoring period. This procedure requires entering the laboratory result of a sample to the online spectrophotometer control unit, then the software adjusts the calibration by matching the online spectral signature of that sample with laboratory results. If the local calibration was conducted, it would allow determination of the concentrations of stormwater quality using the spectrophotometer in real-time. Instead, a post data analysis on the collected data was conducted for the evaluation to determine the feasibility of applying the online instrument as a routine monitoring tool.

Although differences were shown for the online and laboratory measurements, good correlations of the water quality parameters were observed across all the events (Table 1). The Pearson correlation coefficients were also applied, and the results showed significantly positive correlation between the two measurements (online and laboratory) with the r values ranging from 0.603 to 1 and most p < 0.05 except DOC for Event 2. Overall, the -calculated equivalent parameters obtained via the online UV-Vis instrument showed comparable results with the laboratory analyses of stormwater samples for NO₃⁻, DOC and turbidity. These results provide support for the application of the real-time monitoring system for detecting the chemical contaminants present in the stormwater discharges.

Table 1

Differences and correlations between the on-line and laboratory measurements.
	Water quality parameters	Passing–Bablok regression (y = a + bx)	Pearson correlation coefficients (r)
Event 1	NO₃⁻	y = 0.9377 + 1.015x	0.789**
	DOC	y=-5.009 + 0.6887x	0.971**
	Turbidity	y=-75.81 + 1.071x	0.835**
Event 2	NO₃⁻	y = x	0.990**
	DOC	-	0.259
	Turbidity	y=-466.9 + 3.999x	0.603*
Event 3	NO₃⁻	y = x	1.00**
	DOC	y=-3.157 + 0.6483x	0.985**
	Turbidity	y=-99.38 + 1.047	0.941**
Passing–Bablok regression: y scale is for laboratory grab sample and x scale is for online sample. Pearson correlation coefficients: ** means the correlation is significant at p < 0.01 level. * means the correlation is significant at p < 0.05 level.

3.2 Evaluation of correlation among on-line and lab surrogate parameters for storm events

The detail information of on-line surrogate parameters during rainfall events are presented in Figs. 1–3 and the monitored storm events are summarised in Table 2.

The data obtained from the Bureau of Methodology contained total rainfall, rainfall duration and antecedent dry period. Table 2 showed that Events 1 and 2 happened during the wet and winter season in Adelaide, and Event 3 was conducted during the dry and summer season.

Event 1 was captured after a long antecedent dry period of 14 days, while both Event 2 and 3 were captured after 7 days of dry period. Additionally, the data shows that Event 3 had the highest rainfall of 14 mm and medium rainfall duration of 655 min, Event 1 had medium rainfall of 10 mm and the shortest duration of 420 min, Event 2 had the lowest rainfall of 8 mm and the longest rainfall period of 755 min (Huang et al. 2016).

Figures 1–3 show the changes of water levels in the drain during each event. As can be seen, water levels for Event 1 ranged from 9 to 150 mm with an average value of 48 ± 37 mm. The water levels for Event 2 were from 0 to 59 mm with the average result of 14 ± 16 mm and Event 3 had the widest range for water levels from 1 to 194 mm with the average water level of 32 ± 44 mm. Although Event 3 had the highest water level which was up to 194 mm, Event 1 had highest average water level. This is because the rainfall intensities from Event 3 became lower after the halfway mark of the event, leading to the lower average water level for Event 3. Overall, Event 1 and 3 had higher rainfall intensity and presented larger dynamic changes of the water flow conditions while Event 2 had lower rainfall intensity and a more stable and lower flow.

Table 2

Summary of monitored storm events (modify from Huang et al. (2016)).
Rainfall event (m/day)	Rainfall (mm)	Rainfall duration (min)	Antecedent dry period (days)
Event 1 (29/07)	10	420	14
Event 2 (19/08)	8	755	7
Event 3 (25/11)	14	655	7

