Do you know what happens when you smooth a graph. ?
Yes, yes, I do understand. It's a technique I've used many times before and will use again.
You simply go from high point to high point and totally are misled by the graph and can't see what is going on.
No, that's not what happens. As I recall from the last time we discussed this plot, the data is smoothed with a centered 13-point boxcar average. The data point plotted for example for July 2019 would average the values for that month with the 6 months prior and subsequent. Each month from January 2019 to January 2020 would be *equally* weighted in the average. Each value from Jan 2019 to Jan 2020 would be added together and then divided by the total number of values (in this case 13).
The 13-month moving average covers the full year, avoiding the unwanted signal from seasonal variation. It is centered, so it avoids a memory bias. The only disadvantage is that the opposite month (January) is included twice in the average (the previous and subsequent ones). A similar alternative would be a 11-month centered value that would avoid the double counting of the opposite month by omitting it completely. One alternative that might be to use a 12-point average, where the opposite month is included by averaging the prior and subsequent values, i.e., in this example Jan. 2019 and Jan. 2020 would be averaged together and then the result averaged with the 11 months inbetween.
It is a mathematical certainty that the average of a group of data points (no matter the weighting) can not be higher than the maximum in set nor smaller than the lowest in the set. It can only be equal to the maximum (or minimum) if all values are the same. Therefore, the average *cannot* just joint the peaks in the finely resolved data; that is impossible.
IMO it's very important that the temperature record dropped to zero above average in 2008 and 2012. This is not seen on the smoothed graph.
One month fluctuations are not relevant, only the long term trends. The data is smoothed by averaging to avoid getting lost in the short-term fluctuations of the base data set. [I should note that since the data is given as monthly values, that data is almost certainly averaged over all of the days within the month to get the monthly values.]