STORING DATA IN THE NATIONAL WATER INFORMATION SYSTEM
Establishing the LRL is a continuous process with annual re-determination of the LTMDL. Year-to-year re-determination results in changes to the LTMDL and LRL only if a t-test [Erratum] indicates that the standard deviations used to determine the LTMDL are significantly different at the 95-percent confidence level. These updates, when necessary, are announced by the NWQL at the beginning of each water year (October 1). A significant shift in the LTMDL and LRL of an analyte will be reflected in a change in data reportingeither to a higher or lower LRL and LTMDL. Updated LTMDLs and LRLs will not result in changes to historical LTMDLs and LRLs, or any associated sample data, in the data base.
To fully understand and interpret data reported by using the new reporting procedure, one must know the applicable LRL and the LTMDL in addition to the reported concentration. If the LS exceeds the LRL and the reported value is estimated between the LS and LRL, the concentration of the LS must also be known. The current version of the National Water Information System (NWIS) data base does not have the capability to store this information within the record for a single sample. Until this information can be stored in NWIS, the NWQL will maintain information on established LRLs and LTMDLs for each constituent in published method reports, NWQL technical memoranda, and at the NWQL web page http://wwwnwql.cr.usgs.gov. Under the Technical Information section on this web page, click on Long-Term Method Detection Level and Laboratory Reporting Level Information. Data must be referenced by parameter code, method code, and date of analysis.
Enhancements to Remark Code in the National Water Information System
Until the NWIS data base is revised to include qualification codes, users cannot distinguish between values estimated because of the new reporting procedure and those estimated by use of other analytical considerations, such as matrix interference, extrapolation above the calibration standards, effects of sample dilution, and other conditions (see previous discussion on data reporting). Supplemental coding schemes may be available for future versions of NWIS. The capability to associate qualifiers with remark codes will be added. Qualifiers will identify the specific condition that evoked use of the E remark code. With this information, for example, the user can determine if a value was estimated because of a problem that occurred during sample preparation or if the value is between the LS and the LRL. Until these enhancements are made to NWIS, information on why the datum was given an “E” remark code will reside in analyst notes at the NWQL.
Impact on Historical Data
The significance of the effect of the current change in reporting conventions on interpretations by using current and historical data depends, in part, on which of the following two outcomes applies to the data being retrieved:
For the first outcome, no special action is required. Uses or interpretations that require combining historical and new data may require working with multiple reporting levels as described above. For the second outcome, more caution is needed when historical and current data are combined and interpreted. A new reporting level (LRL) that is substantially higher than the historical MRL indicates that the MRL was most likely set too low. Most historical data that are reported substantially below the new LRL should be considered estimated concentrations at best. For information-rich methods, however, detections are considered accurate even if the LRL is substantially higher than the historical MRL. Non-detections that were reported at a lower MRL (reported as <MRL) might be more appropriately censored to the new, higher LRL (<LRL). When project objectives require a high degree of certainty, users may choose to censor all historical data at the new LRL if, for example, the data will be used to determine compliance with environmental regulations at a particular site. These are actions that must be considered and applied by the user after retrieving historical data from the data base. No changes that are based on new LRLs will be made to data already stored in the data base.