Assessing reanalysis quality with early sounders Nimbus-4 IRIS (1970) and Nimbus-6 HIRS (1975)

Abstract This paper revisits the data collected by early sounders Nimbus-4 IRIS (1970) and Nimbus-6 HIRS (1975), after recovery of ageing tapes by NASA GES DISC. New quality controls are proposed to screen out erroneous or suspicious mission data, based on instrument health status data records and other inspection of the data. Radiative transfer coefficients are derived for the fast computation of clear-sky radiative transfer simulations. Atmospheric profiles from ERA-40 and ERA-20C reanalyses are used in input. These spatio-temporally complete datasets are interpolated to each sounding location, using the closest estimate in time. A modern cloud detection method derived for current hyperspectral sounders is applied to IRIS and yields maps of cloud cover that are in line with current knowledge of cloud climatology. For clear scenes, the standard deviation of brightness temperature differences between IRIS observations and simulations from ERA-20C is around 1 K for the lower-peaking temperature channels of the 15 μm CO 2 band, and lower than 1 K for simulations from ERA-40. The IRIS and HIRS instrumental data records are projected in a common sub-space to alleviate issues with different field-of-view resolutions and spectral resolutions. A proxy cloud detection scheme screens out clouds in the same manner in both data records. Considering the month of August, common to both missions, a detailed analysis of the departures from observations suggests that ERA-40 suffers from spurious tropospheric warming, possibly caused by changes in the observation input during the 1970s including a known error in ERA-40 radiance assimilation bias correction. This result, confirmed by considering a climate model integration, demonstrates that it is possible to exploit early sounder data records to derive detailed insight from reanalyses, such as attempting to qualify separately random and systematic errors in reanalyses, even at times when few other independent observation data are available.