Interesting Observations & Notes

• Firstly, this has proved to be a super confusing graphic.
  • I advice viewing the graphic multiple times, each time concentrating on a different thing
    • 1st just look at the temperature number in the middle
    • 2nd just look at the temperature slider on the left
    • 3rd just follow the purple line around
      • It appears in January, disappears for a while, then reappears
    • 4th do the same for the other colours
  • Each frame of the video is based on the temperature for that frame.
    • The first frame of the video is the coldest day in the 1940 -> 2024 Copernicus record
    • The middle frame is the warmest temperature projected for July 2050
    • Each coloured wedge shows what day-of-the-year, the global average temperature (shown in the centre, and in the slider) happens.
      • E.g. the global average temperature of 14.5C :
        • happened in May back in pre-industrial times
        • happened in late April in 1990s
        • happened in early April in the 2010s
        • Will happen in late March in the 2030s (If projected warming of 0.25C / decade)
        • Will happen in mide March in the 2050s (If projected warming of 0.25C / decade)
      • Note, currently we are over 0.3C/Decade since 2010. Hopefully that doesn’t continue!
  • Every frame of the video increases the temperature a bit, until it gets to the highest July temperature projected in 2050, then each frame represents cooling temperatures again
  • You can see that by the time global average temperature is above 16.4 … that never happened in 2010s or earlier
• You can really see the shifting seasons, relative to Pre-Industrial
• Assuming 0.25C / Decade is representative of what we will experience almost a full season shift in temperatures by 2050
• 2010 is 4-6 weeks shifted from Pre-Industrial.
• May 2050 is going to be like July 2010
• Jun – Aug 2050 will have temperatures never seen before
• Given that the gap between 1970 and Pre-Industrial is often smaller than the gap between 1970 and 1990, you can really see how things started to ramp up from 1970
• Note
  • The graphic does use significant smoothing, averaging techniques, but I have tried quite a few different techniques and I fairly happy with the effect, and it being representative of reality.
  • This graphic uses Global Mean Surface Temperatures.
    • This means the temperature changes are more muted, as it covers land and ocean, and the whole globe
    • The high / low temperature range is smaller relative to national (land-only) ranges
• I will be looking to create a national / regional version of this graphic

How Diagram was Created

This graphic was inspired by a video on BlueSky showing days of equal temperature. E.g. December Temperatures are similar to February Temperatures. April Temperatures are similar to October temperatures. I thought, it would be interesting to do a climate change version, E.g. given that the earth is warming, then 2020 January temperatures must be equal to some other month in pre-industrial times.

• Get the Copernicus Data set (See GMST Data Sets)
  • This gives me daily Global Mean Surface Temperatures (GMST) for every day from 1940
• Use the Copernicus 1850-1900 Baseline – Daily GMST Anomaly calculations to get the representative GMST temperatures for each day of the year, representing the period 1850-1900
• Choose which years to plot on the graph: Pre-Industrial, 1970, 1990, 2010
• Using the raw days from a single year is way too jumpy, so apply smoothing
  • I chose to to 20 year averaging, and then loess smoothing, and ignore leap years
    • E.g. for 14th March 1970, take the temperatures on 14th March for each and every year 1960 -> 1979 (Inclusive)
      • and average that value, as representing the temperature on 14th March 1970
    • Do the same for every day of the year, to get a 20-year-centre-averaged value for each day in 1970
    • Now I have representative temperature for every day in 1970
  • Apply Loess Smoothing with a very short window (1 month) to do minor smoothing
• Choose to project into the future. I am choosing 0.25C / Decade from 2010
  • 2030 is two decades after 2010, therefore 2 x 0.25C (0.5C) above the 2010 data
  • 2050 is four decades after 2010, therefore 4 x 0.25C (1.0C) above the 2010 data
• Find the lowest and highest temperatures in all the data I’m using (pre-industrial, 1970, 1990, 2010, 2030, 2050)
• Split the range of lowest -> highest temperatures into 180 temperature increments
  • Initially I just did a linear split, so each temperature increment was equal
  • I later moved to a sin(y) * sin(y), where y = 0 -> 90, to make the increments smaller around January and July
• I now have 180 temperature values, covering the lowest temperate to the highest
• For each temperature (from lowest to highest), find the day of the year, where each year was at that temperature
  • E.g. at the lowest temperature we get 17th Jan Pre-Industrial … but no other years got that low
  • Next lowest temperature we get 18th Jan Pre-industrial
  • Etc…
  • By the 29th temperature increment, (11.8C), 18th Jan 1970 and 48th day of Preindustrial
  • … keep going up to peak temperature in mid July, and then go back down the temperature ranges
  • Note that at the higher temperatures, things have reversed so that Only 2050 (projected with 0.25C / Decade) has a daily GMST temperatures above 16.9C.
• Create an image for each of the 360 temperature increments (180 going hotter, 180 returning colder).
  • Now plot a circle representing Jan-Dec and a temperature slider
    • For each image plot what day of the year that temperature happened in Pre-industrial, 1970, 1990, 2010, 2030, 2050 … plenty of the temperature values won’t have a entry for a given year.
  • Also plot a temperature slider, and also a temperature circle in the middle

Data Sanity Checks

This is just a lightweight check, not using any smoothing, so just checking that the numbers are “about right” and correlate fairly well with the diagram.

Using the data daily temperatures from Copernicus (See GMST Data Sets, and Copernicus 1850-1900 Baseline – Daily GMST Anomaly)

Data Sanity Checks – 1970 Data

Data Sanity Checks – 2010 Data

Data Sanity Checks – 2050 Data