The reasoning behind the conclusion is worth taking a little time to digest as it demonstrates that temperature statistics are often open to interpretation.
Consider the GISS Loti database on which September has a temperature anomaly of 100 (divide by 100 to get changes in degrees C.)
In its 140-year range of monthly data, GISS Loti has only 26 months with an anomaly of 100 or greater of which, with the exception of January 2007, all occur within the past five years.
The most any month has above 100 is four. Some months, June, July have none. May, August, and September have one (102, 102, and 100 respectively). The high-temperature anomaly values occur in the southern hemisphere summer months and show the warming of the South Pacific.
In GISS Loti, the past two Octobers have had anomalies of 102, which to me suggests that the September 2020 data is probably a boundary effect. Nature knows nothing of the ancient decision to have 12 months in a year!
September 2020 is very warm certainly, but if you use it as has been done to infer a global effect then you are using the data to mislead.
In September, Europe, northern Russia, and Siberia certainly had temperatures above the 10-year average. However, near the equator, the temperatures were near or below this average, and it was the same for the southern hemisphere.
The NOAA analysis of this September indicates that overall about 8.49% of the world’s land and ocean surfaces experienced a record September temperature – the second-highest percentage behind 2015.
However, taking the temperature values into account, they say that September globally was the warmest September on record.
This is not a definitive statement if you consider the errors. NOAA says that September 2020 surpassed September 2015 and 2016 by 0.02°C, which is insignificant given the errors are plus or minus about 0.1°C.
Why is it that an organization like NOAA, with its global role concerning climate change, continues to treat numbers in a way that would fail a physics 101 class?
Atmospheric data also shows September to not be among the warmest though again we have the common presentation of data with no errors.
There is another problem with NOAA and numbers. Looking at their 2020 year-to-date temperatures (again no error bars) are shown above. It can be seen that September is the second warmest September on record. Looking at this graph, I will not make any bets as to 2020 being the warmest year ever.
Looking at global maps shows that warming is confined to specific regions, especially northern Siberia. Consider this map of September temperature changes of the past 10 years. It shows that September temperature changes are not global but regional.
Read more at GWPF