Sydney 1998: lessons from a drinking water crisis

2003 © American Water Works Association

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, high concentrations of Cryptosporidium and Giardia were

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

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

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

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

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

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.

BY PETER COX,

IAN FISHER,

GEORGE KASTL,

VEERIAH JEGATHEESAN,

MALCOLM WARNECKE,

MARK ANGLES,

HERI BUSTAMANTE,

TONY CHIFFINGS,

AND PETER R. HAWKINS

he methods for monitoring Cryptosporidium oocysts and Giardia cystsin water are relatively new, with most development occurring in the last15 years. The methods, especially the Information Collection Rule (ICR)method, have been criticized for their technical difficulty and lack ofaccuracy. Although false-positive results can occur, greater problems

have been associated with false-negatives and low and variable recovery ratesdependent on water quality (e.g., turbidity) (Clancy et al, 1999; Dufour et al, 1999;Fricker, 1995; Clancy et al, 1994). Despite these methodological problems, sev-eral hundred reports have been published in recent years (Clancy et al, 2000; Con-nell et al, 2000; Clancy et al, 1999) that have established the patterns of occur-rence of cysts and oocysts in water and have led to substantial improvement inanalytical methods.

The Sydney Water Corporation (SWC), through the laboratories of its subsidiaryAustralian Water Technologies (AWT), has monitored cysts and oocysts in Syd-ney’s water supply since 1992. AWT’s flow cytometric (FC) method was co-developed with Macquarie University in Sydney and built on prior developmentsat Thames Water Utilities Laboratories (TWUL) in Reading, Great Britain. An ini-tial unpublished in-house study prepared by AWT and Macquarie Universityshowed that the FC method, using either flocculation (Vesey et al, 1993) orflatbed filtration (Hansen &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 usefulnessof cyst and oocyst monitoring as a basis for public health decisions (Allen et al,2000). However, a number of environmental microbiologists support the use ofthese data for scientific-based risk assessment (Smith & Rose, 1998) and to mon-itor the performance of barriers using a Hazard Assessment Critical ControlPoint 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.

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

this debate, the American Academy of Microbiology inWashington has continued to endorse monitoring of waterfor key pathogens because these programs provide utili-ties and regulators with useful water quality informationand drive improvement in detection methods (Ford &Colwell, 1996).

BACKGROUND ON SYDNEY WATER CRISISDescription of the Sydney water supply. The configura-

tion of the Sydney water supply network during the cri-sis (Figure 1) has been described (Clancy, 2000; Hawkinset al, 2000). The main source of water is the Warragambacatchment, southwest of Sydney. The storage reservoir,Lake Burragorang, provides 80% of Sydney’s supply. Alarge proportion of Sydney’s water supply was unfiltereduntil 1996, when several new filtration facilities werecommissioned, including the Prospect Water FiltrationPlant, the world’s largest direct filtration plant.

Prior to 1996, Lake Burragorang water was piped tothe 40,000 ML (10,568 mil gal) Prospect Reservoir for dis-tribution. The cold Lake Burragorang inflow (~13oC)maintained year-round thermal stratification in ProspectReservoir, because it was always cooler than the ProspectReservoir water. Because the offtake from Prospect Reser-voir was usually just below the lake surface, the perma-nent thermal stratification prevented short-circuiting andprovided an additional 30 days of storage before distri-bution. Thus, Prospect Reservoir served as an effectivedouble reservoir barrier for Lake Burragorang water.

When the Prospect plant went into operation, Bur-ragorang water was routed directly to the plant, bypass-ing Prospect Reservoir. This route removed the secondarysedimentation benefit that had been provided by ProspectReservoir, but the loss should have been more than com-pensated for by addition of a filtration barrier.

Potential sources of cysts and oocysts in the Warragambacatchment. Approximately 100,000 people live withinSydney’s drinking water catchments, predominantly inthe Warragamba catchment. Sewage from this popula-tion is treated at nine small sewage treatment plants. Theeffluent from these plants could reach Lake Burragorang.There are also a number of residential developments nottied into a municipal sewer system. In addition to thehuman population, numerous domestic livestock andnative and feral animals live in this 9,000 km2 (3,475 sqmi) catchment area (McClellan, 1998). These factors rep-resent potential sources of Cryptosporidium and Giar-dia contamination of Sydney’s water supply.

Data on cysts and oocyst concentrations in the Sydneywater supply system. AWT and SWC performed approx-imately 950 cyst and oocyst analyses on the city’s watersupply prior to the 1998 crisis (Table 1). AWT also per-formed a large number of analyses for other commercialclients in this period. (In this report, all cysts and oocystcounts have been normalized to 100-L volume for ease ofcomparison. This is not intended as a recommendation of

such a transformation prior to data analysis because of thenonuniform distribution of cysts and oocysts in water[Nahrstedt & Gimbel, 1996]).

The 950 analyses included surveys of catchments,assessments of filtration plant efficiency, and surveys ofspecific water distribution systems. These surveys werenot designed to estimate average cyst and oocyst con-centrations across individual systems. Because many sam-ples were collected as immediate followup after cyst andoocyst detections, care must be taken in estimating meanlevels from these data. An analysis of the data collectedprior to 1995 showed the mean concentrations of cystsand oocysts in Sydney’s raw and finished waters weresimilar to those in the United States and the United King-dom (Hutton et al, 1995).

Guidelines for testing drinking water in Sydney in 1998.The November 1997 Memorandum of Understandingbetween SWC and the New South Wales (NSW) Depart-ment of Health committed the utility to meeting the 1996version of the Australian Drinking Water Guidelines(NH&MRC & ARMCANZ, 1996). Although the guide-lines did not recommend routine monitoring for cystsand oocysts and set no action level for cysts and oocystsin drinking water, SWC decided to include cyst and oocystmonitoring in this program. The NSW Department ofHealth audits the performance of SWC against the guide-lines as a condition of the operating license issued by theNSW state government. SWC was obliged to notify thehealth department of any potential public health hazardfrom the water supply. In the draft update of the 1997Interim Drinking Water Quality Incident ManagementPlan, SWC was required to consult with the health depart-ment following the detection of a single cyst or oocystin finished water. SWC could subsequently issue alertsand advisories as necessary to protect consumers.

1998 Sydney drinking water quality crisis and inquiry. InJuly 1998, SWC’s routine surveillance program detectedcysts and oocysts at high concentrations (hundreds per 100L) in the water distribution system. The initial detectionsprompted SWC to progressively escalate the monitoringfrequency and intensity. Episodic detection of cysts andoocysts over the next 10 weeks triggered three boil-wateradvisories (BWAs) for the city. Despite the unprecedentedlevels of contamination reported, no increase in the rateof disease caused by these organisms was detected in theexposed population. The NSW state government estab-lished the Sydney Water Inquiry early in the crisis toreport to the government on the causes of the contami-nation. The inquiry considered advice from many expertsand stakeholders. A panel of advisers led by Colin Fricker,TWUL microbiology and laboratory manager, providedassistance on microbiological issues.

The chronology of the crisis has been outlined previ-ously (Clancy, 2000). Table 2 summarizes key dates andevents. A more detailed description is given in the inquiry’sfinal report (McClellan, 1998). The full reports can be


https://www.researchgate.net/publication/254156130_Sydney's_1998_Water_Quality_Crisis?el=1_x_8&enrichId=rgreq-1cfa26c6-1823-42c6-93bf-abd0dd7ccfb8&enrichSource=Y292ZXJQYWdlOzIzOTc3ODEzMTtBUzo5NzMwNzE0Mzk2NjczNUAxNDAwMjExMjM5NjQ3


accessed at www.premiers.nsw.gov.au by following thelinks to publications and then to Sydney Water Inquiry.

Objective of this article. The aim of this article is todescribe the events of the Sydney water crisis and explainhow the crisis may have arisen. The laboratory methodsused to measure cysts and oocysts during the crisis and theaudits of laboratory performance are reviewed. The ana-lytical results are placed in an operational context throughdescriptions of the hydraulic behavior of the supply sys-tem and the outcomes of a simple mass balance of cystsand oocysts in the reticulated supply.

