One thing I’ve learned in 30 years working with real climate and hydrological data is to always check the maximum available temporal resolution¹. If there’s daily data, don’t just look at monthly averages. If you’ve got monthlies, don’t just rely on annual data. Doing so destroys information, even when the highest resolution data might include excess measurement noise (which is greatly reduced by temporal averaging). Apart from anything else, whatever “clumping” interval you choose is always going to be arbitrary. (What, after all, is so special about a year starting 1 January? Why not one starting on your birthday…)
All of the global temperature estimate series are prepared at monthly resolution, but how many plots of monthly data have you seen? Exactly one, I’d suggest — the one above.
There’s been much recent discussion about a so-called “pause” or “hiatus” in global warming. Look at the chart. See any pause? Plot it monthly, and the claimed pause disappears into the noise. There just isn’t any.
As of now, we’re actually right on the near half-century long trend line.
1. I don’t necessarily mean analyse the data at maximum resolution, but at least look. There can be good reasons not to apply some of the simpler analysis techniques at high temporal resolution (particularly if there is strong autocorrelation).
2. The plot shows monthly global temperature anomalies since 1970 from seven sources spanning three completely different approaches. Some of those are arguably better / more applicable than others, but they don’t actually differ all that much over this interval. The trend shown is the linear regression of the average of the seven monthly estimates against time. The band shown is plus/minus three standard deviations of the residuals.
3. Data sources are discussed at length on the graph page
4. The full record plot is here