Digital Diatribes

A presentation of data on climate and other stuff

January 2009 Update on Global Temperature – UAH

Posted by The Diatribe Guy on January 21, 2009

Note: I realized I screwed up my naming convention. I already had a December update on UAH. This one is January…

uahoverall200812

Sorry it took so long to get this up. I had a couple things I wanted to look at here and since this isn’t my day job, every now and then other priorities get in the way.

The December anomaly from the UAH data is 0.183. This was 0.069 degrees warmer than December 2007, but was 0.068 below the November anomaly. According to UAH, it is the 11th warmest December out of the 30 Decembers in the data. It is ranked 102nd out of 361 overall monthly anomalies. It is the second consecutive warmer anomaly year-over-year after the streak of 14 consecutive cooler anomalies.

uahcooling200812

The longest we can go back in the UAH data without showing a warming trend line is May 1997. Based on the way January seems to be coming in, it would be surprising if this moved forward after this month. It may even move back to April 1997, but we’ll see on that.

I’m not going to show the different trend charts here, because I actually want to point out something else this month. But here’s a quick recap of what is going on with the trend lines:

60-month: -0.309158 represents the third consecutive increase. So, there has been a bit of a moderation in the rate of cooling we have seen lately.

120-month: +0.101070 is the 12th consecutive increase in the 10-year slope value. This has to do somewhat with the moderation in anomalies in recent months, but primarily it is because the front-end of the 10-year period had been the tail end of that huge El Nino in 1998-99. Those high front-end values lowered the slope of the line, and now that those values are passing by, the slope is increasing again. However, my analysis shows that the impact of those high front-end values is almost complete. The next two months will likely show a continued slope increase in the 10-year trend line, and then we will see a fairly rapid decline in the slope. Keeping in mind that the last 12-month average anomaly is less than 0.05, if the next 12 months average 16 or below, we will see a zero or negative 10-year trend line after December 2009. That all said, the current slope value is the highest level it has been since March 2007.

180-month: +0.105952 slope value continues an unabated decline, with no month-to-month reversals at all, since March 2007 in the 15-year slope value. Based on my analysis, the slope value should continue to decline unabated throughout 2009, unless there are some anomalies that suddenly are significantly higher. The current slope value is now at its lowest point since the 15-year period ending January 2002.

240-month: +0.148475 slope value is the steepest positive slope value of the current various trend lines that I review. However, it has generally been in a state of decline, with some month-to-month reversals, and this is expected to continue over the next year.

300-month: +0.146696 slope value continues a steady decline since March 2007. The current value is at its lowest point since June 2006.

360-month: +0.105668 slope value. Not much to say here, since this is only the second 30-year trend point.

Averages:
*The current 2,3,4 and 5 month averages are higher than the lower anomalies in early 2008. The current averages are their highest level since late 2007.
*The current 6 – 12 month averages are all higher than the previosu few months, but all had higher averages at some point for their respective periods ending in 2008.
*15, 18, and 21 month averages, however, are at their lowest levels since the year 2001.
*longer-term averages are also at lower levels since periods ranging from 2002-2006.

So, there are the stats. I now wanted to speak to an observation that jumps out at me in the data.

Let me start by presenting the following chart of 10-year rolling average anomalies:
uah120avg200812

Well, looking at this, the trend is unmistakable. Averaging the temps on a rolling basis produces a filtering of the data that shows a significant warming trend, with the exception of the most recent period.

Another view of the 5-year rolling anomalies still demonstrates this trend:

uah60avg2008121

So, this is pretty clear, right?

Not so fast.

Filtering data and using rolling averages is a very good and useful tool in statistical analysis. But, like any other tool, if it is misapplied, you end up painting the wrong picture. The issue is autocorrelation. Some people so dislike autocorrelation that they avoid filtering altogether. That’s a little overboard, in my opinion. Autocorrelation is not always a bad thing. For example, my charts that show the movement of trend lines have a great extent of autocorrelation in them. In that case, it’s by design. I wanted to see how trend lines shift over time, and the x-month trend lines shift by dropping one month off the front end and adding the next month to the back end. Of the y number of values in a given trend line, y-1 of them were in the previous month’s trend line. This gives me a nice, smooth picture of how the trend lines changed. The reason I don’t see this as inappropriate is because the position of the values change in the regression line. Conceptually, if values changed at the same rate in a linear fashion, then I would get the same slope each month. Thus, the autocorrelation itself is not an issue.

