Skip to content

Latest commit

 

History

History
87 lines (66 loc) · 3.49 KB

README.md

File metadata and controls

87 lines (66 loc) · 3.49 KB
Raw

Folders

Software: /home/olmozavala/Dropbox/MyProjects/EOAS/COAPS/MURI_AI_Ocean/Data_Assimilation/HYCOM-TSIS Outputs: /data/HYCOM/DA_HYCOM_TSIS Summary Results: https://docs.google.com/spreadsheets/d/1EqkI2K6BA5jnXTny7w_U28vN1JUPn_h0J16SLqV1mgA/edit?usp=sharing

Files

0_ComputeMinMaxFromData

This file can read preprocessed or raw data and obtain the min, max and variance values for each of the fields specified in PreprocConfig.py. The outputs are saved in a stats_obs.csv and stats_background.csv file. Important!!! you need to add model in the first line of both files in order to work. Important!!! This file already divides thknss by 9806 so we need to be shure this also happens when we read the data.

1_AnalizeData

This file is used to generate multiple plots from the raw data. Important: this file plots together the fields defined in PreprocConfig parameters:

PreprocParams.fields_names:     ['u-vel.', 'v-vel.','temp', 'salin', 'thknss', 'srfhgt', 'montg1', 'surflx', 'salflx', 'bl_dpth', 'mix_dpth', 'u_btrop', 'v_btrop'],
PreprocParams.fields_names_obs: ['ssh', 'ssh_err', 'sst', 'sst_err', 'sss', 'sss_err', 'uc', 'vc'],

2_PreprocData

This file was used (not anymore) to preprocess the hycom and tsis outputs into cropped netcdf files. It could reduce the number of layers and fields.

3_Train_2D and 3D

This is the main file that trains new models depending on the MainConfig_2D.py configuration file

4_Summary_of_Models

Iterates over all the trained configurations in the Training folder, and for each configuration it saves the model with the lowest validation error into a csv file. This file is useful because it contains a list to the best model for each configuration tested.

4_TestModel

Currently doesn't work!!!!!!. It reads preproc netcdf files. This file is used for two options:

  1. Single model evaluation. In this case the user needs to specify wich weights file to use in MainConfig.py and the configuration inside this file should match the model being tested.

4_TestModel_Whole

It uses the get_preproc_config from PreprocConfig.py to identify which folders to use for reading hycom, tsis, and observation data. This file is used for two options:

  1. Single model evaluation. In this case the user needs to specify wich weights file to use in MainConfig.py and the configuration inside this file should match the model being tested.

6_RMSE_Plot

This file simply plots the RMSE obtained from the 4_TestModel_Whole.py file for all the validation examples.

Data

Start date Jan 1st 2009.

Assimilation every 4 days.

Assimilated variables: T, layer thickness and density

Hycom: /Net/gleam/abozec/HYCOM/TSIS/IASx0.03/hindcast_newtsis/gofs30_withpies/archv*

TSIS: /Net/gleam/abozec/HYCOM/TSIS/IASx0.03/hindcast_newtsis/gofs30_withpies/incup/incupd*

Observations (ssh, sst, and T and S profiles): /data/COAPS_nexsan/people/abozec/TSIS/IASx0.03/obs/qcobs_mdt_gofs/WITH_PIES/tsis_obs*.nc

Variables assimialated with TSIS* are: Temperature (temp), density (calculated from T and S), thickness (thknss), and u and v baroclinic (3D).

Example: For the analysis of day 6 the inputs are:

  • Model state of day 5 (archv.2009_005)
  • Observations of day 5 tsis_obs_ias_2009_0105

TSIS generates file for day 5: incupd.2009_005_00

Available variables in observation's files: ssh, sst, uc, vc, av_ssh (aviso ssh). Also all of them have an error variable associated with them.