Measurements of trace species generally become less certain as concentrations decrease. Data analysts need guidance on the ranges in which particular measurements are meaningful. This guidance is normally stated in the form of detection limits. The International Union for Pure and Applied Chemistry (IUPAC) has defined several parameters to characterize measurement detection limits (Currie, L. A. Pure Appl. Chem. 1995, 67, 1699). The published guidelines envision an ability to prepare reference materials with concentrations close to the detection limits using the same methods as for normal samples. For multianalyte methods such as X-ray fluorescence (XRF), multiple reference materials may be required for each analyte to characterize the effects of interferences. The creation and characterization of such complex reference materials atthe detection limits of modern XRF systems represents a considerable technical challenge. This paper describes an observational approach to estimating the detection limits defined by IUPAC. Our empirical approach takes advantage of collocated (duplicate) measurements that are routinely collected by the Interagency Monitoring of Protected Visual Environments (IMPROVE) network and Speciation Trends Network (STN). The analysis is successfully demonstrated by deriving detection limits at the measurement system level for six elements measured on PM2.5 samples by XRF in both networks. The two networks' detection limits are found to be similar in terms of loading (areal density, ng cm(-2)) on the filters as measured by the XRF instruments despite many differences in sample collection, handling, and analysis. IMPROVE detection limits are an order of magnitude lowerthan STN's in terms of atmospheric concentrations (ng m(-3)), because IMPROVE uses smaller fitters and higher flow rates which lead to more concentrated sample deposits.