There are very few stations in Antarctica. For that reason it is both very volatile with stations changes and it is very easy to spot where a little change in one spot causes a big shift in “trend”. In this case the v3 data have a much stronger “warming” trend. It largely comes from just a few changes in the 1930-1960 part of the curve. That pulls down that section of the averages ( you can see where the range of the yellow dT line becomes wider, but with a small bias on downward excursions then) and the result is a ‘dip’ at that point. By 1910 the orange line is back at zero (pulled up by the yellow dT having more up excursions and fewer down). The net effect is to create a trend upward from the bulk of the past into the present. If you average all those adjustments would it be a net zero? Would that hide the effect on the overall trend?
Notice, too, that the distant past had a group of readings that were inconveniently unchanged from today. Those are omitted in v3.
It is also remarkably striking how the “Ironed Flat” behaviour of recent data really shouts at you from Antarctica. Look at past volatility. +/- 2 C for a total range of 4 C degrees. Look at recent volatility. Barely 1/2 C range. (Though I note that the v3 line is pulled down vs the v1 line even recently, though by a small amount.)
What this chart says to me is that small bias in data shifts (such a a few more down excursions than up) can move a line of trend significanly; and that having adjustments that are just a little shifted in time can make a big shift in trend, even if the “average adjustment” isn’t obviously biasing. (Look at those older high bias data points that put temps back at the zero line in 1910).
Then there is just the effect of “Differential Volatility”. Most of the trend comes from the effect of that very high volatility in the past having a large weight in pulling down that ‘belly of the curve’ while recent low volatility has the line held closer to the zero line. How much of “Global Warming” is really just an artifact of “higher volatility” near the typical baseline period and low volatility now? It was cold in the ’60s and 70s. Nobody doubts that. So if high volatility locations were used then, and low volatility now (as in California) we build in a trend line of “warming” that is really a trend line of “lower volatility” putting the “now” end of the trend line near zero in recent years.
The final point is just to look at how much “trend” changes. The v1 data are dead flat. Run your eye along that thick blue line. It just hugs zero the whole way. Now look at the v3 line. That line has a clearly stronger rising “trend”. If we can get close to 2 C of “warming trend” out of thermometer selection and processing changes between v1 and v3; then how much of “global warming” is in the method and not in the land itself?
Phrased a bit differently:
Which GHCN is a flat out lie? Is it v1 or v3 that’s fraudulent? They can not both be right.