1. Uncertainty Calculation for the Network Leakage Reduction Monitoring Model Dave Lander and Andrew Laughton 3 rd December 2002

2. Overview  Uncertainty estimation now complete  Preliminary report with Shirley Coleman for independent review  No major issues identified  All sources of uncertainty have been included  Standard leakage rates for mains and services (“NLT”)  Pressure  MEG content and its impact on leakage (“MEG”)  Pit/Spun (“Asset”)

3. Standard Leakage Rates (“NLT”)  No change in methodology since last meeting  Standardised on 500,001 trials for final calculation  Mains only  17.6% (90% confidence level)  Services only  0.6% (90% confidence level)

4. MEG addition and impact (“MEG”)  No change in methodology since last meeting  20,001 trials  Overall impact  4.6% (90% confidence level)  Pc and Sc (fraction that is Lead-yarn treated)  0.5 % (90% confidence level)  MEG saturation level  1.2 % (90% confidence level)  Ly (fraction that receive MEG treatment)  2.3 % (90% confidence level)  Fm (MEG leakage saturation factor)  3.6 % (90% confidence level)

5. Pressure  No change in methodology since last meeting  20,001 trials  P (average system pressure)  4.5 % (90% confidence level)

6. Overall uncertainty contributions % Uncertainty (90% confidence) Mains17.6 Services0.6 MEG4.6 Pressure4.5 Overall18.8

7. Pit/Spun cast distribution (“Asset”)  During the 2002 NLT tests a number of pit cast pipes were found to be spun cast  Original database treated spun and pit as cast iron pipe  Probability that pit cast pipe is actually pit cast  Pit cast 3” / spun cast 5”  Pit cast 4-5” / spun cast 5”  Pit cast 6-7” / spun cast 6-7”  Pit cast 8-11” / spun cast >7”  Pit cast >12” / spun cast >7”  Probability 80%  +/-20% (uniform distribution)

8. Overall uncertainty contributions % Uncertainty (90% confidence) Mains17.6 Services0.6 MEG4.6 Pressure4.5 Asset-4.2 Overall18.818.3 Uncertain pit cast is being replaced with less uncertain spun cast

9. Sampling  Building up a network from parts of a large population of pipes  Some parts are leaky, other are not

10. Sampling  Building up a network from parts of a large population of pipes  Some parts are leaky, other are not  Variability is another source of uncertainty  Present even if NLT leakage data were exact  Network size dependent  At LDZ level, the effect is small enough to be ignored  3.6%  Overall uncertainty 18.3 – 18.6% (90% confidence level)

11. Summary  All sources of uncertainty identified and estimated  Overall uncertainty is 18.3% (90% confidence level)  Dominated by uncertainty in the standard leakage rates