The original calibration data of the 47 thermometer was measured
at a relatively low precision, lower than is typically acceptable
today. As a result, efforts to reduce the uncertainties in
temperatures derived from clumped isotopes are limited by a relatively
large calibration error. This may be overcome by reducing
the calibration uncertainties through additional calibration data
points that are analyzed at higher analytical precision. Our laboratory
precipitation experiments were designed to provide such data,
by following the same method of Ghosh et al. (2006a), while using
a wider temperature range (5–65 ◦C), a more even temperature
distribution, and measuring at higher analytical precision (through
longer count time and triplicate analyses). The new data set has a
47–T relationship that is similar, but has a slightly lower slope,
than the original thermometer calibration, with a slope and associated
1 SE uncertainty of 0.0526 ± 0.0025 and an intercept of
0.0520 ± 0.0284. For comparison, the calibration using only the
Ghosh et al. (2006a) data gives a slope and associated 1 SE of
0.0597 ± 0.0083 and an intercept of −0.0239 ± 0.0925. Note that
these values are slightly different than those reported by Ghosh et
al. (2006a) as they reflect an error weighted regression to the original
data and no data averaging as was done in the original study.