R

1
Evaluating the potential causes of excess GI
illnesses observed in the Payments Distribution
System Michèle Prévost Marie Claude
Besner Pierre Payment Drinking Water
Symoisium Yale Medical School, April 20th 2009

2
DS a significant source of waterborne outbreaks?

• Significant portion of WBO caused by DS
  deficiencies (Craun et al. 2001, 2006)
• chemical and microbial contaminants entering the
  DS
• acute breakdown and/or contamination of DS and
  reservoirs.
• reduction in number of WB outbreaks by other
  sources
• Health impacts of non-acute contamination events
  suggested by two successive epidemiological
  studies by Payment et al.

3
Waterborne illness in USA1981-2002 (Regli, 2007)
2001-02 50 DS
4
Waterborne illness in USAPotential Causes of DS
outbreaks1981-2002 (Regli, 2007)
Is intrusion truly a significant source of
waterborne illness?
5
The Payments study distribution system
380,000 consumers Average daily demand 210,000
m3/d 0-45 m elevation 1,590 km of pipes
Ste-Rose WTP (capacity 110,000 m3/d)
Pont-Viau WTP (capacity 135,000 m3/d)
Chomedey WTP (capacity 180,000 m3/d)
6
1st epidemiological study Payment et al., 1991

• WTP using source water impacted by CSO discharge
• conventional treatment ozonation
• system meeting standards for treatment and
  distribution

WTP3
Same water filtered through RO unit
Tap water
WTP1
307 households
299 households
WTP2
15-month period (1988-1989), estimated annual
incidence of GI illness
0.76
0.50

• 35 of reported GI illnesses
• were tap water-related in

7
2nd epidemiological study Payment et al., 1997

• Same WTP using source water with less discharge
• Conventional treatment ozonation higher
  chlorine residuals
• Meeting new low turbidity and CT standards for
  treatment and distribution
• DS with low residence times (lt20 hours) and no
  reservoirs

• extensive disinfection and DS monitoring
• no relationship btw residence time/fixed and
  suspended biomass/Cl2 residual and illness

8
Impact of the Payments studies

• Authors concluded that the DS appeared partly
  responsible
• 1st study pointed out to older DS areas far from
  plant
• No relationship with residuals, biofilm,
  suspended bacteria or microbial indicators
• Concerns over integrity of DS (cross-connections,
  pipe repairs, loss of pressure
• Contamination during normal operations?
• Preliminary transient pressure modeling of this
  DS Kirmeyer et al (2001)
• System appeared especially prone to low/negative
  pressure
• 90 of model nodes with Plt5 psi (power outage
  affecting pumps at all 3 WTPs)
• NAS Committee report (2006)
• Reviewed the Payments epidemiological studies
• Low disinfectant residuals vulnerability of DS
  to pressure transients

9
Follow up studies on theimpact of DS on GI
illnesses

• Melbourne, Aus Hellard et al. (2001)
• 600 families double blind no effect
• Davenport, US Colford et al. (2005)
• 456 families no significant effect (lt11)
• UK, Hunter et al. (2005)
• Survey 423 subjects 15 related to low P
• not a true epi study design
• Nygard study (2007)
• Cohort study, 800 LP events, 1200 households
• Impact of breaks and repairs on GI illnesses

10
NAS 2006 report sources of risk in DS

• Medium priority
• Biofilm growth
• Loss of disinfectant residual
• Low pressure transients and intrusion
• Highest priority
• Cross-connections and backflow
• New and repaired water mains
• Finished water storage

11
Could this DS be a significant source of
waterborne disease?

• Can regulatory and event based monitoring (TC, FC
  and E. coli ) detect transient contamination
  events?
• Does normal DS operation provide opportunities
  for significant contamination?
• Pathways?
• Events? controlled pump startup and shutdown,
  opening/closing of fire hydrants or valves,
  flushing operations, repairs of pipeline breaks
• Can hydraulic transient modeling provide an
  credible estimate of intrusion volumes and health
  risk?

12
Approach to quantify the impact of the DS on
water quality (1995-2009)

• Monitor WQ in the DS
• Investigate the relationship between positive
  coliform samples and data on distribution system
  operations
• Document DS specific contamination sources and
  events
• Monitor and model pressures at full-scale
  monitor actual intrusion at worst potential sites
• Model the intrusions through leek orifices and
  other entry points and compute resulting
  infections risks using QMRA

• Students J. Baudart, M.C. Besner, G. Ebacher, A.
  Carrière, P. Laurent.

