lessons from a drinking water crisis - American Water Works Association

5 downloads 0 Views 679KB Size Report
COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003. 147. Sydney 1998— lessons from a drinking water crisis. From July to September 1998, ...
BY PETER COX,

From July to September 1998, high concentrations of Cryptosporidium and Giardia were

IAN FISHER,

detected episodically in the water supply and distribution systems of Sydney, Australia. The

GEORGE KASTL,

resulting drinking water crisis triggered three consecutive boil-water advisories and a

VEERIAH JEGATHEESAN,

government inquiry into the management of the water supply. The episodic nature of the

MALCOLM WARNECKE,

detections focused attention on the veracity of the laboratory results and triggered an

MARK ANGLES,

investigation of the transport of these pathogens in Sydney’s water supply system. This

HERI BUSTAMANTE, TONY CHIFFINGS, AND PETER R. HAWKINS

article provides information submitted to the Sydney Water Inquiry that explains the episodic occurrence of pathogens in the reticulated water supply, attributing it to rapid fluctuations in the quality of the water reaching the water treatment plant.

Sydney 1998



lessons from a drinking water crisis he methods for monitoring Cryptosporidium oocysts and Giardia cysts in water are relatively new, with most development occurring in the last 15 years. The methods, especially the Information Collection Rule (ICR) method, have been criticized for their technical difficulty and lack of accuracy. Although false-positive results can occur, greater problems have been associated with false-negatives and low and variable recovery rates dependent on water quality (e.g., turbidity) (Clancy et al, 1999; Dufour et al, 1999; Fricker, 1995; Clancy et al, 1994). Despite these methodological problems, several hundred reports have been published in recent years (Clancy et al, 2000; Connell et al, 2000; Clancy et al, 1999) that have established the patterns of occurrence of cysts and oocysts in water and have led to substantial improvement in analytical methods. The Sydney Water Corporation (SWC), through the laboratories of its subsidiary Australian Water Technologies (AWT), has monitored cysts and oocysts in Sydney’s water supply since 1992. AWT’s flow cytometric (FC) method was codeveloped with Macquarie University in Sydney and built on prior developments at Thames Water Utilities Laboratories (TWUL) in Reading, Great Britain. An initial unpublished in-house study prepared by AWT and Macquarie University showed that the FC method, using either flocculation (Vesey et al, 1993) or flatbed filtration (Hansen & Ongerth, 1991) for cyst and oocyst concentrations, resulted in up to 70 times higher recoveries than the ICR method. Despite these technical advances, there is still controversy over the usefulness of cyst and oocyst monitoring as a basis for public health decisions (Allen et al, 2000). However, a number of environmental microbiologists support the use of these data for scientific-based risk assessment (Smith & Rose, 1998) and to monitor the performance of barriers using a Hazard Assessment Critical Control Point approach (Davison & Deere, 1999). Although it recognizes both sides of

T

Sewage from the population living within Sydney’s drinking water catchments is treated at nine sewage treatment plants whose effluent could reach Lake Burragorang—a factor that represents a potential source of contamination of the water supply.

2003 © American Water Works Association COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003

147

this debate, the American Academy of Microbiology in Washington has continued to endorse monitoring of water for key pathogens because these programs provide utilities and regulators with useful water quality information and drive improvement in detection methods (Ford & Colwell, 1996).

BACKGROUND ON SYDNEY WATER CRISIS Description of the Sydney water supply. The configuration of the Sydney water supply network during the crisis (Figure 1) has been described (Clancy, 2000; Hawkins et al, 2000). The main source of water is the Warragamba catchment, southwest of Sydney. The storage reservoir, Lake Burragorang, provides 80% of Sydney’s supply. A large proportion of Sydney’s water supply was unfiltered until 1996, when several new filtration facilities were commissioned, including the Prospect Water Filtration Plant, the world’s largest direct filtration plant. Prior to 1996, Lake Burragorang water was piped to the 40,000 ML (10,568 mil gal) Prospect Reservoir for distribution. The cold Lake Burragorang inflow (~13oC) maintained year-round thermal stratification in Prospect Reservoir, because it was always cooler than the Prospect Reservoir water. Because the offtake from Prospect Reservoir was usually just below the lake surface, the permanent thermal stratification prevented short-circuiting and provided an additional 30 days of storage before distribution. Thus, Prospect Reservoir served as an effective double reservoir barrier for Lake Burragorang water. When the Prospect plant went into operation, Burragorang water was routed directly to the plant, bypassing Prospect Reservoir. This route removed the secondary sedimentation benefit that had been provided by Prospect Reservoir, but the loss should have been more than compensated for by addition of a filtration barrier. Potential sources of cysts and oocysts in the Warragamba catchment. Approximately 100,000 people live within Sydney’s drinking water catchments, predominantly in the Warragamba catchment. Sewage from this population is treated at nine small sewage treatment plants. The effluent from these plants could reach Lake Burragorang. There are also a number of residential developments not tied into a municipal sewer system. In addition to the human population, numerous domestic livestock and native and feral animals live in this 9,000 km2 (3,475 sq mi) catchment area (McClellan, 1998). These factors represent potential sources of Cryptosporidium and Giardia contamination of Sydney’s water supply. Data on cysts and oocyst concentrations in the Sydney water supply system. AWT and SWC performed approximately 950 cyst and oocyst analyses on the city’s water supply prior to the 1998 crisis (Table 1). AWT also performed a large number of analyses for other commercial clients in this period. (In this report, all cysts and oocyst counts have been normalized to 100-L volume for ease of comparison. This is not intended as a recommendation of

