damon k. roth, pe bcee david a. cornwell, phd, pe, bcee€¦ · damon k. roth, pe bcee . david a....

Damon K. Roth, PE BCEE David A. Cornwell, PhD, PE, BCEE

• Trends in Pb levels – Review of datasets – Pb data by season / location / service line type

• Pb profile data • Water Quality Parameters • Takeaways

1) EPA Sequential Database 2) MDEQ Sentinel Database 3) MDEQ Residential Database

15 December 2016

1) EPA Sequential Database – 279 samples across 115 locations – Pb profile data – Samples span 1/28/2016 to 11/15/2016

2) MDEQ Sentinel Database 3) MDEQ Residential Database

15 December 2016

1) EPA Sequential Database 2) MDEQ Sentinel Database

– 3,988 samples across 840 locations – Pb first liter data – Samples span 2/16/2016 to 9/27/2016 – Includes Rounds 1-5 and Extended Rounds

1-5 3) MDEQ Residential Database

15 December 2016


– 24,251 samples across 14,193 locations – Selected for locations in common with EPA

Sequential or Sentinel databases • 4,361 samples across 801 locations in common with

those databases – Pb first liter data – Samples span 9/15/2015 to 9/27/2016

15 December 2016


15 December 2016

The first three databases were combined based on sample location: 1) All EPA Sequential database locations were included 2) All MDEQ Sentinel database locations were included 3) Only includes those MDEQ Residential database

locations where either EPA Sequential or MDEQ Sentinel data were available

– This decision was made to target “higher risk” sites 4) LCR data could not be geocoded, were not included • Result is a common database with 8,596 data across

904 locations.

• Analysis of the common database is dependent on the Pb value used for the EPA Sequential sampling

• For consistency with MDEQ Sentinel and Residential data, we used the Pb level in the first EPA sample (125 mL)

• However using the peak Pb level from the EPA Pb profiles will change maximum Pb distribution

• Analysis based on location-specific maximum Pb values

• The next two plots illustrate this difference ‒ Analysis of EPA profile data presented later

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb Distribution by Database (# of Locations)

EPA-1S (115) Residential (801) Sentinel (840) Action Level

Combined Dataset, Max Pb at Each Location in 2016, EPA 1st Sample Pb Values

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb Distribution by Database (# of Location)

EPA-Peak (115) Residential (801) Sentinel (840) Action Level

Combined Dataset, Max Pb at Each Location in 2016, EPA Peak Pb Values

The following analysis looks at the maximum Pb level at each location (regardless of database source), distributed by: • Season

‒ Winter (Jan/Feb) ‒ Spring (Mar-May) ‒ Summer (Jun-Aug) ‒ Fall (Sept-Nov)

• Before/After May Flush Event • Zip Code

• All plots are presented using EPA-1S data unless otherwise

specified

• Distribution of data appears similar if all locational data is considered, however – # of locations sampled in Winter and Spring

2016 significantly exceeds # of locations sampled in Summer and Fall 2016

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Max Pb Level for each Location, by Season (# of Locations)

Winter (750) Spring (820) Summer (241) Fall (199) Action Level

Combined Dataset, Max Pb at Each Location in Season, EPA 1st Sample Pb Values

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Pb < 15 µg/L 15 µg/L ≤ Pb ≤ 50 µg/L Pb > 50 µg/L

% o

f Loc

atio

ns

Binned Distribution of Seasonal Max Pb Data Winter Spring Summer Fall

Combined Dataset, Max Pb at Each Location in Season, EPA 1st Sample Pb Values

• To better understand how Pb changed at specific locations over time, a subset was created from only those sites with Pb samples collected during all four seasons

• Winter: 750 locations → 128 locations • Spring: 820 locations → 128 locations • Summer: 241 locations → 128 locations • Fall: 199 locations → 128 locations

• Removing non-paired data significantly changed the percentile

distribution for the Winter and Spring data • Summer and Fall data distributions were less affected because a

smaller proportion of data were removed • Unpaired data removed from the Winter and Spring subsets tended

to be low Pb

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Max Pb Level for Locations with Paired Data, by Season (# of Locations)

Winter (128) Spring (128) Summer (128) Fall (128) Action Level

Combined Dataset, Max Pb at Each Location with Paired Data in All Four Seasons, EPA 1st Sample Pb Values

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Pb < 15 µg/L 15 µg/L ≤ Pb ≤ 50 µg/L Pb > 50 µg/L

% o

f Loc

atio

ns

Binned Distribution of Seasonal Max Pb - Paired Data

Winter-Paired Spring-Paired Summer-Paired Fall-Paired

Combined Dataset, Max Pb at Each Location with Paired Data in All Four Seasons, EPA 1st Sample Pb Values

• Paired data distributions show that Pb levels are trending lower over time ‒ If distribution is created using EPA-1S, the Fall data

90th percentile at locations with paired data is <15 µg/L

‒ If distribution is created using EPA-peak, the Fall data 90th percentile is ~equal to 15 µg/L

• However, paired data distributions don’t show what happens at specific locations

• Comparison of data Before/After May Flushing shows most locations improved, some locations saw Pb increase

• Instead of grouping by season, data were grouped by occurrence before or after May Flushing (defined as 5/15/2016)

• Again, significantly more data prior to May Flushing than after

• Comparing these raw data show the distributions before/after the May Flushing are similar – Flushing somewhat improved Pb levels on the high end

of the distribution • 90th percentile before = 55.1 µg/L • 90th percentile after = 38.0 µg/L

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Locational Pb Levels Before/After May Flush (# of Locations)

Before (876) After (259) Action Level


• Again, to better understand how Pb changed at specific locations over time, a subset was created from only those sites with Pb samples collected both prior to and after the May Flush Event

• Before: 876 locations → 232 locations • After: 259 locations → 232 locations

• Removing non-paired data significantly changed the percentile

distribution for the Before data • After data distributions was less affected because a smaller

proportion of data were removed • Unpaired data removed from the Before subset tended to be

low Pb

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Locational Pb Levels Before/After May Flush at Paired Locations (# of Locations)

Before (232) After (232) Action Level

Combined Dataset, Max Pb at Each Location with Paired Data Before/After May 15, 2106, EPA 1st Sample Pb Values

• Paired data distributions don’t show what happened at specific locations over time

• Before May Flushing, 46% of paired data locations exceeded 15 µg/L

• After May Flushing, 22% of paired data locations exceeded 15 µg/L ‒ 62% of the locations above 15 µg/L before May

Flushing dropped below 15 µg/L ‒ 9% of the locations below 15 µg/L before May

Flushing exceeded 15 µg/L after flushing ‒ 11% of locations saw no change in Pb before/after

May Flushing

15 December 2016

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

-200% -100% 0% 100% 200% 300%

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Change in Pb Level after May Flush Event

Impact of May Flushing on Pb Levels by Paired Location

After > Before Lead increased after Flushing

Before > After Lead decreased after Flushing

Combined Dataset, Ratio of Max Pb at Each Location with Paired Data Before/After May 15, 2106, EPA 1st Sample Pb Values

• Locations were classified spatially based on zip code

• Majority of samples collected in the zip code around City center (48503) and the zip code northwest of City center (48504)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Locational Pb Levels Before/After May Flush (# of Locations)

48502 (3) 48503 (328) 48504 (221) 48505 (134) 48506 (65)48507 (120) 48517 (1) 48532 (32) Action Level


Values show 90th percentile of Max Pb at each location sampled in the zip code

115 µg/L

53.7 µg/L

31.5 µg/L

78.6 µg/L

39.3 µg/L

20.3 µg/L

13.8 µg/L

• Maximum Pb levels were plotted based on service line material: – Copper – Pb – Galvanized

• Service line material is based on private-side of meter

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb at Location by Service Line Type, All 2016 (# of Locations)

Copper (592) Galvanized (167) Pb (134)

Combined Dataset, Max Pb at Each Location, EPA Peaks

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb at Location by Service Line Type, Winter 2016 (# of Locations)


Combined Dataset, Max Pb at Each Location during Season, EPA Peaks

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb at Location by Service Line Type, Spring 2016 (# of Locations)



0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb at Location by Service Line Type, Summer 2016 (# of Locations)



0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb at Location by Service Line Type, Fall 2016 (# of Locations)



0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 10 100 1,000 10,000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Maximum Pb Levels at Locations with Pb or Galvanized Service Lines, Paired by Season

(# of Locations) Winter (34) Spring (34) Summer (34) Fall (34)

Combined Dataset, Paired Data Only, Only Galv. and Pb SLs, Max Pb at Each Location during Season, EPA Peaks

EPA Sequential Data - Profile volume where peak Pb occurs

Sampling Round Sample Dates Number of Profiles Number of Houses

Sequential 1 February 2016 94 94

Sequential 2 May 2016 42 41

Sequential 3 July 2016 47 47

Sequential 4 September 2016 51 51

Sequential 5 November 2016 45 45

Total 279 115

15 December 2016

39.1%

17.2%

10.4%

4.7% 4.3% 2.2%

5.0% 2.2%

7.2%

2.9% 1.4% 1.1% 2.5%

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

1st L 2nd L 3rd L 4th L 5th L 6th L 7th L 8th L 9th L 10thL

11thL

12thL

after12th

L

Perc

ent o

f All

Prof

ile P

eaks

Profile Volume

EPA Sequential Data Profile Volume where peak occurred

"2nd L" means peak occurred >1,000 but ≤2,000 mL

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th >12th

Perc

ent o

f Val

ues

Cumulative Profile Volume (L)

EPA Sequential - >15 vs >50 µg/L - Peak Peak >15 µg/L 95 samples Peak >50 µg/L 42 samples

15 December 2016

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th >12th

Perc

ent o

f Val

ues

Cumulative Profile Volume (L)

EPA Sequential - >15 vs >50 µg/L - Any or Peak Peak >15 µg/L 95 samples Peak >50 µg/L 42 samplesAny >15 µg/L 495 samples Any >50 µg/L 114 samples

Peak Lead Values

• First graph compares 1st L in all profiles versus peak lead in all profiles versus maximum peak value for profiles at a given house

• 2nd graph shows peak lead value observed at each house

• 3rd graph shows peak at each profile per season (in all but one case, there was only one profile per house in each season)

15 December 2016

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1 10 100 1000 10000

Obs

erva

tions

Les

s Th

an V

alue

(%)

Peak Lead in a Given profile (µg/L)

Comparison of Peak Lead Values Versus 1st Sample Values (typically 125 mL) - EPA Sequential

1st sample each profile (234 profiles)max of each profile (234 profiles)peak at all profiles at each house (115 household locations)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1 10 100 1000 10000

Obs

erva

tions

Les

s Th

an V

alue

(%)

Peak Lead in a Given Profile (µg/L)

Comparison of Peak Lead in a House EPA Sequential - By Sample Period

Sequential Sequential_2 Sequential_3Sequential_4 Sequential_5

>20% of houses in most recent sampling period have peak >15 µg/L

>15 and >50 µg/L by Houses

85 of 234

37 of 234

55 of 114

34 of 114

0%

10%

20%

30%

40%

50%

60%

>15 >50

Perc

ent o

f Pea

k G

reat

er T

han

Valu

e (%

)EPA Sequential

Percent of Values >15 or 50 µg/LAll Profiles highest value in profiles at a house

Does not include Round 5

• Compares peak lead from all samples in all profiles at a given house

• Black bar is from Flint EPA Sequential (114 houses)

• Blue bar is from Cornwell Engineering Group (CEG) national database (34 houses)

• Green bar is for 25 houses from Del Toral et al. 2013 (ES&T article)

55 of 114

34 of 114

15 of 34

5 of 34

17 of 25

none0%

10%

20%

30%

40%

50%

60%

70%

80%

>15 >50

Perc

ent o

f Pea

ks G

reat

er T

han

Valu

e (%

)EPA Sequential (max values at house)

versus Del Toral et al. 2013 and CEG National DatabasePercent of Peak Lead Values >15 or 50 µg/L

highest value in profiles at a house CEG Nat. database - houses Del Toral et al. 2013 - houses

Does not include Round 5

• EPA provided detailed plumbing breakdown of 77 residences, indicating all plumbing pipes, fixtures, etc. from tap to service line

• Premise plumbing was defined as plumbing from tap to meter • All locations in database have meters located inside

of house • Cumulative volume inside of house based on

calculated volume of plumbing • Analysis compared volume of premise plumbing

to the sample volume where profile peak occurred.

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

10/23/2015 12/12/2015 1/31/2016 3/21/2016 5/10/2016 6/29/2016 8/18/2016 10/7/2016 11/26/2016 1/15/2017 3/6/2017 4/25/2017

Mea

sure

d O

rtho

phos

phat

e (m

g/L

PO4)

Date

2016-2017 Orthophophate

D1/2 D3/D4 D5 D6 D7/D8 D9/D13 D10 D11 R1 R2 S1 CSII Tap

7.0

7.1

7.2

7.3

7.4

7.5

7.6

7.7

7.8

10/23/2015 12/12/2015 1/31/2016 3/21/2016 5/10/2016 6/29/2016 8/18/2016 10/7/2016 11/26/2016 1/15/2017 3/6/2017 4/25/2017

pH

Date

2016 - 2017 pH

D1/D2 D3/D4 D5 D6 D7/D8 D9/D13 D10 D11 R1 R2 S1 CSII Tap

• Raw data can mask extent of Pb levels at higher-risk sites

• Paired data tends to be collected at higher-risk areas

• Removing unpaired data changes percentile distributions from earlier periods, increases median and 90th percentile Pb levels 15 December 2016

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Locational Pb Levels Before/After May Flush

(# of Locations) Before (876) After (259) Action Level

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0.1 1 10 100 1000 10000

Perc

ent o

f Obs

erva

tions

less

than

Val

ue

Pb (µg/L)

Locational Pb Levels Before/After May Flush at Paired Locations

(# of Locations) Before (232) After (232) Action Level

• Pb levels in most locations are decreasing when comparing paired data

15 December 2016

0%

20%

40%

60%

80%

100%

1 10 100 1000 10000

Obs

erva

tions

Les

s Th

an V

alue

(%)


Comparison of Peak Lead in a House EPA Sequential - By Sample Period Sequential Sequential_2 Sequential_3Sequential_4 Sequential_5

• Pb levels in most locations are decreasing when comparing paired data

15 December 2016

0%

20%

40%

60%

80%

100%

1 10 100 1000 10000

Obs

erva

tions

Les

s Th

an V

alue

(%)


Comparison of Peak Lead in a House EPA Sequential - By Sample Period Sequential Sequential_2 Sequential_3Sequential_4 Sequential_5

>20% of houses in most recent sampling period have peak >15 µg/L

15 December 2016

damon k. roth, pe bcee david a. cornwell, phd, pe, bcee€¦ · damon k. roth, pe bcee . david a....

Documents