Is Global Warming Accelerating?

There's a graph circulating on X created by Javier Vinós that is being used to suggest that global warming rates are actually decreasing. The origin of this graph comes from a WUWT blogpost, and it superficially seems convincing. The relevant graph is labeled "Figure 2," and the caption claims, "Evolution of the warming rate for 15-year periods between 1979 and 2022 in °C/decade and its linear trend, from monthly UAH 6.0 satellite temperature data."

Javier Vinós Thinks Global Warming Rates are Decreasing

Javier Vinós was kind enough to explain how he made this graph: "To analyze the evolution of the warming rate, we subtract from each monthly data the previous one to calculate the monthly increase. We then deseasonalize the monthly increase by finding the 12-month moving average to remove a lot of the noise. Finally, we calculate the 15-year average warming rate in °C/decade by calculating the 180-month moving average and multiplying the resulting data by 120." Below I duplicated this process to see if I came up with the same results, and I did with a surprise coming from the data from after 2022 (see the last graph below). However, I can confidently say that Vinós' procedure is unnecessarily complex and in fact, wrong. To calculate 15-year warming rates from UAH-TLT monthly data, you calculate the slopes for successive 180-month (15yr*12mo/yr) increments, and this will give you the trend in °C/yr; you can multiply this value by 10 to get trends in °C/decade. If you do this correctly, you get a graph that looks like this. I'm including here both RSS and UAH satellite TLT data.
15-year Trends from UAH And RSS TLT Data

As you can clearly see, the overall warming rate is increasing in both datasets, and the highest warming rates occur in recent months. The relatively large differences between UAH and RSS in the middle years of this dataset have mostly to do with the way each organization handled satellite issues between 1999 and 2005 or so. 

To be fair, there are two major weaknesses of calculating 15-year trends like this due to the short time frame involved: 1) the results are usually not statistically significant and 2) trends are more influenced by weather, especially ENSO, so they are less reflective of actual climate changes. To minimize these issues, it's much better to use 30-year trends. If you do, you get a graph that looks like this.
30-year Trends from UAH and RSS TLT Data

This shows statistically significant trends, and it clearly shows that warming rates are generally increasing. There is less variability in trend by using 30-years instead of 15-years, so this is more reflective of how climate is changing (because this is showing 30-year trends, the start date has to be about 15 years more recent than the previous graph). 
Comparing Correct (UAH-TLT in blue) and Vinós' (Vinos-UAH in red) 15-Year Trends

Incidentally, above I show what happens when you duplicate Vinós' method and include data since the end of 2022. Even with Vinós' method, there has been acceleration in warming. My guess is Vinós will never update this graph to include the warming of 2023-2024. Nevertheless, Javier Vinós' graph is flawed on multiple levels. If you calculate trends correctly, warming is accelerating in both UAH and RSS datasets, and this holds true even when you use a climate-relevant timescale for trends.





Comments

Popular posts from this blog

The Marketing of Alt-Data at Temperature.Global

Roy Spencer on Models and Observations

Patrick Frank Publishes on Errors Again