13
Where do indicators (egcoliforms) most likely
come from?

• Treatment breakthrough
• Internal production
• Accumulation and amplification in the biofilm
  deposits
• Entry into the DS
• direct contamination
• water main repair/break
• cross connection / backflow
• Intrusion
• Can we actually observe contamination?

14
Methodological detection limitations

• Are bacteria present in the DS detected by
  culture based methods, especially in the presence
  of chlorine residuals?
• Would direct methods allow for the actual
  estimation of viable TC or E. coli
• Can other indicators be used to identify
  intrusion pathways and serve as indicators?

15
Application of the FISH Enterobacteriaceae Probe
to E. coli Cells




FISH with DVC assay Elongated fluorescent cells
Viable Enterobacteriaceae cells
FISH without DVC assay

Oligonucleotide probe designed in Loge et al.
(1999)
16
Enterobacteriaceae Field Survey using FISH probes

Baudart et al., 2002, 2005
57.4 - 100 VBNC
17
Production and contamination of loose deposits in
DS
Carrière et al., 2005

• 4 Canadian DSs
• Deposit composition (metals, and volatile
  content) and quantity depend on
• the accumulation period
• the treated water characteristics (calcium
  content and aggressivity)
• pipe materials.
• Very few TCs (3 samples out of 258 1.2)
• Total cell numbers reached values as high as 1010
  cells/g of deposit.
• Results do not suggest a high level of
  contamination by coliform bacteria

18
Aerobic spore formers as indicators of intrusion
Cartier et al., 2009
Min and Max Concentrations in different types of
water (DS-A)
10000000
1000000
100000
10000
1000
Spores (CFU/L)
100
10
1
0.1
0.01
19
Distribution of ASF in treated water, trench
water, hydrant water and DS water. (Cartier et
al., 2009)
y
100
80
60
Cumulative frequency ()
40
20
0
0,001
0,01
0,1
1
10
100
1000
Aerobic Endospore Concentrations (CFU/100 mL)
DS
FLUSH
REP
REP selected
B.Q.L. (DS)
B.Q.L. (FLUSH)
B.Q.L. (REP)
20
Conclusion microbial tools

• Higher number of positive Enterobacteriaceae
  detected in treated and distributed water with
  FISH probes
• repression of fluorescence by Cl2
• background fluorescence
• wide application of E. coli probe in EC for
  environmental, recreational and DW applications
• Loose deposits were significantly colonized by
  HPCs but very few TCs and no FC/ E. coli
• ASF are promising indicator of intrusion after
  breaks, repairs, etc.

21
Approach to quantify the impact of the DS on
water quality (1995-2009)

• Monitor WQ in the DS
• Investigate the relationship between positive
  coliform samples and data on distribution system
  operations
• Search for associations btw indicators and DS
  operations
• Indirect evidence of intrusion
• DATA MINING
• Document specific contamination sources and
  events
• Monitor and model pressures at full-scale
  monitor actual intrusion at worst potential sites
• Model the intrusions into the DS through leek
  orifices and other entry points

22
Can coliform occurrences be linked to
operations?Tracking the causes and predicting TC
and E. coli MINING THE DATA!!

• Historical compliance data
• Rare events for TC
• Very rare events for FC/E. coli
• Predictive biofilm based models
• Statistical and mechanistic models
• Limited success
• NEW APPROACHES
• 1) DS activity/WQ GIS analysis
• Includes all activities
• A posteriori analysis - historical data
• 2) Extensive event based monitoring
• Increases probability of documenting
  contaminantion
• Research initiative

23
Causes of historical non-compliance events
Development of a GIS Data Integration Approach
Interactive analyzer tool
Databases

• Hydraulic results integration
• Multi criterion multi table
• database query

Hydraulic model
GIS visualization
24
Data Mining Visualization of the query results
Flushed Zone
Sampling Point
July 8-30
Hydrant operation
Valve operation
Flushing
Pipes
Valve operation, July 29
Hydrant operation, July 21
Valve operation, July 14
Sampling point
Valve operation, July 2
4 Samples with

total coliforms

Hydrant operation, July 30
(July 27 to 30, 1998)
25
Summary Events with High Probability of Origin
of the Coliform Occurrences LAVAL
DS n15
Unexplained
WTP and/or DS n22
- Missing DBs? - inaccurate info? - cause
impossible to identify using this type of
approach?
WTP n8
(n88)
26
Data Mining Identification of probable cause
using the Data Query Tool
27
Conclusion Data MiningGIS Query based model

• The occurrence of 140 positive TC samples was
  investigated in five systems
• impact of OM activities on the occurrence of TCs
  varies from one system to another
• explains minimum of 9 and up to 45.
• 48 HPC events were evaluated in four systems
• 217 customer complaints were investigated in
  three systems.
• 1/3 of the customer complaints could be
  associated with OM work, while its association
  with HPC events was negligible in three systems
  out of four.
• Improvement of prediction with better operations
  records

