MAY 1997 UPDATE OF THE QADATA DATA BASE

The Branch of Technical Development and Quality Systems (BTD&QS) operates the Blind Sample Project (BSP). One of the goals of the Blind Sample Project is to document variability and bias for the NWQL and Ocala QWSU through the use of double-blind quality-assurance samples. The quality-assurance results are stored in a SAS database, on the BTD&QS Windows NT under a program named "QADATA". This program is accessed via a DG/UX terminal and available to all WRD personnel. QADATA instructions are available on the BTD&QS home page.

The Blind Sample Project has not sent a QADATA summary since April 1996 due to development & programming of the new QADATA system. We apologize for any inconvenience that this delay may have caused. Starting with this summary the BSP intends to send the QADATA summary bi-monthly.
The QADATA Summary is the Blind Sample Project's cursury review of the NWQL and QWSU results on the quality assurance samples. The data reviewed is for the time period of 11/1/96 - 5/1/97, released from the labs as of 5/20/97. For this 6-month time period, 96 determinations at the NWQL and 40 determinations at the Ocala lab were quality assured. The summary is intended to be a timely, initial review of the laboratories' performance on blind quality- assurance samples. District personnel can improve the quality of their data by requesting reruns from the NWQL or QWSU; both labs provide free re-analysis of determinations. WRD projects can document NWQL and QWSU variability and bias by tailoring a QADATA retrieval to parallel their project's sampling period.

This report was compiled by Amy Ludtke and Mark Woodworth. The control plots, precision plots, and statistical summaries used for the QADATA summary are available on request. Please address any questions, comments, or changes in distribution to:

Amy Ludtke

e-mail: asludtke@usgs.gov

phone: (303) 236-1870, ext. 314

NWQL PERFORMANCE ON BSP SAMPLES:

NUTRIENTS:

• Nitrite + Nitrate-N, LIS, Dis., Colorimetric (labcode=1979, parmcode=00631G) - displaying high variability with 5 of 13 (39%) samples outside the 2-sigma control limits. The standard reference samples (SRS) do not appear to be the problem because both SRS are being analyzed by labcodes 1975 and 1990 at the NWQL which are not displaying similar variability problems.
  

NWQL Response:

Although N-49 shows a low bias by all methods for NO2 + NO3, there appears to be higher variability for this method than for methods 1979 and 1990. The bias and variability has been and is under investigation. It is also suggested that the MPV for N-49 be re-evaluated by BTD&QS.

• Phosphorus-P, WWR, JKA, (labcode=1984, parmcode=00665D)
• Phosphorus-P, WWR, EPA, (labcode=1993, parmcode=00665E)
• Phosphorus-P, DIS, JKA, (labcode=1983, parmcode=00666D)
• Phosphorus-P, DIS, EPA, (labcode=1992, parmcode=00666E) - all 4 determinations appear to be having analytical problems when in fact the SRS are most likely unstable. The BSP will review the phosphorus data with the help of the SRS project and the labs and will summarize the findings in an upcoming QADATA summary.
• Orthophosphorus-P, DIS., COL (labcode=1974, parmcode=00671H) - The last summary noted a problem with a standard reference sample. That reference sample was determined to be of poor quality so all BSP orthophosphorus data associated with the sample was deleted.
  

MAJOR IONS:

• Calcium, ICP, LIS, Dis. (labcode=1273, parmcode=00915E) - displaying both a positive bias and high variability with 4 of 7 (57%) samples outside the 2-sigma control limits.
  

NWQL Response:

It appears that there could be a problem with the control limits for calcium in P-24. Because of the limited amount of QC available for these blind sample charts, it is difficult to do a thorough evaluation at this time. In addition, the Plasma group does not presently use the same QC samples submitted through the blind QC program for the on-line ICP QC. However, to better evaluate the blind QC submitted. the Plasma group will start accumulating data for the same QC submitted for the blind QC program. See the Mn response for rationale.

• Fluoride, Dis., ISE (labcode=31, parmcode=00950B) - the April 1996 summary noted a high variability with 8 of 40 (20%) samples outside the 2-sigma control limits for this determination. The high variance has improved significantly with only 1 of 119 (<1%) samples outside the 2-sigma control limits.
  