However, there are other instances where autocorrelation creates huge problems with the presentation and conclusion of the data set. Let me provide a very simple illustration of this.

The following chart represents some series of observations.
example

It’s pretty clear that there is no upward trend here. What is fairly obvious is that the situation changed somehow at a particular point in time. In this case, the first 25 observations are all a value of 5, and observations 26-50 are a value of 15. A step occurred at the 26th observation. And while it is reasonable to say that the value today is 3 times as much as it was 25 observations ago, you would be playing fast and loose with the facts if you were to imply that this occurred evenly over a period of time.

However, someone could filter that data to present exactly that argument. The following chart adds a black line, which is a rolling 25-observation average of the data:

example2

If someone supplied the chart with the black line only, it would be very easy to conclude that there has been a very consistent increasing trend in the observations. Only by observing the raw data will you realize that this is entirely misleading, and flat-out incorrect. Now, this does not say that it isn’t reasonable to question why the step occurred. There may be legitimate debate about the mechanism involved at point 26, which then perpetuated forward. But it is simply not a trend. In this case, even the best-fit trend line over the long term is very misleading. The better trend line is actually determined by the shorter observations, not the long-term fit. People can’t seem to ever get past the argument that short-term trends may in fact be better tools in analysis than long-term trends. And while that certainly is not the case with steadily increasing trends, it may well be the case where step functions exist.

So why the lesson on autocorrelation and step functions? Very simple. I re-present the latest flat/cooling trend:
uahcooling200812

Since May 1997, the line is essentially flat. But look at the period from the inception of the data up to that point. What do you see? Here… let me assist:
uahsplit200812

It is important to note that the anomalies here are in terms of 0.01 degrees Celsius.

The trend line fit to the period from December 1979 to April 1997 is essentially flat (+.0265 slope. The negative trend line from May 1997 is -0.0095.) Notice the b value in the linear equation for each chart. Prior to May 1997, the b value was negative 6.3439, and since May 1997 it is positive 22.7340! We see a step of over a quarter of a degree Celsius (29.0779), and only a small portion of that is attributable to the increased slope from 1979 – April 1997, since the slope is miniscule. Over the 209 months preceding May 1997, the contribution amounts to a calculated 5.5385 using the slope from the observed trend. Thus, there was a step that occurred for the succeeding period of about 23.5. That’s huge.

So, it would appear that temperature jumped to another level and stayed there. In recent years, it has been dropping down again. This is just me musing, but it surely would seem that the super El Nino event of 1998 and subsequent El Nino activity helped prod that jump. Our atmosphere does exhibit a greenhouse effect, which is a good thing for us, unless you want -200 degree temperatures during the winter and huge temperature swings between day and night. What it also means is that a huge event can elevate temps, and they may stay there a while. Add a few other similar events, and they’ll perpetuate. Eventually, this will dissipate, as we are finally seeing. It is my contention, then, that the separate analysis of the trend lines split before and after 1997 are more meaningful than an overall trend line over the entire period.

I’ve taken a look at the ENSO index, and the longer-term persistency of it. I’ve done the same with the PDO, and also with sunspot counts. All these cases tend to point to the persistence of cyclical events. These are also the kinds of things that can change suddenly. Carbon Dioxide may be a nice theory for long=term, gradual trends, but it makes no sense at all as an explanation for a step function. A review of the Mauna Loa Carbon Dioxide data, for example, simply shows no great leaps and bounds in the concentration of carbon dioxide in our atmosphere.

Food for thought.

About these ads

13 Responses to “January 2009 Update on Global Temperature – UAH”

  1. [...] bookmarks tagged consistent December 2008 Update on Global Temperature – UAH saved by 9 others     Ivoryshadows bookmarked on 01/21/09 | [...]

  2. Layman Lurker said

    Interesting observations. While the super el nino seems to conicide with the timing of the step, distribution of volcanic and ENSO events during the time period may play into the magnitude of the step. Are you going to look at any of the other metrics?

