Float profiles should be also classified to chose the same class from reference dataset

• Load float data (argopy)
• add float data to dataset
• create function add_floatdata
• delete profiles already in reference dataset
• make classification

• Interpolation standard levels: using argopy functions
• Spatial and temporal selection: using a .txt file with max and min lat and lon for each box
• Read more than one .mat file
• Create a .py with functions

Use OW ellipses to select reference data before classification

• data selection using simple ellipses
• read longitude_large and latitude_large from ow_config.txt
• change season option to get training dataset
• use PV
• read paths in ow_config.txt

Test method with float 3901928, drifting in the Southern Ocean and which cross the polar front.
https://fleetmonitoring.euro-argo.eu/float/3901928

• make optimal classification
• use OW with class selection
• figure number of profiles in ref by profile
• try k =5 and k =6

Try correction with float 3901915 drifting in south african coast.
https://fleetmonitoring.euro-argo.eu/float/3901915

• create .mat file
• classification
• correction in OW

Hi @AndreaGarciaJuan @cabanesc ,
Where should I start ?
I understood that the float to be validated data are processed in OWC matlab in a specific manner.
Where is this explained please ?

I anticipate that if argopy would mimick this processing, it could be made available with the xarray accessor:

fetcher = ArgoDataFetcher(mode='expert').float(2901623).load()
ds = fetcher.data
ds_processed = ds.argopy.to_owc()

https://fleetmonitoring.euro-argo.eu/float/6900230

• notebook: preproc_gdac_files_before_pyxpcm.ipynb

Test results only usind CTD data with float 3901915 (Agulhas current)

Plot of argo reference dataset used for classification (before interpolation to standard levels)

Create a first notebook to apply the PCM to the argo reference database

• Example: Nord Atlantic
• Output: .txt with classes
• Option: argo, CTD, argo +CTD (use the wmo_boxes file)
• use different depths (z=500m now)
• BIC plot
• Time plot
• add seasonal selection

• change structure of notebook: create trianing dataset and create prediction dataset sections
• use codes in BIC_calculation.py
• random profile in each grid
• clasification figures including quantiles

Create same figure as in Maze et al 2017

• Find reference profiles choosen for profile 78 and profile 80 (temperature)
• Create a notebook to make plot in exemples folder
• Plot standard reference envelop using mean and standard deviation
• PLot with AVISO data
• add TS diagrams
• Plot standard reference envelop using distance weighted mean

Class selection after the 250 profiles are selected

• make a classification using ellipses
• ow using class selection before
• ow using class selection after

• distance calculation (sklearn.metrics.pairwise.haversine_distances) for all points first
• iterations until all points are far enough
• use mask
• how long it takes?

First exemple using float 4900136. As a first step we select profiles in the same class (next step: modify weights)
https://fleetmonitoring.euro-argo.eu/float/4900136

• include float profiles in classification
• test results with k=2

First try using .txt to select only profiles in the same class

• create .mat with data
• good configuration with paths to ref data
• use PCM_classes_NAtlac1_argo.txt file

this could be fixed by adding

# remove duplicates from list and sort into ascending order
boxes_list = sorted(list(set(boxes_list)))

at line 83

In the pcm_owc_config.ini file the MAPSCALE_AGE variable can be only set as the integer value.
It would be good to change it to double.

applying OWC

updating salinity mapper
Traceback (most recent call last):
File "/users/argo/dm_qc/SO_assesment/DMQC-PCM-Python/so_dmqc_master.py", line 244, in
obj.run()
File "/users/argo/dm_qc/SO_assesment/DMQC-PCM-Python/so_dmqc_master.py", line 173, in run
calibration.update_salinity_mapping(str(float_WMO), self.OWC_CONFIG,
File "/users/argo/dm_qc/SO_assesment/DMQC-PCM-Python/argodmqc_owc/pyowc/calibration.py", line 83, in update_salinity_mapping
map_age_large = int(config['MAPSCALE_AGE_LARGE'])
ValueError: invalid literal for int() with base 10: '0.69'

plot the probability of the float profile to be in its class

Compare correction results using the classification and the SAF option in OW (float 3901928)