Difference between revisions of "HadCET CUSUM analysis"
(Created page with "CUSUM analysis - a statistical technique borrowed from the quality control discipline - is deceptively simple to calculate by spreadsheet or even by hand yet a powerful tool f...") |
m (Include New Scientist article) |
||
| (One intermediate revision by the same user not shown) | |||
| Line 6: | Line 6: | ||
# and so on | # and so on | ||
# final value = 0 ''by definition of average'' must be so | # final value = 0 ''by definition of average'' must be so | ||
| − | The thing to look for is abrupt kinks in the graph, which shout 'something changed here' | + | The thing to look for is abrupt kinks in the graph, which shout 'something changed here'. The chart below - original work by FLUKE - appears to show four distinct trend periods: 1772 to 1892, 1892 to 1932, 1932 to 1988 and 1988 to present |
| + | |||
| + | It is interesting to compare these results - which are purely a result of statistical analysis of temperature at a single location - with the broader world-wide climate trends over the last 120 years. For example the [https://www.newscientist.com/article/dn9912-timeline-climate-change/ New Scientist climate-change timeline] which identifies the period from 1940 to 1970 as one of global ''cooling'' which interrupted the earlier 'dust bowl' epoch and the present day heating periods | ||
| + | |||
| + | [[File:Hadcet3.png|800px]] | ||
Latest revision as of 11:22, 10 August 2022
CUSUM analysis - a statistical technique borrowed from the quality control discipline - is deceptively simple to calculate by spreadsheet or even by hand yet a powerful tool for screening out 'noise' and spotting trend changes
- calculate the deviations from mean average
- first value = 0
- second value = first value + first deviation
- third value = second value + second deviation
- and so on
- final value = 0 by definition of average must be so
The thing to look for is abrupt kinks in the graph, which shout 'something changed here'. The chart below - original work by FLUKE - appears to show four distinct trend periods: 1772 to 1892, 1892 to 1932, 1932 to 1988 and 1988 to present
It is interesting to compare these results - which are purely a result of statistical analysis of temperature at a single location - with the broader world-wide climate trends over the last 120 years. For example the New Scientist climate-change timeline which identifies the period from 1940 to 1970 as one of global cooling which interrupted the earlier 'dust bowl' epoch and the present day heating periods