such a transformation prior to data analysis because of the nonuniform distribution of cysts and oocysts in water [Nahrstedt & Gimbel, 1996]). The 950 analyses included surveys of catchments, assessments of filtration plant efficiency, and surveys of specific water distribution systems. These surveys were not designed to estimate average cyst and oocyst concentrations across individual systems. Because many samples were collected as immediate followup after cyst and oocyst detections, care must be taken in estimating mean levels from these data. An analysis of the data collected prior to 1995 showed the mean concentrations of cysts and oocysts in Sydney’s raw and finished waters were similar to those in the United States and the United Kingdom (Hutton et al, 1995). Guidelines for testing drinking water in Sydney in 1998. The November 1997 Memorandum of Understanding between SWC and the New South Wales (NSW) Department of Health committed the utility to meeting the 1996 version of the Australian Drinking Water Guidelines (NH&MRC & ARMCANZ, 1996). Although the guidelines did not recommend routine monitoring for cysts and oocysts and set no action level for cysts and oocysts in drinking water, SWC decided to include cyst and oocyst monitoring in this program. The NSW Department of Health audits the performance of SWC against the guidelines as a condition of the operating license issued by the NSW state government. SWC was obliged to notify the health department of any potential public health hazard from the water supply. In the draft update of the 1997 Interim Drinking Water Quality Incident Management Plan, SWC was required to consult with the health department following the detection of a single cyst or oocyst in finished water. SWC could subsequently issue alerts and advisories as necessary to protect consumers. 1998 Sydney drinking water quality crisis and inquiry. In July 1998, SWC’s routine surveillance program detected cysts and oocysts at high concentrations (hundreds per 100 L) in the water distribution system. The initial detections prompted SWC to progressively escalate the monitoring frequency and intensity. Episodic detection of cysts and oocysts over the next 10 weeks triggered three boil-water advisories (BWAs) for the city. Despite the unprecedented levels of contamination reported, no increase in the rate of disease caused by these organisms was detected in the exposed population. The NSW state government established the Sydney Water Inquiry early in the crisis to report to the government on the causes of the contamination. The inquiry considered advice from many experts and stakeholders. A panel of advisers led by Colin Fricker, TWUL microbiology and laboratory manager, provided assistance on microbiological issues. The chronology of the crisis has been outlined previously (Clancy, 2000). Table 2 summarizes key dates and events. A more detailed description is given in the inquiry’s final report (McClellan, 1998). The full reports can be

2003 © American Water Works Association

148

MAY 2003 | JOURNAL AWWA • 95:5 | PEER-REVIEWED | COX ET AL

Major features of the Sydney water supply

Co x Lake 's Lyell

Hawke sb

Lithgow

Warragamba Dam

e ury R i v

upper c a na

Ne pe an

W

amba rag ar s line e p pi

Sydney Central business district

Potts Hill Reservoir

g

Mittagong

Broughtons Pass

t tl

South Pacific Ocean

Metropolitan Reservoirs

iver

Woll on d Riv illy er

Burragorang

Natta iR ive r

m

Werriberri Creek e R Li

u Ko w m

Riv er

r Rive

ba River um ed

Blue Moutains Lake

Palm Beach

Prospect WFP

K

Katoomba

r

l

FIGURE 1

Illawarra

Nepean

Winge caribe e River

Goulburn

W

ol

lon

dilly Catchment areas Warragamba Upper Nepean 0

10

20

30

40

N

Kilometres scale

The two main catchments for Sydney are Warragamba (gray) and Upper Nepean (white). Both of these catchments supply the Prospect Water Filtration Plant (WFP), which treats 80% of Sydney’s water supply. Pipelines extend from the Prospect facility to the northern and eastern parts of the city. The supply from Prospect passes to the twin Potts Hill Reservoirs before continuing via two tunnels to Sydney’s central business district. The western and southern areas of Sydney are supplied by smaller systems.

accessed at www.premiers.nsw.gov.au by following the links to publications and then to Sydney Water Inquiry. Objective of this article. The aim of this article is to describe the events of the Sydney water crisis and explain how the crisis may have arisen. The laboratory methods used to measure cysts and oocysts during the crisis and the audits of laboratory performance are reviewed. The analytical results are placed in an operational context through descriptions of the hydraulic behavior of the supply system and the outcomes of a simple mass balance of cysts and oocysts in the reticulated supply.

