Post on 14-Dec-2015
Rafael Bastos (1), Paula Bevilacqua (2), Richard Gelting (2), Demétrius Viana (1) , João Pimenta (1)
(1) University of Viçosa, Brazil
(2) US Center for Disease Control and Prevention
Seasonal fluctuations in source Seasonal fluctuations in source water quality and health related water quality and health related
risks. risks. A QMRA approach applied to Water A QMRA approach applied to Water
Safety Plans.Safety Plans.
Water Safety Conference 2010
Water Safety Plans
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Davison et al. (2006)
Introduction
Frequency/likelihood consequence/severity
matrix
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Deere et al. (2006)
WSPWSP
Hazards / hazardous events identification
/ prioritization
Risk characterization
Control measures
Qualitative / semi-quantitative approach
subjective judgement
High risk
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
WSP QMRAWSP QMRA
Quantitative Microbial Risk Assessment (QMRA)
Exposure model + Dose-response model
Risk estimates
Objective / quantitative input for risk assessment and management in WSP
(Smeets et al., 2010; Medema & Ashbolt 2006)
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Quantitative Microbial Risk Assessment
Hazardous events Seasonal fluctuations in source water
quality (rainfall)Water treatment performance
Risk estimatesLong and shorter-
terms Annual, seasonal,
daily
Objectives
“provide opportunities for improved risk management, with an incentive to reduce the occurrence and impact of event-driven peaks” (Signor & Ashbot , 2009).
10-4 pppy
10-6 pppd
Methods
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
≈ 650 m; 20º 45' 14" S; 42º 52' 53" W
≈ 70,000 inhabitants (90% urban)
10º C (winter) - 30º C (summer)
rainy season (November – March); dry season (April - October)
Viçosa – Minas Gerais
(Southeast Brazil)
UFV (1926)
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
DW supply system
WSP
WTP 1 (100 L/s)
WTP UFV(50 L/s)
WTP 2(100 L/s)
São Bartolomeu Stream
Turvo River
Rainy season
70% SB + 30% TR
Dry season
70% TR + 30 SB
150 km
Viçosa DW water supply systemViçosa DW water supply system
Lagoa 1
Lagoa 2
WTP UFV(50 L/s)
Storage reservoir
WTP 1 (100 L/s)
8 km
Storage reservoir
UFV DW water supply systemUFV DW water supply system
Rainy season ≈ 200L/s
Dry season ≈ 100 L/s
SB Catchment
(≈ 2000 ha)
Conventional treatment
UFV WTPUFV WTP
coagulation (aluminum sulphate), hydraulic
rapid mixture and flocculation, conventional
sedimentation, rapid sand filtration, and
disinfection with chlorine. Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
QMRA model
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
d = dose
C =Cryptosporidium concentration in source water (oocysts /L)
r = recovery fraction of the oocysts enumeration method (%)
R = oocysts removal efficiency (log) (filtration)
V = volume of water consumed per day (L/d)
Exposure model
d = C x (1/r) x R x V
QMRA model
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Dose – response model (exponential) (Haas et al. , 1999)pd = 1 - exp (-θd) (daily)
pΣ = 1- (1- pd)n [seasonal: prain and pdry; and annual)
risk of infection (pd) in an individual following
ingestion of a single pathogen dose d, i.e. per exposure event (daily risk)
total probability of infection over n exposures to the single pathogen dose d
Methods – Results
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Cryptosporidium concentration in source water (oocysts /L)
PDF : β distribution
Monitoring (five previous studies, 2002-2008)
r = recovery fraction of the oocysts enumeration method (%)
30-60% (uniform distribution)
Methods – Results
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
R = oocysts removal efficiency (log) (filtration)
log10 removal Cryptosporidium oocysts = 0.9631 log10
removal turbidity + 1.009(Nieminsky & Ongerth, 1990)
turbidity removal 0.29 to 3.79 logRdry = 0.29 to 2.72 log - Rrain = 0.5 to 3.8 log
Oocysts removal 1.38 to 4.76 logRdry = 1.38 to 3.72 log - Rrain = 1.58 to 4.76 log
Triangular distribution ≈ pilot experiment s
Methods – Results
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
θ = 0.042 ± 25% - variation in susceptibility (as most existing dose-response models derive from oral challenge data from healthy adult volunteers)
Uniform distributionV = volume of water consumed per day (L/d)
Poisson (λ=0.87 L/day) (Australian)
Methods – Results
Stochastic modelling –
Monte Carlo Simulation
50,000 iterations
Variability and Uncertainty
Methods – Results
0,0
00
00
,28
