Cross-year comparison and possible reasons for large discrepancy

[leewujung]

Below are notes taken during meeting with @brandynlucca

Cross-year comparison table

1995:

• use pre-consolidated files for NASC input (age-1 removed, extrapolated)
• very large differences in the transect-level biomass density
  • likely due to bio data being different
• cannot plot transects:
  • some transects have very few points (1?) so cannot be plotted
  • due to the duplicated lat/lon points in the transect-level NASC files

1998:

• cannot plot transects
  • some transects have very few points (1?) so cannot be plotted
• somehow the transect-level differences are very large, but kriging outputs are surprisingly similar, especially on biomass (which comes from biomass density)

2003:

• bio data are different!

2005:

• cannot plot transects
  • some transects have very few points (1?) so cannot be plotted
• length bin count in haul data has a specific haul that is very different
• length-age abundance breakdown has the same shape but the absolute numbers are different
• unaged abundance has a large difference, this is at the same length range the length bin counts are very different

2011:

• bio data are different!
  • Echopop input (master spreadsheet) has a gap from haul 50-72
  • EchoPro inputs (3 excel files) does not have anything beyond haul 72
  • length bin counts per haul are different
    • small fish has different dominant length-age grid
    • more contiguous coverage in EchoPro data

2011:

• bio data are different!
  • EchoPro goes up to haul 143
  • Echopop does not have anything going beyond haul 72
  • Echopop also has some missing hauls

[leewujung]

Updates below:

1995:

• acoustic export input file (instead of using what's in proc_*, use the version in another folder in old_*)
• transect results were bigger in Echopop before; now the results are bigger in Echopop
• for biodata: the difference is in the stratum assignment, but the underlying biodata is the same

1998:

• assuming all missing vessel log entries are 0.5 nmi would lead to much closer output values compared with EchoPro

2003:

• acoustic export input file (also there was another version, using the old_* version)
• for biodata: the difference is in the stratum assignment, but the underlying biodata is the same

2005:

• EchoPro somehow produces a very low biomass estimates (200 kmt) that is not consistent with the input data, and also the lowest of all years
• so something is different in the EchoPro implementation specific to 2005

2011:

• for biodata: the difference is in the stratum assignment, but the underlying biodata is the same

cc: @aliciabillings-noaa @rebeccathomas-NOAA

[brandynlucca]

Side-note: for most years where biomass densities were not large offshore toward the western-most extent of the transects, kriging using the default uniform adaptive search method vs FEAT-specific did not yield substantially large differences. Years like 2001, 2003, 2009, 2015, and 2017 appeared more sensitive on the decision to use one over the other.

[brandynlucca]

Looking into things more:

2012
When inspecting the consolidated spreadsheet manually, US haul numbers (ship id = 19) go up to 74; CAN haul numbers (ship id = 499) go up to 44. This differs from the split US (biodata_specimen.xlsx/biodata_length.xlsx) and CAN (biodata_specimen_Canada.xlsx/biodata_length_Canada.xlsx) whose hauls go up to 72 and 44, respectively.

• Biodata mismatch
  • Hauls 32, 70, 66, 68, 73, 74 do not exist in the consolidated spreadsheet
  • Hauls 57, 61, 62, 65 have different counts
• The Canadian data has two different survey IDs for the specimen (201205) and length (201303) datasets.

[aliciabillings-noaa]

Are you using the newly created consolidated spreadsheet? This was a problem tow and the main reason I rebuilt the spreadsheet.

Can you send me the separated biodata files so I can check them against those that I have in our records?

quick data check for reference on what is on the new sheet
US data:
There are no hake in tows 
15, 31, 48, 60 ,63, 66, 70, 73, 74
No hake in biodata_length for tows
4, 5, 15, 17, 21, 31, 32, 40, 43, 48, 49, 53, 56, 60, 63, 64, 66, 68, 70, 72, 73, 74
No hake in biodata_specimen for tows
15, 31, 48, 60, 63, 66, 70, 73, 74

CAN data:
No hake in tows
2, 3, 9, 11, 20, 23, 25, 27, 28, 32, 35, 40
No hake in biodata_length for tows
1-11, 13, 18-21, 23-32, 34-44
No hake in biodata_specimen for tows
2, 3, 9, 11, 16, 20, 23, 25, 27, 28, 32, 35, 40

[brandynlucca]

@aliciabillings-noaa

When using the updated spreadsheet that runs through 2025 these are the observed differences:

biodata_length

• Hauls with different length counts from split US:
  • 57
• Tallied differences:
  • US: 126 fish

biodata_specimen

• No mismatch

biodata_catch

• Haul with different summed weights from split US:
  • 32
• Tallied differences:
  • US: 33.4 kg