REVIEW OF MATERIALS AND METHODS USED AWT analytical methods for cysts and oocysts in water in 1998. All samples were collected by AWT staff and transported to the laboratory at ambient temperature within 4 h of collection. Samples consisted of 100 L of finished water or 20 L of raw water collected in 20-L carboys. These containers were either previously unused or had been carefully cleaned according to the following procedure. Containers were scrubbed with detergent and

rinsed in tap water. This was followed by a 10-min wash with 12.5% volume per volume (v/v) hypochlorite and rinses with sodium thiosulfate and reverse osmosis (RO) water. As the sampling load increased during the crisis, new carboys were purchased in bulk, and reuse ceased. At all times, seeded positive control samples were prepared in separate, labeled black carboys. The Laboratory Information Management System assigned a unique code number to each sample at the point of collection to enable tracking throughout the handling and analysis process. In late 1997, monoclonal antibodies1,2 developed specifically for the detection of cysts and oocysts in water were introduced into the routine AWT immunofluorescent method described previously (Hawkins et al, 2000). A modified FC method 3 was used to sort cysts and oocysts from other debris. This method was able to sort particles from each sample onto a single 13 mm (0.51 in.) diameter 0.8-µm-pore-size polycarbonate membrane. This allowed the concentrate from each sample to be assessed microscopically on a single membrane. However, this procedure precluded routine direct differential interference contrast (DIC) microscopy. AWT had employed DIC microscopy during the development of the method with Macquarie University and found it unnecessary for routine identification of cysts and oocysts. Membranes were scanned at 200× magnification, and cysts and oocysts were identified and counted at 400× magnification. Cryptosporidium oocysts were defined as spherical, 4–6 µm in diameter, with apple-green fluorescence predominantly around the edge and with a suture line occasionally visible under immunofluorescence assay (IFA). Giardia cysts were defined as oval or spherical, 10–16 µm in length or width, showing applegreen fluorescence under IFA. SWC contracted AWT to supply results as IFA counts only. The entire sample pellet was routinely counted to avoid the substantial error associated with partial counting methods. On rare occasions, when difficult or ambiguous objects were detected in samples, expert staff and equipment were available at Macquarie University to analyze duplicate samples using phase contrast and DIC microscopy. This protocol was used particularly at the start of the crisis for confirmation. Turbid samples processed at AWT did not undergo immunomagnetic separation (IMS) before FC. Some concentrates from turbid samples were sent to Macquarie University and further processed using IMS by a modified method4 developed at the university. Otherwise, concentrates from turbid samples were passed carefully through the flow cytometer without IMS. In early September 1998, TWUL staff introduced into the AWT laboratory use of a heat-enhanced 4´,6-diamidino-2-phenylindole (DAPI) staining of cysts and oocysts developed at TWUL. This DAPI method was then used intermittently during the incident to gain more informa-

2003 © American Water Works Association COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003

149

TABLE 1

Summary of AWT Cryptosporidium and Giardia analyses for SWC on Sydney water distribution systems prior to the 1998 crisis*

Sampling Period

Cryptosporidium oocysts/100 L†

AWT 1993 report on Orchard Hills system

May–June 1993

Raw water (Warragamba pipeline): mean = 23, maximum = 160 All filtered water: mean = 0.05

ND§

Raw water: 60 (73%) Filtered water: 60 (3%)

Raw water (10-L samples) and distribution water (100-L samples) analyzed by flocFC

AWT 1994 report on North Richmond system

February 1993– October 1993

Filtered water: 0.94; maximum = 0.7

Not calculated because of low number of positive samples

Filtered water: 11 (36%)

Filtered water (1,000 L) analyzed by filtFC

Hutton et al, 1995**

May 1992– August 1993

Six major storages: mean = 240 Eight minor storages: mean = 66 Maximum in 14 storages = 4,290 Warragamba Dam: mean = 69 Hawkesbury–Nepean River: mean = 87 Distribution system: mean = 2.1

Six major storages: mean = 0.6 Eight minor storages: mean = 1 Warragamba Dam: mean ND Hawkesbury–Nepean River: mean = 1.2 Distribution system: mean ND

Major storages: 42 All storages: (60%) Distribution system: 27 (30%)

Raw water (10-L samples) analyzed by flocFC Finished water (1,000 L) and distribution water (100 L) analyzed by filtFC

AWT 1996 report

August 1993– Lake Burragorang: ND November 1994

Lake Burragorang: ND

46 ND

Analyzed by flocFC

AWT 1996 report

February 1993– June 1994

1 cyst detected in single sample

Raw: 73 (15%)

Analyzed by flocFC

AWT 1996 report

November 1993– Orchard Hills raw water: May 1994 single sample positive at 58,660 Orchard Hills filtered water: ND Warragamba–Prospect Reservoir raw water: range = 10–20 Reservoir outlet: range = 20–40

Orchard Hills raw water: ND Warragamba– Prospect Reservoir raw water: ND Reservoir outlet: 1 in one sample

Orchard Hills raw water: 59 (1.7%) (single positive sample) Orchard Hills filtered water: 2 ND Warragamba– Prospect Reservoir raw water: 89 (3%) Reservoir outlet: 49 (6%)

Raw water and Prospect Reservoir outflows analyzed by flocFC Finished and filtered water analyzed by filtFC