28
Approach to quantify the impact of the DS on
water quality (1995-2009)

• Monitor WQ in the DS
• Investigate the relationship between positive
  coliform samples and data on distribution system
  operations
• Document DS specific contamination sources and
  events
• Monitor and model pressures at full-scale
  monitor actual intrusion at worst potential sites
  
• Model the intrusions through leek orifices and
  other entry points and compute resulting
  infections risks using QMRA

29
Types of events investigated

• Flushing for maintenance
• Pipe replacement
• Pipe repairs
• Type I Emergency repair
• Major pipe break, loss of pressure
• Type II Emergency repair
• Pipe break with pressure reduction
• Type III Planned repair
• Pipe leak with pressure maintained
• Difficult to sample unplanned transient events as
  waves travel at high velocities (1000 m/s)

30
Worst case of potential contaminationSoil and
water samples from pipe trenches and water
samples during repairs
31
Monitored parameters

• Excavation
• (soil et dirty water)
• E. coli ( total coliforms)
• Clostridium perfringens
• coliphages
• Human enteric viruses
• Aerobic sporeformers

• DS water
• (Houses and flushed hydrant)
• Total coliforms E. coli
• Aerobic sporeformers
• Free Chlorine/T/pH
• Turbidity
• Pressure (1-20 readings / second)

Besner et al., 2007
32
Methodology Planned repairs of water
main leaks
Excavation
Hydrant
House outside
House inside
LEAK
Isolated main
Upstream hydrant
Downstream hydrant
Flushed hydrant
Flow direction
Besner et al., 2007
33
Impact of water main repairs on water quality
coliform analyses/ positive samples
House isolated segment
Houses outside
Flushed hydrant
House SL
Before valve closure (T0)
38/0
When valves are closed
5/0
Before valve opening
20/0
E. coli
During flushing
Total coliforms
279/16 (5.7)
Total coliforms
Total coliforms
ASF
ASF
Prior to hydrant closure
49/0
After hydrant closure
33/1 (3.0)
Total coliforms
Total number of coliform analyses positive
coliform samples
71/1 1.4
189/5 2.6
155/11 7.1
9/0 0
424/17 4.0
Besner et al., 2007
34
Sampling of air valve chambers
35
Microbial contaminantionRecovery of faecal
indicators of contamination and human enteric
viruses
NA
36
Level of microbial contamination
8
8
Coliformes totaux
Spores aérobies
7
7
6
6
5
5
(log CFU/100ml)
Cconc. in positive samples
(log CFU/100ml)
Cconc. in positive samples
4
4
3
3
2
NA
NA
2
1
1
Karim et al. (2003)
Payment (2003)
Payment (2003)
Karim et al. (2003)
0
Cette étude
Cette étude
0
8
8
C. perfringens
E. coli
7
7
6
6
5
5
(log CFU/100ml)
Cconc. in positive samples
(log CFU/100ml)
Cconc. in positive samples
4
4
3
3
2
2
1
1
0
0
Sol
Sol
Sol
Sol
Eau usée
Eau usée
Eau brute
Eau brute
Eau tranchée
Eau tranchée
Eau tranchée
Eau tranchée
Eau CV inondée
Eau CV inondée
37
Level of contamination
8
Coliphages
7
6
5
(log CFU/100ml)
Cconc. in positive samples
(log CFU/100ml)
4
Cconc. in positive samples
3
2
1
0
Sol
Sol
Eau usée
Eau brute
Eau tranchée
Eau tranchée
Eau CV inondée
Levels of faecal contamination similar to those
observed in Des Prairies river water
38
Automatic water sampler (June 2007)

• Located in area susceptible to negative pressures
• Connected to pressure sensor
• Start sampling up to 150L of water following low
  or negative pressure

39
PathwaysContamination of trenches, pipes and
valve chambers

• Event based investigations showed that operations
  (pipe repairs, flushing, fire drills, etc.) were
  not associated with an increased presence of
  indicators of microbial contamination such as
  total coliform bacteria, E. coli, aerobic spores,
  coliphages and other water quality indicators.
• Few indicators of faecal contamination in trench
  water soil indicators present
• Submerged valve chambers may be a significant
  pathway
• higher frequency of faecal contamination
• source runoff and infiltration/sewers some
  positive for animal, one for human Bacteroides
• offer greater potential of significant intrusion
  (volume and quality) 2dia orifice!!

