ScienceBlogs on the acceleration of rising sea levels

Tim Lambert at ScienceBlogs debunks the idea that rising sea levels are slowing or have stopped.

So if you were unscrupulous, and wanted to try to make it look like sea level rise had decelerated what could you do? You could split the series at a point where sea level was above the trend line and compare trends before and after. this is what Klaus-Eckart Puls did (green line added by me):



  1. Next year, you can release the 94-04 and 04-14 series and it will look like sea level rise is significantly accelerating decade over decade.

    At which point, both statements are true, that sea-level significantly decelerated in 03-13 compared to 93-03 and that sea-level rise significantly accelerated in 04-14 compared to 94-04.

    At which point the average person loses all faith in climate science.

    Unscrupulous indeed.

  2. They didn’t even screw up right.  The second correlation should have a y-intercept right where the first one left off, so it actually looks like a perfect deceleration.  Freakin’ amateurs.

  3. I’m confused. Of course if you fit straight lines to different data sets you will get different results. You can’t prove that the rate of increase ISN’T declining by fitting a straight line to the entire set. What they really need to do is confirm whether the most recent drop can be deemed statistically significant when compared against the earlier data. I’m thinking something like a basic control chart would do the trick.

    1. It’s not a “different” data set, it’s a cherry-picked portion of a large data set. It’s like trying to show that crime is decreasing in your neighborhood by starting your graph on a day that had multiple homicides.

      1. He split it directly down the middle, it’s hardly cherry picked. And besides, I’m still arguing that it doesn’t prove much, just not for the equally-as-simplified reason that the second author proclaims.

  4. When you “fit” a straight line to a data set you are doing a linear regression using a least squares calculation.  Not just drawing a line where you want it to be.  The least squares calculation helps determine the best fit straight line that minimized the sum of the squares of the delta of each data point off that line.  I know 0_o

    If you want to move the fit line you need to modify the data set.  This guy could very well have run a calculation to determine the best fit combination that would maximize the appearance of deceleration assuming decade long data sets.  Then again he could have taken the data he had and split it down the middle.


Comments are closed.