SWC surveillance, response, and project work

June 1995– June 1998

Raw water: maximum = 8 Finished water: maximum = 3

Raw water: 185 (8%) Finished water: 406 (4%)

All samples analyzed by modified filtFC

Data Source

North Richmond system Raw water (Hawkesbury–Nepean River): mean = 14; maximum = 6,700

Raw water: maximum = 2,738 Finished water: maximum = 758

Giardia cysts/100 L†

Number of Samples (Percent of Samples With Cryptosporidium Above Detection Limit)

Analytical Method‡

*AWT—Australian Water Technologies, SWC—Sydney Water Corporation. The Hawkesbury–Nepean River is affected by sewage. The Orchard Hills Water Filtration Plant is supplied directly from Warragamba Dam via a 900 mm (35.4 in.) pipeline. †Concentrations are not adjusted for measured or inferred recovery unless otherwise noted. ‡FlocFC—flocculation (Vesey et al, 1993) and flow cytometry (Vesey et al, 1994), filtFC—flatbed filtration (Hansen & Ongerth, 1991) and flow cytometry §ND—not detected **Concentrations are adjusted for measured recovery. Data on Warragamba raw waters are from an Orchard Hills 1993 study; both the Hawkesbury–Nepean River raw data and the finished water data are from a North Richmond Water Filtration Plant study of 1994. Giardia was rarely found in this study.

tion about the cysts and oocysts being counted. Briefly, the method of Grimason and co-workers (1994) was modified by warming the slide containing the filter to 50oC for 1 min, followed by the addition of DAPI (0.05 mg mL–1 in 1:20 v/v of methanol to deionized water) and incubation at room temperature for 5 min. AWT staff training and proficiency. AWT laboratory staff had been trained in the identification of cysts and oocysts in water since 1992, beginning with the laboratory’s collaboration with Macquarie University. Staff training records were maintained as part of the laboratory quality assurance system. Staff members were rotated through

the performance of positive and negative control samples, and their proficiency on such samples was monitored. Since December 1997, AWT had been a participant in the quarterly United Kingdom’s Laboratory Environmental Analysis Proficiency interlaboratory staff proficiency trials (operated at the time by Yorkshire Water Ltd. in Bradford) involving 13 laboratories. In the four assessments before the crisis, no staff member had produced an outlying result. Quality control (QC) samples. Cryptosporidium oocysts were prepared from bovine fecal material (C. parvum, Camden strain) using sucrose flotation and Percoll–Per-

2003 © American Water Works Association

150

MAY 2003 | JOURNAL AWWA • 95:5 | PEER-REVIEWED | COX ET AL

coll density gradient centrifugation. Giardia cysts5 were routinely imported; during the crisis, Giardia was also sourced from TWUL when the existing stock suddenly failed to stain with the modified monoclonal antibodies. Positive QC seeds comprised a mixture of cysts and oocysts, the concentration of which was determined by hemocytometer counts. Seed was stored at 4–8oC in phosphate-buffered saline with 0.05% weight per volume (w/v) sodium azide. Seeds were vortex mixed for 2 min before use. The standard 50-µL aliquot QC seeds used during the crisis contained an average of 260±70 oocysts (mean and 95% confidence interval) and 230±77 cysts. Seeds were used to perform daily QC samples on the flow cytometer to check recovery over this stage of the process. This QC also provided information on the integrity of the seed. Cyst and oocyst recovery across the entire process was assessed by running a positive control (seeded water) after approximately every eight samples, immediately followed by a negative control sample. To check the effectiveness of cleaning procedures, positive seed samples were always followed in the sample train by a negative control of distilled water. Between each sample, the concentration equipment was cleaned according to the following procedure. Equipment was cleaned by recirculating 12.5% v/v hypochlorite solution for 5 min through the equipment, rinsing with RO water, recirculating 5% w/v sodium thiosulfate for 5 min, and rinsing with RO water. If a negative control sample returned a positive result, the laboratory manager consulted with staff. At a minimum, remedial actions included investigation of the source of contamination and subsequent performance of a satisfactory set of positive and negative controls by the oper-

TABLE 2

ator. The client was also informed of QC failures. The AWT QC system also provided several internal process triggers for the close examination of all results, including the validation of work sheets before reporting to the client. AWT laboratory response to the increasing analytical workload. Cyst and oocyst analytical results were normally provided within four working days. AWT maintained a core competency unit (including one analyst with six years of experience) capable of 8 to 10 cyst and oocyst analyses per week. Backup expertise and equipment were available from experts based at Macquarie University. The maximum capacity of the laboratory was assessed as 16 samples per day to maintain quality assurance. As the crisis escalated, demand for analytical results also rose, peaking at 61 samples in a single day in late August 1998 (Figure 2). Laboratory hours were extended to 7 days a week, 24 hours a day. Experienced laboratory staff members were transferred from TWUL to AWT, bulk water concentrates were sent to Macquarie University for FC and microscopy, and 12 new analysts were hired and trained at AWT. Confirmation of positive results. The episodic nature of the cyst and oocyst detections and the initial difficulty in ascribing sources of the contamination triggered intense scrutiny of AWT laboratory results. A number of external audits were undertaken, and at the instigation of the Sydney Water Inquiry, many samples were dispatched to other laboratories to confirm the AWT results (McClellan, 1998). At the inquiry’s request, 43 water concentrates collected between July 24 and Sept. 4, 1998, were tested at Macquarie University using a fluorescent in situ hybridization (FISH) method designed to be specific for C. parvum