Pearson Correlation was applied to evaluate if any correlations existed among water quality parameters and water level (Table 3) during the events. The results show that there is a strong negative correlation between NO₃⁻ and water level for Event 1 and 3 which had higher rainfall intensities. Event 2, which had low rainfall intensity and flow rate, presented a strong positive correlation to NO₃⁻. In this study, high rainfall intensity could lead to decreased NO₃⁻ concentration in the drain as the water level increased, while low rainfall intensity resulted in increased NO₃⁻ concentration when the water level was subsiding. The decrease in NO₃⁻ concentration in the stormwater drain could be due to the dilution effect from the high rainfall intensity. Another possible reason why a negative correlation is shown between NO₃ concentrations and water level is because there is a time delay between the two factors. McCarthy et al. (2012) reported that rainfall intensity needed to be adjusted to capture the time of concentration for each event. In MaCarthy’s study, the measured hydrologic data was “shifted” (by ± 15 min) to capture the inherent variability in the timing of pollutographs with respect to hydrologic patter (McCarthy et al. 2012). The result from this study is different from other studies, but not contradictory. For instance, Mallin et al. (2009) presented that there was not a corresponding positive correlation between rainfall and NO₃⁻. Dillon andChanton (2005) reported that NO₃⁻ concentrations were elevated in stormwater relative to rainwater in an urbanised coastal environment, but this trend was not statistically significant. In addition, the NO₃⁻ results from Events 1 and 3 indicate that the NO₃⁻ levels were high in the sump water before the event started, which could be due to the impact of the fertilizer runoffs from farming activities. Furthermore, NO₃⁻ showed a significantly positive correlation with DOC, indicating NO₃⁻ and DOC were discharged into the system at the same time during the rainfall events. Moreover, NO₃⁻ indicated strong negative correlations with turbidity and TSS through all events, although Nebbache et al. (2001) revealed no clear correlation between turbidity and NO₃⁻ concentration, whether positive or negative during the rainfall event. These negative correlations can suggest different sources for NO₃⁻, turbidity and TSS.

Additionally, there is a strong positive correlation between water level, and turbidity and TSS except Event 2 which indicates a weak negative relationship. In this study, high rainfall intensity led to increased turbidity and TSS in the drain as the water level increased, while low rainfall intensity and flow rate resulted in increased turbidity and TSS when the water level decreased. Mallin et al. (2009) reported that turbidity was significantly higher during rain events compared to non-rain periods and a positive correlation occurred between rainfall and turbidity. Other studies also discovered a significantly positive correlation between the rainfall intensity and TSS concentrations (Mallin et al. 2009, McCarthy et al. 2012). Furthermore, turbidity was an indication of the concentration of colloids and suspended particulates (Huang et al. 2016). Our study indicated a significantly positive relationship between turbidity and TSS during the events. Furthermore, the correlations between the water level and DOC were similar to NO₃⁻. Event 2 with the lowest rainfall showed a strong positive correlation with DOC and Event 1 with the highest rainfall intensity indicated a strong negative correlation with DOC, although a weak positive correlation for Event 3. Other studies indicated that there was a significant increases in DOC concentrations during storm events (Dalzell et al. 2005, Vidon et al. 2008). In addition, DOC also presents negative relationships with the turbidity and TSS except for Event 3. Once again, these negative correlations can suggest different origins between DOC, turbidity and TSS.

The study indicated that the chemical and physical parameters were characteristic of changing water quality through the drain that was not just related to the velocity of the flow. The factors that are affecting the response of NO₃⁻, turbidity and TSS in this study are clearly more complex than just the velocity of water passing through the sump. These changes could include the rainfall intensity and the distance from the stormwater drain affecting the time at which the peak levels arrived at the detectors, time required to re-suspend particles, the nature of the catchment area, and the size and character of the particles (clay, sand, soot, etc.).

Table 3

Pearson correlation coefficients among water quality parameters and water level.
		NO₃⁻	DOC	Turbidity	TSS	Water level
Event 1	NO₃⁻		0.828**	-0.211*	-0.194*	-0.572**
	DOC	0.828**		-0.046	-0.039	-0.263**
	Turbidity	-0.211*	-0.046		0.999**	0.435**
	TSS	-0.194*	-0.039	0.999**		0.408**
	Water level	-0.572**	-0.263**	0.435**	0.408**
Event 2	NO₃⁻		0.811**	-0.757**	-0.760**	0.165*
	DOC	0.811**		-0.193**	-0.190**	0.251**
	Turbidity	-0.757**	-0.514**		1.000**	-0.052
	TSS	-0.760**	-0.510**	1.000**		-0.049
	Water level	0.165*	0.251**	-0.052	-0.049
Event 3	NO₃⁻		0.362**	-0.328**	-0.364**	-0.206**
	DOC	0.362**		0.119	0.122	0.085
	Turbidity	-0.328**	0.119		1.000**	0.816**
	TSS	-0.364**	0.122	1.000**		0.814**
	Water level	-0.206**	0.085	0.816**	0.814**
* means p < 0.05 and ** means p < 0.01.