REVIEW OF MATERIALSAND METHODS USED

AWT analytical methods for cysts and oocysts in waterin 1998. All samples were collected by AWT staff andtransported to the laboratory at ambient temperaturewithin 4 h of collection. Samples consisted of 100 L of fin-ished water or 20 L of raw water collected in 20-L car-boys. These containers were either previously unused orhad been carefully cleaned according to the followingprocedure. Containers were scrubbed with detergent and

rinsed in tap water. This was followed by a 10-min washwith 12.5% volume per volume (v/v) hypochlorite andrinses 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 pre-pared in separate, labeled black carboys. The Labora-tory Information Management System assigned a uniquecode number to each sample at the point of collection toenable tracking throughout the handling and analysisprocess.

In late 1997, monoclonal antibodies1,2 developedspecifically for the detection of cysts and oocysts in waterwere introduced into the routine AWT immunofluores-cent method described previously (Hawkins et al, 2000).A modified FC method3 was used to sort cysts andoocysts from other debris. This method was able to sortparticles 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 beassessed microscopically on a single membrane. However,this procedure precluded routine direct differential inter-ference contrast (DIC) microscopy. AWT had employedDIC microscopy during the development of the methodwith Macquarie University and found it unnecessary forroutine identification of cysts and oocysts.

Membranes were scanned at 200× magnification, andcysts and oocysts were identified and counted at 400×magnification. Cryptosporidium oocysts were definedas spherical, 4–6 µm in diameter, with apple-green fluo-rescence predominantly around the edge and with asuture line occasionally visible under immunofluores-cence assay (IFA). Giardia cysts were defined as oval orspherical, 10–16 µm in length or width, showing apple-

green fluorescence under IFA. SWC contracted AWT tosupply results as IFA counts only.

The entire sample pellet was routinely counted to avoidthe substantial error associated with partial countingmethods. On rare occasions, when difficult or ambiguousobjects were detected in samples, expert staff and equip-ment were available at Macquarie University to analyzeduplicate samples using phase contrast and DICmicroscopy. This protocol was used particularly at thestart of the crisis for confirmation.

Turbid samples processed at AWT did not undergoimmunomagnetic separation (IMS) before FC. Some con-centrates from turbid samples were sent to MacquarieUniversity and further processed using IMS by a modifiedmethod4 developed at the university. Otherwise, concen-trates from turbid samples were passed carefully throughthe flow cytometer without IMS.

In early September 1998, TWUL staff introduced intothe AWT laboratory use of a heat-enhanced 4´,6-diami-dino-2-phenylindole (DAPI) staining of cysts and oocystsdeveloped at TWUL. This DAPI method was then usedintermittently during the incident to gain more informa-


GoulburnW

ollondilly

Winge caribeeRiver

Riv

er

Wol

lond

illy

Nepean

Riv

er

Ko

wm

umg

River

Katoomba

Burragorang

Lake

Kedumba

River

BlueMoutains

Nat

tai R

iver

Illawarra

BroughtonsPass

MetropolitanReservoirs

Little

River

Mittagong

Werrib

erri Creek

WarragambaDam Prospect

WFP

Haw

kesb

ury R iver

W

arragamba

pipelines PottsHillReservoir

Nep

ean

SydneyCentral business district

PalmBeach

SouthPacificOcean

N

Lithgow

LakeLyell

Cox's

Catchment areasWarragambaUpper Nepean

0 10 20 30 40

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.

up

per

can

al

FIGURE 1 MMaajjoorr ffeeaattuurreess ooff tthhee SSyyddnneeyy wwaatteerr ssuuppppllyy


tion about the cysts and oocysts being counted. Briefly, themethod of Grimason and co-workers (1994) was modi-fied by warming the slide containing the filter to 50oC for1 min, followed by the addition of DAPI (0.05 mg mL–1

in 1:20 v/v of methanol to deionized water) and incuba-tion at room temperature for 5 min.

AWT staff training and proficiency. AWT laboratory staffhad been trained in the identification of cysts and oocystsin water since 1992, beginning with the laboratory’s col-laboration with Macquarie University. Staff trainingrecords were maintained as part of the laboratory qual-ity assurance system. Staff members were rotated through

the performance of positive and negative control sam-ples, and their proficiency on such samples was moni-tored. Since December 1997, AWT had been a partici-pant in the quarterly United Kingdom’s LaboratoryEnvironmental Analysis Proficiency interlaboratory staffproficiency trials (operated at the time by Yorkshire WaterLtd. in Bradford) involving 13 laboratories. In the fourassessments before the crisis, no staff member had pro-duced an outlying result.

Quality control (QC) samples. Cryptosporidium oocystswere prepared from bovine fecal material (C. parvum,Camden strain) using sucrose flotation and Percoll–Per-


Number of Samples(Percent of Samples

Sampling Cryptosporidium Giardia With Cryptosporidium Analytical Data Source Period oocysts/100 L† cysts/100 L† Above Detection Limit) Method‡

AWT 1993 report May–June 1993 Raw water (Warragamba ND§ Raw water: Raw water (10-L on Orchard Hills pipeline): mean = 23, 60 (73%) samples) and system maximum = 160 Filtered water: distribution water

All filtered water: 60 (3%) (100-L samples)mean = 0.05 analyzed by flocFC

AWT 1994 report February 1993– Filtered water: 0.94; Not calculated Filtered water: Filtered water on North October 1993 maximum = 0.7 because of low 11 (36%) (1,000 L) analyzed Richmond number of positive by filtFCsystem samples

Hutton et al, May 1992– Six major storages: Six major storages: Major storages: 42 Raw water (10-L1995** August 1993 mean = 240 mean = 0.6 All storages: (60%) samples) analyzed

Eight minor storages: Eight minor Distribution system: by flocFCmean = 66 storages: mean = 1 27 (30%) Finished water Maximum in 14 Warragamba Dam: (1,000 L) and storages = 4,290 mean ND distribution water Warragamba Dam: Hawkesbury–Nepean (100 L) analyzedmean = 69 River: mean = 1.2 by filtFCHawkesbury–Nepean Distribution system:River: mean = 87 mean NDDistribution system: mean = 2.1

AWT 1996 report August 1993– Lake Burragorang: ND Lake Burragorang: ND 46 ND Analyzed by flocFCNovember 1994

AWT 1996 report February 1993– North Richmond system 1 cyst detected in Raw: 73 (15%) Analyzed by flocFCJune 1994 Raw water single sample

(Hawkesbury–Nepean River): mean = 14; maximum = 6,700

AWT 1996 report November 1993– Orchard Hills raw water: Orchard Hills raw Orchard Hills raw water: Raw water and May 1994 single sample positive water: ND 59 (1.7%) (single Prospect Reservoir

at 58,660 Warragamba– positive sample) outflows analyzed Orchard Hills filtered Prospect Reservoir Orchard Hills filtered by flocFCwater: ND raw water: ND water: 2 ND Finished and Warragamba–Prospect Reservoir outlet: Warragamba– filtered water Reservoir raw water: 1 in one sample Prospect Reservoir analyzed by filtFCrange = 10–20 raw water: 89 (3%)Reservoir outlet: Reservoir outlet: range = 20–40 49 (6%)

SWC surveillance, June 1995– Raw water: Raw water: Raw water: 185 (8%) All samples response, and June 1998 maximum = 2,738 maximum = 8 Finished water: analyzed by project work Finished water: Finished water: 406 (4%) modified filtFC

maximum = 758 maximum = 3

*AWT—Australian Water Technologies, SWC—Sydney Water Corporation. The Hawkesbury–Nepean River is affected by sewage. The Orchard Hills Water FiltrationPlant 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 Riverraw data and the finished water data are from a North Richmond Water Filtration Plant study of 1994. Giardia was rarely found in this study.

TABLE 1 SSuummmmaarryy ooff AAWWTT CCrryyppttoossppoorriiddiiuumm aanndd GGiiaarrddiiaa aannaallyysseess ffoorr SSWWCC oonn SSyyddnneeyy wwaatteerr ddiissttrriibbuuttiioonn ssyysstteemmss pprriioorr ttoo tthhee 11999988 ccrriissiiss**


coll density gradient centrifugation. Giardia cysts5 wereroutinely imported; during the crisis, Giardia was alsosourced from TWUL when the existing stock suddenlyfailed to stain with the modified monoclonal antibodies.