Summary of key developments during the 1998 Sydney water crisis

Date Reported

Description*

July 21

0 C/3 G—Prospect plant distribution chamber

July 22

0 C/1 G—Sydney Hospital tap sampling point. SWC declares a Drinking Water Incident. Investigative sampling begins.

July 23

43 C/19 G—Sydney Hospital tap. Surrounding sites return nondetects.

July 24–25

More sites in vicinity of hospital tested up to local pumping station and return positives including two >100 G. Hydrant flushings give high readings.

July 26

Further high local readings. City tunnel returns low positive. Prospect plant and key suburban sites return nondetects.

July 27–28

Eastern central business district BWA issued by SWC. Low-level positive results in eastern central business district

July 29

Positive IMS test result from Prospect plant clear water tank sediment. BWA area extended

July 30

End of distribution system high positive (Palm Beach, 365 C/151 G). BWA extended to entire Prospect distribution system

August 4

Progressive lifting of BWA completed after consistent nondetects. End of first event†

August 14

Positive results at Prospect plant inlets and outlets

August 24

Positive results at Prospect plant inlet

August 25

High positives across entire Prospect distribution system. Second BWA issued. Beginning of second event

September 1–4

Progressive lifting of second BWA under way.

September 5

High and widely distributed positives return. BWA reinstated for two weeks. Beginning of third event

September 17

Final BWA lifted

*All counts have been adjusted to 100-L sample volume equivalents (for details, see McClellan, 1998). BWA—boil-water advisory, C—Cryptosporidium, G—Giardia, IMS—immunomagnetic separation, SWC—Sydney Water Corporation †The New South Wales government inquiry divided the drinking water incident into three events, each associated with a BWA.

2003 © American Water Works Association COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003

151

tration plant was logged at 15-min intervals. These high-frequency data were complemented by in situ profiles of 70 temperature, turbidity, pH, and con60 ductivity and by microbiological analysis (including cyst/oocyst content) of 50 water samples collected from the epil40 imnion, metalimnion, and hypolimnion, as described by Hawkins and col30 leagues (2000). During the crisis, water 20 was sampled daily at one lake station near the offtake and in the plant sup10 ply. Samples were also collected less 0 frequently at other locations in Lake Burragorang, particularly during the period when floodwaters were traversing the reservoir. Data collected Staff Jan. 1998 End Jul. End Aug. Mid-Sept. End Sept. from the raw water supply pipelines Full-time 5 5 10 13 13 Casual 0 0 4 4 4 were normalized to account for the 8Equipment Jan. 1998 Aug. 21 Aug. 27 Sept. 18 Sept. 25 h travel time from the Lake BurragoFlow cytometer 1 1 1 2 2 rang offtake to the Prospect facility. Flatbed filters 2 3 3 5 6 Microscopes 2 2 3 3 3 Specialized study 2: input–output analysis of oocysts in the reticulated sysAWT—Australian Water Technologies. Period shown is from Jan. 1 to Oct. 10, 1998. The contamination crisis began on July 21. tem. At the end of the third BWA, an input–output analysis of Cryptosporidium oocysts entering and exiting the ribosomal rRNA (Vesey et al, 1998). In mid-August, sevProspect distribution system was performed. The aim was eral hundred slides that had been processed at AWT were to measure the load of oocysts entering and leaving this sent for recounting to the TWUL laboratory in Reading. distribution system from measurements of their concentraResults from 103 slides were reported to AWT in time for tion at several key locations. Giardia cysts were not included presentation to the inquiry. In addition, water samples because of their greater susceptibility to disinfection and were shipped to TWUL as well as two US laboratories lower probability of survival in traversing the system. (Clancy Environmental Consultants, St. Albans, Vt., and The load for any period was calculated as the concenCH Diagnostics and Consulting Services, Loveland, Co.). tration of oocysts/litre measured for that period multiLaboratory audits in response to the crisis. In August plied by the flow (in litres) for that period (Figure 3). The 1998, Colin Fricker of TWUL conducted an external total input load was then the cumulative total of input audit of the AWT and Macquarie University laboratoloads over the time of intensive measurement (late July ries, and the findings were presented directly to the Sydto mid-September). The total output load was calculated ney Water Inquiry. An internal AWT audit of the laborasimilarly for oocysts measured at the extremities of the tory was conducted Sept. 2, 1998, and a second external distribution system. It was assumed that the presence of audit commissioned by AWT in the aftermath of the crisuburban storage reservoirs did not interfere with the sis was conducted over two-and-a-half days between Sept. comparison of load-in versus load-out, because there were 25 and Sept. 29, 1998. The substance of this last audit and no long-term alterations in storage levels of these reservoirs. a commentary has since been published (Clancy, 2000). From late July to mid-September, oocyst concentraSpecialized study 1: limnology of the source water. The tions were measured almost daily in either the distribution behavior of the supply reservoir (Lake Burragorang), in chamber or the clear-water tank immediately downstream terms of the change in the thermal structure of the water of the Prospect plant. Raw water was monitored in the column adjacent to the plant supply offtake, was monipipelines from Lake Burragorang to the Prospect faciltored by thermistors suspended from the dam wall. These ity. Sampling of raw water started in late July, a few days sensors were permanently positioned at 3 m (9.8 ft) interafter the sampling downstream of the treatment plant. vals, from 7 m (23 ft) below the full supply level down to Samples were taken at multiple points at the extremities 60 m (197 ft) below full supply. Water temperatures over of the distribution system supplied by the plant. Where no these ranges of depths were recorded every 15 or 30 min. readings were taken, a zero score was interpolated to The turbidity in the twin raw water supply pipelines that provide conservative estimates of loads. Cumulative loads carry water from Lake Burragorang to the Prospect filwere determined as shown in Figure 4. Number of cyst and oocyst samples processed per day at AWT during the incident

