Common use of DETECTION LIMITS Clause in Contracts

DETECTION LIMITS. As described by ▇▇▇▇ (2005), the instrument response to particle-free air (a “dynamic zero” test) was routinely measured by placing a Teflon filter mounted in a Teflon filter housing in the sample line immediately upstream of the denuder. Dynamic zero testing was performed during the bimonthly (twice a month) maintenance with at least three 10-minute cycles run with the Teflon filter installed. Standard deviations for the study-mean dynamic zero tests were in the range 0.2-0.3 μg/m3 for the three instruments deployed over the period February 2002 through July 2005, corresponding to an MDL of 0.6-0.9 μg/m3. For semicontinuous instruments, dynamic zero tests have been used both to estimate method detection limits (from the standard deviation of the replicated measurements) and estimate an offset to be applied to the data. For the R&P 8400N, however, the interpretation of dynamic zero tests is not clear. ▇▇▇▇▇▇▇▇ et al. (2004) conducted dynamic zero tests with a HEPA filter at the inlet of an 8400N. They postulate that nitric acid and ammonia were likely adsorbed by the HEPA filter and the observed response resulted from these precursor gases desorbing from the denuder into the ammonia- and nitric acid-free sample stream (presumably to form ammonium nitrate). Thus, in terms of the ambient PM2.5 data, the dynamic zero response might provide insights into a positive bias from sharp decreases in ambient concentrations of the precursor gases which would promote denuder off gassing. This scenario assumes that ammonium nitrate formation from denuder off gassing is both thermodynamically and kinetically favorable. While this is a plausible interpretation of dynamic zero testing, more work is needed to determine whether this is correct. In the interim, the estimated MDL of 0.6-0.9 μg/m3 should be used with caution. For example, this may be an upper bound since the dynamic zero response exhibited seasonal behavior (▇▇▇▇, 2005) and the standard deviations used to estimate the MDL include such variation Precision. Six weeks of collocated data was collected in January-February 2005 to quantify the collocated precision in the measurements (Figure 4-2). The units showed excellent agreement, and the scatter provides insights into the collocated precision of the 8400N measurements at hourly resolution. ▇▇▇▇▇▇’▇ regression was performed on the data using on the data using record-specific uncertainties of the form σi = a + b×Ci, with a = 0.30 μg/m3 and b = 0.05; Ci is the hourly-averaged nitrate concentration in μg/m3. The regression coefficients are reported in Figure 4-2.