Positive QC seeds comprised a mixture of cysts andoocysts, the concentration of which was determined byhemocytometer counts. Seed was stored at 4–8oC in phos-phate-buffered saline with 0.05% weight per volume(w/v) sodium azide. Seeds were vortex mixed for 2 minbefore use. The standard 50-µL aliquot QC seeds used dur-ing 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 flowcytometer to check recovery over this stage of the process.This QC also provided information on the integrity ofthe seed. Cyst and oocyst recovery across the entire processwas assessed by running a positive control (seeded water)after approximately every eight samples, immediately fol-lowed by a negative control sample. To check the effec-tiveness of cleaning procedures, positive seed sampleswere always followed in the sample train by a negativecontrol of distilled water. Between each sample, the con-centration equipment was cleaned according to the fol-lowing procedure. Equipment was cleaned by recirculat-ing 12.5% v/v hypochlorite solution for 5 min through theequipment, rinsing with RO water, recirculating 5% w/vsodium 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 mini-mum, remedial actions included investigation of the sourceof contamination and subsequent performance of a satis-factory set of positive and negative controls by the oper-

ator. The client was also informed of QC failures. TheAWT QC system also provided several internal processtriggers for the close examination of all results, includingthe validation of work sheets before reporting to the client.

AWT laboratory response to the increasing analyticalworkload. Cyst and oocyst analytical results were nor-mally provided within four working days. AWT main-tained a core competency unit (including one analyst withsix years of experience) capable of 8 to 10 cyst and oocystanalyses per week. Backup expertise and equipment wereavailable from experts based at Macquarie University.The maximum capacity of the laboratory was assessed as16 samples per day to maintain quality assurance.

As the crisis escalated, demand for analytical results alsorose, peaking at 61 samples in a single day in late August1998 (Figure 2). Laboratory hours were extended to 7days a week, 24 hours a day. Experienced laboratorystaff members were transferred from TWUL to AWT,bulk water concentrates were sent to Macquarie Univer-sity for FC and microscopy, and 12 new analysts werehired and trained at AWT.

Confirmation of positive results. The episodic nature of thecyst and oocyst detections and the initial difficulty inascribing sources of the contamination triggered intensescrutiny of AWT laboratory results. A number of externalaudits were undertaken, and at the instigation of the Syd-ney Water Inquiry, many samples were dispatched to otherlaboratories to confirm the AWT results (McClellan, 1998).

At the inquiry’s request, 43 water concentrates col-lected between July 24 and Sept. 4, 1998, were tested atMacquarie University using a fluorescent in situ hybridiza-tion (FISH) method designed to be specific for C. parvum


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.

TABLE 2 SSuummmmaarryy ooff kkeeyy ddeevveellooppmmeennttss dduurriinngg tthhee 11999988 SSyyddnneeyy wwaatteerr ccrriissiiss


ribosomal rRNA (Vesey et al, 1998). In mid-August, sev-eral hundred slides that had been processed at AWT weresent for recounting to the TWUL laboratory in Reading.Results from 103 slides were reported to AWT in time forpresentation to the inquiry. In addition, water sampleswere shipped to TWUL as well as two US laboratories(Clancy Environmental Consultants, St. Albans, Vt., andCH Diagnostics and Consulting Services, Loveland, Co.).

Laboratory audits in response to the crisis. In August1998, Colin Fricker of TWUL conducted an externalaudit of the AWT and Macquarie University laborato-ries, and the findings were presented directly to the Syd-ney Water Inquiry. An internal AWT audit of the labora-tory was conducted Sept. 2, 1998, and a second externalaudit commissioned by AWT in the aftermath of the cri-sis was conducted over two-and-a-half days between Sept.25 and Sept. 29, 1998. The substance of this last audit anda commentary has since been published (Clancy, 2000).

Specialized study 1: limnology of the source water. Thebehavior of the supply reservoir (Lake Burragorang), interms of the change in the thermal structure of the watercolumn adjacent to the plant supply offtake, was moni-tored by thermistors suspended from the dam wall. Thesesensors were permanently positioned at 3 m (9.8 ft) inter-vals, from 7 m (23 ft) below the full supply level down to60 m (197 ft) below full supply. Water temperatures overthese ranges of depths were recorded every 15 or 30 min.The turbidity in the twin raw water supply pipelines thatcarry water from Lake Burragorang to the Prospect fil-

tration plant was logged at 15-minintervals.

These high-frequency data werecomplemented by in situ profiles oftemperature, turbidity, pH, and con-ductivity and by microbiological analy-sis (including cyst/oocyst content) ofwater samples collected from the epil-imnion, metalimnion, and hypolim-nion, as described by Hawkins and col-leagues (2000). During the crisis, waterwas sampled daily at one lake stationnear the offtake and in the plant sup-ply. Samples were also collected lessfrequently at other locations in LakeBurragorang, particularly during theperiod when floodwaters were tra-versing the reservoir. Data collectedfrom the raw water supply pipelineswere normalized to account for the 8-h travel time from the Lake Burrago-rang offtake to the Prospect facility.

Specialized study 2: input–outputanalysis of oocysts in the reticulated sys-tem. At the end of the third BWA, aninput–output analysis of Cryptosporid-ium oocysts entering and exiting the

Prospect distribution system was performed. The aim wasto measure the load of oocysts entering and leaving thisdistribution system from measurements of their concentra-tion at several key locations. Giardia cysts were not includedbecause of their greater susceptibility to disinfection andlower probability of survival in traversing the system.

The load for any period was calculated as the concen-tration of oocysts/litre measured for that period multi-plied by the flow (in litres) for that period (Figure 3). Thetotal input load was then the cumulative total of inputloads over the time of intensive measurement (late Julyto mid-September). The total output load was calculatedsimilarly for oocysts measured at the extremities of thedistribution system. It was assumed that the presence ofsuburban storage reservoirs did not interfere with thecomparison of load-in versus load-out, because there wereno long-term alterations in storage levels of these reservoirs.

From late July to mid-September, oocyst concentra-tions were measured almost daily in either the distributionchamber or the clear-water tank immediately downstreamof the Prospect plant. Raw water was monitored in thepipelines from Lake Burragorang to the Prospect facil-ity. Sampling of raw water started in late July, a few daysafter the sampling downstream of the treatment plant.Samples were taken at multiple points at the extremitiesof the distribution system supplied by the plant. Where noreadings were taken, a zero score was interpolated toprovide conservative estimates of loads. Cumulative loadswere determined as shown in Figure 4.


FIGURE 2 NNuummbbeerr ooff ccyysstt aanndd ooooccyysstt ssaammpplleess pprroocceesssseedd ppeerr ddaayy aatt AAWWTT dduurriinnggtthhee iinncciiddeenntt

70

60

50

40

30

20

10

0

Sam

ple

s/d

ay

Januar

y 1

Januar

y 31

March

2

April 1

May 1

May 31

June 3

0

July

30

August 29

Septe

mber

28

StaffFull-timeCasual

EquipmentFlow cytometerFlatbed filtersMicroscopes

Jan. 199850

End Jul.50

End Aug. 104

Mid-Sept. 134

End Sept. 134

Jan. 1998122

Aug. 21132

Aug. 27133

Sept. 18253

Sept. 25263

AWT—Australian Water Technologies. Period shown is from Jan. 1 to Oct. 10, 1998.The contamination crisis began on July 21.

https://www.researchgate.net/publication/254156130_Sydney's_1998_Water_Quality_Crisis?el=1_x_8&enrichId=rgreq-1cfa26c6-1823-42c6-93bf-abd0dd7ccfb8&enrichSource=Y292ZXJQYWdlOzIzOTc3ODEzMTtBUzo5NzMwNzE0Mzk2NjczNUAxNDAwMjExMjM5NjQ3


Measurements at the system extremities were oftentaken at a less-than-daily frequency. However, when apositive reading was returned, it typically triggered resam-pling and retesting within one day. Where no sampling wasdone on the previous day, a zero reading was interpo-lated to avoid bias of load calculations by the variable timeelapsed after the previous sampling. No hydrant or swab-bing samples were included, because these were obtainedunder abnormal flow conditions and contained substan-tial resuspended solids. Results from such samples werenot directly comparable to samples obtained under nor-mal flow conditions. If more than one sample was ana-lyzed per day, the average of all concentrations was used.

REVIEW OF RESULTSCyst and oocyst counts during the crisis. Table 3 sum-

marizes AWT results for 861 water samples collectedfrom reservoirs, bulk raw water systems, plant processes,and finished water distribution systems during the crisis.The highest cyst and oocyst concentrations were found inthe raw waters. A sample of one of the two bulk waterpipelines from Lake Burragorang to the Prospect plant onAugust 27 had the highest Cryptosporidium and Giardialevels per 100 L, with counts of 12,080 and 7,620, respec-tively. The presence of Cryptosporidium was significantlycorrelated with the presence of Giardia (p < 0.01) in allsample types, except the plant backwash waters, whereoocysts were present twice as often as cysts and at num-bers an order of magnitude higher.