28

9

m

be r

t2

Se

pt e

30

Au gu s

Ju ly

31

30

Ju ne

ay

1

M

ay

M

2

pr il 1

A

31

ar ch

M

y

nu ar

Ja

Ja

nu ar

y

1

Samples/day

FIGURE 2

2003 © American Water Works Association

152

MAY 2003 | JOURNAL AWWA • 95:5 | PEER-REVIEWED | COX ET AL

REVIEW OF RESULTS Cyst and oocyst counts during the crisis. Table 3 summarizes AWT results for 861 water samples collected from reservoirs, bulk raw water systems, plant processes, and finished water distribution systems during the crisis. The highest cyst and oocyst concentrations were found in the raw waters. A sample of one of the two bulk water pipelines from Lake Burragorang to the Prospect plant on August 27 had the highest Cryptosporidium and Giardia levels per 100 L, with counts of 12,080 and 7,620, respectively. The presence of Cryptosporidium was significantly correlated with the presence of Giardia (p < 0.01) in all sample types, except the plant backwash waters, where oocysts were present twice as often as cysts and at numbers an order of magnitude higher. Independent validation of AWT counts. On July 25, 1998, AWT reported counts of hundreds of cysts and oocysts per 100 L in samples from Sydney’s central business district. On July 26, James Smith of Montana Microbiological Services in Bozeman, who was working in Sydney at the time, examined several slides prepared at AWT. Smith had no knowledge of and was not given any information about the origin of the samples. Under IFA examination, many particles of the appropriate size and shape of Cryptosporidium oocysts and Giardia cysts were visible. DAPI staining by Smith showed clear nuclei typical of cysts and oocysts present in these organisms. These slides were photo documented. Smith also reported the presence of considerable numbers of large, autofluorescent chlorophyllous planktonic algae in the finished water samples that was indicative of raw surface water contamination. Graham Vesey, then an employee of Macquarie Research Limited in Sydney, provided another independent opinion on the identification and counts of cysts and oocysts from slides prepared at AWT from mid-August 1998 onward. Both of these experts concurred with AWT identification of cysts and oocysts on the slides. FISH results. Of 38 samples containing oocysts (as determined by IFA) that were tested using FISH, 20 samples (53%) were positive to a C. parvum rRNA probe (McClellan, 1998). The proportion of positive FISH results

Schematic of input–output analysis for Cryptosporidium oocysts taken across the distribution system during the water quality incident

FIGURE 3

Q i , Ci i

Warragamba Prospect Water Qi n ,Ci n Filtration Plant Pipeline

Distribution system

Qin is the flow rate out of Prospect Water Filtration Plant, Cin is the input concentration of oocysts, Qi is the flow rate at an extremity (i) of the distribution system, Ci is the oocyst concentration measured at extremity i. Flows and concentrations at the extremities were used to calculate the total loads exiting the system.

Method for determining cumulative counts of oocysts

FIGURE 4

Oocyst Concentration—number/L

Measurements at the system extremities were often taken at a less-than-daily frequency. However, when a positive reading was returned, it typically triggered resampling and retesting within one day. Where no sampling was done on the previous day, a zero reading was interpolated to avoid bias of load calculations by the variable time elapsed after the previous sampling. No hydrant or swabbing samples were included, because these were obtained under abnormal flow conditions and contained substantial resuspended solids. Results from such samples were not directly comparable to samples obtained under normal flow conditions. If more than one sample was analyzed per day, the average of all concentrations was used.

C11

Data point C10

C5

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15

Zero counts are interpolated where no sample was taken. Cumulative loads are determined by adding the areas under the concentration-versus-time curve. In this example, the cumulative load (concentration * time) was determined as equal to [1/2 * C5 * 2 (days + 1/2 * C10 * 1 (day) + (C10 + C11)/2 * 1 (day) + C11/2 + 1 (day)] which is [C5 + C10/2 + (C10 + C11)/2 + C11/2], which is C5 + C10 + C11.

increased as the crisis progressed. The Sydney Water Inquiry divided the crisis into three events, each associated with a BWA. During the first event, 5 of 17 results were positive, 9 of 14 in the second event, and 6 of 7 in the third event. In all 38 samples tested, 7% of the IFA positive oocysts were FISH-positive. Slides recounted at TWUL. When TWUL recounted IFAstained slides supplied by AWT laboratories, there was “good agreement in general” (Fricker, 1998). These slides were at least three weeks old when they were reread in Great Britain, and the slides were disturbed when their cover slips were lifted for DAPI staining. Results of 103 slide recounts were presented to the Sydney Water Inquiry. As expected, the TWUL Giardia and Cryptosporidium results were lower than the AWT counts (37% and 56%, respectively). Regression analysis showed that these differences between the laboratories were highly consistent (R2 = 0.93 and 0.92, respectively;

2003 © American Water Works Association COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003

153

Comparison of AWT and TWUL IFA counts

TWUL Results—oocysts/slide

2,000

Cryptosporidium y = 0.56x 2 R = 0.92

1,500

Giardia

2,000

TWUL Results—cysts

FIGURE 5

1,000

500

y = 0.37x 2 R = 0.93 1,500

1,000

500

0

0 0

500

1,000

1,500

2,000

2,500

3,000

3,500

0

AWT Results—oocysts/slide

500

1,000 1,500 2,000 2,500 3,000 3,500 4,000

AWT Results—cysts/slide

AWT—Australian Water Technologies, TWUL—Thames Water Utilities Laboratories, IFA—immunofluorescent assay. The figure compares IFA counts of Cryptosporidium oocysts and Giardia cysts on 103 slides prepared and initially counted by AWT, then recounted by TWUL up to six weeks later. Prior to the recount, TWUL removed each cover slip and additionally stained each slide with DAPI.

Figure 5). This suggests that the differences were due to the recounting procedure and not because the laboratories were counting substantially different objects. TWUL reported 83% of AWT’s IFA-positive slides were positive and 99% of AWT’s IFA negative slides were negative. The divergence in positive counts occurred almost entirely with slides that AWT had recorded as having low counts (15. The TWUL DAPI analysis of the slides showed 66% of AWT’s IFA-positive slides were positive by DAPI staining for either Cryptosporidium or Giardia. Laboratory audit 1. In August, Colin Fricker reported directly to the Sydney Water Inquiry that all personnel at AWT and Macquarie University involved in analyzing samples for cysts and oocysts were well-trained and competent. However, Fricker’s audit and that of AWT on Sept. 2, 1998, identified several QC issues including inadequate labeling of reagents and monitoring of refrigeration temperature at AWT during the crisis. The documentation made note of the deficiencies, including the need to control forms and finalize standard operating procedures, check data entry, and document QC methods, ongoing QC requirements, and actions taken to remedy QC failures. Dates for reporting on remedial actions were to be set after the crisis had ended. Laboratory audit 2. The second external audit, carried out by Jennifer Clancy of Clancy Environmental Consultants in late September, raised a number of concerns about the AWT cyst and oocyst data. Major concerns centered on (1) the potential for misidentification of cysts and oocysts as algae because AWT did not routinely use phase contrast or DIC microscopy; (2) recov-

eries of cysts and oocysts from seeded positive control samples were often below 50%, and 73% of performance samples failed to meet this criterion (the auditor may have misinterpreted a staff training criterion as the operational QC criterion, explained in detail subsequently); (3) the failure to finalize documentation for seed preparation and assessment, acceptance criteria for positive control samples, and remedial actions taken in cases of QC failures; (4) the lack of experience of AWT’s parasite-testing staff at the onset of the crisis; and (5) the failure to adequately train new staff recruited during the crisis. The auditor acknowledged that external experts had identified cysts and oocysts in water samples collected during the crisis but hypothesized that AWT had possibly contaminated these samples with material from the positive control stock because of poor laboratory practices. This hypothesis was subsequently reported as fact (Allen et al, 2000). In an article that included the audit findings (Clancy, 2000), the laboratory was described as being in “QC failure.” The auditor recommended to SWC and the Sydney Water Inquiry that the laboratory results should not be used to make public health decisions until remedial actions had been taken by the laboratory and validated by an external party. QC during the crisis. During the period of the crisis, i.e., July 15 to Sept. 18, 1998, 292 QC samples were performed at the AWT laboratory, including some samples that were concentrated at AWT and subsequently processed at other laboratories. These included 123 positive controls, 123 negative controls, and 46 training samples processed by new staff. Although positive and negative control sets were normally performed after every

