Review krill inversion code

[leewujung]

Review krill inversion code.

Main goals are to clarify the below components:

• [x] What input files are required - answer
• [x] What output files are generated - answer
• [x] What are the data pre-processing steps before inversion? - answer
• [x] What scattering model(s) are used - answer
• [x] What inversion algorithm(s) does the code use - answer

References here are:

• Matlab code for 2020 processing
• Basic documentation of processing steps: https://docs.google.com/document/d/10PsW2ZiQvZAQGtDJU-_1xYX6Hh8PC7pW/edit

[leewujung]

What inversion algorithm(s) does the code use

Matlab lsqnonlin

[leewujung]

What are the data pre-processing before inversion?

• Restructure Echoview NASC exports --> save as organized .mat files with names like US-euph-2019 hake survey.mat
• Extract relevant data values for inversion
  • Output data in xlsx: Sv mean, std, median, std of median (med_std), layer depth, layer thickness, effective volume, lat/lon
  • First round of clean up:

    for tr in transects:  # loop through all transects (files)
    for f in freq:  # loop through all available frequencies
      if data at f kHz exists:
         # extract relevant data

  • Remove inconsistent Sv samples at certain frequencies
    • there are three choices: "Accept", "Automated", and "Manually"
    • goal is to retain data that is consistent with how krill backscatter should look like
    • observation from running the code to produce plots:
    • when "Automated" is selected, the 18 kHz and 38 kHz data points are corrected using the function automated_data_proc. From a quick look, the correction is an attempt to remove potential "contamination" from mesopelagic fish based on krill scattering model prediction
    • when "Accept" is selected, all data points are accepted as is
  • Save outputs as excel files with names like input_para_data_hake2017-US.xlsx

[leewujung]

• What input files are required
• What output files are generated

• Echoview NASC exports
  • input to Display_Proc_exported_transect_data.m
  • x10_18_0.5nmi_10m_cells_50to300m.csv (from the fname field in the organized .mat file)
  • x10_18_10m_cells_50to300m.csv (from the file_ID field in the organized .mat file)
  • the exports are organized using script Display_Proc_exported_transect_data.m and save to a sheet called ("hake_YEAR")
  • input_para_data_hake2019-US_bkup.xlsx contains many years of data in different hake_YEAR sheets
• central input excel file
  • input to the main inversion script KrillSvInversion_simu_data_YEARDATE.m
  • with names like input_para_data_hake2017-US.xlsx
  • initial spreadsheets generated from Echoview NASC exports
  • manually added (?) spreadsheets for model parameters
  • not sure where the sheet bio_data come from, but these look like simple numbers to parameterized the scattering model
• outputs from the main inversion script KrillSvInversion_simu_data_YEARDATE.m
  • input to the "analysis" scriptinversion_results_analysis_YEARDATE.m
  • in folder output hake survey - euphausiids
  • have names like 2013 US Regs1-98_mean_data_inversion_results_09-May-2020_21-0-16.xlsx
• outputs from the "analysis" script inversion_results_analysis_YEARDATE.m
  • default output name is Generic_Output_File.xlsx
  • content of Generic_Output_File.xlsx is copied to a specific filename 2013_CAN_inversion_output50-300m-additional.xlsx

[leewujung]

What scattering model(s) are used

• function DWBAbscat1: weakly scattering uniformly bent finite inhomogenous body, with segmented g and h along the body
• with orientation average
• with length average

See: https://github.com/uw-echospace/EchoPro_krill_inversion/blob/main/inversion/functions/SVpredictionALL.m#L56 and https://github.com/uw-echospace/EchoPro_krill_inversion/blob/main/inversion/Scat_models/zoo_bscat.m#L59-L60

[leewujung]

I'll close this now as I've transferred the content to the matlab krill repo README.