Ratio of surveyed area to total survey area in 2001 GOA survey

[MargaretSiple-NOAA]

Data product requested:
An estimate of the survey footprint covered by the actual survey and the total survey area, so we can make a ratio and check it against the value they've been using in the FHS stock assessment. I (Megsie) can self-assign this but may not get to it promptly so I'm posting it here in case anyone else has already done it or knows exactly how to.

Output type:
Single value (s) in txt or just the number.

Species: Technically for FHS but could be any species.

Region : GOA

Research team making the request:
@MargaretSiple-NOAA for @mkapur-noaa

More info: In 2001 the GOA survey wasn't completed. Different assessments have different approaches for handling this. Maia would like an area of coverage (like, the area covered by the survey points) to make sure the constant used as a multiplier is appropriate. Janky image of 2001 survey below. I think we just need the area covered by the red points compared to the total survey area.

[mkapur-noaa]

A first step to this request would be to clarify/confirm that the current design-based expansion method does not appropriately deal with this already. My naïve intuition would be that the DB approach should have a unique number of "surveyed" cells for each year, which would be a smaller number in this case, and the resultant biomass estimate would already be corrected for the reduced survey footprint (for all species).

If that is not the case (if, for example, the number of cells used to compute the expansion) doesn't change every year, it'd be useful to make that clear here and it is then likely necessary to explore this footprint calculation as a one-off. Thanks.

[zoyafuso-NOAA]

I need to subscribe to this repo. I totally missed this and sorry for the lack of action on here @mkapur-noaa @MargaretSiple-NOAA .

The 2001 survey was reduced to entirely exclude the Yakutat and SE INPFC areas and to only go down to 500 m. From the data report (page 3, Tech Memo):

"The total area the [2001] survey represents is approximately 220,569 km² (Table 1). In most years the Gulf of Alaska survey is continued to include the Yakutat and Southeast Alaska INPFC areas which extend all the way to the southern end of Alaska to Dixon Entrance. Survey depths also usually extend out to 1,000 m, making the usual survey area about 320,000 km². "

Our data products for regional biomass/sizecomp/agecomp estimates from the 2001 GOA survey does not correct for the non-sampled area in the Eastern GOA. The regional estimates are simply truncated and as a result are not very comparable to other years.

[mkapur-noaa]

OK, this is really helpful to know, since to my knowledge folks have been doing ad-hoc adjustments - mine was 0.9. I will point our group to this thread. Let's leave it open for now in case people have other thoughts.

My initial confusion remains, and perhaps reveals a major misunderstanding on my part -- can you (briefly) explain why the design-based biomass estimates (what gets provided as a data product) would not account for this change? My instinctive thought was that the expansion protocol accounts for something like number of cells, a proxy for area survey, when providing the total biomass for a given year. What did I miss?

[MargaretSiple-NOAA]

Thank you Zack for that info from the TM. I will make an attempt at answering your question Maia since we talked about this before too. If I were redoing that in the assessment for FHS, I would investigate using an adjustment closer to that ratio, though it makes sense that it could be different for less uniformly distributed species. I do think this is important to talk about in the assessment world, though I was not at NOAA in 2001. Maybe they discussed it at length in Plan Team that year and the institutional knowledge slowly dissipated.

I will make an attempt at answering your question Maia since we talked about this before too.

Maybe this will clarify why the design-based estimator doesn't automatically account for the different coverage in 2001: The way the estimator is calculated is by first calculating stratum biomasses, then summing them to get total GOA biomass. The strata are based on INPFC areas (Yakutat and SE are two of these) among other things. Since in 2001 the survey stopped in the middle and just wasn't finished, there are several strata in those more eastern areas that have no hauls at all. If we had, for example, sparse survey data in those strata, we would have an estimate (because the CPUEs would be expanded!) but since they're missing entirely, there's no data to expand to the stratum area, and those stratum biomasses are NAs in the calculations. Therefore, the index for 2001 should be thought of as a total biomass index for the surveyed area which is about 70% of the total area.

If I were snatched from my survey chair and tasked with the assessment for FHS, since it's pretty uniformly distributed, I would investigate using an adjustment closer to that ratio, though it makes sense that it could be different for other spps with different distributions.

Maia, did that answer your question about the db estimator?

[mkapur-noaa]

Thanks Megsie, that is helpful regarding the DB estimator's treatment of those cells for the biomass value -- what about the uncertainty? I'd think the reduction in sampled cells would lead to (rightfully) increased CVs, but with the total absence of certain strata I could also see it being a wash (no perceptible change in CV for most species).

Thanks for your comments about FHS particularly. FWIW, the 2022 assessment found very little sensitivity to this value (it's nearly a generation ago at this point) but it's good to have a record of this conversation.

[MargaretSiple-NOAA]

Yep, the CV is the same way.
For 2001 it is pooled from the strata that are covered by the survey that year and the CVof stratum biomass in the un-surveyed area is NA and not included in the CV calculation for total biomass.