3.3 Assessment of correlation amongst microbiological and chemical parameters

Figures 4a, 4b and 4c showed the water levels in the stormwater monitoring sump overlaid with the sample points analysed for E. coli with counts as Most Probable Number (MPN) per 100mL. The results presented the highest average concentrations of E. coli (93166 ± 89283 count) were found at Event 3, followed by Event 2 with 6442 ± 3850 counts, and Event 1 had the lowest average E. coli concentration of 4687 ± 5684 counts. This could be due to a more intense rainfall that occurred during the first eight hours of Event 3, introducing a larger amount of E. coli into the drain. Furthermore, temperature was another important factor for the E. coli numbers. In Adelaide, the temperature was higher in the dry season than in the wet season and E. coli could be persisting or growing in the aquatic environments during periods of warmer temperature, contributing to higher initial concentrations and peaks during subsequent runoff events. But this is also different from Mallin et al. (2009) where the authors reported that wet-period faecal coliform bacteria counts were significantly higher than dry-period counts (no rainfall in the previous 72 h) for the individual creek in Hanover County, North Caroline, U.S.A., although the effect of temperature was not discussed between the wet and dry-period.

In the cases of Events 1 and 2, the distribution indicates that bacterial numbers did not correlate with the water level, with the highest values detected late in the event with no increase due to water level rises. Although other studies reported that the incidence of rainfall and rainfall intensity were highly significantly correlated with faecal coliform bacteria (Davies &Bavor 2000, Haydon &Deletic 2006, Kelsey et al. 2004, Mallin et al. 2009). At the same time, McCarthy et al. (2012) also reported E. coli was less correlated to hydrologic parameters, such as rainfall and runoff variables, for urban runoff. Event 3, with a different, more intense onset, did exhibit the highest E. coli numbers with the first flush. Additionally, most of the timing of peak E. coli concentrations was randomly scattered with respect to the timing of the peak water level, indicating that E. coli did not consistently experience a first flush and that concentrations were probably not affected solely by rainfall or catchment depletion processes. Many researchers have reported that pollutants can exhibit higher levels in the first flush of a stormwater runoff event, but these strategies may be deficient as the first flush of typical stormwater pollutants, such as sediment, nutrients and others, is not always observed in urban runoff (Duncan 1995, McCarthy 2009). McCarthy (2009) determined that cumulative mass versus volume curves related to the first flush phenomenon was not consistently present, the presence and magnitude of a first flush varied considerably between each site and poor consistent correlations were shown between E. coli and the first flush effect. This could also explain the contrary to the chemical parameter observations and introduces the likelihood that different mechanisms of transport are operating in the stormwater drain. Furthermore, while the chemical parameters potentially enter the stormwater by rapid dissolution or dilution of a concentrated source, bacterial transport may require re-suspension before subsequent flows can flush the cells from the catchment to the stormwater drain.

NO₃⁻ is the logical choice for surrogate microbial detection, as it is a representative of nutrient content (Whitehead &Cole 2006) and therefore possible raw sewage contamination. In addition, NO₃⁻ is detected in the low wavelength (< 220nm), high energy region giving excellent sensitivity to changes and the lowest detection limits. Thus, calculated equivalent NO₃⁻ concentration was the most likely candidate as a surrogate for microbial numbers. The monitored NO₃⁻ concentrations during the captured flow events in July, August and November are shown in Figs. 41a, 4b and 4c respectively. Our study indicated a negative correlation between NO₃⁻ and E. coli numbers through all the events, but this relationship was not significant (Table 3). To confirm the relationships with E. coli, other calculated equivalent parameters (DOC, turbidity and TSS) from the time scale of the sample collection from all the captured events were also plotted against E. coli counts. Correlation factors were very low in almost all cases indicating very little potential for relating these parameters to stormwater microbial loadings (Table 4). A strong correlation was found with turbidity (r = 0.916, p < 0.01) and TSS turbidity (r = 0.913, p < 0.01) but only within Event 3 indicating this relationship may not be reproducible, although faecal bacteria were found frequently associated with turbidity and/or TSS (Mallin et al. 2009). With NO₃⁻, the direction of the trend also indicated that if a relationship existed it was the reduction of the parameter that encouraged E. coli proliferation which is counter-intuitive if its role as a surrogate for growth nutrients is considered. The possible reasons for this lack of correlation are believed to be related to the nature of the stormwater catchment rather than any failing of the instrumentation or procedures. Because the water quality parameters from the laboratory analysed grab samples were also plotted against E.coli counts within the three events. The disparity of trends between the chemical parameters (NO₃⁻, DOC and turbidity) and E. coli counts was visually apparent, and the poor correlations were also present except for Event 3 (Table 2). Compared to calculated equivalent, limited results were provided by the grab samples and the strong correlations among E.coli counts and NO₃⁻, DOC and turbidity for Event 3 might be coincident. Unlike a permanent water body where solutes and nutrients are constantly accessible to waterborne organisms, the stormwater catchment is typically dry until the onset of a rain event. As a result, the bacteria and chemical nutrients are likely to be spatially separated until they come into contact during the event and detection of either would therefore be unrelated. In addition, while NO₃⁻ does promote bacterial proliferation, it is not usually a growth limiting nutrient and its absence does not preclude the detection of significant bacterial numbers. Furthermore, faecal microbial contamination in urban areas could come from various sources, including domestic animals and urban wildlife including pets, waterfowl, pigeons, rats and others (McCarthy 2009). Additionally, NO₃⁻ concentrations could be influenced by non-point sources, especially for agriculture runoffs. The complicated sources from the urban areas could contribute to a weak correlation between NO₃⁻ and E. coli (James &Joyce 2004).