  3. The Diatribe Guy said

    My to-do list is quite long in this area. I really want to take a look at all the oceanic oscillation data I can find (I’ve already downloaded it), look at solar activity, and also the greenhouse gas measures and do a full minimum bias procedure that reduces cross-biases between all these things.

    For example, PDO and ENSO likely have influence on each other’s anomalies, so without tryuing to eliminate the “double counting” you get too little/too much influence calculated for each factor.

    Is there some kind of an index for volcanic activity (more importantly, a measure of substance in the air from volcanic activity)? Aerosols would also be interesting to take a look at.

    It would be nice to know we had good data going back two centuries, because the more metrics you look at the less credible the sample of temperature data points becomes. However, I think I’ll probably be able to simply eliminate some of the metrics that have no definitive contribution to the temperature.

    Anyway, yes, I plan to really di into that. I’ve been planning on it for some time now. I keep getting pulled in other directions, partly because it’s a big project and I’d like to spend a lot of time on it in chunks rather than bit by bit.

  4. The Diatribe Guy said

    Oh, and I wasn’t necessarily offering the ENSO explanation as anything definitive. Just my rudimentary first reaction to the data. But there is clearly a step which, to me, rules out CO2 as the primary driver in the temperature change.

  5. Layman Lurker said

    “Stratospheric Aerosol Optical Thickness” data I think might be what you are looking for.
    http://data.giss.nasa.gov/modelforce/strataer/

  6. Jeff Id said

    I didn’t leave a comment the first time I read. This was a very nice post. It also points out that the spike in 97-98 is largely responsible for the recent downtrend.

    People who don’t read graphs (and even some who do) don’t understand some of these subtleties.

    Actually, that’s another point about the anomaly graphs. Even the word anomaly assumes something varying from normal. What is normal? The layperson sees +0.5 they don’t know what that means except up. Scientists read it and see nothing. The polyscienticians in charge of AGW blend the two.

  7. Jeff Id said

    Oops, rewrite.

    I didn’t leave a comment the first time I read. This was a very nice post. It also points out that the spike in 97-98 is largely responsible for the recent downtrend.

    People who don’t read graphs (and even some who do) don’t understand some of these subtleties.

    Actually, that’s another point about the anomaly graphs. Even the word anomaly assumes something varying from normal. What is normal? The layperson sees +0.5 they don’t know what that means except up. Scientists read it and see nothing but trend. The polyscienticians in charge of AGW blend the two.

  8. Flanagan said

    Nice post – it looks like the evidences are accumulating in favor of a “concentrate and then release” type of warming, with stepwise increases as you mentioned. On the other hand, the January anomaly will more than probably be similar to that of 2006 or 2005, given the UAH data (see the DISCOVER UAH site), which is in fact quite high.

  9. docattheautopsy said

    Great post. I like the explanation of data filtering– really well done, and I think it’s perfect for people without a solid grounding in data processing and analysis.

    Congrats on the increased # of hits, too. They all come from my site. :) (Well, 3 of them, anyway.) It’s a testament to the work you’re putting in. Keep it up!

  10. The Diatribe Guy said

    Thanks, Doc!

    Hopefully this month I can finally get into more of the Oceanic Oscillation work.

  11. [...] that the smooth increase over time has to do with autocorrelation of the data. As I discussed here, a single step in the data will give an appearance of a gradual trend when a plot is presented in [...]

  12. [...] (5) A demonstration of the fallacy of a singular trend-line fit in the recent temperature data, and … [...]

  13. P.J. Maher said

    Very interesting post. Keep up the good work and I’ll check back from time to time. The concept of sudden, large changes rather than steady gradual change is one I’ve been entertaining for some time. I’ve seen clear evidence for it in evolution and geology, disciplines in which I am far more comfortable. In fact, I remember in my early college geology classes arguing with the professor over uniformitarianism and catastrophism. It’s the same idea really, just different disciplines. As professor Trent was so fond of saying, “if a concept applies to another field of science, it more than likely applies in all fields of science. It may not be easy to find but it’s likely there”
    It’s nice to find I may have been on to something all along. Nature likes to do some things gradually but from time to time it does tend to shake things up a bit. Makes things interesting.
    Thanks for the good work

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

%d bloggers like this: