Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Options

Laplacian/Dt

David Mahler suggested to divide the Laplacian of the surface temperatures by dt:

Laplacian/Dt 2013

Laplacian/Dt 2010

Back to Paul's comment, it is clear that there is something going on on the Himalayas! under that region North East of India seems to be mostly 0 flux/flow.

Gosh we need a super hero to explain this to me

Dara

• Options

Please take these computations with a grain of salt, I am re-checking the code.

I got divide by 0 for the division, so I Fudged by adding 0.0001 and then most of the numbers were too large, so I re-normalized.

all need to be checked again

Dara

• Options

So I am thinking: Most of the areas are almost small amount of flux/flow, there though these very select areas of large flux/flow.

But the surrounding areas of these large flux regions are very low in flux and therefore no connectivity to any place to drain into or leak.

Then I must conclude that these high flux regions are connected, since the Himalays are not on the water, then there has to be a upper atmospheric connection, but this connection is not on or off, it is like a wind tunnel of somekind. I could 100% be off the mark :)

Dara

• Options

Hi Dara, It might also be interesting to see the scatter plot of Laplacian[T] vs dT/dt. That may help understand if the peaks are just due to small dT/dt or something more intersting.

• Options

Hmmmm how do you do scatter plot of two 4D datasets? have you seen an example? I could give it a shot D

• Options

Oh you mean x1 vs. x2 each from different data set?

• Options

I meant getting a (Laplacian, dt) pair for each (lat, lon, time) and just plotting the whole set on a single graph. They should be strongly correlated and the slope should give us the alpha of the heat equation governing the system.

• Options

Ok, let me hack some code... almost dawn and another unruly night a at Gotham, the cape crusader nowhere in sight...

• Options

3D scatter plot

x=lat, y=lon, z=t, then either use colour maps for the relative size of Laplacian/dt or change the radius of spheres proportionally.

• Options

I will try also the traditional 2D scatter plots of pairwise vars

• Options

For the 3-d plot I think x=Laplacian, y=dT/dt, z=time, with colors/shapes/sizes indicating location might be more informative

x=lat, y=lon, z=t, then either use colour maps for the relative size of Laplacian/dt or change the radius of spheres proportionally.

I think that will reproduce the Laplacian/dt slicer view.

• Options

x=Laplacian, y=dT/dt, z=time ... Ok in a few hours