Independent validation of AWT counts. On July 25, 1998,AWT reported counts of hundreds of cysts and oocysts per100 L in samples from Sydney’s central business district.On July 26, James Smith of Montana MicrobiologicalServices in Bozeman, who was working in Sydney at thetime, examined several slides prepared at AWT. Smithhad no knowledge of and was not given any informa-tion about the origin of the samples.

Under IFA examination, many particles of the appro-priate size and shape of Cryptosporidium oocysts andGiardia cysts were visible. DAPI staining by Smith showedclear nuclei typical of cysts and oocysts present in theseorganisms. These slides were photo documented. Smithalso reported the presence of considerable numbers oflarge, autofluorescent chlorophyllous planktonic algaein the finished water samples that was indicative of rawsurface water contamination. Graham Vesey, then anemployee of Macquarie Research Limited in Sydney, pro-vided another independent opinion on the identificationand counts of cysts and oocysts from slides prepared atAWT from mid-August 1998 onward. Both of theseexperts concurred with AWT identification of cysts andoocysts on the slides.

FISH results. Of 38 samples containing oocysts (asdetermined by IFA) that were tested using FISH, 20 sam-ples (53%) were positive to a C. parvum rRNA probe(McClellan, 1998). The proportion of positive FISH results

increased as the crisis progressed. The Sydney WaterInquiry divided the crisis into three events, each associatedwith a BWA. During the first event, 5 of 17 results werepositive, 9 of 14 in the second event, and 6 of 7 in the thirdevent. In all 38 samples tested, 7% of the IFA positiveoocysts were FISH-positive.

Slides recounted at TWUL. When TWUL recounted IFA-stained slides supplied by AWT laboratories, there was“good agreement in general” (Fricker, 1998). These slideswere at least three weeks old when they were reread inGreat Britain, and the slides were disturbed when theircover slips were lifted for DAPI staining.

Results of 103 slide recounts were presented to theSydney Water Inquiry. As expected, the TWUL Giardiaand Cryptosporidium results were lower than the AWTcounts (37% and 56%, respectively). Regression analysisshowed that these differences between the laboratorieswere highly consistent (R2 = 0.93 and 0.92, respectively;


Prospect WaterFiltration Plant

Warragamba

Pipeline

Qin,Ci n

Qi , Ci

Distributionsystem

i

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.

FIGURE 3 SScchheemmaattiicc ooff iinnppuutt––oouuttppuutt aannaallyyssiiss ffoorrCCrryyppttoossppoorriiddiiuumm ooooccyyssttss ttaakkeenn aaccrroossss tthhee ddiissttrriibbuuttiioonnssyysstteemm dduurriinngg tthhee wwaatteerr qquuaalliittyy iinncciiddeenntt

Oo

cyst

Co

nce

ntr

atio

n—

num

ber

/L

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Data point C11

C10

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.

C5

FIGURE 4 MMeetthhoodd ffoorr ddeetteerrmmiinniinngg ccuummuullaattiivvee ccoouunnttss ooff ooooccyyssttss


Figure 5). This suggests that the differences were due to therecounting procedure and not because the laboratorieswere counting substantially different objects. TWULreported 83% of AWT’s IFA-positive slides were positiveand 99% of AWT’s IFA negative slides were negative. Thedivergence in positive counts occurred almost entirely withslides that AWT had recorded as having low counts (<15).On two slides (2%), TWUL recorded a nil count, whereasAWT recorded a count >15. The TWUL DAPI analysisof the slides showed 66% of AWT’s IFA-positive slideswere positive by DAPI staining for either Cryptosporidiumor Giardia.

Laboratory audit 1. In August, Colin Fricker reporteddirectly to the Sydney Water Inquiry that all personnel atAWT and Macquarie University involved in analyzingsamples for cysts and oocysts were well-trained and com-petent. However, Fricker’s audit and that of AWT onSept. 2, 1998, identified several QC issues including inad-equate labeling of reagents and monitoring of refrigera-tion temperature at AWT during the crisis. The docu-mentation made note of the deficiencies, including theneed to control forms and finalize standard operatingprocedures, check data entry, and document QC methods,ongoing QC requirements, and actions taken to remedyQC failures. Dates for reporting on remedial actions wereto be set after the crisis had ended.

Laboratory audit 2. The second external audit, carriedout by Jennifer Clancy of Clancy Environmental Con-sultants in late September, raised a number of concernsabout the AWT cyst and oocyst data. Major concernscentered on (1) the potential for misidentification ofcysts and oocysts as algae because AWT did not rou-tinely use phase contrast or DIC microscopy; (2) recov-

eries of cysts and oocysts from seeded positive controlsamples were often below 50%, and 73% of perfor-mance samples failed to meet this criterion (the auditormay have misinterpreted a staff training criterion as theoperational QC criterion, explained in detail subse-quently); (3) the failure to finalize documentation forseed preparation and assessment, acceptance criteria forpositive control samples, and remedial actions taken incases of QC failures; (4) the lack of experience of AWT’sparasite-testing staff at the onset of the crisis; and (5)the failure to adequately train new staff recruited duringthe crisis.

The auditor acknowledged that external experts hadidentified cysts and oocysts in water samples collectedduring the crisis but hypothesized that AWT had possiblycontaminated these samples with material from the pos-itive control stock because of poor laboratory practices.This hypothesis was subsequently reported as fact (Allenet 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 andthe Sydney Water Inquiry that the laboratory resultsshould not be used to make public health decisions untilremedial actions had been taken by the laboratory and val-idated 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 per-formed at the AWT laboratory, including some samplesthat were concentrated at AWT and subsequentlyprocessed at other laboratories. These included 123 pos-itive controls, 123 negative controls, and 46 training sam-ples processed by new staff. Although positive and neg-ative control sets were normally performed after every


FIGURE 5 CCoommppaarriissoonn ooff AAWWTT aanndd TTWWUULL IIFFAA ccoouunnttss

2,000

1,500

1,000

500

00 500 1,000 1,500 2,000 2,500 3,000 3,500

Cryptosporidium

y = 0.56xR2 = 0.92

AWT Results—oocysts/slide

TW

UL

Res

ult

s—o

ocy

sts/

slid

e

2,000

1,500

1,000

500

00 500 1,000 1,500 2,000 2,500 3,000 4,0003,500

Giardia

y = 0.37xR2 = 0.93

AWT Results—cysts/slide

TW

UL

Res

ult

s—cy

sts

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.


eight experimental samples, sometimes for logistic reasonsfewer than eight client samples were analyzed in the batchbetween controls.

The positive control samples yielded average recover-ies of 59% for Cryptosporidium oocysts (n = 123, stan-dard deviation = 37) and 40% for Giardia cysts (n = 123,standard deviation = 35). Both the method of seed prepa-ration (dilution followed by hemocytometer estimation ofconcentrations) and variations in the quality of watersamples tested (e.g., the raw waters during the crisis hada wide range of turbidity levels) contributed to the rela-tively high variation in calculated recoveries.

During this same period, 8.6% of positive controlsamples gave zero recoveries for either cysts or oocysts.The majority of these failures were attributable to thesudden loss of detectability of Giardia in the seed stockand a delay in obtaining replacement cysts from over-seas. With the 123 negative control samples, 120 yieldednegative results. Of the remainder, two samples containedone Cryptosporidium oocyst and zero Giardia cysts, andone sample contained one Giardia cyst and zero Cryp-tosporidium oocysts. Two of these failed negative controlswere concentrated at AWT and subsequently processed atMacquarie University, making it difficult to assess wherethe contamination may have occurred.

Specialized study 1: limnology of Lake Burragorang. Priorto the crisis, Lake Burragorang was last at full supply inJanuary 1992. Several years of below-average rainfallhad resulted in progressive drawdown of the reservoir to58% of capacity (13 m [43 ft] below the spillway) in July1998. Fecal matter from native and domestic animalswould have accumulated on catchment riverbanks and inthe emerging drawdown zone of the reservoir during thisdrought period (McClellan, 1998).