33
0,5
66
50
,84
98
1,1
33
01
,41
63
1,6
99
51
,98
28
2,2
66
0Valores em Milésimos
0,0%
14,3%
28,6%
42,9%
57,1%
71,4%
85,7%
100,0%
highly skewed risk probability distributions
typical of long-term variability in which the overall mean value is highly sensitive to the rarely occurring but relatively ‘extreme’ higher risk periods
Results – risk estimates (pooled data)
0,0
00
00
,28
33
0,5
66
50
,84
98
1,1
33
01
,41
63
1,6
99
51
,98
28
2,2
66
0
Valores em Milésimos
0,0%
14,3%
28,6%
42,9%
57,1%
71,4%
85,7%
100,0%
0,0
00
00
0,0
70
38
0,1
40
75
0,2
11
13
0,2
81
50
0,3
51
88
0,4
22
25
0,4
92
63
0,5
63
00
0,0%
14,3%
28,6%
42,9%
57,1%
71,4%
85,7%
100,0%
5.6x10-4 1.1x10-3 1.6x10-3 2.2x10-30
0 7x10-2 2.1x10-1 3.5x10-1 5.6x10-14.2x10-1
Pdaily
Pannual
50% = 2 x 10-6 (Signor & Ashbolt, 2009)
95% = 2.2 x 10-3
50% = 6.9 x 10-4
(EPA)
95% = 5.6 x 10-1
Results – risk estimates (dry season)
Pdaily
50% = 4.6 x 10-6
95% = 2 x 10-3
50% = 8.3 x 10-4
(EPA)
95% = 3.1 x 10-1
0,0
00
00
,33
45
0,6
69
01
,00
35
1,3
38
01
,67
25
2,0
07
0
Valores em Milésimos
0
2000
4000
6000
8000
10000
12000
14000
0,0%
14,3%
28,6%
42,9%
57,1%
71,4%
85,7%
100,0%
Pdaily
0,0
00
00
0,0
61
24
0,1
22
48
0,1
83
72
0,2
44
96
0,3
06
20
0
5
10
15
20
25
30
35
40
0,0%
12,5%
25,0%
37,5%
50,0%
62,5%
75,0%
87,5%
100,0%
6x10-2 1.2x10-1 1.8x10-1 2.5x10-1 3.1x10-10
3x10-4 6.7x10-4 1x10-3 1.4x10-3 1.7x10-3 2x10-30
Pseason
0,0
00
0,0
47
0,0
94
0,1
41
0,1
88
0,2
35
0,2
82
0,3
29
0,3
76
0
5
10
15
20
25
30
0,0%
16,7%
33,3%
50,0%
66,7%
83,3%
100,0%
0,0
00
0,5
24
1,0
48
1,5
72
2,0
96
2,6
20
Valores em Milésimos
0,0%
16,7%
33,3%
50,0%
66,7%
83,3%
100,0%
Results – risk estimates (rainy season)
Pdaily
50% = 1.9 x 10-5
95% = 2.6 x 10-3
50% = 3.4 x 10-3
95% = 3.8 x 10-1
Pseason
5.2x10-4 1.1x10-3 1.6x10-3 2.1x10-3 2.6x10-30
9.4x10-2 1.9x10-1 2.8x10-1 3.8x10-10
Results – sensitivity analysis
Variable
Spearman rank correlation coefficient (rs)
Dry seasonRain
seasonPolled data
Occurrence of Cryptosporidium in source water – C (oocysts/L)
+0.30 +0.20 +0.23
Recovery of the oocysts enumeration method – r (%)
-0.02 -0.01 -0.02
Cryptosporidium oocysts removal in the WTP – R (log)
-0.16 -0.27 -0.20
Drinking-water consumption – V (L/day) +0.84 +0.85 +0.84Dose response parameter - θ +0.03 +0.02 +0.02
Sensitivity of probability of infection to variation in input random variables
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
need of data collection on drinking-water consumption in Brazil the importance of reliable data on oocysts occurrence/removal and properly specifying statistical distributions for these variables.
Results – sensitivity analysis
VariableDaily/annual risks
(polled data)
Daily/seasonal risks
Rainy season Dry season
Occurrence of Cryptosporidium in source water
Highest + 1.8 + 0.9 + 2.2
Lowest- 0.8
-0.9 -1.0
Cryptosporidium oocysts removal in the WTP
Highest - 1.7 - 1.5 -1.1
Lowest + 1.7 + 1.7 + 1.3
Sensitivity analysis : log10- values of the decreased or increased median risk compared to when the total distribution is used
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Results – Scenario analysis
Variable
Annual risks Seasonal risks
>10-1 <10-2 <10-3 <10-4
Dry season Rain season
>10-1 <10-2 <10-3 <10-4 >10-1 <10-2 <10-3 <10-4
C +1.28 83.3%
+0.6874.6%
R -0.84 17.9%
-0.63 26.6%
-0.81 18.5%
V +1.07 94.2%
-1.07 41.8%
-1.07 41.8%
+1.07 94.2%
-1.07 41.8%
Scenario analysis results: combinations of inputs which lead to risk of infection targets
Figures within parenthesis: percentile of the subset median of the input variable in the
complete distribution; figures outside parenthesis difference between the subset and the
overall medians divided by the standard deviation of the original simulation; the higher
this number, the more significant is the input variable in achieving the output target value.
Water Safety ConferenceNovember 2-4 2010, Kuching, Malaysia
Conclusions
Seasonal fluctuations in source water quality
(rainfall) and treatment performance ►
Hazardous events ► WSP (Signor et al., 2005;
Signor & Ashbolt, 2009; Smeets et al., 2010).
Seasonal risk fluctuations seems to be
attenuated over the annualized estimates.
Case for shorter-term risk estimates (seasonal,
daily) ►acceptable targets (Signor & Ashbolt,
2009).
Conclusions
QMRA ► objective quantitative input ► WSP
(Smeets et al., 2010).
QMRA models :
pathogens in source water : reliable data,
PDF (variability & uncertainty)
pathogens removal : indicators (turbidity) ???
Critical limits
Thank you !!!!!!!!Thank you !!!!!!!!
rkxb@ufv.br
Water Safety Conference 2010