NWQL Response:

Problems with the fluoride ion selective electrode meter were resolved in April, 1996. The fluoride system has been running well since that time.

• Sulfate, Dis., IC (labcode=1572, parmcode=00945G) - the April 1996 summary noted a negative bias with 13% of samples falling below the (-2) sigma control limit. This bias is no longer evident and only 2 of 119 (2%) samples fall outside the 2-sigma control limits.
  

TRACE METALS:

• Arsenic, Dis., and WWR, Hydride, AA (labcodes 112 & 118, parmcodes 01000B & 01002B) - as noted in the April 1996 summary, both of these determinations continue to display high bias and variability. Labcode 112 has 12 of 75 (16%) samples above the 2-sigma control limit; and labcode 118 has 10 of 20 (50%) samples above the 2-sigma control limit. The BSP submits many of the same SRS to the Ocala arsenic method and the NWQL arsenic EPA method neither of which are displaying a high bias or variability.
  

NWQL Response:

Troubleshooting efforts have been on-going since November 1996 to determine the cause of the bias and variability shown by arsenic analysis at the NWQL. Sample results have not been released unless data for all of the on-line QC samples were within acceptance limits. We are exerting considerable effort to transfer the arsenic analysis to the graphite furnance atomic absorption method, which is much more precise and reliable.

• Cadmium, Dis., ICP (labcode=673, parmcode=01025D) - displaying a high variability with 5 of 46 (11%) samples outside the 2-sigma control limits.
  

NWQL Reponse:

The variation shown is the level of precision typically demonstrated by the ICP-OES technique. Most blind samples submitted have concentrations and limits that are too tight for the ICP-OES technique and are more applicable to the ICP-MS or GFAA techniques. This can be demonstrated by the results for Mix #50135-50137, which includes SRWS T-135 which has a cadmium concentration more than five times greater than the other SRWS's shown on the chart. These data for Mix #50135-50137 have no points outside of +/- 2 standard deviations. The ICP-OES chart demonstrates similar precision in past BTD&QS charts.

• Copper, Dis., GFAA (labcode=1559, parmcode=01040F) - the April 1996 summary noted a high variability with 6 of 37 (16%) samples outside the 2-sigma control limits. The high variance has improved significantly and none of the 43 samples analyzed were outside the 2-sigma control limits.
  
• Lead, Dis., ICP (labcode=646, parmcode=01049C) - continues to display a positive bias and high variability with 11 of 46 (24%) samples outside the 2-sigma control limits.
  

NWQL Reponse:

The variation shown is the level of precision typically demonstrated by the ICP-OES technique. The limits and concentrations submitted for Pb determination by ICP are too tight for ICP-OES: the control limits appearto be biased by the more precise analytical systems for Pb such as GFAA and ICP-MS. The ICP-OES chart demonstrates similar precision compared to past BTD&QS charts.

• Manganese, Dis., ICP (labcode=648, parmcode=01056C) - displaying high variability from Sept. 1996 through mid-Jan. 1997. With 22 of 113 (20%) samples outside the 2-sigma control limits. Beginning in mid-Jan. 1997, the determination is back in control.
  

NWQL Response:

All of the results outside of minus two standard deviations are from amix that includes SRWS T-141. This SRWS has demonstrated erratic behavior for manganese and therefore cannot be considered when assessing analytical control for manganese. Documentation regarding this problem was recently sent to BTD&QS. It is recommended that a mix containing SRWS T-141 not be used for the QA of Mn.

BSP Comment:

The BSP will discuss this issue at an upcoming NWQL Inorganic Quality Assurance meeting. At issue will be why these same mixes came back into control beginning in Jan. 1997. The results of any findings will be posted in an upcoming QADATA summary.

• Mercury, WWR, CVAA (labcode=227, parmcode=71900B) - displaying high variability with 5 of 20 (25%) samples outside the 2-sigma control limits. The variability did improve from mid-Feb. 1997 through the beginning of May 1997.
  

NWQL Reponse:

Values that were outside the 2-sigma control limits were due to transcription errors. The present manual data transfer program for WWR mercury determinations automatically enters default values for all the samples and then the analyst must edit the default values. We are in the process of revising the data entry program. This new procedure will prompt the analyst to enter all values twice as performed in other manual data entry processes. This will minimize data entry errors.

