Easy reference to some of the mistakes / improvements I have made.

Errata – 2024 – September

My original version of 1000 Days above 1.5C before Trend passes 1.5C? had a few issues.

• As per section lower down, I have improved my calculations for the copernicus data relative to 1850-1900. This meant that the second version of the graphic has improved numbers, but someone looking closely comparing v1 (March 2024) and v2 (September 2024), they would notice that some of the numbers don’t go up as one might expect.
• The “Climate passes” values were a bit controversial, and used a home-grown non-standard smoothing function I came up with. For v2, I have replaced this with Loess smoothing with a 30 year window. This means the years I have posted for passing various anomalies has changed a bit (not much), but is it more standard.

Errata – 2024 – September – Poor labelling of “Quadratic” / “Linear”

Looking to fix some loose language on the website (more to come).

I labelled some of the trend lines as “Quadratic” or “Linear”, whereas they were actually: “centre-moving-best-fit using quadratic / linear” … E.g. given a data point, take x-years of data from either side, and calculate the best linear / quadratic fit for that date. Then for the next data point start again.

This gave some nice fitting lines, except for the ends, and the first/last points in the trend really were pure linear or pure quadratic best fit. E.g. if the centre-moving part was using 20 years (10 years either side), then start/end were just using 10 years-after (for the first point), and 10-years-before for the last point.

I think I’ll move to Loess smoothing from now on, as that is a standard, and not semi-home-grown.

Errata – 2024 – Summer – Improved data handling

After trying to work out why my Copernicus calculations were often slightly different from that posted by the Copernicus official account (we’re talking < 0.05C level of difference)… I spent ages trying to get my results (which use the copernicus data, and the copernicus methods), to match those published by Copernicus. See rather dull article Copernicus 1850-1900 Baseline – Daily GMST Anomaly

The next time I update the 1000 Days above 1.5C before Trend passes 1.5C? graphic, it will use this improved method. This adjusts some “daily temperature anomalies relative 1850-1900” by under 0.05C, and therefore around some of the numbers from the first version of this graphic, won’t appear to match well with the update. E.g. the number of days over X may not go up as much as maths would suggest, when comparing the graphics.