Jump to content

Archived

This topic is now archived and is closed to further replies.

BLAST FROM THE PAST

Why Record Breaking Shannon Entropy?

Recommended Posts

Hello folks, we have recently been introduced by the term Shannon Entropy and that the current models are at or near record breaking levels.

Shannon entropy is the average unpredictability in a random variable - named after Claude E Shannon after his paper in 1948.

Thus models are at their most unpredictable. Now I start this thread to both ask and find out why?

Lets look at what we have - stratospheric warming - we've had many of them before. MJO, GWO in phases that exist eg we not in an undiscovered phase 12 or something. so we've had them phases. ENSO - nothing unusual there, -ve PDO - we've had them before. Cohen research regarding Siberian snowcover, well documented and proposed atmospheric responses - is snowcover experienced well different to ever before? Arctic HPs, Atlantic LPs, Shortwaves etc etc we've had them all before

Models have been upgraded and advanced - so why are they struggling SO MUCH.

So I ask for anyone to join in with suggestions, reasons and answer Why are the models currently at such a high level of variability?

BFTP

Share this post


Link to post
Share on other sites

Im thinking because of the SSW and positive NAO, they just dont know if the jet will buckle allowing the cold to come down.

Share this post


Link to post
Share on other sites

Im thinking because of the SSW and positive NAO, they just dont know if the jet will buckle allowing the cold to come down.

But has that situation never been had before?

BFTP

Share this post


Link to post
Share on other sites

My guess would be that the numerical models are none-too-adept at predicting the effects of SSWs: they are interpreting the data as if they were originating from other (already understood) phenomena?

Caveat: my own 'understanding' may also be falling victim to Shannon entropy!good.gif

Share this post


Link to post
Share on other sites

lnformation Theory Lecture

http://videolectures.net/mlss09uk_mackay_it/

It's actually quite entertaining, and I did study it 20 years ago as a student although I remember little. The context here is machine learning (heuristics) which is suggestive to me of 'learning from the past' which logically we might expect to be a part of modern nwp models. Given that only recently have the models begun to use data up to the top of the Strat that this indeed is why the uncertainty is so high.

Complete speculation on my part, but there you go.

Share this post


Link to post
Share on other sites

Rybris is close to the mark methinks.

GIGO rules apply:

There are a few studies which show that the Stratopause and Mesopause coupling mechanisms are not well understood. SSW events are shown to influence different longitudinal zones in markedly different and complex ways. i.e. thermal, gravitational and dynamic regimes.

For example there is no stable temperature response to the SSW forcing in some longitudinal zones whereas in others the correlation is verified.

Couple this with temporal discrepancies caused by varying inertia over different longitudinal zones together with missing grid data and the uneven distribution of the spatial sample grid and reduced grid resolution with increasing height, (phew) then the picture becomes increasingly desperate.

Dynamic forcing and propagation from the upper to lower atmospheric model can therefore be full of inaccuracies.

The seed data for the Troposphere is therefore incomplete at best and at worst incorrect during active SSW events with the rate of change increasing error.

It is only when the layers reach relative equilibrium with stable temporal differentials, can the lower atmospheric modelling get to grips.

It is my belief that these are the primary reasons why the various models differ widely and lurch from extremes.

Perhaps others would care to comment?

ffO.

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×