Friday, February 26, 2016

Pause? Slowdown? Whence? Wither?

The Pause has been in the news again. There is a new paper in Nature Climate Change (pdf here) by Fyfe et al, with numerous notable authors, saying that there really was a slowdown, but not a stop, in global warming. This is partly in response to a paper last year from Karl et al, basically saying it was an artefact of ship/buoy bias, and disappeared after appropriate adjustment. There is an extensive summary in Scientific American here, and good coverage by Sou here. Also Tamino

I'm with Gavin, as quoted in SciAm. I think it's a pointless argument. Yes, warming was a bit slower for a few years. Unless someone can show some statistical significance of something, I don't see why it should even have a name.

However, that seems to be a minority view. I write a lot about temperature anomaly series and their trends, and I argue a lot about Pause matters, so I should say something. What I want to do is to show how the Trend Viewer can be used to show how pauses originate, and to give an idea how unusual is the one under discussion.

Here is a typical plot which I used in discussion at WUWT. It shows the history of the NOAA index since 1960. I suggest that you open up the Viewer here. The active features make it much easier to follow what is going on. The values of trend for all possible subintervals of 1960-2016 are shown on a triangle, plotted in color with x-axis the end time of the interval, and y-axis the start. I actually don't show trends of less than a year; short trends are along the hypotenuse, and the axis there is where a trend of zeo length would be. Length of trend goes with distance from the hypotenuse, and is marked with white lines. Here is that plot:



You'll find description in the text of the viewer page, and links to earlier posts. But I'd urge you to use the interactivity to help figure it out. There is an associated time series plot:



You'll see a red and blue dot, indicating the ends of a trend line for a chosen interval - here Jan 2001 to Dec 2013. Below the graph is a block of information about the chosen interval:



You can choose the interval in different ways - with the colored bars or nudgers on the time series plot (see page for details) or by clicking on the triangle. The dots will move and the text will change. The text info is a good way to work out the triangle. Just click for response.The rest of the space has buttons that you can click to choose different datasets, time intervals, or style. With style, instead of just trend you can have trend colored where significant (different from 0 at 95%), or you can choose to plot a CI or a t-value (see page info).

The plot uses rainbow shading for colors, except for brown at zero, and grey at 1.7°C/Century, chosen as a representative AGW value. If you look along the hypotenuse, there are sharply alternating reds and blues, with some labelled events. These correspond to peaks and dips, which are actually located where red and blue met. On the left of a peak, short trends are positive (red), and afterward, blue. For a dip, it's the opposite. Peaks are usually ENSO related, dips usually ENSO or volcano.

Where there is a peak, the red tends to propagate downward. That is where you hold the end point on the peak, but extend the start backward. Trends tend to be positive, but less when longer. And the blue propagates right, as the start is held on the peak, but the end moves forward. Dips behave oppositely.

You'll see the recent Pause region marked. It's a greenish/yellow area, with trends generally positive but closer to zero than surroundings. And you can see why. Blues on the diagonal are extending cross cool bars, with a deficit of reds above which could extend balancing warm downward. Somewhere between the blue and green you'll see a sketchy brown line. On the blue side of that, trends are zero. Sometimes there are brown-edged islands.

So you can look around for other pauses or slowdowns. You should look for cool colors a long way from the hypotenuse. That distance is a measure of the trend duration. And, depending on what impresses you, you can look for a long slowdown or a shorter actual hiatus. But you should also check either the CI's in the text info or show the plot with significance mask to see whether your pause actually stacks up.

If you do this, you will see that the recent pause is larger than most in that triangle. There are cool colors bottom left, the fringe of the Big Pause from about 1950-1970. You can see this better if you switch to the time range starting 1901.

So you can see where the current Pause came from, and why it is ending. The current El Nino makes a red blob at the top, and that extends down. The green area is being closed off. Remember, the right edge is the current month. Some say that after the El Nino the Pause will return. I think that is fairly unlikely. You'll see that cool regions tend to form islands. For a long time the 1997/8 El Nino has anchored the Pause. I think this El Nino will now be the anchor for whatever slowdowns we have in the future.



12 comments:

  1. It may not be a minority view. I suspect that people who think it is a pointless argument are less likely to comment.

    ReplyDelete
  2. The pause could have persisted. Had it, there would have been something wrong with the theory. That's why GS at RC wrote a post several years ago about the length of time required with no warmest year. We are in an experiment.

