Here’s the water data we have so far

January 14, 2014 by Ken Ward Jr.

Many readers have contacted me expressing interest in the water sampling data we’ve received so far from the Tomblin administration. So here it is:

Water sample results dated Jan. 12, 2014

Water sample results dated Jan. 14, 2014

 Water sample results dated Jan. 15, 2014

As of 6 p.m. Wednesday, Jan. 15, 2014, the Department of Military Affairs and Public Safety is posting the results on their Homeland Security Division’s website here.

2 Responses to “Here’s the water data we have so far”

  1. Albert Donnay, Consulting Toxiclogist says:

    Thanks for posting these data.

    The links dated Jan 12 and Jan 15 open reports that do not identify the
    testing lab and so are not worth analyzing.

    The link dated Jan 14 –which opens an Excel file dated Jan 13– identifies results by lab for samples collected from Jan 10 at 12:30pm until Jan 13 at 6am. These files do not identify any samples collected on Jan 9 when the spill was discovered.

    My analysis of the data in the JAN 14 file is presented below.
    The last point is most important as it shows none of the labs to which split samples were sent could reliably reproduce each other’s results, casting doubt on the accuracy of all the reported data.

    1. Six labs were involved across four days in testing water at WVAWC plant and in the community for MCHM. They are identified as [WV]DHHR, DUPONT, MATRIC, TEST AMERICA, DC CST, and OH CST.

    2. No lab was used on all four days. The CST labs were each used for only one day and the others for three days each. There is no explanation.

    3. Of the 5 labs used to test water at the WVAWC plant, two were used on Jan 10, three on Jan 11 and 12, and just one [MATRIC] on Jan 13.

    4. The limit of detection [LOD] is not specified for any lab and only DC CST consistently reported its results in parts per million [ppm] to 2 decimal places, suggesting an LOD of 10 parts per billion. DUPONT and OH CST reported their results with 0 or 1 decimals, DHHR with 1 or 2, and the others used 1, 2, 3 or even 4, suggesting a detection limit of a tenth of a part per billion. In contrast, WV Gov Toblin is quoted in WV Gazette as saying the LOD is 0.1 ppm.

    5. The inlet and outlet pipes of the WVAWC plant were tested for MCHM from Jan 10-13 by 5 of the labs, but only 1 to 3 labs were used each day.

    Higher levels of MCHM were found at the outlet than the inlet in
    – 45% of 11 paired samples collected on Jan 10,
    – 59% of 22 paired samples Jan 11,
    – 71% of 14 paired samples on Jan 12,
    – and 20% of 5 paired samples on Jan 13 [last sample = 6am].

    The ever higher levels coming out from Jan 10-12 show that the primary source of the contamination on these days was not river but the WVAWC plant itself. The company managed to remove much of the MCHM from the plant on the night of Jan 12-13.

    6. Hydrants and other taps in the community were tested from Jan 11-13 by 4 different labs, but only 3 per day.

    Higher levels of MCHM were found at these taps than was coming out of the WVAWC plant on Jan 11, 12 and 13.

    Just as the WVAWC plant was more contaminated than the river on Jan 11 and 12, the company’s distribution pipes were more contaminated than their water treatment plant from Jan 11 to Jan 13.

    7. To check for agreement between labs, some of the samples collected on Jan 10, 11 and 12 were split and sent to two labs. None of the samples were sent to three or more labs.

    -Dupont ppm level = 31 to 103% of OH CST level [all from Jan 10]
    -Dupont ppm level = 21 to 141% of DC CST level [all from Jan 11]
    -DHHR ppm level = 68 TO 199% of Dupont level [all from Jan 11]
    -DHHR ppm level = 0 TO 100% of Matric level [all from Jan 12]

    These ranges of agreement are very wide, suggesting the labs could not consistently replicate each others’ results. Given this variability, additional split samples should have been sent to each of the labs to see which could consistently replicate their own results.

    # # #
    P.S. I have not yet compared these data with those posted later by Department of Military Affairs and Public Safety at but will post here again if I find the results are substantially different.

  2. Albert Donnay, Consulting Toxicologist says:

    I’ve now reviewed all the other water testing data posted at
    Operational Sampling Results as of January 21.
    The links for these reports are identified as:

    Areas in WV American Water Service Area 1
    Areas in WV American Water Service Area 2
    Areas in WV American Water Service Area 3
    WV American Water Plant
    Initial Sampling Results
    Spot Checks
    Sample Data as of 21 January 2014

    These data are presented in various printout formats — no more Excel spreadsheets — and show many significant problems, both in terms of the results themselves and the formats in which they are reported.

    1) Results are no longer identified by lab, and split samples are identified only as Lab 1 and Lab 2, even though 6 different labs were used from Jan 10-13 according to

    2) None of these reports are complete. All include some results that are missing some important information, such as the date and time of sampling and/or the sample number. The results are not sorted by sample number or by date, or within date by time, making it hard to evaluate the change in results over time.

    None of the reports identify what chemical is being measured [pure MCHM or the crude MCHM mixture that actually spilled?];
    none specify what type of method or instrument they used;
    and none specify if the reported values are the results of single test or the average of several replicates.

    3) The Limit of Detection is specified only in the most recent report [dated Jan 21]. This says in footnote 4 that:

    “Any Lab result below 10 parts per billion (ppb) is reflected as Non Detected (ND).”

    But every non-ND value in this report is given with three digits beyond the decimal [as 0.XXX ppm]. This suggests a LOD of 1ppb, not 10ppb.

    4) All the other testing reports posted earlier by DHSEM include results reported as either ND or with 2, 3, 4, 5 or even 6 digits after the decimal. This suggests that some of the unnamed labs could detect as little as 1 part per trillion. Two consecutive pairs of in/out results from the water treatment plant on Jan 16 and 17 are even reported as 0.000000 in and ND out, which suggests the people involved do not understand that 0.000000 and ND in this context are synonymous.


    5) Some split samples continue to show poor correlation, eg:
    0.216 vs 0.085ppm in sample C003 collected Jan 17

    but many show almost perfect correlation. eg:
    0.025 vs 0.028 in sample C033, also collected on Jan 17

    Both of these example are in

    6) Levels above 1ppm continued to be reported from hydrants in the community as late as Jan 17, eg:

    – Route 62 Buffalo, sample #0605, at 1.22ppm
    – Hydrant 6650 in both a double split sample #5540, at 0.83 and 1.46ppm,
    and in a rare triple split sample #0003, at 0.75, 1.08 and 1.80ppm

    Note again the poor correlation between these splits.

    These examples are from

    7) One entry in one of the reports for a sample collected on Jan 16 from
    “POND GAP BOOSTER, ZONE 20” specifies that the result is:

    “.676 <.7 standard"

    which begs the question of who set this lower than 1ppm standard, when and why.


    # # #

Leave a Reply