Two heavy rainfall events occurred in the water sup-ply catchments in August 1998. These rain events causedsewage treatment plants in the catchment to overflowand swept accumulated contamination from the catchmentinto Lake Burragorang (McClellan 1998). The first rainoccurred August 7–9 and resulted in a rise in the stored

water volume from 58 to 83%. The second rain occurredAugust 16–19 and lifted the lake level to 100% capacity.

Various statistics have determined the probability ofthese events recurring. The Annual Recurrence Interval(ARI) for the total August rainfall was 2.5 years, whereasthe compound probability for the rapid two-step filling ofLake Burragorang that occurred in August 1998 had anARI of 30 years. Most important, the ARI for an inflowof 700,000 ML (184,940 mil gal) from the Warragambacatchment into Lake Burragorang, which occurred inAugust 1998, was 4.5 years.

Lake Burragorang was thermally mixed in July 1998(midwinter), and the temperature of the water columnwas a uniform 13oC at the dam wall. Floodwater from thefirst rain entered the reservoir as an underflow and estab-lished a hypolimnetic water mass, with a temperature,pH, alkalinity, turbidity, and color signature distinctlydifferent from the overlying lake water. The “old” lakewater formed the epilimnion and was conveniently demar-cated from the turbid floodwater by the 12oC isotherm.The position of this isotherm and, therefore, the positionof the floodwater were resolved within ±1.5 m (4.9 ft) bythe thermistors at the dam wall (Figure 6).

According to 1 m (3.3 ft) interval vertical profiles oftemperature, pH, and turbidity taken at locations through-out the reservoir during the flood period, there was onlylimited mixing between the “new” floodwater and “old”lake water. Most important, floodwater from each raintraveled through the reservoir to the dam wall withinseven days. Subsequent numerical modeling simulationsof the inflows (as described by Antenucci, 2001) sup-ported the field observations of limited mixing and seven-day travel time.

The first floodwater mass arrived at the dam wallbetween August 14 and August 16. This coincided withrenewed detection of cysts and oocysts at the Prospectfiltration plant (Table 2). The second floodwater masswas slightly warmer than the original and traversed thelake as a new layer between the “old” lake water andthe first floodwater. This second flood arrived at the dam


Number Cryptosporidium Crypto- Crypto- Crypto- Giardiaof percent sporidium sporidium sporidium percent Giardia Giardia Giardia

Location Samples positive† maximum minimum mean positive† maximum minimum mean

Storage lakes 143 21 8,600 0 186 18 1,822 0 66

Bulk supply 200 33 12,080 0 338 25 7,620 0 143

Treatment plants 343 18 9,445 0 60 17 3,500 0 30

Water filtration 36 28 2,687 0 304 14 285 0 13plant backwash

Reticulation 139 11 273 0 5 10 109 0 2system

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

TABLE 3 SSuummmmaarryy ooff 886611 ssaammpplleess pprroocceesssseedd dduurriinngg tthhee ccrriissiiss ((JJuullyy 2211 ttoo SSeepptt.. 1188,, 11999988))**


wall around August 23 as strong westerly winds alignedwith the main axis of the lake and established a largeamplitude internal wave. The subsequent wave activitycaused periodic upwelling of floodwater that reached theofftake (Figure 6).

The effect of internal wave activity on the water qual-ity of the Prospect supply was evident as rapid changes inthe water temperature at the offtake (Figures 6 and 7)and in turbidity (Figure 7) and cyst and oocyst concen-trations (Figure 8) in the plant pipeline. These indicatorsconfirmed a pattern of episodic abstraction of turbid, cyst-and oocyst-contaminated water from Lake Burragoranginto the raw water supply pipeline from August 17 onward.

The turbidity of the WFP supply from Lake Burrago-rang was normally <2 ntu. Significant turbidity spikes (>5ntu) were initially detected in the pipeline on August 22(Figure 7). This incident coincided with the second BWA(Table 2). Over the following days, pipeline turbidity lev-els periodically exceeded 15 ntu and exceptionally highoocyst concentrations (up to 10,000/100 L) were detectedin the Prospect supply (Figure 8). These contaminantswere probably associated with the second floodwater massthat arrived at the dam wall around August 22.

The abstraction depth for the plant water supply fromLake Burragorang can be selected anywhere between 0 and55 m (0 and 180 ft) below the full supply line. Typicallywater is drawn from two separate offtakes at any time.The offtake position during the crisis is shown in Figure6 as the height of the offtake aperture in the wall of thedam. This simple approximation of the “abstraction

zone” underestimates the depth belowthe offtake from which water can bedrawn. During the crisis, thermal strat-ification was weak, and the abstrac-tion zone would probably haveextended substantially above andbelow the offtake aperture.

Immediately after the first BWA,the deepest offtake was closed, andthe supply was drawn exclusively fromthe shallowest offtake (4–11 m [13–36ft] below the surface). The 13 m (43 ft)rise in lake level following the floodsand the internal waves meant that thefloodwater moved into the proximityof this offtake (Figure 6). The rela-tionship between internal wave activ-ity, offtake position, and occurrenceof cysts and oocysts in the water sup-ply was not immediately recognized. Inhindsight, the episodic contaminationof the Prospect water supply by turbidfloodwater can be seen as a series ofmore than 10 turbidity peaks, mea-sured by online turbidity meters in thepipeline during the crisis (Figure 7).

Concerted management of the offtake level to avoidfloodwater began on August 31 when the offtake wasraised. The offtake was raised progressively to within 4m (13 ft) of the lake surface (Figure 6). The periodicingress of turbidity and cysts and oocysts to the plantsupply continued during September (Figures 7 and 8).The large internal waves (45 m [148 ft] amplitude)observed during this period probably caused turbulentmixing at the dam wall and mixed floodwaters into theepilimnion. Starting Sept. 11, 1998, cyst and oocyst con-centrations in floodwater collected from Lake Burragorangand the plant pipeline declined rapidly. This was proba-bly attributable to the sedimentation of cysts and oocystswithin the lake. Subsequent monitoring of inflow eventsin Lake Burragorang showed Cryptosporidium oocystsand Giardia cysts sedimented at 5–10 m/d (16–33 ft/d)(Hawkins et al, 2000).

Specialized study 2: input–output analysis of oocysts. Thisanalysis was carried out to determine whether there wasa net accumulation or release of Cryptosporidium oocystsin the system over the 10-week period of intensive mon-itoring. It also provided an overall assessment of the inter-nal consistency of the oocyst data set measured at theentry and exit of the Prospect filtration plant and at theextremity of the distribution system, rather than focusingon the accuracy of individual counts.

The cumulative levels of Cryptosporidium measuredin each of these three locations during the crisis are shownin Figure 9. The outgoing load was a substantial propor-tion of the incoming load, over a period similar to the


FIGURE 6 PPlloott ooff iinntteerrnnaall wwaavvee aaccttiivviittyy aatt tthhee ddaamm wwaallll iinn LLaakkee BBuurrrraaggoorraanngg dduurriinnggtthhee ccrriissiiss ppeerriioodd

12

13

14

15

12

13

14

15

01 05 09 13 17 21 25 29 02 06 10 14 18 22 26 30August 1998 September 1998

Tem

per

atu

re a

t War

rag

amb

a D

am W

all—

oC

Time—days

Dep

th F

rom

Fu

ll S

up

ply

Lev

el—

m (

ft)

60 (197)55 (180)50 (164)45 (148)40 (131)35 (115)30 (98)25 (82)20 (66)15 (49)10 (33)5 (16)0 (0)

5 (16)

Internal wave activity was plotted using high-frequency thermistor data. Cold turbid floodwater contaminated with Cryptosporidium and Giardia and with temperature <12˚C (black) flowed under the existing Lake Burragorang water and created thermal stratification in the lake. (Water temperature scale is indicated by the bar on the right of the figure.) The floodwater arrived at the dam wall between Aug. 9 and Aug. 13, 1998. An internal wave was established on August 20, and the wave activity caused oscillations in the depth of the floodwater at the offtake. The operating offtake (white boxes) were progressively raised during the crisis. When 12˚C water crossed the operating offtake, turbidity increased, and Cryptosporidium and Giardia were detected in the raw water supply. The "closed" offtakes are shown in gray.


known transit time of water throughthis system, i.e., 12–17 days. Figure 9also shows there was a clear and con-sistent relationship between the cumu-lative load curves in terms of the tim-ing of rises and the known times ofenvironmental disturbance (e.g., inflows,internal waves) in the main supply reser-voir (Lake Burragorang); from this, itcan be concluded that random contam-ination was not a significant influence onthe data set as a whole. Furthermore,there was also consistency in the sizes ofincreases in cumulative load over par-ticular time periods. These increaseswere largest in the raw water, interme-diate in the treated water, and smallestin the distribution system extremities.This pattern does not support thehypothesis of significant random con-tamination in the analytical procedure.