    2004 0.5375 - EL Nino
    2005 0.679167 - El Nino (warmest year)
    2006 0.625 - El Nino
    ---------
    2007 0.649167 - La Nina
    2008 0.525833 - La Nina
    2009 0.639167 - El Nino
    2010 0.709167 - El Nino - La Nina (weak warmest year)
    ------------------
    2011 0.5925 - La Nina
    2012 0.625 - La Nina
    2013 0.6475 - neutral
    2014 0.738333 - neutral
    2015 0.861667 - El Nino


    Essentially, from 1998 through 2005 there were three El Nino events in a row, and a sharp increase in warming (30-y-T = .18 ℃ PD, plus .01 ℃.) This is ignored in the idiotic claims that there has been no warming since 1998. There was warming; there was a lot of it.

    From 2005 through 2012 there was a clear dominance of La Nina and ENSO neutral over El Nino, and there was a sharp decrease, a cooling ((30-y-T = .16 ℃ PD, minus .02 ℃.)

    So to me, the current 30-y-T is .169 ℃. The pause can fog a mirror; in the General Franco sense, it's not completely dead. I think normal flipping from La Nina to El Nino will kill it. The cause of the pause, imo, was Matt England's anomalous winds. They have ceased. If the next trade winds are normal, the pause will be forgotten in history. If Matt England's intensified trade winds come right back, the pause will gain back some ground. The additional upwelling of very cold water caused by intensified trade winds would result in a lengthy slowdown in AGW. The moment-to-moment physics of a dynamic system can matter; not all wiggles always average to zero in the relevant time frame.

    ReplyDelete
  3. The time series for ocean heat content and sea level rise show no slowdown in the past several decades, indicating no slowdown in the TOA radioactive imbalance. A slowdown in the atmospheric temperature series must therefore be due to non-uniform heat exchanges between the high heat capacity ocean and the much lower heat capacity atmosphere. Over a period of time, heat accumulating in the oceans will equilibrate with the atmosphere.

    ReplyDelete
  4. It is sad that such an article can be published in 2016. In Nature Climate Change.

    ReplyDelete
  5. Y'know, rather than just say "this is/isn't statistically significant", it'd be better to report probabilities. Like, "there is a 90% chance of a change in trend", or "with the new data, there is a 85% chance of a change in trend".

    That's more useful, to be honest. It also gives us an idea of the significance of papers like Karl et al (2015), and how much of a shift the new data represents.

    TL;DR: Instead of focusing on the binary question of "this is/isn't real", focus on how likely it is. It's the approach we already generally use with climate change (e.g., probability distributions of ECS).

    ReplyDelete
  6. "It is sad that such an article can be published in 2016. In Nature Climate Change."

    Hopefully that will change. I think that soon the
    old guard of atmospheric scientists such as Lindzen,
    Curry, Salby, Spencer, and Pielke will disappear and
    it will make room for scientists that actually have
    something to offer.

    I have been studying QBO and found that Lindzen and
    Salby have accomplished very little in advancing the
    understanding of the behavior. I was amazed how inept
    they are at interpreting the time-series data that is
    available. Too bad the rest of the climate science
    community let them get away with their research.


    http://contextearth.com/2016/02/13/qbo-model-validation

    http://imageshack.com/a/img921/2124/k0xw17.png

    ReplyDelete
  7. "Unless someone can show some statistical significance of something, I don't see why it should even have a name."

    Agreed. It is interesting to examine anomaly data in the context of characterizing the 'observations' over time, but there is something deeply discomforting about where one is sifting through different testing approaches when formulating a convincing, useful hypothesis rooted at some level in physical reasoning--the same problem at a deeper level--seems slippery.

    From another perspective: uncertainty cuts in all directions

    ReplyDelete
  8. Nick, you should consider changing the size your NCEP graph. +1.137°C is now too much for your scale.

    ReplyDelete
  9. Huh. When your NCEP reanalysis graph hit 0.997 and then dropped for two days, I thought that was the turning point, and it'd stay under 1.0. Now it's at 1.137, and as Keri points out, it's broken the top of your chart.

    Very curious what GISTEMP et al. will report for February. The NCEP record for Feb. is astonishing.

    ReplyDelete
    Replies
    1. By my calculation, based on your NCEP daily numbers, GISS LO for February should be 1.3 +- 0.1. That would be an increase of 0.3C over the pre-2015 monthly record (0.95, in Jan 2007).

      Delete