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.

Code:

midTroposphere = False; (* True/False *)

whereString = If[midTroposphere, "TMT", "TLT"];
a = Import[
"http://www.remss.com/data/msu/monthly_time_series/RSS_Monthly_MSU_\
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]