DISCUSSIONOther information available to decision-makers. At the

onset of the crisis (the period before the first BWA), someirregularities in the processes of the Prospect filtrationplant had taken place during a routine maintenance pro-gram in mid-July 1998. The Sydney Water Inquiry(McClellan, 1998) noted that these irregularities includedthe release of sediment from the inlet chamber duringflow surges, the loss of dilution water (potentially reduc-ing the effectiveness of the coagulation process), prob-lems with the backwash procedure, problems associatedwith the use of a bypass channel during the cleaning of theclear-water tanks, the maximizing of filter runs associ-ated with lowering of chemical dosing, and suggestions offluctuations in pH that might adversely affect coagulationand flocculation processes.

Laboratory results rapidly verified and additional infor-mation provided. When AWT reported high levels of cystsand oocysts in repeat samples collected after initial detec-tions during the routine surveillance, James Smith wasasked to independently inspect a set of positive slidesprovided by AWT. He insisted he be given no knowledgeof their status or origin. Smith confirmed the AWT IFAcounts and provided additional confirmation using DAPI.He also identified other material in the samples indicativeof surface water contamination. These samples containedno detectable total or fecal coliforms. SWC did not findany evidence of cross-connections, negative pressureevents, or broken pipes that may have allowed surfacewater ingress in the Sydney central business district.

The FISH test supplied by Macquarie University atthe request of the Sydney Water Inquiry was not fullyvalidated. However, supportive evidence that C. parvumoocysts were present in the samples was provided when

more than half the samples tested showed FISH-positiveoocysts. It would be unlikely that 100% of IFA oocystswould be positive using FISH because IFA is a genus-spe-cific test and FISH is designed to be species-specific. Inaddition, the RNA in many environmental oocysts wouldhave degraded. Published data available at the timeshowed a high correlation between in vitro excystation anda positive FISH result (Vesey et al, 1998). Therefore,repeated positive FISH results suggested that a proportionof the C. parvum oocysts detected were likely to be viable.The strong correspondence between the AWT counts onthe slides and the later rereading at TWUL provided fur-ther support for the validity of the AWT analyses.

Insights from understanding of limnology of Lake Bur-ragorang. The physical, chemical, and microbiologicalmonitoring of the catchment and reservoir waters clearlyshowed that floodwater transported the cysts and oocystsduring the second and third BWA events. Floodwater tra-versed the reservoir and reached the offtake within sevendays. An exceptionally large internal wave facilitatedfloodwater contamination of the Prospect plant supply andrendered the reservoir ineffective as a barrier against thetransport of contaminants into the reticulated supply.

Input–output analysis supported laboratory analyses.Additional assurance that cysts and oocysts were presentin the water supply was independently derived from theconsistency in the timing of the rise of the cumulativecurves (Figure 9) and the progressive reduction in cumu-lative oocyst levels observed as water passed from thesource through the filtration plant and to the extremi-ties of the distribution system.

Laboratory support for system management during the cri-sis. Within a week of the start of the Sydney water crisis,all credible service providers in Australia were supplyingcyst and oocyst analyses at their maximum capacity as


FIGURE 7 PPlloott ooff tteemmppeerraattuurree ooff WWaarrrraaggaammbbaa ooffffttaakkee wwaatteerr vveerrssuuss ttuurrbbiiddiittyyooff tthhee PPrroossppeecctt wwaatteerr ssuuppppllyy dduurriinngg tthhee sseeccoonndd BBWWAA

0

5

10

15

20

Water temperature

Turbidity

7/19/98 7/26/98 8/2/98 8/9/98 8/16/98 8/23/98 8/30/98 9/6/98 9/13/98 9/20/98 9/27/9811

12

13

14

15

16

Wat

er T

emp

erat

ure

—oC

Turb

idit

y—n

tu

BWA—boil-water advisory. Temperature was measured at the dam wall by a thermistor chain. Turbidity was measured by online meters in the pipeline from the dam to the Prospect Water Filtration Plant. The timing of the turbidity data was adjusted to compensate for transit through the pipeline. Breaks in the temperature trace in September coincide with the raising of the offtake level. Turbidity spikes coincide with falls in the offtake water temperature as colder floodwater moved into the abstraction zone.


water utilities throughout the country decided to testtheir own supplies. It was not until late in the crisis thatthe associations between the internal wave behavior,pipeline turbidity, and cyst and oocyst counts in rawwaters were determined and provided the only useful pre-dictor of the presence of cysts and oocysts in the water sup-ply. Therefore, in the absence of surrogates and lacking suf-ficient capacity for rapid cyst and oocyst analysis elsewherein Australia, SWC decided that for operational purposeswhile the crisis lasted, it required a number of prioritizedanalyses that were beyond the AWT laboratory’s statedcapacity. This decision was made despite advice from thelaboratory that such action could adversely affect dataquality (McClellan, 1998). AWT scaled up its cyst andoocyst analysis capacity and reduced turnaround time inorder to deliver results as requested. This allowed SWCto meet its objective of determining which areas of itsdistribution system were affected during each contami-nation episode (Quill, 2000).

Second external audit—IFA versus DAPI versus DIC andthe misidentification of algae. Repeated assertions weremade throughout the crisis and afterward that the highcyst and oocyst counts were attributable to misidentifi-cation of algal cells as cysts or oocysts (Allen et al, 2000;Clancy, 2000; McClellan, 1998). The monoclonal anti-bodies used in the AWT method were both IgG1 isotypesdeveloped specifically for use in water analysis (ratherthan IgM isotypes developed for fecal analysis). Com-pared with the use of less-selective monoclonal antibod-ies, the use of these more-selective monoclonal antibod-ies in combination with FC and IFA narrowed the rangeof nontarget organisms that might be confused with cystsor oocysts (Ferrari et al, 1999).

DIC microscopy or DAPI staining can provide furtherinformation to assist the identification of oocysts (Clancyet al, 1999; Grimason et al, 1994). Neither method wasin regular use in the AWT laboratory at the commence-ment of the crisis. However, a number of environmental

samples collected during each stage ofthe crisis in which AWT had detectedoocysts by IFA were submitted to otherlaboratories that used both techniques.After DAPI staining, presumptiveoocysts with the nuclear staining typi-cal of Cryptosporidium oocysts andGiardia cysts were found in many ofthese samples, providing confirmationthat both Cryptosporidium and Giardiawere present.

On Sunday, August 30, at therequest of the Sydney Water Inquiry,five experts from the United Kingdomand Sydney together examined slidesreported by AWT as positive for cystsand oocysts. This group submitted tothe inquiry that there was “unequivo-

cal presence of Cryptosporidium and Giardia in Sydney’streated water at concentrations that are of public healthconcern.” This finding was based on the slides the grouphad seen with all the analytical techniques available,including DAPI staining, DIC, and phase contrastmicroscopy (McClellan, 1998). It is improbable that twoalgae (one mimicking oocysts and the other mimickingcysts) both of which cross-react with the antibodies used,would repeatedly be synchronously present, then absent,in the raw and finished waters in the dramatic fashion seenover the 70 days of the crisis.

Throughout the crisis, the AWT laboratory regularlydetected cysts and oocysts in floodwater collected fromcatchment streams, the underflow within Lake Burrago-rang, and the Prospect supply, when it was contaminatedby floodwater. These stormwater samples were the mostturbid raw water samples provided to the laboratory, butthey contained few algal cells because they arose fromthe catchment runoff. The water samples that did containan abundance of native algal cells and were most likely toproduce misidentification because of algae were from theepilimnion of Lake Burragorang or from the Prospectplant supply when it was not contaminated by floodwa-ter. High concentrations of cysts or oocysts were neverdetected in these samples (see Figure 8).

