Get stuck in using Cal_IT_Fric.

[Jiangchao3]

Get stuck in using Cal_IT_Fric.

[krober10nd]

Dear @Jiangchao3,

Perhaps you could pause the code when it appears "stuck" and see where it is? How long did you wait? This routine can become very costly.

[Jiangchao3]

Hi @krober10nd , thanks for your kindly reply.

This time, I wait for about 5 minutes, pause it and enter the debug mode, find the code stuck in the following functions:

Calc_IT_Fric:

Gridded_to_Mesh_SeaBed_DepthAveraged:

fillmissing:

I will continue to test the code.

[Jiangchao3]

Hi @krober10nd, I think I found what the error lies:

in the function of Gridded_to_Mesh_SeaBed_DepthAveraged, it stuck in the iteration of zvalue = 82;

J is calculated as the following:

the result from the 54 line is a 12*1 NaN vector：

it seems that line 58 and 59 can not work when the input is a full NaN vector:

nothing can be fillmissing.

That may be why the Calc_IT_Fract stuck and can not move forward.

I also test the iteration zvalue = 81;

J:

It's clear that the N_interp{zvalue}(J) is not a full NaN vector, so fillmissing in line 59 can work.

Hope the above description is clear to present the problem, thanks.

Finally, maybe the depth for each node is critical to calculate the IT fraction.

[krober10nd]

Thanks Jiang. I think it's clear what the problem is now. When the interpolation returns back all missing values, it becomes stuck in an infinite loop. It would be helpful if @WPringle took a look at this since he is the creator of this script.

[Jiangchao3]

Yeah, waiting for @WPringle have a look!

[WPringle]

@Jiangchao3 I see, so I guess the problem is partly with N data being passed to the function, that there is too many NaNs at certain zcontour values. Perhaps then we need to fill NaNs in the N data in the vertical direction too which could be done just using nearest neighbor from the zcontour value above. Can you try this?

The reason for this can be that the depth of the N data at a certain location say goes to 1000 m but the mesh actually is at 1100 m, so at 1100 m the data has a NaN value but we want a value at the 1100 m contour on the mesh. A fix then is to extrapolate all the N values down the column using the nearest non-NaN value above. I think this may work better.

[Jiangchao3]

@WPringle, thanks very much for your kindly reply. I am out of my office in recent days, I will check it after I come back!

[Jiangchao3]

Dear @WPringle, today, I followed your suggestion to fix my Gridded_N_values.mat dataset. This is my code:

clearvars; clc;
load('H:\QJC\8-Github\OceanMesh2D_Jch\datasets\Gridded_N_values.mat')

%%% bangladesh
% lon from 60 to 110, (i from 960 to 1160)
% lat from 0 to 30, (j from 361 to 501）

for i = 960:1160
for j = 361:501
temp_vertical = N(i,j,:);
temp_vertical_new = fillmissing(temp_vertical,'previous');
N(i,j,:) = temp_vertical_new;
end
end

o_name = 'Gridded_N_values_new';
save([o_name '.mat'],'lon','lat','z','N');

Now, the Cal_IT_Fric function seems able to work well.

[WPringle]

Nice, thanks for letting me know!
I would welcome a PR to add this change into the function

[Jiangchao3]

Sure, I will create a function called Calc_IT_Fric_fix_N.m and then add it into the unitities folder and add a note into the Calc_IT_Fric.m to remind users who meet with similar problem to know how to fix it.

[WPringle]

Could you just make the change directly to `Gridded_to_Mesh_SeaBed_DepthAveraged'?

[Jiangchao3]

`Gridded_to_Mesh_SeaBed_DepthAveraged'?

Ok, let me have a try.

[Jiangchao3]

Hi @WPringle , I have created a PR #280, please review.