Guest Post by Willis Eschenbach
The satellite-based atmospheric temperature dataset is one of the better datasets in climate science. Drs. Roy Spencer and John Christy have long been scientific heroes of mine because of the quality of their work in the creation, analysis, corrections, and curation of the dataset. It is kept at the University of Alabama Huntsville (UAH), and it is based on measurements taken by a series of satellite-based instruments called “microwave sounding units” (MSU). One part of it has to do with the temperature of the lower troposphere, called “TLT”.
So of course, it is called the UAH MSU TLT dataset
I noticed that the new 6.0 beta version of the UAH MSU dataset was now available, to replace the current version 5.6 of the dataset. And of course, after doing my own analysis below, I found out that Dr. Roy has been there already with a most excellent and detailed discussion of the new dataset here.
To get the UAH MSU data, I went to the marvelous KNMI climate data access portal. One of the less obvious beauties of the KNMI portal is that after you’ve chosen whatever dataset you are interested in (e.g. the UAH MSU v5.6), on the very bottom of the page that comes up it says:
If you really want to get it here, UAH MSU v5.6 Tlt anomaly is available as a netcdf file (size 17.2871 MB).
For me that’s great, I’m quite fond of netcdf files because they contain all of the metadata (e.g. dimensions, units, starting times, coverage) and they store the data typically as a three-D array (rows are latitude, columns are longitude, layers are months or years or days). But unfortunately, near as I can tell the UAH doesn’t offer a gridded dataset as a netcdf file … but that doesn’t matter when KNMI does it.
So at KNMI I snagged both the older version 5.6 of the UAH MSU dataset for the temperature of the lower troposphere (TLT), the layer down near the ground, and the newer UAH MSU 6.0 beta 2 TLT version as well (also about 17 MBytes or so).
In order to highlight the differences between the two UAH MSU datasets, I made a map of the decadal temperature trends on a gridcell by gridcell basis. Figures 1 and 2 show the versions 5.6 and 6.0 beta2 of the UAH MSU data:
Figure 2a. Several commenters asked for a graph showing the 6.0 beta 2 information using the same color range as was used in Figure 1. This is that graph.
It was most interesting to see both the commonalities and the differences of the two datasets. One of the first things that I noticed in both maps was that despite warming in most areas of the planet over the 36 years, there are large areas of the Pacific, the Southern Ocean, the North Atlantic, and Antarctica that have actually cooled over the period. If ever there were a graph to emphasize the complexity of the climate, Figure 2 is a candidate.
Next, if I had to choose between the two versions based solely on what I see above, it would be version 6.0 all the way. To explain why, look at say India in both maps. It is well understood and verified that when there is a change in conditions the land generally warms or cools both faster and more than the ocean. We see this on a daily, monthly, and annual basis.
As a result, it is unlikely that India would warm or cool at the same rate as the ocean around it, as is shown by v5.6. In the v6.0 results, on the other hand, India is shown as warming at a different rate than the ocean. The same can be seen in western Australia, central Africa, and all over South America.
(In passing, let me note that the above graphs were made from the UAH MSU data. This data comes from KNMI at a 5° by 5° gridcell size. I resampled them to a 1° x 1° gridcell size, using the R function “resample” in the package “raster”. I was concerned about the accuracy of such a radical change in resolution … but when I look at say Australia, I gotta say that their “bilinear interpolation” method handled the resampling much better than I expected. The colors line up very well with the black lines everywhere on the map … and the colors are from the resample while the black lines are from the mapping program.)
There are a couple other differences between the two datasets. The overall global decadal trend has decreased by ~ three hundreds of a degree per decade. Also, the range of the trends has decreased by about 60%, from a range of 1.3°C (-0.5 to +0.8 degrees) per decade in the earlier version to a range of 0.8°C (-0.3 to +0.5 degrees) per decade in the later version.
Finally, I note that much of the central tropical Pacific has either cooled or stayed about the same for 36 years. Here’s how I read that situation. I’ve described elsewhere how the Nino/Nina pumping action is a major part of the global temperature regulation system. When the Pacific starts overheating we get an El Nino, and warm water piles up in the eastern Pacific as shown in the left half of Figure 3. Then during the subsequent La Nina, increasing trade winds pump the warm surface waters westward across the Pacific and from there they flow polewards.
Figure 3. 3D section of the Pacific Ocean looking westward alone the equator. Each 3D section covers the area eight degrees north and south of the equator, from 137° East (far end) to 95° West (near end), and down to 500 metres depth. Click on image for larger size. SOURCE http://www.pmel.noaa.gov/tao/jsdisplay/
Notice in the right half of Figure 3 how the strong La Nina trade winds have hollowed out the surface by pumping away the warm surface water. In addition to moving the warmth polewards where it can radiate away more easily, there is another important effect of the Nino/Nina Pump—it exposes the cool underlying waters to the atmosphere.
Now, bear in mind that for all practical purposes there is an unending reservoir of cold water underlying the tropical Pacific. As I discussed in the post Things in General, the simplified circulation of the Pacific looks like this:
Because the cold bottom water is constantly being replaced from the poles, and because the overturning time is half a millennium or more, the supply of ascending cool water in the tropical mid-Pacific can be thought of as infinite.
So IF we assume for the sake of discussion that the Nino/Nina pump is a part of the temperature regulatory system, then let’s look at what might happen during the time of a general temperature rise. Due to the need to move increasing amounts of energy polewards, I’d expect to see increased Nino/Nina pumping, with a consequent greater exposure of the cool underlying Pacific waters.
So I can certainly see how the central tropical Pacific might be cooling or staying about the same while the rest of the world is warming, as shown in Figure 2. As long as the wind is removing the warm water from that part of the ocean surface, the amount of upwelling cool water will determine the surface temperature.
And what is the explanation for the other area of cooling shown in Figure 2, in the Southern Ocean and Antarctica?
How do you say in your language, “I don’t have no stinkin’ clue”?
Regards to everyone, thanks again to Drs. Christy and Spencer,
The Perennial Request: If you disagree with someone, please quote their exact words that you object to, so that we can all understand the exact nature of your disagreement.
Data and Code: It’s in a 14 Mbyte zipped folder here. It contains R code, the functions, and the two MSU datasets (5.6 and 6.0).