Abstract

The Intergovernmental Panel on Climate Change Assessment Report 5 (IPCC AR5, 2013) discussed bulk atmospheric temperatures as indicators of climate variability and change. We examine four satellite datasets producing bulk tropospheric temperatures, based on microwave sounding units (MSUs), all updated since IPCC AR5. All datasets produce high correlations of anomalies versus independent observations from radiosondes (balloons), but differ somewhat in the metric of most interest, the linear trend beginning in 1979. The trend is an indicator of the response of the climate system to rising greenhouse gas concentrations and other forcings, and so is critical to understanding the climate. The satellite results indicate a range of near-global (+0.07 to +0.13°C decade−1) and tropical (+0.08 to +0.17°C decade−1) trends (1979–2016), and suggestions are presented to account for these differences. We show evidence that MSUs on National Oceanic and Atmospheric Administration’s satellites (NOAA-12 and −14, 1990–2001+) contain spurious warming, especially noticeable in three of the four satellite datasets.

Comparisons with radiosonde datasets independently adjusted for inhomogeneities and Reanalyses suggest the actual tropical (20°S-20°N) trend is +0.10 ± 0.03°C decade−1. This tropical result is over a factor of two less than the trend projected from the average of the IPCC climate model simulations for this same period (+0.27°C decade−1).

5. Summary

We performed this intercomparison study so as to document differences among the four microwave satellite temperature datasets of the bulk atmospheric layer known as TMT. While all datasets indicated high levels of agreement with independent data, UAH and RSS tended, in broad terms, to exhibit higher levels of agreement than NOAA and UW. This conclusion does not apply however to the test-statistic of the trend where UAH tended to agree most closely with independent datasets.

One key result here is that substantial evidence exists to show that the processed data from NOAA-12 and −14 (operating in the 1990s) were affected by spurious warming that impacted the four datasets, with UAH the least affected due to its unique merging process. RSS, NOAA and UW show considerably more warming in this period than UAH and more than the US VIZ and Australian radiosondes for the period in which the radiosonde instrumentation did not change. Additionally the same discrepancy was found relative to the composite of all of the radiosondes in the IGRA database, both global and low-latitude. While not definitive, the evidence does support the hypothesis that the processed satellite data of NOAA-12 and −14 are characterized by spurious warming, thus introducing spuriously positive trends in the satellite records. Comparisons with other, independently-constructed datasets (radiosonde and reanalyses) support this hypothesis (Figure 10). Given this result, we estimate the global TMT trend is +0.10 ± 0.03°C decade−1.

The rate of observed warming since 1979 for the tropical atmospheric TMT layer, which we calculate also as +0.10 ± 0.03°C decade−1, is significantly less than the average of that generated by the IPCC AR5 climate model simulations. Because the model trends are on average highly significantly more positive and with a pattern in which their warmest feature appears in the latent-heat release region of the atmosphere, we would hypothesize that a misrepresentation of the basic model physics of the tropical hydrologic cycle (i.e. water vapour, precipitation physics and cloud feedbacks) is a likely candidate.