postprocessing

[Emmanuel Hanert]

This topic deals with all the issues related to the model results postprocessing.

[Ruth Patterson]

Hi, I’m trying to compute tidal components using the GitLab page example. The FS.fourier_at_point command returns this error.

================= Start Reading Files =================
Reading file output/eta/eta-000100.idx …
Fatal : Unable to open import file ‘output/eta/eta-000100.idx’

I understand this is because my ‘output’ folder and file structure is different to the one that is inbuilt. However there doesn’t seem to be a way to change it. My ‘output’ file contains partitioned components of each iteration. E.g for the 10th iteration:

‘slim_0010_00000.vtu’
‘slim_0010_00001.vtu’
‘slim_0010_00002.vtu’
…etc…

What type of processing do I have to do to get ‘eta-000100.idx’? Can I use an ‘eta-000100.txt’ which is output from the export time series script?
Is there a way to change the file name (‘eta-000100’) and directory (output/eat/) if I stored eta time series elsewhere?

My script is below (please feel free to recommend any other changes – I’m not sure if the period and time-frame nominated below is sensible. My only other experience is using t-tides in matlab and this requires 1 month of time series eta data) :

import slimPost

data_dir_base = ‘data_7/’
mesh = data_dir_base+ »/0/mesh.msh »
path_input = « output/ »

#Fourier Series
initial_index = 100
final_index = 148 # 12 hour period
initial_time = 90000 #init =100. 15 min timesteps. 100/4=25. 25 h *3600 s = 90000
n_period = 1
n_iter = 200
n = 5 #number of components to compute
period = 21600

FS = slimPost.Fourier_series(mesh, path_input, initial_index, final_index, initial_time, n_period, n_iter, n, period)

x = 235292
y = 8719371

iComp = 1

FS.fourier_at_point(x, y, fig_name = « test », variable= »eta ») #script stops here
FS.fourier_map(iComp, « eta », « Range », « Phase »)

[Valentin Vallaeys]

Dear Ruth,

You are right, there was a missing argument in the Fourier_series function in order to use your outputs.
We have updated the code. The only change you have to make in your script is replacing this line

FS = slimPost.Fourier_series(mesh, path_input, initial_index, final_index, initial_time, n_period, n_iter, n, period)

by the following ones:

format_input = "vtk"
FS = slimPost.Fourier_series(mesh, path_input, format_input, initial_index, final_index, initial_time, n_period, n_iter, n, period)

Otherwise, your script seems OK. I would just set n_period = 2 as you set the period as 6h and the time frame is 12h.

Best,
Valentin

[Ruth Patterson]

Thanks Valentin, it works.

[Ruth Patterson]

Hi,

I would like to sum the u v current components across the whole domain for a range of different time periods (say, 2 weeks, 1 month etc).

My model output are partitioned (for up to 16 cores) and the output format is .vtu.

Is there something already built for this?

Cheers, Ruth

[jonathan]

Hi Ruth,
The easiest solution is probably to do it directly in paraview. You can achieve this using two filters :
– Temporal->Extract Time Steps : to extract the time period you want.
– Temporal->Temporal Statistics : to compute the mean value (and optionaly min, max, and stddev) over the extracted time period.
After that use File->Save Data to save the result.
Cheers,
Jonathan

[Jodie Schlaefer]

Hello,

I was just wondering what the default value of 2e^-4 m^0.85 /s for the coefficient of the Okubo scheme in the slimPost.create_okubo_map function is derived from?
Does it come directly from Okubo (1971)? If not, can you please provide a reference?

Thanks for your help!

[Valentin Vallaeys]

Hi Jodie,

Sorry for the late reply.

It comes from the original paper of Okubo. He gave K = 0.0103 l^1.15 [cm^2/s].
Converting cm in m, it gives : 0.0103 cm^0.85/s = 2.0551e^-4 m^0.85/s.

Cheers,
Valentin

[Jodie Schlaefer]

Awesome, thanks Valentin!

[Author]

Posts