Thursday, June 09, 2011 ... Deutsch/Español/Related posts from blogosphere

RSS AMSU: all cooling and warming trends through May 2011

There is still a lot of confusion and misinformation concerning the question whether there has been a cooling trend or a warming trend between some moment XY and the most recent months.

To lift the confusion, I (and Wolfram Mathematica 7.0.1) have calculated 3500 linear regressions and the result is offered to you in this truly tall PDF file:

RSS AMSU trends in °C per century
The file above allows you to pick any month between January 1979 and April 2011 as the starting month. The last month is May 2011. The table offers you all trends - calculated accurately - expressed in Celsius degrees per century. The line for May 2011 contains the actual most recent temperature anomalies.

In the table, you find not only the trends of the global temperature but also the trends of the regional temperatures observed by RSS AMSU, a satellite team. Apologies that I picked RSS - Roy Spencer et al. are doing a great job but I just want to avoid possible accusations that I have picked the skeptics and the most cooling dataset etc.

Let me just select two lines among the nearly 400 lines that the table offers you. The warming or cooling trends between January 1979 and May 2011 have been
  • +1.43 °C / century: globally
  • +1.32 °C / century: tropics
  • +2.23 °C / century: North extratropics
  • +0.69 °C / century: South extratropics
  • +3.40 °C / century: Arctic
  • -0.19 °C / century: Antarctica
  • +1.63 °C / century: contin. USA
  • +1.90 °C / century: North Hemisphere
  • +0.93 °C / century: South Hemisphere
You see it's been mostly warming in the 31+-year period. However, let's write the same numbers with January 2001 - the beginning of the new century - as the initial month. Note that we're not trying to include the El Nino year 1998: instead, we just pick the most natural beginning of the centtury. It's been more than 10 years and the linear regression in this period gives us:
  • -0.40 °C / century: globally
  • -1.16 °C / century: tropics
  • +0.22 °C / century: North extratropics
  • -0.19 °C / century: South extratropics
  • +3.83 °C / century: Arctic
  • -1.27 °C / century: Antarctica
  • -4.84 °C / century: contin. USA
  • -0.23 °C / century: North Hemisphere
  • -0.58 °C / century: South Hemisphere
These are decidedly different numbers! The globe has been slightly cooling since January 2001 although the trend has been less than half a degree per century. The tropics saw more than one degree of cooling per century. The Arctic has seen a significant continued warming by four degrees per century, the Antarctica experienced a cooling by a degree per century.

The biggest figure (when it comes to the absolute value) that you may see anywhere in the tables above is the trend since 2001 in the United States of America. The U.S. has been cooling by nearly 5 Celsius degrees per century since 2001. The global trend stays negative if you pick any month of 2001, 2002, or 2003 (except for December 2003) as the initial month. Quite generally, negative and positive trends are pretty much equally represented in the recent part of the table.

Obviously, you can't or you shouldn't extrapolate any of the figures. The shorter periods we investigate, the bigger trends we typically obtain. For example, the global cooling trend since May 2010 has been cooling by sixty degrees Celsius. ;-) But even when you look at the 31+-year trends, they show a huge variability. The "same" portions of the Northern and Southern Hemisphere display very different warming or cooling trends which imply that the (inevitable) changes are not "global" in any nontrivial sense.

midTroposphere = False; (* True/False *)

whereString = If[midTroposphere, "TMT", "TLT"];
a = Import[
AMSU_Channel_" <> whereString <>
"_Anomalies_Land_and_Ocean_v03_3.txt", "Table"]; Length[a]

labels = {"year", "month", "-70/ +82.5", "-20/ +20.0", "+20.0/ +82.5", "-70/ -20.0", "+60.0/ +82.5", "-70/ -60.0", "Cont. USA",
"0.0/ +82.5", "-70/ 0.0"}

b = a[[4 ;;]]; bwith = Prepend[b, labels]; Grid[bwith, Frame -> All]

(* Linear trends in bwith *)

LMfit[v_] :=
LinearModelFit[Transpose[{Table[i, {i, 1, Length[v]}], v}], x, x];

howmanyrows = Length[bwith] - 1
btrendsPREP = Table[
Round[1200 * D[Normal[LMfit[bwith[[m ;; howmanyrows + 1, column]]]], x], 0.01] , {m, 2, howmanyrows}, {column, 3, 11}];
btrends = bwith;
btrends[[2 ;; howmanyrows, 3 ;; 11]] = btrendsPREP;

Grid[btrends, Frame -> All]

Add to Digg this Add to reddit

snail feedback (4) :

reader Boonton said...