QC issues. The second external auditor misinterpreteda training criterion specifying 50% recovery in positivecontrol samples as the laboratory routine performancecriterion on spiked sample matrixes. Such a high recov-ery criterion for routine samples is unrealistic given cur-rent methods. Prior to and during the crisis, AWT had nowritten acceptance criteria for recovery in positive controlsamples for cyst and oocyst analyses. All results werereviewed and assessed by laboratory management as theywere obtained. Often, however, no record was kept ofresponses to QC failures. At least two of the examplescited by the second external auditor (#98071497 and#98071498) as laboratory QC failures were performed by


FIGURE 8 CCoorrrreellaattiioonn bbeettwweeeenn ccyysstt//ooooccyysstt lleevveellss aanndd ttuurrbbiiddiittyy iinn tthhee ssuuppppllyyttoo tthhee PPrroossppeecctt ppllaanntt

0

5

10

15

20

0

1

10

100

1,000

10,000

100,000

7/19/9

8

7/26/9

8

8/2/98

8/9/98

8/16/9

8

8/23/9

8

8/30/9

8

9/6/98

9/13/9

8

9/20/9

8

9/27/9

8

Turb

idit

y—n

tu

Cys

ts a

nd

oo

cyst

s/10

0 L

Cysts and oocysts were measured in grab samples from the supply. Turbidity was measured by an online meter in the pipeline.

CryptosporidiumGiardia

Inlet


new staff as training seed samples, not client samples.Twelve additional personnel were hired and trained dur-ing the crisis. During the training phase, the new staffmembers often failed one of their three required seededtraining matrixes. As a result, they were required to repeattraining until they performed satisfactorily.

The three (2.4%) low-level positive results in nega-tive control samples did not trigger corrective action dur-ing the crisis. These levels of cross-contamination wereinsignificant compared with the positive sample readingsin the same batch (often hundreds to thousands per 100L) and were never repeated with consecutive control sets.In addition, the results were being used to aid managementof a water supply system, rather than to make publichealth decisions based on each test result.

Interpreting results from external labs used to validateAWT. Because of space limitations, it is not possible toreview here the results from samples sent to external lab-oratories during the crisis. These laboratories all useddifferent methods for analysis of the water samples orconcentrates sent to them. The environmental sampleswere rarely subsampled, stored, and shipped togetherwith appropriate seeded controls to allow direct com-parisons of recoveries or counts. At least one laboratory(Clancy Environmental Associates) noted that some sam-ples arrived in poor condition and were processed onlybecause of the urgent need of the Sydney Water Inquiry.The inquiry weighed all the results from water samples andconcentrates from local and overseas laboratories. Over-all, it found too many confounding influences affecting theresults from these samples to draw strong conclusions. Theinquiry did conclude that the samples analyzed by over-seas laboratories could not be used to prove the originalIFA counts were overestimates (McClellan, 1998).

When IFA counts by AWT were compared with DAPIcounts from other laboratories, significant differenceswere found. This was hardly surprising because only a sub-set of IFA-positive environmental cysts and oocysts wouldbe expected to show clear DAPI-stained nuclei. Addi-tionally, DAPI staining was frequently performed afterdays or weeks of storage. It also required removal of thecover slip, removal of excess DAPI after incubation, andalteration of buffering conditions. An example of thelarge variation between an AWT IFA count and a subse-quent TWUL DAPI count (sample #98068786) was citedby Clancy (2000).

AWT slide counts assessed as cross-contamination. Partof the explanation proposed by Clancy (2000) for thecrisis was that the cysts and oocysts identified by exter-nal experts were Cryptosporidium and Giardia seed stockthat contaminated water samples in the AWT laboratorybecause of poor analytical practice. Given that the crisislasted 70 days and encompassed more than 1,200 analy-ses, two pieces of analytical evidence suggest that thiscould not have happened on the scale necessary to affectthe credibility of the counts. First, only 3 of the 123 neg-

ative control samples were contaminated, and these werecontaminated with single cysts or oocysts. Second, if seedswere carried over from the positive control to environ-mental samples in the same batch because of poor clean-ing of equipment, it would constitute a dilution process.No batch runs showed a dilution pattern consistent withcross-contamination. High readings did not occur after thepositive control and were usually interspersed between lowor nondetect readings.

Therefore, the cross-contamination theory is incon-sistent with the other available water distribution systeminformation. It requires that samples were contaminatedin order that the balance of cysts and oocysts into and outof the system matched when tallied at the conclusion ofthe crisis. It also requires that the contamination beepisodic and that each contamination episode correspondwith a floodwater incursion into the Prospect supply in thecomplex pattern described previously. In fact, the AWTlaboratory results were consistent with the limnologicalevents in Lake Burragorang following the flood. Highcyst and oocyst concentrations were found in samplescollected from floodwater, low concentrations were foundin lake water with a long residence time, and high con-centrations detected in finished waters during the secondand third BWAs coincided with entry of floodwater intothe Prospect supply.

Issuing BWAs. Three BWAs were issued during the cri-sis. Following the initial detection (and confirmation) oflarge numbers of cysts and oocysts of unknown infectiv-


8

7

6

5

4

3

2

1

0

Cry

pto

spo

rid

ium

(C

um

ula

tive

To

tal)

—✕

1010

7/22/9

8

7/29/9

8

8/5/98

8/12/9

8

8/19/9

8

8/26/9

8

9/2/98

9/9/98

9/16/9

8

9/23/9

8

The cumulative load in treated water is initially higher than that of the source water because sampling of the latter started later, after most of the initial rise in treated water.

Raw waterTreated waterDistribution system water

FIGURE 9 CCuummuullaattiivvee llooaadd ooff ooooccyyssttss mmeeaassuurreedd eenntteerriinnggaanndd lleeaavviinngg tthhee PPrroossppeecctt WWaatteerr FFiillttrraattiioonn PPllaannttaanndd eexxiittiinngg tthhee ddiissttrriibbuuttiioonn ssyysstteemm


ity in the distribution system, the potential for suboptimalwater treatment and the implications of the floods in thecatchment were also recognized. This set of circumstancesfulfilled the Centers for Disease Control guidelines forissuing BWAs (CDC, 1997). The Sydney Water Inquiryconcluded that the BWAs were justified (McClellan, 1998).The nature of cyst and oocyst testing at the time meantthat detailed information on the health implications of thecontamination could not be concurrently supplied. Withup to 4 million consumers exposed to the contaminatedwater and experts debating the significance of the avail-able data, refusal to issue the BWAs would have consti-tuted a high-risk strategy.

The reasons why an outbreak of cryptosporidiosis didnot occur during the crisis may never be known. Perhapsthe oocysts were not sufficiently viable or infective tohumans. The prompt issuance of BWAs and the episodicnature of the contamination may have prevented con-sumers from ingesting sufficient oocysts to cause disease.

Explanation of the crisis. The causes of the first event inthe 1998 Sydney water crisis will probably never be iden-tified conclusively. As to the causes of the second andthird events, the authors propose the following scenario,which requires a minimum number of assumptions:

• Drought-breaking rains after five dry years resultedin catchment runoff contaminated with cysts and oocysts.The floodwater moved rapidly through the reservoir andreached the offtake with minimal loss of fine suspendedmaterial.

• The timing of the flood during the reservoir’s briefannual mixing produced weak thermal stratification,which enhanced the amplitude of internal waves.

• Delay in raising the offtake level compromised thereservoir barrier.

• The rapid fluctuation in the quality of the rawwater supply reduced filtration efficiency in such a way

that some cysts and oocysts passed through the watertreatment plants.

• Insufficient numbers of viable and/or sufficientlyinfective cysts and/or oocysts were ingested by consumersprior to or during the BWAs to cause a detectable increasein disease rates.

Managing and monitoring barriers to contamination. TheSydney water crisis emphasized the importance of vigilantmanagement of all barriers to the entry of contaminantsinto the water supply system. The introduction of thewater filtration plant at Prospect in 1995 added an effi-cient barrier. However, the new hydraulic route reducedthe detention time for raw water in the system, and therare, but not exceptional, floods delivered an unexpectedchallenge to this filtration barrier.

The crisis highlighted the need to manage the catchmentand storages to improve the effectiveness of the barriersupstream of the filtration facility. Since the crisis, themanagement of the raw water supply has been ceded toa new organization, the Sydney Catchment Authority.The authority has broadly expanded the reservoir mon-itoring network by installing thermistor chains in allmajor storages. In addition, a customized software pack-age6 has been developed to rapidly transmit field dataand graphically display internal wave and inflow behav-ior in real time (Hawkins & Simmons, 2000). An exam-ple of the output from this software is shown in Figure 6.This improved monitoring capability has provided a bet-ter understanding of the dynamics of reservoir stratifi-cation and permits managers to respond more quickly tochanges in the system.

