Merged Recent climate observations disagreement with projections

Wangler said:
So what you do is average over the smallest interval that makes sense, so you can have your cake and eat it too.

Where does the 30-year number come from?

I hear "30 years is climate" and "less than 30 years is weather"....

but who decides? What is the basis? Anyone have a reference?
Click to expand...

30 years is the consensus view. It's not judgement handed down. 30 years spans a few solar cycles and slightly more ENSO cycles. A period long enough to take out significant amounts of noise.
 
lomiller said:
What mhaze really really wants to do is look at shorter trend and draw conclusions from it. If one looks at the last 10 years you get a warming trend of ~0.05 deg/dec, but the error bars are +/-0.3 deg. This time period simply cannot tell you if the long term trend of ~0.2 deg/dec has stopped, the natural variation is still to large.
Click to expand...

I think the deniers will remember this decade as their Glory Days. A fortuitous decade with no marked surface warming trend, quite in line with model projections. They've nailed their flag so securely to this happy chance that they've no chance of recovery.

Time is not on their side.
 
Wangler said:
No, aren't we talking about a 30 year average, like thus (for example):

point 1: average of 1951-1980
point 2: average of 1952-1981
point 3: average of 1953-1982
...
...
Click to expand...

Not really. None of those three are very independent.

A proper series of data points in this context would be

point 1 - trend 1951-1980
point 2 - trend 1981-2010
 
Last edited:
Geckko said:
Not really. None of those three are very independent.

A proper series of data points in this context would be

point 1 - trend 1951-1980
point 2 - trend 1981-2010
Click to expand...

NO.

Sliding windowed running averages are standard for analyzing such things in just about any field.

There is no "trend" 1951-1980. Just an average. One point on that graph. It becomes a trend when you add the window before and the window after. It becomes a useful trend when you add a few dozen of such windows to the graph. An average such as this, of course, is millions of points of data, so any periodic variation smooths out.
 
Last edited:
BenBurch said:
NO.

Sliding windowed running averages are standard for analyzing such things in just about any field.

There is no "trend" 1951-1980. Just an average. One point on that graph. It becomes a trend when you add the window before and the window after. It becomes a useful trend when you add a few dozen of such windows to the graph. An average such as this, of course, is millions of points of data, so any periodic variation smooths out.
Click to expand...


I would suggest that nothing is "standard", rather you should just use approriate statistics.

You could indeed use rolling average trends to produce a lot more "data points", but you need to analyse them on the basis that you have greatly reduced degrees of freedom.

In effect, a measured trend from 1952-1981 tells you very little more than a measured trend between 1951-1980.
 
Last edited:
Geckko said:
I would suggest that nothing is "standard", rather you should just use approriate statistics.

You could indeed use rolling average trends to produce a lot more "data points", but you need to analyse them on the basis that you have greatly reduced degrees of freedom.

In effect, a measured trend from 1952-1981 tells you very little more than a measured trend between 1951-1980.
Click to expand...
Why bother, when leading climate scientists pick a group of proxy temperature measurements out of hundreds, such that it supports their preconceived notions?
 
Geckko said:
I would suggest that nothing is "standard", rather you should just use approriate statistics.

You could indeed use rolling average trends to produce a lot more "data points", but you need to analyse them on the basis that you have greatly reduced degrees of freedom.

In effect, a measured trend from 1952-1981 tells you very little more than a measured trend between 1951-1980.
Click to expand...

Sorry, just not how science is done. I've done similar work in characterizing huge sensor arrays doing real science. Analysis is all about carefully and intelligently reducing degrees of freedom. Otherwise all you see is noise,
 
Geckko said:
I would suggest that nothing is "standard", rather you should just use approriate statistics.

You could indeed use rolling average trends to produce a lot more "data points", but you need to analyse them on the basis that you have greatly reduced degrees of freedom.

In effect, a measured trend from 1952-1981 tells you very little more than a measured trend between 1951-1980.
Click to expand...

What's a "rolling average trend"?
 
You must log in or register to reply here.

Back
Top Bottom