Table 4

Pearson correlation coefficients (r) among water quality parameters and *E.coli*.
	Event 1	Event 2	Event 3
NO₃⁻	-0.535	-0.143	-0.608
DOC	-0.470	-0.119	0.572
Turbidity	-0.284	0.007	0.916**
TSS	-0.286	0.010	0.913**
Water level	-0.163	0.186	0.257
* means p < 0.05 and ** means p < 0.01.

Furthermore, temporal variations were apparent for the detection of parameters in the stormwater flow events, with the dissolved chemical parameters peaking early in the first flush and decreasing rapidly throughout the monitored event. Microbiological parameters such as E. coli and coliform counts were lower in the early period of the event and peaked during subsequent increases in water flow indicating that different mechanisms of transport are exerting influence. It is believed that while chemical species are easily dissolved and carried with the primary flow, microorganisms require re-suspension from solid substrates before the later flow increases can transport them through the catchment area and into the stormwater drain. In addition, this can be because the bacterial comes from animal faeces that slowly disssolves in fields and on roads before entering the stormwater system.

This study has applied an on-line UV-Vis spectrophotometer in the field to collect real-time continuous data for various water quality parameters (nitrate, DOC, turbidity and total suspended solids) during three storm events in Mannum, Adelaide, Australia. This study demonstrated that the trends for on-line and comparative laboratory analysed samples were complimentary through the events. The study not only investigated the correlations among various water quality parameters (nitrate, DOC, turbidity and TSS) and water levels, but also investigated the correlations among water quality parameters and E. coli. The results showed that water levels had a strong negative correlation with NO₃⁻ and DOC, a strong positive correlation with turbidity and TSS during high rainfall intensity. While water levels presented a positive correlation with NO₃⁻ and DOC, and negative correlation with turbidity and TSS during low rainfall intensity. Microbiological parameter - E. coli counts was lower in the early period of the event and peaked during subsequent increases in water flow indicating that different mechanisms of transport are exerting influence. Our study also indicated that NO₃⁻, DOC, turbidity, TSS and water level did not show strong correlations with E. coli. The poor correlations among the microbiological data and on-line parameters can be due to the nature of the stormwater catchment outside of rain events, or due to the different sources of bacteria and nutrients that end up into the stormwater drain. This case study represented a trial of technologies in a prototype self-contained monitoring station for the ability to act as a hazard detection tool for stormwater inputs into drinking water sources. While solutions to a number of issues could not be resolved within the duration of the project, the monitoring should be continued to collect more valuable data which will help guide future research.

Acknowledgements:

This project was financially supported by Water Quality Research Australia (Project No.1020-09), and a SA Water Capital Project for setup the monitoring system. The authors are grateful to Ms. Gretchen Schroeder, Water Data Service, DCM Process Control Ltd and Water Treatment and Microbiology Research units, Australian Water Quality Centre for technical and analytical support.

Ethical Approval:

Not applicable.

Consent to Participate:

Not applicable.

Consent to Publish:

Not applicable.

