1. An Examination of Chlorine Demand of the Catskill and Delaware Supplies of the NYC Water Supply System 2005-2006 Charles R. Cutietta-Olson, Deputy Chief Watershed Water Quality Operations, BWS, DEP 9/10/09 New York City Department of Environmental Protection Water Quality

2. Acknowledgements Thanks to Lori Emery, Salomé Freud and Lin Lu for looking at early graphics and output and providing correction, direction and encouragement. Thanks to the Early Warning Remote Monitoring Group for providing the continuous monitoring data files. Thanks to Ralph Marchitelli, Dan Massi and the supervisors at Shaft 18 for providing documentation and patiently responding to questions. New York City Department of Environmental Protection Water Quality

3. Why examine chlorine demand? Quantifying chlorine demand and understanding patterns could lead to better system operations. The investigator speculated that different turbidity sources may be associated with different levels of chlorine demand, and 2005- 2006 includes periods higher turbidities associated with storms in the Catskills and storm events local to Kensico Reservoir, the Source Water for both systems. Most of the data were available as an electronic file of continuous on-line measurements. Chlorine demand of the water supply should be a natural function of water quality and thus empirically quantifiable rather than something that must be examined through probability and statistics. New York City Department of Environmental Protection Water Quality

4. Presentation Overview Site locations and source data Cl demand = initial FCR – FCR at first treated site Estimating initial Free Chlorine Residual ( initial FCR) Data frequency plots, Cat vs. Del Chlorine demand during two weather events proximal to the Source Water Discussion of upper 0.2%ile of Chlorine Demand “events” Data frequency plots of upper 25 th %ile of Chlorine Demand Multiple linear regression models

5. Data were gathered from four sites: CATLEC Shaft 18 Eastview Shaft19 The equation for calculating disinfection compliance requires four parameters: free chlorine residual (FCR), contact time, temperature and pH. DEP uses the values recorded at CATEV, Uptake 1, DEL19 and Uptake 2 for compliance purposes. Cl demand = initial FCR – FCR first treated site

6. Continuous On-line Data Time interval in minutes of data parameters recorded in 2005-06. (n/r = not recorded) SiteTurbidityFlow Free chlorine residualpHtemp Specific conductance DEL18515n/r15 DEL1930n/r530 n/r CATLEFF55n/r5 5 CATEV30n/r530 n/r Data had gaps which had to be filled. Data fill sources included compliance Source Water turbidity data and Kensico Laboratory Data. New York City Department of Environmental Protection Water Quality

7. Estimating Chlorine Dose Concentration ( initial FCR) Cl Demand = initial FCR – FCR first treated site initial FCR is not measured, but must be estimated. Dose-based initial FCR is derived from System Operations’ records of dose changes in lbs/MGD. Use-based initial FCR is derived from records of pounds of chlorine gas used by each system and the amount of flow it was applied to on an hourly basis. To compare Use-based and Dose-based estimates of initial FCR, the noon time moment for each day for the 2005-2006 period was plotted for each system. New York City Department of Environmental Protection Water Quality

9. Kendall’s tau correlation coefficients and number of observations (n) for dose-based initial FCR (dFCR initial ) and use-based initial FCR (uFCR initial ) and for first treated site. (p-value for all coefficients <0.0001) dFCR initial uFCR initial FCR CATEV dFCR initial tau = 1 n = 659 0.691 654 0.686 655 uFCR initial 1 683 0.587 678 FCR CATEV 1 683 dFCR initial uFCR initial FCR DEL19 dFCR initial tau = 1 n = 682 0.543 679 0.538 675 uFCR initial 1 709 0.323 702 FCR DEL19 1 705 Catskill System: Delaware System:

10. Use estimates should be a better predictor than dose estimates. Both tables list a stronger tau coefficient between FCR first treated site and the dose-based (dFCR initial ) rather than use-based (uFCR initial ) initial FCR estimate. The tau coefficients suggest that dFCR initial is at least not a worse predictor of FCR first treated site than uFCR initial. Selected Dose-based Estimate of initial FCR New York City Department of Environmental Protection Water Quality

11. Any Patterns Visible in Time Series Plots? Complete data sets of ~200,000 observations were pared down based on frequency of treated pH monitoring (30 min). Data sets of ~32,000 observations were plotted as time series, but patterns were not obvious. The following frequency distribution plots display Catskill System data on the left and Delaware System data on the right. New York City Department of Environmental Protection Water Quality

15. The next two slides display time series data from two local weather-driven turbidity events.

16. Note that Cl Demand increases with flow reduction and residence time increase New York City Department of Environmental Protection Water Quality

17. Fluctuation of chlorine demand around turbidity spike. New York City Department of Environmental Protection Water Quality

18. Examination of Upper 0.2%ile of Chlorine Demand Catskill System, Cl Demand ≥ 1 mg/L: 38% of the 74 observations were associated with aqueduct shutdowns (28 obs). Other instances appeared to be “blips” which could result from power spikes or equipment maintenance. September 2006 event comprises 26% of this data set (19 obs). Delaware System, Cl Demand ≥ 0.9 mg/L: 85% of the 82 observations occur in the 12/19-20/05 period. Some instances appear to be “blips”. New York City Department of Environmental Protection Water Quality

19. Looking for Water Quality Characteristics Associated with High Chlorine Demand Working only with treated and raw pH, treated and raw temperature, and raw water turbidity. The following plots display ~32,000 observation data set on left vs. data subsetted for approximately upper 25% percentile of Cl Demand data on right. New York City Department of Environmental Protection Water Quality

22. New York City Department of Environmental Protection Water Quality

25. New York City Department of Environmental Protection Water Quality

26. Multiple Linear Regression Models ParameterEstimatePartial R 2 Treated temp0.018230.248 Raw pH0.218340.071 Treated pH0.172350.042 Turbidity0.065200.028 Raw temp0.004340.001 Catskill System All data n=24,765 total model R 2 = 0.390 ParameterEstimatePartial R 2 Treated pH0.235140.089 Turbidity0.095990.020 Raw temp0.011650.058 Raw pH0.102030.005 Treated temp0.004240.004 25 th %ile Cl Demand of above data n=6,856 total model R 2 = 0.176

27. Multiple Linear Regression Models Delaware System ParameterEstimatePartial R 2 Treated temp0.020910.200 Treated pH0.186860.075 Raw pH0.119810.012 Raw temp0.011650.010 Turbidity0.010700.001 ParameterEstimatePartial R 2 Raw pH0.215240.181 Turbidity0.065810.035 Treated temp0.002080.012 Treated pH0.028400.005 All data set n=27,793 total model R 2 = 0.296 25 th %ile Cl Demand of above data n=7,256 total model R 2 = 0.234

28. Conclusions Chlorine demand of both systems was normally distributed around 0.5 ppm over the period of investigation Fluctuations in chlorine demand appeared most closely related to changes in temperature as measured at the treated sites. Unfiltered continuous–monitoring data require considerable work before analysis. The Catskill System may be more influenced by ambient environmental factors than the Delaware System. New York City Department of Environmental Protection Water Quality