I'm not quite getting the point here. You have 31 yrs of data and that regression shows warming. Yet you 'discover' you can achieve cooling by selecting smaller data sets (such as the last ten years).

What you're missing is a confidence interval. With confidence intervals, I think you would discover that a positive warming trend based on 31 years of data will produce a tighter interval with both the high and low range positive. When you choose smaller sets of data, like the last ten years, you'll end up with a larger confidence interval that will very likely include both positive and negative values in your range.

What this means is that your larger data set would indicate a high degree of confidence in warming while your small data sets will more likely produce cooling estimates that you cannot be confident in.

reader Luboš Motl said...

Well, I surely agree. You may derive a smaller error margin of the slopes measured over long times. The trends coming from the longer time scales are generally slower, too. But this fact is true regardless of the existence of a "warming signal".

However, it's also true that the data unequivocally show that in 10-15 years, the warming trend is undetectable even with the finest techniques. It means even if there were a "lasting trend", it wouldn't matter if one waits for 10-15 extra years. The trend, whatever its sign is, is negligible.

Quite generally, the statement that you're not getting the point is insane. The point was exactly what the table says - to list the trends for any interval whose end point is May 2010. The data matter. Maybe for you, only "points" and religious interpretations that are independent of the data actually matter. But for me, the religious interpretations are just rubbish and what matters *are* the data. That's the point.

Many things can be extracted from the tables, including the fact that the trend since 2001 has been in the cooling direction, so the models and statements that would predict warming "in each decades" are dead. It means many other things that can be directly extracted from the data. If someone wants to hide his head into the sand as an ostrich, it's a different issue.


reader cary.boyce said...

2000 is the natural start of the millennium. 2001 is the mathematical start. :-)

reader Amr said...

I have the pleasure to brief you on our Data Visualization software
"Trend Compass".

TC is a new concept in viewing statistics and trends in an animated
way by displaying in one chart 5 axis (X, Y, Time, Bubble size &
Bubble color) instead of just the traditional X and Y axis. Discover
trends hidden in spreadsheets. It could be used in analysis,
research, presentation etc. In different business sectors, to name a
few we have Deutsche Bank, NBC Universal, RIM, Vanguard
Institutional Investor, Ipsos, Princeton University as our clients.

NBC presentation on TED using Trend Compass exported Videos. Now on
CNN Money / Fortune (Dec 21st,2010):

The following link is for our new Geographical Trend Compass
(Earthquake in Japan - Mag vs Depth vs Time):

Link on OPEC oil indicators:

Link on Ads Monitoring on TV Satellite Channels during April 2008.
Pick Duration (Ads daily duration) vs Repeat (Ads repetition per day).

Link on UK Master Card vs Visa performance :

Links on Funds:

Link on other KPIs :

Link on Chile's Earthquake (Feb 27th 2010):

This a link on weather data :

This is a bank link to compare Deposits, Withdrawals and numbers of
Customers for different branches over time ( all in 1 Chart) :

Misc Examples :

This is a project we did with Princeton University on US unemployment :

A 3 minutes video presentation of above by Professor Alan Krueger
Bendheim Professor of Economics and Public Affairs at Princeton
University and currently Chief Economist at the US Treasury using
Trend Compass :

You can download a trial version. It has a feature to export
EXE,PPS,HTML and AVI files. The most impressive is the AVI since you
can record Audio/Video for the charts you create.

All the best.

Amr Salah
Trend Compass Team
Epic Systems

(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) })(window,document,'script','//','ga'); ga('create', 'UA-1828728-1', 'auto'); ga('send', 'pageview');