Authors Contributions:

Jianyin Huang: data analysis and manuscript writing. Christopher W.K. Chow: experimental design, data analysis and manuscript writing. Zhining Shi: data analysis and manuscript writing. Rolando Fabris: experimental design and data analysis. Amanda Mussared: experimental design and data analysis. Gary Hallas: experimental design and data analysis. Paul Monis: experimental design, data analysis and manuscript writing. Bo Jin: experimental design and data analysis. Christopher P. Saint: experiment design, data analysis and manuscript writing.

Funding:

Not applicable.

Competing Interests:

The authors declare no competing interests.

Availability of data and materials:

Not applicable.

APHA (2005): Standard methods for the examination of water and waste water, 21st edn. American Public Health Association, Washington, DC
Bricker SB (2007): Effects of nutrient enrichment in the nation's estuaries: a decade of change: National Estuarine Eutrophication Assessment update.
Briggs R, Melbourne K (1968): Recent advances in water quality monitoring. Water Treat. Exam 17, 107-120
Brito RS, Pinheiro H, Ferreira F, Matos JS, Lourenço N (2014): In situ UV-Vis spectroscopy to estimate COD and TSS in wastewater drainage systems. Urban Water Journal 11, 261-273
Causse J, Thomas O, Jung A-V, Thomas M-F (2017): Direct DOC and nitrate determination in water using dual pathlength and second derivative UV spectrophotometry. Water research 108, 312-319
Chow C, Dexter R, Sutherland-Stacey L, Fitzgerald F, Fabris R, Drikas M, Holmes M, Kaeding U (2007): UV Spectrometry in Drinking Water Quality Management. WATER-MELBOURNE THEN ARTARMON- 34, 63
Chow C, Fabris R, Dixon M (2008): Case studies using S:: CAN on-line monitoring system. Adelaide
Chow C, Sweet V, Adams K, Mosisch T, Shephard M, R D (2009): Implementation of a Real Time Early Warning System for Water Quality Incidents, OzWater 09, AWA, Melbourne Convention and Exhibition Centre, Melbourne
Dalzell BJ, Filley TR, Harbor JM (2005): Flood pulse influences on terrestrial organic matter export from an agricultural watershed. Journal of Geophysical Research: Biogeosciences 110
Davies C, Bavor H (2000): The fate of stormwater‐associated bacteria in constructed wetland and water pollution control pond systems. Journal of Applied Microbiology 89, 349-360
Davis E, Casserly D, Moore J (1977): BACTERIAL RELATIONSHIPS IN STORMWATERS 1. JAWRA Journal of the American Water Resources Association 13, 895-906
Dillon KS, Chanton JP (2005): Nutrient transformations between rainfall and stormwater runoff in an urbanized coastal environment: Sarasota Bay, Florida. Limnology and Oceanography 50, 62-69
Duncan H (1995): A review of urban stormwater quality processes.
Duncan H (1999): Urban stormwater quality: a statistical overview. CRC for Catchment Hydrology
El Khorassani H, Theraulaz F, Thomas O (1998): Application of UV spectrophotometry to the study of treated wastewater discharges in rivers. Acta hydrochimica et hydrobiologica 26, 296-299
Eriksson E, Baun A, Scholes L, Ledin A, Ahlman S, Revitt M, Noutsopoulos C, Mikkelsen PS (2007): Selected stormwater priority pollutants—a European perspective. Science of the total environment 383, 41-51
Ferguson C, Husman AMdR, Altavilla N, Deere D, Ashbolt N (2003): Fate and transport of surface water pathogens in watersheds.
Gnecco I, Berretta C, Lanza L, La Barbera P (2005): Storm water pollution in the urban environment of Genoa, Italy. Atmospheric research 77, 60-73
Göbel P, Dierkes C, Coldewey W (2007): Storm water runoff concentration matrix for urban areas. Journal of contaminant hydrology 91, 26-42
Greenway M, Le Muth N, Jenkins G (2002): Monitoring Spatial and Temporal Changes in Stormwater Quality through a Series of Treatment Trains. A Case Study—Golden Pond, Brisbane, Australia, Global Solutions for Urban Drainage, pp. 1-16
Gruber G, Bertrand-Krajewski J-L, Beneditis JD, Hochedlinger M, Lettl W (2006): Practical aspects, experiences and strategies by using UV/VIS sensors for long-term sewer monitoring. Water Practice and Technology 1
Hamilton AJ, Boland A-M, Stevens D, Kelly J, Radcliffe J, Ziehrl A, Dillon P, Paulin B (2005): Position of the Australian horticultural industry with respect to the use of reclaimed water. Agricultural Water Management 71, 181-209
Haydon S, Deletic A (2006): Development of a coupled pathogen-hydrologic catchment model. Journal of Hydrology 328, 467-480
Huang H, Chow CW, Jin B (2016): Characterisation of dissolved organic matter in stormwater using high-performance size exclusion chromatography. Journal of Environmental Sciences 42, 236-245
James E, Joyce M (2004): Assessment and management of watershed microbial contaminants. Critical Reviews in Environmental Science and Technology 34, 109-139
Kelsey H, Porter D, Scott G, Neet M, White D (2004): Using GIS and regression analysis to evaluate relationships between land use and fecal coliform bacterial pollution. Journal of Exp. Marine Biology and Ecology 298, 197-209
Lepot M, Torres A, Hofer T, Caradot N, Gruber G, Aubin J-B, Bertrand-Krajewski J-L (2016): Calibration of UV/Vis spectrophotometers: A review and comparison of different methods to estimate TSS and total and dissolved COD concentrations in sewers, WWTPs and rivers. Water Research 101, 519-534
Mallin MA, Johnson VL, Ensign SH (2009): Comparative impacts of stormwater runoff on water quality of an urban, a suburban, and a rural stream. Environmental monitoring and assessment 159, 475-491
McCarthy D, Mitchell V, Deletic A, Diaper C (2007): Escherichia coli in urban stormwater: explaining their variability. Water Science and Technology 56, 27-34
McCarthy D (2009): A traditional first flush assessment of E. coli in urban stormwater runoff. Water Science and Technology 60, 2749-2757
McCarthy D, Hathaway J, Hunt W, Deletic A (2012): Intra-event variability of Escherichia coli and total suspended solids in urban stormwater runoff. Water research 46, 6661-6670
Mrkva M (1975): Automatic UV-control system for relative evaluation of organic water pollution. Water Research 9, 587-589
Nebbache S, Feeny V, Poudevigne I, Alard D (2001): Turbidity and nitrate transfer in karstic aquifers in rural areas: The Brionne Basin case-study. Journal of environmental management 62, 389-398
Olivieri VP (1977): Microorganisms in urban stormwater, 1. Environmental Protection Agency, Office of Research and Development …
Petney T, Taraschewski H (2011): Waterborne parasitic diseases: Hydrology, regional development, and control.
Selvakumar A, Borst M (2006): Variation of microorganism concentrations in urban stormwater runoff with land use and seasons. Journal of Water and Health 4, 109-124
Shi Z, Chow CW, Fabris R, Liu J, Jin B (2020): Alternative particle compensation techniques for online water quality monitoring using UV–Vis spectrophotometer. Chemometrics and Intelligent Laboratory Systems 204, 104074
Thayanukul P, Kurisu F, Kasuga I, Furumai H (2013): Evaluation of microbial regrowth potential by assimilable organic carbon in various reclaimed water and distribution systems. Water research 47, 225-232
Thomas O, El Khorassani H, Touraud E, Bitar H (1999): TOC versus UV spectrophotometry for wastewater quality monitoring. Talanta 50, 743-749
Torres A, Bertrand-Krajewski J-L (2008): Partial least squares local calibration of a UV–visible spectrometer used for in situ measurements of COD and TSS concentrations in urban drainage systems. Water Science and Technology 57, 581-588
van den Broeke J (2007): On-line and in situ UV/Vis spectroscopy. AWE International March, 55-59
Vidon P, Wagner LE, Soyeux E (2008): Changes in the character of DOC in streams during storms in two Midwestern watersheds with contrasting land uses. Biogeochemistry 88, 257-270
Whitehead R, Cole J (2006): Different responses to nitrate and nitrite by the model organism Escherichia coli and the human pathogen Neisseria gonorrhoeae. Portland Press Limited
WHO 2006: Guidelines for Drinking Water Quality, World Health Organisation, Geneva, Switzerland

Stormwater Monitoring Using On-line UV-Vis Spectroscopy for Surrogate Measurements of Chemical Contaminants and Compared With Microbiological Data

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials And Methods

2.1 On-line monitoring instrumentation and sampling system

2.2 Laboratory analytical methods

2.2.1 Chemical

2.2.2 Microbiological

2.3 Statistical analyses

3. Results And Discussion

3.1 Validation of online stormwater quality monitoring

3.2 Evaluation of correlation among on-line and lab surrogate parameters for storm events

3.3 Assessment of correlation amongst microbiological and chemical parameters

4. Conclusion

5. Declarations

6. References

Status:

Version 1