40
Approach to quantify the impact of the DS on
water quality (1995-2009)

• Monitor WQ in the DS
• Investigate the relationship between positive
  coliform samples and data on distribution system
  operations
• Document DS specific contamination sources and
  events
• Monitor and model pressures at full-scale
  monitor actual intrusion at worst potential sites
• Model the intrusions through leek orifices and
  other entry points and compute resulting
  infections risks using QMRA

41
Field investigation of low/negative pressure

• Transient pressures may originate from
• Pump starting / stopping, rapid opening or
  closing of valves, hydrant flushing, etc.
  (Kirmeyer et al., 2001)
• Continuous field pressure monitoring showed the
  occurrence of transient low or negative pressures
  (Kirmeyer et al., 2001 LeChevallier et al.,
  2004 Gullick et al., 2004-2005 Hooper et al.,
  2006)
• It is suspected that a high potential risk of
  introduction of pathogens exists when transient
  pressure events occur (LeChevallier et al., 2003)

42
Methodology

• 2 types of pressure monitoring
• Targeted monitoring during
• Planned repairs of water mains
• DS flushing
• Hydrant flow tests
• Pump operation
• Continuous monitoring
• 12 DS sites for 80 days
• 18 DS sites during transmission main closure

43
Transient pressures due to valve/hydrant
operation during repairs of water mains

• 10 planned repairs of pipe leaks (6-8 inches)
• 11 sites monitored (1 site with 2 loggers)

44
Example Power failure at the WTP on Aug 22, 2006
Resulted in 4 DS sites with negative P (1min)
Site 2 (2min)
Site 12 (2min)
0psi
Site 11 (1min45s)
Site 10 (1min45s)
45
Transmission main closure during period of
monitoring
Normal operation (DS area supplied by WTP A and B)
Repair conditions (DS area supplied by WTP B
only)
46
Water quality sampling during pipe closureat
planned pipe replacement sites

• At 3 DS sites (2 hydrants, 1 routine sampling
  point)
• Turbidity on line
• 33 litres of water analyzed for the presence of
  total coliforms and E. coli
• Only 1 positive sample TC collected 24 h after
  the end of the repair
• At worst point, 3 litres of water collected
  directly after loss of pressure (6 min and 7 min)
  
• No TC
• Free chlorine residual (0.18 - 0.20 mg/L)

47
Approach to quantify the impact of the DS on
water quality (1995-2009)

• Monitor WQ in the DS
• Investigate the relationship between positive
  coliform samples and data on distribution system
  operations
• Document DS specific contamination sources and
  events
• Monitor and model pressures at full-scale
  monitor actual intrusion at worst potential sites
• Model the intrusions through leek orifices and
  other entry points and compute resulting
  infections risks using QMRA

48
Simulation of 3 pump shutdowns at the WTP
Modeled pump shutdown corresponding to measured
pressure profile
9-12 sensors (Telog)
P en sortie dusine
49
Simulation result without cavitation head
P lt -14.35 psi Physically impossible!
Aug 22nd, c100, CH-1000, AVno, t300,
PSLT100
50
Modification of wave speed in cast iron mains
(30 of original value)
22 août 2006

• Better fit in some areas
• Pmin better
• Less nodes in neg P

50
51
Number of nodes experiencing negative pressure
Aug 22nd, c100, CH-1000, AVno, t300,
PSLT100
Typical wave speed in cast-iron (1060 m/s) (2163
nodes)
Aug 22, c30, CH-1000, AVno, t300, PSLT100
30 of typical wave speed in cast-iron 318 m/s
(504 nodes)
52
How can this fit be improved

• Modeling uncertainties
• Actual inflow equations
• Local dissipation
• Wave speed in pipes
• Demands
• Distribution of demands
• Air content in water
• Air-vacuum valves and orifice diameters
• Level of skelettonisation

53
Estimation of intrusion volumes

• A. Intrusion through pipe leaks

54
Estimation of intrusion volumes Aug 22nd event
Worst-case scenario

• Intrusion through submerged air-vacuum valves
• Assumptions
• - 45 flooded vaults
• Orifice diameter 50 mm
• (2)

Intrusion through pipe leaks Assumptions External
head 1.4 m leakage 20
Kirmeyer et al., 2001
55
Intrusion volumes through submerged air-vacuum
valves (Aug 22nd)
Number of vaults where intrusion takes place
(Total 45 flooded vaults)
2000
16
1800
15
1600
1400
15
1200
Intrusion volume (L)
1000
800
8
600
400
3
200
0
0
0.2
0.4
0.6
0.8
1
1.2
of original wave speed in cast-iron
CH-1000, AVno, t300, PSLT100
56
Intrusion volumes through pipe leaks (Aug 22nd)
CH-1000, AVno, t300, PSLT100
Avg of 100mL/node in negative P
57
Conclusions

