TRADE WASTE COMPOUND DETECTION AND SEWER PROCESS MODELLING USING ONLINE, IN-SITU SPECTROSCOPY Adrian Malyon 1, Vicky Whiffin 1, Mark Angles 1, Roger Wood 1 1. Sydney Water Corporation, NSW, Australia ABSTRACT Sydney Water Corporation has deployed on-line instruments in a three year study of trade waste sites in wastewater networks. The purpose of this study was to ascertain whether target substances could be isolated from the background chemical matrix via online measurements, including online UV-Vis spectroscopy. Several substances were identified and calibrations developed from the spectral data to detect them, including a robust phenol calibration and an alarm function which can detect the presence of a number of volatile hydrocarbons. The use of several conventional electrodes alongside the spectral component also allowed for in-situ mapping and modelling of several key wastewater-quality processes. INTRODUCTION Current assessment of trade waste discharges to the wastewater system is predominantly by routine grab samples. Nevertheless, by being based on one-off sampling of limited catchment sites at a relatively low frequency, this sampling approach is restricted in its ability to fully characterise the wastewater catchments, and particularly identify short-term variations and unpredictable discharge events. Sydney Water has been carrying out validation of the S::CAN series of online instruments, testing their ability to detect in near to real-time the various substances of concern likely to be discharged in a trade waste environment. Several online monitoring electrodes were installed at a wastewater pumping station impacted by significant trade waste inputs, after undergoing extensive laboratory validation. The electrodes measured several conventional parameters, Oxidation Reduction Potential (ORP), Ammonia, ph and Conductivity (EC), and in addition also measured UV-Vis absorbance via the principal instrument, the Spectro::lyser. UV-Vis absorbance is a common laboratory technique that can be employed to characterise molecules that contain one or more degrees of unsaturation, mostly in the form of double or triple bonds on a carbon atom. The absorbance measured is the amount of light absorbed by certain molecules as electrons are excited to a higher energy state, the light is then emitted at a different wavelength as the electrons sink back to the lower energy state. The frequency of both the absorbed light and the emitted light has a specific signature, molecules with similar functional groups around the carbon atoms will tend to absorb at similar wavelengths, and absorption follows Beers law in that it is proportional to concentration. This allows, in particular, the characterisation of a wide variety of organic molecules. In the laboratory environment, these tests are conducted by using specific reagents, which are used to create a colorimetric reaction by which individual substances can be isolated. An in-situ device such as the Spectro::lyser cannot employ such reagents, however, so the absorbance spectra must be carefully analysed against parallel grab sample analytical data using Gas Chromatography-Mass Spectroscopy or other comparable quantitative technique. By this method, the Spectro::lyser can effectively be calibrated over a period of time for each water matrix. If the instrument is moved to a different location, the instrument must effectively be recalibrated to the new water matrix over a period of 1-4 months, depending on the similarity of the water matrices of the two locations. This allows all of the natural variation within the water matrix to be taken into account in the sampling regime. The calibrations for a particular site, or scan-point, can be saved in the instrument controller for later use. As well as calibrating to certain parameters, alarm limits can be set to trigger an autosampler to collect a sample for laboratory analysis. Via this procedure, unusual events can be used to constantly refine the calibration of the instrument. The use of online monitoring in this fashion has the potential to be employed effectively in several fields and applications, including: switching to an eventbased monitoring system, catchment special investigations, monitoring the compliance of the output of problematic trade waste customers, process optimisation and modelling of sewer networks and processes.
the unit required manual cleaning every 2-3 weeks, using the cleaning brush the unit operated successfully for three months without needing the measuring area cleaned, a comparatively sturdy result. Figure1. Unit installed at a wastewater pumping station, in parallel with autosampler. METHODOLOGY Sensor validation Prior to deployment, the objective accuracy of the Spectro::lyser unit was tested using a series of standard laboratory grade solutions of known concentration. The solutions were equivalent to those used in AS 3753-2001 to verify the wavelength accuracy, position and linearity of signals from laboratory spectrophotometers. Solutions containing substances of known concentrations were then run though the Spectro::lyser unit in a controlled laboratory environment. The analytes chosen were those that were both of interest to Sydney Water in a trade waste environment and that were predicted to be able to be identified by the instrument. This analysis was carried out so that when the instrument was validated in the field, analytes could be chosen with a reasonable expectation that they could be identified within the more complex wastewater matrix. Field study The cabinet unit was installed at a wastewater pumping station known to be impacted by significant organic trade waste, as well as higher than normal levels of Ammonia (measured as NH 3 - N). The instruments were installed and commissioned by DCM Process Control, who also assisted in the calibration and verification of the electrodes. Samples were delivered to the Spectro::lyser by a pneumatic lift pump (DCM), which is specially designed to deliver a turbulencefree sample to the measuring point. The air supply is also used to clean the measuring area of the Spectro::lyser, either through a blast of air used to purge fouling, or by driving a pneumatic cleaning brush. Over the course of the