2003 © American Water Works Association

154

MAY 2003 | JOURNAL AWWA • 95:5 | PEER-REVIEWED | COX ET AL

eight experimental samples, sometimes for logistic reasons fewer than eight client samples were analyzed in the batch between controls. The positive control samples yielded average recoveries of 59% for Cryptosporidium oocysts (n = 123, standard deviation = 37) and 40% for Giardia cysts (n = 123, standard deviation = 35). Both the method of seed preparation (dilution followed by hemocytometer estimation of concentrations) and variations in the quality of water samples tested (e.g., the raw waters during the crisis had a wide range of turbidity levels) contributed to the relatively high variation in calculated recoveries. During this same period, 8.6% of positive control samples gave zero recoveries for either cysts or oocysts. The majority of these failures were attributable to the sudden loss of detectability of Giardia in the seed stock and a delay in obtaining replacement cysts from overseas. With the 123 negative control samples, 120 yielded negative results. Of the remainder, two samples contained one Cryptosporidium oocyst and zero Giardia cysts, and one sample contained one Giardia cyst and zero Cryptosporidium oocysts. Two of these failed negative controls were concentrated at AWT and subsequently processed at Macquarie University, making it difficult to assess where the contamination may have occurred. Specialized study 1: limnology of Lake Burragorang. Prior to the crisis, Lake Burragorang was last at full supply in January 1992. Several years of below-average rainfall had resulted in progressive drawdown of the reservoir to 58% of capacity (13 m [43 ft] below the spillway) in July 1998. Fecal matter from native and domestic animals would have accumulated on catchment riverbanks and in the emerging drawdown zone of the reservoir during this drought period (McClellan, 1998). Two heavy rainfall events occurred in the water supply catchments in August 1998. These rain events caused sewage treatment plants in the catchment to overflow and swept accumulated contamination from the catchment into Lake Burragorang (McClellan 1998). The first rain occurred August 7–9 and resulted in a rise in the stored

TABLE 3

water volume from 58 to 83%. The second rain occurred August 16–19 and lifted the lake level to 100% capacity. Various statistics have determined the probability of these events recurring. The Annual Recurrence Interval (ARI) for the total August rainfall was 2.5 years, whereas the compound probability for the rapid two-step filling of Lake Burragorang that occurred in August 1998 had an ARI of 30 years. Most important, the ARI for an inflow of 700,000 ML (184,940 mil gal) from the Warragamba catchment into Lake Burragorang, which occurred in August 1998, was 4.5 years. Lake Burragorang was thermally mixed in July 1998 (midwinter), and the temperature of the water column was a uniform 13oC at the dam wall. Floodwater from the first rain entered the reservoir as an underflow and established a hypolimnetic water mass, with a temperature, pH, alkalinity, turbidity, and color signature distinctly different from the overlying lake water. The “old” lake water formed the epilimnion and was conveniently demarcated from the turbid floodwater by the 12oC isotherm. The position of this isotherm and, therefore, the position of the floodwater were resolved within ±1.5 m (4.9 ft) by the thermistors at the dam wall (Figure 6). According to 1 m (3.3 ft) interval vertical profiles of temperature, pH, and turbidity taken at locations throughout the reservoir during the flood period, there was only limited mixing between the “new” floodwater and “old” lake water. Most important, floodwater from each rain traveled through the reservoir to the dam wall within seven days. Subsequent numerical modeling simulations of the inflows (as described by Antenucci, 2001) supported the field observations of limited mixing and sevenday travel time. The first floodwater mass arrived at the dam wall between August 14 and August 16. This coincided with renewed detection of cysts and oocysts at the Prospect filtration plant (Table 2). The second floodwater mass was slightly warmer than the original and traversed the lake as a new layer between the “old” lake water and the first floodwater. This second flood arrived at the dam

Summary of 861 samples processed during the crisis (July 21 to Sept. 18, 1998)*

Number of Samples

Cryptosporidium percent positive†

Storage lakes

143

Bulk supply

200

Treatment plants Water filtration plant backwash Reticulation system

Location

Cryptosporidium maximum

Cryptosporidium minimum

Cryptosporidium mean

Giardia percent positive†

Giardia maximum

21

8,600

0

186

18

1,822

0

66

33

12,080

0

338

25

7,620

0

143

343

18

9,445

0

60

17

3,500

0

30

36

28

2,687

0

304

14

285

0

13

139

11

273

0

5

10

109

0

2

Giardia Giardia minimum mean

*Counts are adjusted to per 100 L. Mean is arithmetic mean. †Positive by immunofluorescence assay

2003 © American Water Works Association COX ET AL | PEER-REVIEWED | 95:5 • JOURNAL AWWA | MAY 2003

155

zone” underestimates the depth below the offtake from which water can be drawn. During the crisis, thermal stratification was weak, and the abstrac5 (16) tion zone would probably have 0 (0) 5 (16) extended substantially above and 15 10 (33) below the offtake aperture. 15 (49) Immediately after the first BWA, 20 (66) 14 the deepest offtake was closed, and 25 (82) the supply was drawn exclusively from 30 (98) 13 35 (115) the shallowest offtake (4–11 m [13–36 40 (131) ft] below the surface). The 13 m (43 ft) 45 (148) 12 rise in lake level following the floods 50 (164) and the internal waves meant that the 55 (180) floodwater moved into the proximity 60 (197) 01 05 09 13 17 21 25 29 02 06 10 14 18 22 26 30 of this offtake (Figure 6). The relaAugust 1998 September 1998 tionship between internal wave activTime—days ity, offtake position, and occurrence Internal wave activity was plotted using high-frequency thermistor data. Cold turbid floodwater contaminated with Cryptosporidium and Giardia and with temperature of cysts and oocysts in the water sup