Determination and significance of upper-tropospheric humidity
Abstract. We present a novel retrieval for upper-tropospheric humidity (UTH) from HIRS channel 12 radiances that successfully bridges the wavelength change from 6.7 to 6.5 µm that occurred from HIRS 2 on NOAA 14 to HIRS 3 on NOAA 15. The jump in average brightness temperature (T12) that this change caused (about −7 K) could be fixed with a statistical intercalibration method (Shi and Bates, 2011). Unfortunately, the retrieval of UTHi based on the intercalibrated data was not satisfying at the high tail of the distribution of UTHi. Attempts to construct a better intercalibration in the low T12 range (equivalent to the high UTHi range) were either not successful (Gierens et al., 2018) or required additional statistically determined correctionsto the measured brightness temperatures (Gierens and Eleftheratos, 2017). The new method presented here is based on the original one (Soden and Bretherton, 1993; Stephens et al., 1996; Jackson and Bates, 2001), but it extends linearisations in the formulation of water vapour saturation pressure and in the temperature-dependence of the Planck function to second order. To achieve the second-order formulation we derive the retrieval from the beginning, and we find that the most influential ingredient is the use of different optical constants for the two involved channel wavelengths (6.7 and 6.5 µm). The result of adapting the optical constant is an almost perfect match between UTH data measured by HIRS 2 on NOAA 14 and HIRS 3 on NOAA 15 on 1004 common days of operation. The method is applied to both UTH and UTHi, the upper-tropospheric humidity with respect to ice. For each case retrieval coefficients are derived. We present a number of test applications, e.g. on computed brightness temperatures based on high-resolution radiosonde profiles, on the brightness temperatures measured by the satellites on the mentioned 1004 common days of operation. Further we present time series of the occurrence frequency of high UTHi cases and we show the overall probability distribution of UTHi. The two latter applications expose clear indications of moistening of the upper troposphere over the last 35 years. Finally, we discuss the significance of UTH. We state that UTH algorithms cannot be judged for their correctness or incorrectness, since there is no true UTH. Instead, UTH algorithms should fulfil a number of usefulness-postulates, that we suggest and discuss. In the course of this discussion an alternative method to estimate the weighting function is presented.