National Water-Quality Assessment (NAWQA) (public)
National Analysis of Trace Elements (public)
Trace Element National Synthesis internal homepage - QA/QC

Blank Sample Analysis and Summary for FY91 Study-Unit Trace-Element Data

[This summary was transmitted as an attachment to TE News v.12, 09Feb2000.]

To: Study-Unit staffs, NAWQA Program
From: Michael T. Koterba, Quality Assurance Chief, NAWQA Program
Subject: Blank Sample Analysis and Summary for 1992-1995 (FY91 Study-Unit) Trace-Element Data
Date: February 8, 2000

• Summary of 1992-1995 blank results
• Elements with no apparent contamination/bias and Excel spreadsheet
• Elements with measured contamination/bias
• Sources of trace element contamination in blank samples
• Factors that affect identification of sources of bias
• Conclusions and implications

Results from field blank data collected by the 1991 study units and analyzed for trace elements are presented here. The procedures we currently use are in many ways similar to those used by the 91 SUs. Therefore, your data are likely to have these same limitations. Inadequacies in the coding and completeness of ancillary information, which prevent the exact causes of the 91 contamination from being known, are also still occurring.

For the FY91 study-units, radon and uranium were the only trace elements included in routine NAWQA ground-water data collection. Neither element required blanks. However, some study units collected additional trace element (TE) data at their own discretion; most study units used NAWQA TE schedule 2703 for these additional data, guided by the Division's Parts-per Billion Sampling Protocol for Low Level Inorganic Constituents (OQW Tech Memo 93.11). Each subunit survey (SUS) or land-use survey (LUS) collecting trace elements was required to collect field evaluation blanks and routine field blanks, to be analyzed using SC172, an ICP-MS low-level blank schedule similar to SC2703.

There were no guidelines given on the type and minimal number of TE quality-control blanks to be collected during surface-water sampling. Blanks generally were collected at the discretion of the study-unit sampling team. If collected, they generally were taken at the time of other SW blanks--upon initially arriving at a SW sampling site (BFS or IFS), after SW sampling, cleaning, and storing equipment at a previous surface-water site (so as to include travel).

For most TEs, the blank database generally consists of about 90 to 100 field blanks for GW, and about 25 to 50 field blanks for SW. Too few blanks (~20) appeared for boron (B), which was an "add-on" TE, to make inferences concerning the quality of its data for either ground or surface water. Although not a trace element, Fe was part of the major ions schedule (SC2750 or SC2751). Analyses for Fe were not run on blanks. Because of similarities in chemical behavior, the quality of Fe environmental data, at least at this point, should be considered similar to that of Mn.

Summary of 1992-1995 Blank Results

For each element, blank results generally were similar for GW and SW. Hence, all subsequent remarks apply to either medium.

Elements with no apparent contamination/bias

For ten trace elements (As, Sb, Mo, Co, Se, Ni, Pb, Ag, Cd, and Cr), no contamination was detected in 90-100% of the blanks. Overall, there appears to be no systematic contamination (or false-positive bias) in the corresponding GW or SW environmental data. The environmental data for these 10 elements generally may be interpreted to low levels (1-2 ug/L). For each of these elements, any blank concentrations that exceeded method reporting levels generally had low values (commonly 1 ug/L or less). High values were quite infrequent.

Elements with measured contamination/bias

For five elements (Ba, Mn, Cu, Zn, and Al, in order of increasing severity), there is an increase over time in both the frequency and the magnitude of detectable TE concentrations in blanks. For example, for Ba, about 10% of the blanks had detectable values, which generally were never greater than 1-2 ug/L; whereas for Al, over 90% of the blanks had detectable values, almost a third on the order of 10 ug/L. (See the excel blank-data spreadsheet http://water.usgs.gov/nawqa-only/trace/qaqc/91blanks_gwsw.xls
for details on bias data for each element.)