• Strontium, Dis., ICP (labcode=652, parmcode=01080B) - displaying a high bias with 49 of 56 samples reported above the Most Probable Value (MPV). The same reference materials are being submitted to the Ocala laboratory which is not displaying a similar high bias.
  

NWQL Response:

Around October-November 1996, strontium determinations by ICP-OES began to exhibit a slightly high bias but within the control limits. Action was taken at that time which seemed to shift the overall results for strontium slightly downward. The actions taken did not have the results anticipated, therefore, the matter is still under investigation and will continue to be until the control charts demonstrate a more acceptable bias or no bias at all.

OCALA PERFORMANCE ON BSP SAMPLES:

NUTRIENTS:

• Phosphorus-P, WWR, COL. (labcode=1984, parmcode=00665D)
• Phosphorus-P, DIS, COL, (labcode=1983, parmcode=00666D) - Both determinations appear to be having analytical problems when in fact the SRS are most likely unstable. The BSP will review the phosphorus data with the help of the SRS project and the labs and will summarize the findings in an upcoming QADATA summary.
  
• Orthophosphorus-P, DIS. COL (labcode=1974, parmcode=00671H) - The last summary noted a problem with a standard reference sample. That reference sample was determined to be of poor quality so all BSP orthophosphorus data associated with the sample was deleted.

MAJOR IONS:

• Calcium, Dis., ICP (labcode=650, parmcode=00915D) - the April 1996 summary noted a high variability with 23 of 90 (26%) samples outside the 2-sigma control limits. The high variance has improved with 3 of the 33 (9%) samples outside the 2-sigma control limits.
  
• Chloride, Dis., IC (labcode=292, parmcode=00940F) - displaying a high bias with all of the 36 samples reported above the MPV. Many of the same SRS are being submitted to the NWQL which is not displaying a similar high bias.
  

OCALA Response:

We are currently checking our stock standards. Our internal control charts show a slight high bias. This could also be a problem with the computer program that integrates the chromatograms.

• Magnesium, Dis., ICP (labcode=663, parmcode=00925C) - the April 1996 summary noted a high variability with 20 of 90 (22%) samples outside the 2-sigma control limits. The high variance has improved with 1 of the 33 (3%) samples outside the 2-sigma control limits.
  
• Sodium, Dis., AA (labcode=59, parmcode=00930B) - displaying a negative bias with all of the 34 samples reported below the MPV. Many of the same SRS are being submitted to the NWQL which is not displaying a similar low bias.
  

OCALA Response:

New commercial stock standards were ordered. When new working standards were made, the bias was no longer there. The working standards will be made more often.

• Silica, Dis., ICP (labcode=667, parmcode=00955D) - displaying a negative bias with 22 of 24 samples reported below the MPV.
  

OCALA Response:

Our internal control charts do not show a low bias for silica. The low bias seen in the blind samples could be due to the dilution when they are acidified. Most are less than 2% below the MPV. We will check this further.

TRACE METALS:

• Arsenic, Dis. (labcode=112, parmcode=010020) - the April 1996 summary noted a high variabiltiy with 6 of 20 (30%) samples outside the 2-sigma control limits. The high variance has improved significantly and none of the 13 samples analyzed were outside the 2-sigma control limits.
  
• Cadmium, WWR, GFAA (labcode=1555, parmcode=01027F) - the April 1996 summary noted a high variabiltiy with 5 of 20 (20%) samples outside the 2-sigma control limits. The high variance has improved significantly and none of the 14 samples analyzed were outside the 2-sigma control limits.
  
• Strontium, Dis., ICP (labcode=3049, parmcode=010801) - the April 1996 summary noted a high variabiltiy with 10 of 40 (25%) samples outside the 2-sigma control limits. The high variance appears to have improved after mid-January 1997.
  

OCALA Response:

The variance in the strontium data was due to the rounding algorithm used on our computer system. The raw data did not show any significant variance. When the algorithm was changed, the variance improved.

Return to the QADATA Summary Listing

Return to BTD&QS Home Page