• The transient model overestimates negative
  pressures
• Could be used to identify zones of the system
  susceptible to negative pressures BUT
• Less adequate for estimation of intrusion volumes
• Directly linked with duration and magnitude of
  negative pressure
• Field pressure data are necessary to validate the
  results from the model
• At this time, we should be careful about using
  the intrusion volumes obtained as input in a risk
  analysis model

58
Risk modelling - QMRA

• Evaluate the risk and fraction of exposed
  population in the DS

At each node
Weighted risk for exposed population
59

• What about the Payment studies?

60
Critical review of the vulnerability of the
Payments DS to low/negative pressure (2007-09)

• Georeference cases in the Payment study
• look at general trends
• Apply SOA transient modelling
• extended period simulation using InfoSurgeTM
• analyze specific events of low pressure
• Compare field results with model results
• Adjust model parameters to better describe actual
  events
• Estimate intrusion volume and health risk

61
Georeferenced distribution of illness and
negative pressures in the Payment DS.

• Participants spread over whole DS
• Spatial cluster of negative pressures (higher
  elevation)
• 1st study suggested more GI cases in 2 sub-areas
  (but low n)
• AND NO spatial distribution of GI cases in 2nd
  study

62
Are negative pressures really common in the
Payment DS?

• Yes but not as pronounced as with the model
  output
• When attempting to calibrate a transient model
  with field data, it is most important to keep in
  mind that the system is underdetermined.
• Evaluation on intrusion volumes must be done with
  great caution until technical pitfalls are
  resolved
• On going work
• Actual measurement of intrusion in valve chambers
• Further modelling improvements to determine the
  range of values for model parameters to estimate
  intrusion volumes.

63
Conclusion

• Intrusion in DS is a hot topic
• Concern over intrusion is the result of the
  Payments epi studies
• The Payments DS is prone to negative pressures
  in elevated areas
• Extended event monitoring did not reveal
  significant pathways and moderate levels of
  contamination
• Transient modelling of the system overestimates
  low pressures but revealed that air valves may be
  a critical control point
• Still searching for supporting evidence

64
Intrusion, is it a concept? Reality of ageing
infrastructure
Winnipeg 1,9 car/year!!!
65
www.polymtl.ca/chaireeau
66
DS research On going work 2008-09

• Finalize the analysis of low pressure events
• Compare simulation models
• Expand characterization of valve chamber
  contamination field meaurements 2009
• Conduct simulations of intrusion volume once
  modelling is adjusted
• Calculate risk estimates for intrusion with and
  w/out chlorine residual using QMRA
• New collaboration with USEPA on overall risk
  evaluation (Dr. Besner)

67
Methodology

• High speed pressure transient data loggers
• Radcom RDL 1071L/3 model
• Up to 20 pressure readings per second, (-15 to
  225 psi)
• Installation on transmission mains, washstand
  taps inside buildings, fire hydrants
• Telog HPR-31 model
• Up to 4 pressure readings per second (-15 to 200
  psi)
• Installation on fire hydrants

68
Results of continuous monitoring Distribution
system pressures
Pressures lt 20 psi recorded at 11 DS sites out of
12 Negative pressures recorded at 7 DS sites out
of 12
Negative pressure events 4 at power failure at
WTP 4 at repair of a transmission main 2 at
repair of isolated mains
898 days of logger data collected by 12 loggers
69
Difference between modeled and measured pressures
spatial representation
Aug 22nd, c100, CH-1000, AVno, t300,
PSLT100
70
Modeling of transient pressure events
Hydraulic simulations
Picture of the system under a specific condition
(max summer demand)
Operation of the system 24-hour simulations
Impact of sudden changes in flow
conditions Few-minute simulations
71
Greater energy dissipation was observed in the
real distribution system

• Skeletonization of the model reduces the number
  of nodes and pipes, thereby decreasing energy
  losses by reflections and friction
• Uncertainty regarding the initial distribution of
  flows and pressures
• Demand, and leaks, are important dissipative
  mechanisms
• The wave speed depends on the pipe diameter, but
  it is also a function of the pipe material, wall
  thickness, and restraint, and of the fluid
  density, elasticity, temperature, and air and
  solids content. Many of these are unknown.
• Presence of air in pipes

72
Simulation results for pressure in DS with
cavitation (Aug 22nd event)
Cavitation head -10.1 m (-14.35 psi)
c100, CH-10.1, AVno, t300, PSLT100