SRS Project Header

U.S. GEOLOGICAL SURVEY STANDARD REFERENCE SAMPLE PROJECT -
PERFORMANCE EVALUATION OF ANALYTICAL LABORATORIES

INTRODUCTION
Since 1962, the U.S. Geological Survey (USGS) has operated the Standard Reference Sample Project (SRSP) to evaluate the performance of USGS, cooperator, and contract analytical laboratories that analyze chemical constituents of environmental samples. The laboratories are evaluated by using performance evaluation samples called Standard Reference Samples (SRSs). SRSs are submitted to labs biannually for inter-laboratory comparison purposes. For the Spring 2017 Round Robin study, 87 laboratories were evaluated for their analytical performance on six SRSs for inorganic and nutrient constituents. As part of the SRSP, a surplus of homogeneous, stable SRSs are maintained for purchase by USGS offices and participating laboratories for use in continuing quality-assurance and quality-control activities.
Statistical evaluation of the laboratory results provides information to compare the analytical performance of the laboratories and to determine possible analytical deficiencies and problems. SRS results also provide information on the bias and variability of different analytical methods used in the SRS analyses.
Although the SRSP is not a certification program, participation is required for all laboratories that provide water-quality data for the USGS (Office of Water Quality Technical Memorandum 2016.06). The participating laboratories are identified only by a confidential ID number. Laboratory personnel select the SRSs to analyze and must include the analyses that are routinely performed for USGS investigations. Analytical methods of choice can be used by these laboratories to analyze the SRS constituents.

TYPES AND PREPARATION OF THE STANDARD REFERENCE SAMPLES
USGS personnel prepare the following types of SRSs from different natural-matrix water sources along the front range and plains of Colorado:

The SRSs are prepared by pumping the water through a series of 0.45- and 0.2-micrometer filters and ultraviolet sterilizer into a polyethylene drum. When appropriate, the water is either acidified with nitric acid or chlorinated to 5 parts per million free chloride or both. After determining the background analyte concentrations, selected constituents are spiked in the water with reagent-grade chemicals. The water is then circulated for at least 24 hours. During SRS bottling, the water continues to pass through filters and the ultraviolet sterilizer. The bottles and caps used for SRSs are acid leached, deionized water rinsed, and autoclave sterilized.

ANALYSIS AND REPORTING OF THE SRS RESULTS
The results of the SRSP inter-laboratory comparison studies are entered into a data base and evaluated by using non-parametric statistics. Data for each SRS constituent are displayed in ascending order (by method) and the median (midpoint of all the values) is considered the most probable value (MPV). The midpoint of the values greater than the median is called the upper hinge (Uh); the midpoint of the values less than the median is called the lower hinge (Lh). The upper and lower hinges are similar to quartiles but are not mathematically equivalent. The range of data between Uh and Lh, which includes the middle 50 percent of the data, is called the hinge spread and is used to calculate the sample data deviation or F-pseudosigma (Fps = (Uh-Lh)/1.349). The Fps is equivalent to the sample standard deviation (parametric statistic) when the sample data have a normal statistical distribution. Non-parametric statistics are best used when the sample data include outlier values (Hoaglin and others, 1983). A report is prepared for each SRSP inter-laboratory evaluation and includes graphs and tables that summarize laboratory performance. For each SRS constituent, the following type of graphical plot is prepared:


The upper and lower boundaries in the above graph at 6.68 and 8.46 milligrams per liter (mg/L) are at +/- 3 F-pseudosigmas from the MPV (7.57 mg/L). The MPV (solid horizontal line) and the z-values (number of Fps the value is from the MPV) are determined from the constituent-pooled data.

REFERENCES CITED

 

For additional information, contact:
USGS Quality Systems Branch
Standard Reference Sample Project
Denver Federal Center, Bldg. 95, MS401
P.O. Box 25046
Denver, CO  80225

 

USGS banner