SUMMARYThere is compelling evidence that the 1998 Sydney

water crisis actually involved the contamination of theSydney water supply system with Cryptosporidium and


REFERENCESAllen, M.J.; Clancy, J.L.; & Rice, E.W., 2000. The

Plain, Hard Truth About Pathogen Moni-toring. Jour. AWWA, 92:9:64.

Antenucci, J., 2001. Lake Burragorang Dynam-ics: The Sydney “Boil Water Alerts.”Water, 28:34.

Centers for Disease Control and Prevention(CDC), 1997. Cryptosporidium and Water: APublic Health Handbook. Working Groupon Waterborne Cryptosporidiosis. Atlanta.

Clancy, J.L., 2000. Sydney’s 1998 Water QualityCrisis. Jour. AWWA, 92:3:55.

Clancy, J.L. et al, 2000. New Approaches forIsolation of Cryptosporidium and Giardia.AWWA Res. Fdn. and AWWA, Denver.

Clancy, J.L. et al, 1999. USEPA Method 1622.Jour. AWWA, 91:9:60.

Clancy, J.L.; Gollnitz, W.; & Tabib, Z., 1994. Com-mercial Labs: How Accurate Are They?Jour. AWWA, 86:5:89.

Connell, K. et al, 2000. Building a Better Proto-zoan Data Set. Jour. AWWA, 92:10:30.

Davison, A. et al, 2000. Sources of Cryp-tosporidium and Giardia in Sydney’sWater Supply Catchment. Enviro2000WaterTech, Sydney (Apr. 10–12, 2000).

Davison, A. & Deere, D., 1999. Safety on Tap!Microbiol. Australia, 20:28.

Dufour, A.P. et al, 1999. Criteria for Evaluation ofProposed Protozoan Detection Methods.US Environmental Protection Agency.Ofce. of Res. and Devel. Cincinnati.

Ferrari, B.C. et al, 1999. Comparison of Cryp-tosporidium-specific and Giardia-spe-cific Monoclonal Antibodies for Moni-

toring Water Samples. Water Res.,33:1611.

Ford, T.E. & Colwell, R.R., 1996. A GlobalDecline in Microbiological Safety ofWater: A Call for Action. American Acad.of Microbiol., Washington.

Fricker, C., 1998. Thames Water LaboratoryReport on AWT Cryptosporidium and Giar-dia Tests. Thames Water Utilities Micro-biol. Lab., Reading, Great Britain.

Fricker, C., 1995. Detection of Cryptosporidiumand Giardia in Water. Protozoan Parasitesand Water (W. Betts et al, editors). RoyalSociety of Chemistry, Cambridge, England.

Grimason, A. et al, 1994. Application of DAPIand Immunofluorescence for EnhancedIdentification of Cryptosporidium spp.Oocysts in Water Samples. Water Res.,28:733.


Giardia. A range of independent experts repeatedly con-firmed laboratory identification of cysts and oocysts in avariety of samples for more than two months. After thefirst BWA, there was a temporal correlation between thearrival of contaminated, turbid floodwaters at the Prospectplant water intake and the pulses of cysts and oocystsdetected in the distribution system. An input–outputanalysis performed using the data available at the end ofthe crisis showed that a substantial proportion of theCryptosporidium oocysts identified entering the systemwere identified leaving it.

The AWT laboratory mounted a proactive responseto meet the needs of the water supplier in a crisis. Cyst andoocyst results during the crisis were not used in isolationto make public health decisions. However, these data didstimulate intensive additional monitoring and carefulreview of all other water quality data available. The cystand oocyst measurements were consistently applied suc-cessfully to the management of the water distributionsystem through a period of tumultuous events, providinga sound basis for important public health decisions. Thiswas their primary purpose during the crisis.

Evidence from the crisis and subsequent investigationsof Sydney’s catchments indicates the major risk to con-tamination of the city’s water supply is from high-flowevents in the catchment (Davison et al, 2000; Hawkins etal, 2000). In response to the crisis, Sydney water utilitieshave developed a comprehensive monitoring network toprovide real-time information on the location of flood-waters throughout the system.

ACKNOWLEDGMENTThe authors thank Duncan Veal of Macquarie Uni-

versity and Julie Dinsmor of Sydney Water Corporationfor reviewing the manuscript. Australian Water Services(constructors and operators of the Prospect Water Fil-

tration Plant) provided some of the turbidity data shownin Figures 7 and 8.

ABOUT THE AUTHORS:Peter Cox7 is project director for the Sydney (Australia)Water Corporation (SWC), 51 Hermitage Rd., West

Ryde, NSW 2114, Australia, [email protected]. Heholds BVSc and PhD degrees from theUniversity of Sydney and has morethan a decade of experience with Aus-tralian water technologies, primarilyin developing rapid methods formicrobial detection. Ian Fisher is prin-cipal scientist and George Kastl is

senior chemical engineer at SWC. Veeriah Jegatheesanis a lecturer in water and wastewater engineering atJames Cook University in Townsville, Australia. Mal-colm Warnecke is a lab supervisor, Mark Angles is lab-oratory manager and senior consultant, and Heri Bus-tamante is a principal consultant at SWC. TonyChiffings is manager of innovation, research, anddevelopment for EQS Environmental Systems in Guild-ford, Western Australia. Peter R. Hawkins is a princi-pal consultant at SWC.

FOOTNOTES1Cry104, Becton Dickinson Pty. Ltd., Sydney, Australia2G203, PanBio, Brisbane, Australia3FACSCalibur, Becton Dickinson Pty. Ltd., Sydney, Australia4Ausflow IMS, Macquarie Research Ltd., Sydney, Australia5 Waterborne Inc., New Orleans, La.6ResMan, Australian Water Technologies Pty. Ltd., Sydney, Australia7To whom correspondence should be addressed


If you have a comment about this article, please contactus at [email protected].

Hansen, J. & Ongerth, J.E., 1991. Effects ofTime and Watershed Characteristics onConcentration of CryptosporidiumOocysts in River Water. Applied & Envir.Microbiol., 57:2790.

Hawkins, P. & Simmons, B., 2000. “Real Time”:Reservoir Water Quality Monitoring inthe New Millennium. Proc. 12thIWA–ASPAC Regional Conf. & Exhibition,Changmai, Thailand (Nov. 5–9).

Hawkins, P.R. et al, 2000. Understandingthe Fate of Cryptosporidium and Giardiain Storage Reservoirs: A Legacy ofSydney’s Water ContaminationIncident. Jour. Water Supply: Res. Technol.–AQUA, 49:289.

Hutton, P. et al, 1995. Cryptosporidium and Giar-dia in the Aquatic Environment of Sydney,Australia. Protozoan Parasites and Water

(W. Betts et al, editors). Royal Society ofChemistry, Cambridge, England.

McClellan, P., 1998. Sydney Water Inquiry,Final Rept., Vol. 2. New South WalesPremier’s Department,www.premiers.nsw.gov.au/pubs.htm,(accessed throughout 1999 and 2000).

Nahrstedt, A. & Gimbel, R., 1996. A StatisticalMethod for Determining Reliability of Ana-lytical Results in Detection of Cryptosporid-ium and Giardia in Water. Jour. Water Sup-ply: Res. Technol.–AQUA, 45:101.

NH&MRC & ARMCANZ (National Health andMedical Research Council and Agricul-ture and Resource Management Councilof Australia and New Zealand), 1996.Australian Drinking Water Guidelines,National Water Management Strategy.Canberra, Australia.

Quill, R., 2000. Sydney Water Crisis SpursMore Debate (letter to the editor). Jour.AWWA, 92:11:18.

Smith, H. & Rose, J., 1998. Waterborne Cryp-tosporidiosis Current Status. Parasitol.Today, 14:14.

Vesey, G. et al, 1998. Use of a Ribosomal RNATargeted Oligonucleotide Probe for Fluo-rescent Labeling of Viable Cryptosporid-ium parvum Oocysts. Jour. AppliedMicrobiol., 85:429.

Vesey, G. et al, 1994. Application of Flow Cyto-metric Methods for Routine Detection ofCryptosporidium and Giardia in Water.Cytometry, 16:1.

Vesey, G. et al, 1993. Concentration of Proto-zoan Parasites in Water by Calcium Car-bonate Flocculation. Jour. AppliedBacteriol., 75:82.

Sydney 1998: lessons from a drinking water crisis

Documents