deployment, it became clear that the cleaning brush is a far superior method to air cleaning. Using air cleaning, An initial calibration period of 4 months was required to establish a baseline spectral matrix for the wastewater in this location. In addition, over a period of four weeks a series of composite samples were collected via an autosampler deployed on site. The samples were collected in order to establish the baseline chemical matrix, and to, if possible, correlate this with the parallel spectral readout. None of the initial target analytes were present in the matrix, however there were significant levels of Total Petroleum Hydrocarbons, including up to 25.2 mg/l of C15-C28 compounds. This indicates that there is likely a high level of oils and greases entering the wastewater system in this area. The nature of TPH makes it impossible to calibrate for these compounds specifically, these compounds are undetectable via UV-Vis absorbance. However, alarm limits set to certain wavelengths should give the ability to detect the volatile benzene or benzene-like susbstances which are likely to be in wastewater which has experienced a large inflow of petroleum-based substances. Also observed in the composite samples were high levels of acetone, as high as 3.2mg/L over a week-long composite. An attempt was made to calibrate the instrument for acetone detection, as this was unsuccessful, it was assumed that acetone would not interfere with the other calibrations The spectral parameters were calibrated on-site by spiking wastewater with known concentrations of analytical-grade compounds. This is otherwise known as a calibration by standard additions, and is the optimal way to calibrate an instrument where matrix effects are a concern. Wavelength outputs were stored after each addition of analyte to determine their respective spectral fingerprints. A sample was also taken after each addition to submit to the analytical laboratory. By matching up the stored wavelength to the laboratory result, a calibration curve was established. Where a global calibration already exists on the instrument for the target analyte, a local calibration can be performed relatively simply to fit the calibration curve to the lab measurements. It is unlikely that the global calibration will be accurate without a local calibration (NO 3 and H 2 S Global calibrations excepted in some cases). Where a new global calibration must be created, operators must use an in depth mathematical process to generate a calibration algorithm.
RESULTS/DISCUSSION Laboratory validation While the spectrophotometer component marginally failed to meet the stipulations of AS 3753-2001, it was determined that the absolute accuracy of the instrument was of secondary importance to its wavelength precision, or repeatability, over time. The instrument output did not drift over the time period of the project. The spectral fingerprints of toluene, phenol (Figure1.) and PAH compounds were able to be sufficiently characterised in the laboratory, indicating the potential for site-specific calibration of the instrument and the establishment of effective detection limits for these substances in wastewater. In the laboratory testing, the peak absorbance values for both phenol and toluene were able to be detected easily and showed good linearity across the tested range. The tests which were carried out in a deionised water matrix should give the best detection limit possible by the instrument. Figure 2. Arrows showing the derivitised absorbance values increasing/decreasing proportional to concentration of phenol in calibration samples (sewer matrix). In a wastewater matrix the absorbance of individual substances will be obscured somewhat by the absorbance of other substances, particularly when the target analyte is at lower levels, however this testing indicated that these substances would most likely be able to be identified. The other volatile hydrocarbons which were tested, o-xylene and ethyl benzene, gave response curves of poorer linearity in the laboratory testing. While these substances did absorb, the linearity was too variable to be used for a firm calibration. This is likely because the less polar nature of the molecules made them less soluble, creating an unstable concentration. Since the matrix of wastewater is likely to be more polar in nature, the more appropriate way to detect such substances would be to use a gas-phase detection system. An alarm parameter set at the appropriate wavelength could still be useful to detect these substances in the liquid phase, however, if backed up by an autosampler programmed to capture a sample for lab analysis. An optimal system would possibly contain both a gas and liquid phase detection system for total coverage. Field testing As can be seen in Figures 2 and 3 the instrument suite has the ability to deliver a unique capability to detect changes in wastewater chemical composition in real time, which can provide decision makers with interesting and usefull insights into in-situ chemical processes occurring in the wastewater matrix. In addition to the ability to continuously monitor several wastewater quality parameters simultaneously and in-situ, the results showed clear trends arising from events such as: ammonia discharge from trade waste customers indicating sub-optimal on-site treatment, rainwater intrusion events, H 2 S spike events and nitrification/denitrification/re-nitrification cycles. This information can be particularly usefull when combined with the calibrations for NO 3 and H 2 S from the Spectro::lyser unit, which require no calibration, as they each reference a well defined area of the spectrum which do not suffer interferences (Sutherland-Stacey et al., 2007). This information can then be used to monitor wastewater system processes such as nitrification/denitrification/re-nitrification cycles, Sulphate/H 2 S ratios and other sewer events, such as rainwater/stormwater ingress. Several of these processes were mapped, but more work remains to extract and interpret the full implications that can be drawn from this wealth of data. As well as modelling the in-situ processes occurring within the sewer, the research has shown that the concentrations of several trade waste