Another dot on the graph

So last week was the annual release of the temperature records from NASA, NOAA and Berkeley Earth. The Copernicus ERA5 data was released a few days ago, and the HadCRUT data will follow soon. Unlike in years past, there is no longer any serious discrepancy between the records – which use multiple approaches for the ocean temperatures, the homogenization of the weather stations records, and interpolation.

Depending on the product, 2021 was either the 5th, 6th or 7th warmest year, but in all cases, it is part of the string of warm years (since 2015) that have all been more than 1ºC warmer than the late 19th C.

Controversy is so 2000 and late

Many of the issues that exercised the blogosphere a decade ago have been put to rest. Despite flailing attempts by a couple of diehards to resurrect talk of a ‘new pause’ (no warming since the last record warm year!), and the never-ending insistence of some solar enthusiasts that a dramatic cooling is right around the corner, these are not serious issues. Discussion too of the ‘irrelevance’ of the global mean changes to ‘normal people’ has also faded as the contribution of the overall warming has become more and more obvious in the incidences of extreme heat waves, intense precipitation, coastal flooding and wildfire intensity. Transparency of the data and methods has all increased (though note that for GISTEMP, the code and the publically available data has been available since 2007), but since that came at the same time that as a large increase in the digitized raw data available and the strength of the trends, the importance of the specifics of the coding has diminished. Additionally, we now have much better quantification of the uncertainties in these estimates (see below). The ‘polar hole’ issue is almost completely put to be bed, with HadCRUT5 and the NOAA (Interim) products coming around to interpolation of one sort or another.

So given this increased maturity of these analyses, are there any scientifically interesting issues left? Yes, but they are perhaps a little more subtle than before.

Southern Ocean trends

The least visited part of the ocean are the waters around Antarctica. They are not on many trade routes, scientific expeditions are infrequent, and the Argo buoys have trouble near the sea ice edge. Nonetheless, we have enough information to know that these waters are not warming at the surface as much as the rest of the planet, and indeed, at around 60ºS, some indications seem to suggest that they have cooled in the last few decades. The size of this cooling varies in the records, most of all in the satellite-derived AIRS v7 data, where the cooling is quite pronounced, and not at all in the ERA5 reanalysis.

As you can see the southern oceans are the area of greatest divergence between GISTEMP, ERA5 and the AIRS versions. Both ERA5 and GISTEMP are relying on SST data here (either ERSSTv5 or HadISST2 (I think)), so may suffer from observational data sparsity. But the AIRS data may have systematic issues as well, perhaps associated with changes in surface type from sea ice to open ocean and vice versa. In contrast, CMIP6 model hindcasts/forecasts for this time period don’t show any cooling at all at these latitudes.

Why does this matter? If the GISTEMP or AIRS data is qualitatively correct, then that would point to a systematic difference with the models, suggesting a missing process or forcing. There is a candidate for this, which is the impact of anomalous freshwater from Antarctica (see for instance, Rye et al., 2020). This in turn suggests that Antarctic melt could be an important negative feedback on southern hemisphere warming – which is one of the regions with the most important cloud feedbacks in the high sensitivity CMIP6 models (Zelinka et al., 2020).

We should probably try and sort this out at some point…

Ensembles of uncertainty

As mentioned above, the characterization of the total uncertainty in these products has improved a lot. Uncertainties arise in the raw data directly, in any homogenizations that are performed, in the data interpolation, in corrections for non-climatic effects etc. and getting a total uncertainty that includes almost all of these aspects is now available for HadCRUT, GISTEMP, Berkeley Earth and NOAA timeseries. The results across the products are comparable, though they differ, particularly in the earlier parts of the record.

Nonetheless, there are still improvements that can be made. Notably, standard approaches to the uncertainty are difficult to apply to errors that are correlated over time. For instance, data sparseness only changes slowly so differences between two adjacent years might be less uncertain than if the calculated uncertainty was treated as an independent variable each year. An approach that would do better would be one based on a Monte Carlo ensemble of possible estimates – each one including both systematic and stochastic uncertainties. Then these could be sampled to generate relevant statistics. HadCRUT, ERSST and GHCN have constructed ensembles for this purpose and there are efforts to do this for more products.

UAH vs the world

This is an older issue, but it remains the case that the UAH TLT trends are the outlier amongst all the related datasets. It isn’t as dramatic a difference as in the 1990s when the UAH record suggested that the world had been cooling since 1979, but as you can see below, it stands out. The MSU/AMSU records would also benefit from a Monte Carlo approach to sample the structural uncertainties in their construction. For now the growing difference between the RSS record and the UAH one acts a (unsatisfying) stand-in for the structural uncertainty, but differences this large make it difficult to conclude much from MSU data/model comparisons.

Volcanic wild cards

Too soon to tell (today at least), but volcanic impacts on temperature are well known from the record. Whether Hunga Tonga will affect temperatures depends a lot on how much SO₂ is emitted (which you can track here – the 15 Jan image will be telling). A big eruption will potentially cool the planet for a year or a few and postpone further increases in temperature. More on this soon if indeed it looks to be important.

Updating the model-observation comparisons

This is not really an scientific issue, but we will try and get this done over the next week. We will also add a new comparison to the CMIP6 models… so stay tuned.

References

1. 
   C.D. Rye, J. Marshall, M. Kelley, G. Russell, L.S. Nazarenko, Y. Kostov, G.A. Schmidt, and J. Hansen, “Antarctic Glacial Melt as a Driver of Recent Southern Ocean Climate Trends”, Geophysical Research Letters, vol. 47, 2020. http://dx.doi.org/10.1029/2019GL086892
2. 
   M.D. Zelinka, T.A. Myers, D.T. McCoy, S. Po?Chedley, P.M. Caldwell, P. Ceppi, S.A. Klein, and K.E. Taylor, “Causes of Higher Climate Sensitivity in CMIP6 Models”, Geophysical Research Letters, vol. 47, 2020. http://dx.doi.org/10.1029/2019GL085782
3. 
   N.J.L. Lenssen, G.A. Schmidt, J.E. Hansen, M.J. Menne, A. Persin, R. Ruedy, and D. Zyss, “Improvements in the GISTEMP Uncertainty Model”, Journal of Geophysical Research: Atmospheres, vol. 124, pp. 6307-6326, 2019. http://dx.doi.org/10.1029/2018JD029522