Environmental data for these five elements are possibly biased. This implies data interpretations for these elements, particularly at low levels, require caution. For example, definitions of reference or other conditions derived from environmental data using these elements are likely to be positively biased. Reference conditions are lower than the levels we are measuring and reporting. Also the frequency of environmental samples with concentrations that exceed some fixed-value standard or criterion (be it regulatory, biological, or otherwise) will be positively biased if the criterion is near the contamination level, or the exceedance is within contamination levels from the criterion. The blank data indicate that the positive bias is not constant, but varies in magnitude. Thus, it could introduce additional variability in concentration data that masks true differences in concentrations among different sites, or increases the variability in the concentrations measured through time at a site.

Sources of Trace Element Contamination in Blank Samples

No systematic sources for the blank contamination bias have been clearly identified. What is known is that the Ocala inorganic blank water is not the culprit. Lot analysis of blank water issued in 1992-1995 indicates that with few exceptions (mainly B), the Ocala blank water did not regularly contain any TEs at detectable concentrations (See the Ocala Lot Analysis worksheet within the data spreadsheet cited above).

For each of the 5 suspect TEs, the bias occurs in each year of data collection at about the same frequency and range of intensity. The bias also occurs in more than one study-unit's data sets during the 1992-1995 period. A limited number of field equipment blanks and solution blanks (latter have no exposure to sampling equipment) obtained by some study-unit teams indicate the bias could be the result of one or more of the following:

• equipment carryover, in the field or laboratory,
• use of contaminated equipment -- filters, sample bottles, or preservative
• the analytical method used.

NAWQA guidance at the time called for the use of the low-level schedule 172 for blanks. However, some study units selected analytical methods other than ICP-MS (used by SC172) having lower precision and higher reporting limits. This may have resulted in the reporting of detectable values for a blank analysis which were in fact false positives resulting from instrument noise (Brooke Connor, NWQL, written communication). Too few blanks were analyzed with SC172 to substantiate that the environmental data for the five suspect elements are unbiased.

Factors That Affect Identification of Sources of Bias

Based on comments in the individual sample-records, and electronic memoranda from study-unit teams, some blanks appear not to have been collected or coded in accordance with the NAWQA protocols. For example,

• the FA sample bottles were not acid-soaked and pre-rinsed
• the encapsulated filters used were not from Ocala (but, for example, GEOTECH, and differ in internal construction and external design).
• the filters were not pre-rinsed with Ocala-grade water
• blanks were not preserved with NWQL Ultrex acid, but acid from another source
• blank source water used was not Ocala inorganic grade, or was Ocala grade, but possibly stored many months before use, or regardless of source and storage time, was employed in a such sparing manner that too little was passed to adequately rinse equipment before blanks were collected.
• Blank types other than the specified type in protocols for NAWQA Program were coded in the exact same manner as the required type.
• Environmental samples have likely been miscoded as blanks.
• Ancillary information on the blank, such as blank type, water source, and water lot, was incomplete or incorrectly coded.
• No information is available (this info was not initially requested) on what equipment was blanked, such as type of sampling pump.

Conclusions and Implications

Due to the above data-quality or coding problems, and a general lack of ancillary data on many blank samples, it is unlikely that the actual source(s) of this apparent bias can be clearly identified from further analysis of the FY91 QC blank data. Analysis of the 1994 blank data is proceeding.

To insure that there is more information on contamination sources in current and future NAWQA sampling, all NAWQA Field Evaluation and Routine Field blank data should be collected and coded in accordance with NAWQA sampling protocols and coding formats. As there are several documents where these protocols are listed, they will be collated and summarized in one place, and transmitted in an upcoming memo.

Yet to be determined by the Trace-Element Synthesis Team is whether the blank data for these five elements can be used to determine which environmental sample data might be flagged (V-coded) to indicate to sample contamination, or re-censored to better represent what is known about the ambient concentrations.

Trace Element National Synthesis internal homepage - QA/QC
National Analysis of Trace Elements (public)
National Water-Quality Assessment (NAWQA) (public)