Revisiting radiosonde upper air temperatures from 1958 to 2002

Abstract

HadAT is a new analysis of the global upper air temperature record from 1958 to 2002 based upon radiosonde data alone. This analysis makes use of a greater number of stations than previous radiosonde analyses, combining a number of digital data sources. Neighbor buddy checks are applied to ensure that both spatial and temporal consistency are maintained. A framework of previously quality controlled stations is used to define the initial station network to minimize the effects of any pervasive biases in the raw data upon the adjustments. The analysis is subsequently expanded to consider all remaining available long‐term records. The final data set consists of 676 radiosonde stations, with a bias toward continental Northern Hemisphere midlatitudes. Temperature anomaly time series are provided on 9 mandatory reporting pressure levels from 850 to 30 hPa. The effects of sampling and adjustment uncertainty are calculated at all scales from the station series to the global mean and from seasonal to multidecadal. These estimates are solely parametric uncertainty, given our methodological choices, and not structural uncertainty which relates to sensitivity to choice of approach. An initial analysis of HadAT does not fundamentally alter our understanding of long‐term changes in upper air temperature changes.

Internal variability in simulated and observed tropical tropospheric temperature trends

“…We use the most recent versions of the following six radiosonde-based compilations of tropospheric temperatures: RATPAC-A [Free et al, 2005], HadAT2 [Thorne et al, 2005], RAOBCORE [Haimberger, 2007;Haimberger et al, 2012], RICH- [Haimberger, 2007;Haimberger et al, 2012], RICH-obs [Haimberger, 2007;Haimberger et al, 2012], and IUKv2 [Sherwood and Nishant, 2015]. The data set HadCRUT4.5 [Morice et al, 2012;Osborn and Jones, 2014] is used as observational estimate for surface temperatures.…”

Recent slowdown of tropical upper tropospheric warming associated with Pacific climate variability

“…The difference in the T TUT trends between the observations and models cannot be assessed easily because uncertainties in the T TUT trends based on the radiosonde and satellite observations and GCM simulations [Thorne et al, 2005;Sherwood et al, 2005;Randel and Wu, 2006;Christy et al, 2007;Haimberger et al, 2008;Allen and Sherwood, 2008] are large compared with the difference between them [Santer et al, 2008;Thorne et al, 2011;Seidel et al, 2012]. However, the latest observational data suggest that the models have systematic warming biases in recent (since the late 1990s) T TUT variation ( Figure S1; see section 3.1) [Fu et al, 2011, Figure 3;Lott et al, 2013, Figure 4].…”

How do we tell which estimates of past climate change are correct?

“…Efforts to quantify atmospheric warming from inhomogeneous radiosonde data, in particular, have nearly all done this (Lanzante et al, 2003;Thorne et al, 2005b;Haimberger, 2007) and at least one recent comparison of two efforts to homogenize satellite temperatures has picked a winner, in part, on the basis of such a metric (Christy and Norris, 2006). It is our hope that the present discussion will bring about a reassessment of such assumptions of transferability between different measures of success.…”

“…Recently, to test the robustness of the procedure used previously by Thorne et al (2005b) with neighbour-based composites, thereby identifying and adjusting breaks. It therefore tends to minimize the RMS differences between station series that are relatively close together.…”