NATIONAL WATER QUALITY LABORATORY TECHNICAL MEMORANDUM 93-03 December 28, 1992 To: Assistant Chief Hydrologist, PC&TS Regional Hydrologists Chief, Office of Water Quality Assistant Chief, Office of Water Quality Deputy ACH for PC&TS for NAWQA Area Assistant Regional Hydrologists District Chiefs Regional Water-Quality Specialists Area Assistant Regional Hydrologists for NAWQA District Water-Quality Specialists Chiefs, NAWQA Study-Units Chief, Ocala Project Office Chief, Yucca Mtn. QA Group Chief, Branch of Quality Assurance Employees, National Water Quality Laboratory From: Chief, National Water Quality Laboratory Subject: Change in method for the determination of chloride in water due to data quality problems. Authors: Charlie Patton, Gary Cottrell, and Jeff Pritt. Revision: Does not supercede any previous Tech Memo. Scope This memo describes problems of data quality for chloride determinations by ion chromatographic method (Lab Code 1571) and announces the reinstitution of the approved method for determination of chloride by automated colorimetry (Lab Code 15). Data showing precision and accuracy for the colorimetric method for Standard Reference Water Samples (SRWS), SRWS mixtures, and a National Institute of Standards and Technology-certified reference material are included. ________________________________________________________________________________ A problem surfaced concerning data quality for chloride ions determined by the ion chromatographic (IC) method (Lab Code 1571, WATSTORE code 640J), which has been in routine operation at the National Water Quality Laboratory (NWQL) since April 1990. Investigation by a troubleshooting team--consisting of Inorganic Program chemists and technicians, Quality Management Group personnel, Methods Research and Development Program chemists, and Branch of Quality Assurance (BQA) personnel--found both a lack of precision and a slight positive bias in the method. This bias was less than one standard deviation based on the BQA Standard Reference Water Sample (SRWS) program. As a result of this investigation, the method was determined to be out of control and operation of the IC method for chloride determinations was discontinued on October 20, 1992. Please note that these data quality problems do not affect the Low Level chloride IC method. Modifications to correct problems with the current IC method for chloride are underway but expected to take several months. Therefore, the decision was made to return to the colorimetric method that was in use prior to the IC method. The colorimetric method for chloride (TWRI method I-2187-85) is identified as lab code 15 (WATSTORE code 940E). We decided that re- implementing the colorimetric chloride method provides the best means to maintain operations and to ensure data quality while work continues on the IC method. In conjunction with BQA, the troubleshooting team has revalidated the colorimetric method using modern equipment. Because this is an approved method, revalidation consisted of determining precision and accuracy for 12 SRWSs and SRWS mixes used in the BQA blind sample program. Ten replicates of each reference sample were determined on 6 consecutive days. The NWQL will begin using the colorimetric method for chloride determinations on December 16, 1992. Accuracy and precision data from the revalidation experiments are presented in Table 1. The data indicate a small apparent positive bias for the higher concentrations of SRWSs and SRWS mixtures. A National Institute of Standards and Technology (NIST)-certified reference material with a more precisely determined chloride concentration (1000 +/-5 mg/L certified concentration) was analyzed to determine bias. The NIST data in Table 2 show the bias is not associated with the colorimetric chloride method. Because of the method change, please delete Lab Code 1571 and add Lab Code 15 on the Analytical Services Request form when you submit samples that include chloride determinations. Appropriate lab code changes for samples in the current backlog have been made here at the Lab. Please note the parameter code in your sample results to determine which method was used for chloride determinations (IC -- 940J, Colorimetry -- 940E). These changes do not affect requests for Low Level chloride analysis. Over the next several months, the data from the colorimetric method will be compared with data from two types of IC separation column methods. This will allow the NWQL to choose the method for chloride determinations that best suits customers' needs. John Turk of the Colorado District is assisting us by determining what the chloride data quality needs actually are for the Division. Issues concerning long-term data quality will be addressed in a follow-up Tech Memo as more information becomes available. Please call Gary Cottrell (303 467-8104) or Charlie Patton (303 467-8084) if you have additional questions or concerns. ________________________________________________________________________________ TABLE 1.-- Summary of Precision and Accuracy Data for the Colorimetric Chloride Method. ________________________________________________________________________________ SRWS [1] BQA MPV BQA SD N [4] Measured Measured (mg/L) [2] (mg/L) [3] Mean (mg/L) Std. Dev. [5] (mg/L) [6] ________________________________________________________________________________ P13 0.1 0.2 60 0.12 0.03 P15 1.4 0.1 60 1.34 0.02 P11 2.2 0.3 60 2.25 0.06 M120 7.6 0.7 60 7.52 0.05 M106 13 1 60 13.83 0.09 50:50 DI:M112* 23 1.3 60 22.6 0.2 M122 56 2.5 60 56.3 0.4 25:75 M114:M104 66 3 60 67.0 0.3 25:75 M110:M112 101 4 60 102.4 0.6 50:50 M110:M118 161 7 60 162.6 0.9 25:75 M120:M110 201 8 60 205 1.1 M110 266 10 60 269 1.2 ________________________________________________________________________________ [1] -- SRWS refers to the named Standard Reference Water Sample or SRWS mixture. [2] -- BQA MPV refers to the most probable value based on the mean calculated from the Branch of Quality Assurance Round Robin studies of close to 100 participating laboratories. [3] -- BQA SD is the calculated standard deviation for each SRWS based on regression equations. Branch of Quality Assurance uses +/- 2X this value for control charts. [4] -- N is the number of determinations. [5] -- Method Mean is the calculated mean for N determinations by the colorimetric method (Lab code 15). [6] -- Method Std. Dev. is the calculated pooled standard deviation for N determinations over all the analytical runs by the colorimetric method (Lab code 15). * 50:50 refers to the mixture volume ratio; DI is deionized water. ________________________________________________________________________________ TABLE 2. -- Summary of Colorimetric Chloride Method Bias Using National Institute of Standards and Technology Certified Reference Sample 3182 (Certified Concentration - 1000 +/-5 mg/L) ___________________________________________________________________ No. of Measured Mean Expected Value Bias mg/L (%) Replicates (Std. Dev.) (+/- Tolerance) [2] mg/L mg/L [1] ___________________________________________________________________ 4 250.6 (0.9) 250 (2.5) 0.6 (0.2%) 4 201 (1.1) 200 (2.3) 1 (0.5) 4 101.8 (0.3) 100 (1.4) 1.8 (1.8) 4 50.3 (0.3) 50 (0.8) 0.3 (0.6) 4 24.4 (0.1) 25 (0.5) -0.6 (-2.4) 4 11.9 (0.1) 12.5 (0.3) -0.6 (-4.8) ___________________________________________________________________ [1] -- Dilutions of SRM 3182 were required to prepare these concentrations. The tolerance value (mg/L) refers to propagating the dilution uncertainty and the uncertainty associated with the SRM to make the lower concentration samples. [2] -- Bias = Mean value-Expected value. Percent = (Bias/Expected Value) x 100. Supersedes: None Key Words: Chloride, Method, Data Quality Distribution: See Above plus QWTALK.