Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Geoff Sherrinton's comment


Please do not use bad science to impugn geostatistics.
Yes, extrapolation from one point to another and interpolation between points are common methods in geostatistics and other methods.

However, those sample points have conditions precedent before they can be used properly.
In work familiar to me, and now talking only geostatistics, one does not interpolate between different media, as from sea to air. Or ice to adjacent water. Boundaries matter.

Further, there has to be some knowledge of the properties assumed for or known about the points in a given medium. In rock work it is common to process different major rock types separately because they can have different fabrics with different alignments, leading to different ‘solids of search’ for later weighting and other complications.

Now taking a vertical sea profile containing a buoy, do we have the equivalent of different rock types through the profile? Yes we do, especially in fine detail. The very surface of still water has sub mm layers impacted by long wave IR, different to lower down, being evaporated, special effects on T. The top 500 mm or so of sunlit water is often at higher T, by a deg C or more, than lower down. Proceeding down, you can meet thermoclines and ipsoclines before 100m down, the depth used by some to express overall surface sea temperatures. Therefore, such SST are by definition an average for some sort of T whose variation in that profile is large compared with the effect often sought, namely the T difference between one profile site and another. Even day/ night sea cases are different.

It is mathematically wrong to use geostatistics when the within-sample static variation is much larger than between sample, let alone including dynamics on time scales of making a measurement. Yet, that is being done. Mixing by Nature can make results seem better, but they are not actually better unless the pre-mixing T distribution is known in detail so that the appropriate sub- sample can be compared site to site, apples to apples, later in the process. What part of a variable sea T profile should be compared to air T above? How do you know if you have captured it? Given the size of T variation down a profile, this is a fundamental impediment. Sure, you can grope around and get some general figures but these will not usually be good enough even for government work. It is stuff to kid yourself with.

A further problem happens when air T is compared with sea T. Their thermal inertias differ. Some heating or cooling effects work to different time patterns. You cannot interpolate between air and sea because of this, except with huge assumption errors.

For reasons like these, the Karl pause buster paper is invalid. The Cowtan & Way fiddles in the Arctic are wrong at Kindergarten level and should be retracted before doing more harm. Rohde from BEST might like to address some of these points to justify his recent revisionist work about the hottest evah. I had hoped he would have done better. Others like satellite T people should refrain from overextension of ideas linking air T to sea T. Again, it is invalid unless given a huge and correct error from non- physical assumptions.

It is our attitude toward free thought and free expression that will determine our fate. There must be no limit on the range of temperate discussion, no limits on thought. No subject must be taboo. No censor must preside at our assemblies.

–William O. Douglas, U.S. Supreme Court Justice, 1952

Users browsing this thread: 1 Guest(s)