pollutants, including phenol and some polycyclic aromatic hydrocarbons (PAH s), can be accurately calibrated against the spectral output of the instrument. It should be noted, that the outputs from the calibrations are open to misinterpretation. Often multiple substances may absorb in the same region. For this reason it is important to keep an autosampler in parallel to any installed online monitoring station, so that high results or triggered alarms can be used to capture a sample. The sample can then be taken for further lab analysis to confirm the result. If the results differ from the present calibration, a further tuning of the calibration may be necessary. The calibrations applied to the instrument were used to capture events via autosampler and laboratory analysis, which also served to provide data for further optimisation of the calibration. The phenol calibration successfully captured a labconfirmed result of 2.2mg/L combined phenols (being 1.0mg/L phenol and 1.2mg/L di-ethyl
phenol). The ammonia electrode, once calibrated, captured a result of 88 mg/l Ammonia (Figure 3.) overnight, which was confirmed by follow up interview with the identified trade waste customer. As well as modelling the in-situ processes occurring within the wastewater system, the research has shown that the concentrations of several trade waste pollutants, including phenol and some polycyclic aromatic hydrocarbons (PAH s), can be accurately calibrated against the spectral output of the instrument. With this capability, it is possible, by installing online units at strategic points in the trade waste network, to develop a time-resolved model for estimating the source of these pollutants. The model shown below (Figure 4.), illustrates how such a network could be set up to monitor the concentration of pollutants throughout the network, using a flow/ water quality integrated model. 1 α 3 γ 2 β 4 ε this nature, it is likely that the source of the pollution can be identified. CONCLUSION Having applied successful calibrations to the Spectro::lyser instrument, Sydney Water is now able to detect in-situ several organic trade waste substances of interest in a wastewater environment, using UV-Vis spectral fingerprint data. The calibrations for these parameters could still potentially yield false positives, but if the parameters are combined with more generalised alarm limits, further refining of the calibrated parameters can be done over time. With further work, more substances may be identified and isolated from the background spectra. In addition, the ability to monitor the absolute and relative changes of several parameters simultaneously and continuously promises the ability to more accurately model various chemical interactions within a wastewater environment. This ability will be crucial to any future attempt to move towards a smart model of wastewater network management, and a shift from the reliance of routine monitoring to event-based sampling. The development of such a network would likely involve a large initial investment; being comprised of extensive modelling, data collection and sampling, as well as calibration of online parameters to the local matrix. The dividends from such an investment would include: odour and corrosion management, using integrated water quality/flow data to manage organic loads on wastewater treatment plant digesters and bioreactors, ability to trace recurring pollutants back to source and estimate their initial concentration, and the ability to protect digesters and reactors from high levels of contaminants which could poison the digestion bacteria and thereby stall the plant. STP Figure 4. Diagram showing a possible deployment model for multiple online Spectro::lyser instruments (1-4 ) in a sewer network.. Following a modelling period, integration of both flow and water quality data into each branch of the network will generate concentration/flow calibration coefficients (α-ε),for each analyte,allowing estimation of initial concentration and origin point of target pollutant through a flow/concentration resolved regression. For example: if phenol were detected at point 3 at concentration = x, but not detected at point 1, A regression of xα against measured flow would be able to give an approximation of source point between instruments 1&3 and an estimation of the initial concentration of phenol. Since there are usually limited sources of large scale pollutants of ACKNOWLEDGMENT This research was made possible by the efforts of several departments across Sydney Water Corporation, including Analytical Services, Field Sampling &Testing, Air Quality and Atmospheric Monitoring, and Instrumentation and Special Projects. All the S::CAN instrumentation as well as the sampling pump was supplied and commissioned by DCM Process Control
REFERENCES Langergraber, G., Fleischmann, N, Van Der Linden, F., Wester, E., Weingartner, A. Hofstaetter., F. (2002) In Situ Measurement of Aromatic Contaminants in Bore Holes by UV/VIS Spectrometry Field Screening Europe. Sutherland-Stacey, L., S. Corrie, A. Neethling, I. Johnson, Gutierrez, Dexter, R., Yuan, Z. Keller, J. Hamilton,G. (2007) In Situ Continuous Measurement of Dissolved Sulfide in Sewer Systems. S::CAN media Library (http://www.scan.us/medialibrary/publications/p_2007_01.pdf)
Table 1. List of initial target analytes using Spectro::lyser with estimated detection limits based on regression of calibrated values to 99% confidence intervals of normal absorbance values for this location. Substance Estimated Detection Limit Potential Surrogates Phenol 0.1-0.5 ppm Various Phenols i.e. diethyl Phenol Ethyl benzene 10+ppm benzene, other aromatics Toluene 1ppm Similar aromatics O-Xylene 5ppm Xylenes Napthalene 0.1-0.5 ppm various PAH's Aromatic Alarm parameter 2-4ppm Napthalene, Toluene, O-Xylene, Ethyl benzene 12 10 z value 8 6 4 2 0-2 -4-6 NH-3 ORP EC H2S ph Figure 2. Graph showing the inter-relationships between normalised values of NH3, ORP, EC, H 2 S and ph for two days during a period of high H 2 S events
8 6 4 2 0-2 z value -4-6 normec normph normnh3 normorp -8-10 -12. Figure 3. Graph showing the inter-relationships between normalised values of EC, ph, NH 3 -N and ORP during a two day period following a high ammonia event.