WGI AR5 Fig2-23

Figure 2.23 Vertical weighting functions for those satellite temperature retrievals discussed in this chapter (modified from Seidel et al. (2011)). The dashed line indicates the typical maximum altitude achieved in the historical radiosonde record. The three SSU channels are denoted by the designated names 25, 26 and 27. LS (Lower Stratosphere) and MT (Mid Troposphere) are two direct MSU measures and LT (Lower Troposphere) and *G (Global Troposphere) are derived quantities from one or more of these that attempt to remove the stratospheric component from MT.

The major global radiosonde records extend back to 1958, with temperatures, measured as the balloon ascends, reported at mandatory pressure levels. Satellites have monitored tropospheric and lower stratospheric temperature trends since late 1978 through the Microwave Sounding Unit (MSU) and its follow-on Advanced Microwave Sounding Unit (AMSU) since 1998. These measures of upwelling radiation represent bulk (volume averaged) atmospheric temperature (Figure 2.23). The ‘Mid-Tropospheric’ (MT) MSU channel that most directly corresponds to the troposphere has 10 to 15% of its signal from both the skin temperature of the Earth’s surface and the stratosphere. Two alternative approaches have been suggested for removing the stratospheric component based on differencing of view angles (LT) and statistical recombination (*G) with the ‘Lower Stratosphere’ (LS) channel (Spencer and Christy, 1992; Fu et al., 2004). The MSU satellite series also included a Stratospheric Sounding Unit (SSU) that measured at higher altitudes (Seidel et al., 2011).

At the time of AR4 there were only two ‘global’ radiosonde data sets that included treatment of homogeneity issues: RATPAC (Free et al., 2005) and HadAT (Thorne et al., 2005). Three additional estimates have appeared since AR4 based on novel and distinct approaches. A group at the University of Vienna have produced RAOBCORE and RICH (Haimberger, 2007; Haimberger et al., 2008, 2012) using ERA reanalysis products (Box 2.3). Sherwood and colleagues developed an iterative universal kriging approach for radiosonde data to create IUK (Sherwood et al., 2008) and concluded that non-climatic data issues leading to spurious cooling remained in the deep tropics even after homogenization. The HadAT group created an automated version, undertook systematic experimentation and concluded that the parametric uncertainty (Box 2.1) was of the same order of magnitude as the apparent climate signal (McCarthy et al., 2008; Titchner et al., 2009; Thorne et al., 2011). A similar ensemble approach has also been applied to the RICH product (Haimberger et al., 2012). These various ensembles and new products exhibit more tropospheric warming / less stratospheric cooling than pre-existing products at all levels. Globally the radiosonde records all imply the troposphere has warmed and the stratosphere cooled since 1958 but with uncertainty that grows with height and is much greater outside the better-sampled NH extra-tropics (Thorne et al., 2011; Haimberger et al., 2012), where it is of the order 0.1°C per decade.

For MSU, AR4 considered estimates produced from three groups: UAH (University of Alabama in Huntsville); RSS (Remote Sensing Systems) and VG2 (now no longer updated). A new product has been created by NOAA labelled STAR, using a fundamentally distinct approach for the critical inter-satellite warm target calibration step (Zou et al., 2006a). STAR exhibits more warming/less cooling at all levels than UAH and RSS. For MT and LS, Zou and Wang (2010) concluded that this does not relate primarily to use of their inter-satellite calibration technique but rather differences in other processing steps. RSS also produced a parametric uncertainty ensemble (Box 2.1) employing a Monte Carlo approach allowing methodological inter-dependencies to be fully expressed (Mears et al., 2011). For large-scale trends dominant effects were inter-satellite offset determinations and, for tropospheric channels, diurnal drift. Uncertainties were concluded to be of the order 0.1°C per decade at the global mean for both tropospheric channels (where it is of comparable magnitude to the long-term trends) and the stratospheric channel.

SSU provides the only long-term near-global temperature data above the lower stratosphere, with the series terminating in 2006. Some AMSU-A channels have replaced this capability and efforts to understand the effect of changed measurement properties have been undertaken (Kobayashi et al., 2009). Until recently only one SSU data set existed (Nash and Edge, 1989), updated by Randel et al. (2009). Liu and Weng (2009) have produced an intermediate analysis for Channels 25 and 26 (but not Channel 27). Wang et al. (2012g), building on insights from several of these recent studies, have produced a more complete analysis. Differences between the independent estimates are much larger than differences between MSU records or radiosonde records at lower levels, with substantial inter-decadal time series behaviour departures, zonal trend structure, and global trend differences of the order 0.5°C per decade (Seidel et al., 2011; Thompson et al., 2012; Wang et al., 2012g). Although all SSU data sets agree that the stratosphere is cooling, there is therefore low confidence in the details above the lower stratosphere.

In summary, many new data sets have been produced since AR4 from radiosondes and satellites with renewed interest in satellite measurements above the lower stratosphere. Several studies have attempted to quantify the parametric uncertainty (Box 2.1) more rigorously. These various data sets and analyses have served to highlight the degree of uncertainty in the data and derived products.