Table S1 of Ganley et al. (2019): mean dive and surface intervals, percent time at the surface, and median availability, from 86 focal follows of North Atlantic right whales in Cape Cod Bay in 2016 and 2017. These are real measurements, unlike example_dive_intervals.
Format
A tibble with 4 rows and 8 columns:
- month
Month, as an ordered factor from January to April.
- percent_surface_time
Percent of time at the surface. Not the ratio of the two interval columns; see above.
- mean_dive
Mean diving interval, seconds.
- mean_surface
Mean surfacing interval, seconds.
- availability
Median availability, \(a(x)\).
- availability_variance
As labelled in the source; see above.
- n_follows
Number of focal follows. January has 7 and April 48: the month with the lowest availability is also the least measured, because the weather that keeps the food deep also keeps observers on the ground.
- hours_followed
Total follow time, hours.
The pooled "All sightings" row of Table S1 is stored as the
all_sightings attribute, not as a fifth row, so it is not summed or
plotted with the months by mistake.
Source
Ganley, L.C., Brault, S. and Mayo, C.A. (2019) What we see is not what there is: estimating North Atlantic right whale Eubalaena glacialis local abundance. Endangered Species Research 38:101-113. doi:10.3354/esr00938 , Table S1. Open access under CC-BY.
Details
Availability runs from 0.27 in January to 0.91 in April, as the copepods
the whales feed on move up through the water column. This threefold change
within one season is measured evidence against a constant g(0).
Three things not to do with these columns
- Do not treat
percent_surface_timeas \(E(s)/(E(s)+E(d))\) The two do not match. January lists 16%, but its mean intervals (48 s surface, 533 s dive) give \(48/(48+533)\) = 8.3%. The mismatch occurs every month and in both directions: April's intervals give 90% against a listed 55%. The percentage appears to be a mean of per-follow percentages, while the interval columns are means of intervals (a mean of ratios against a ratio of means). Using it as instantaneous availability would be wrong by up to 35 percentage points.
- Do not expect
availability()to reproduceavailability The reported value is a median over bootstrap replicates, and availability depends on the time in view, which varies with distance. Working back from the tabulated means, January's 0.27 implies a window near 122 s and April's 0.91 a window near 8 s, so the months were not evaluated at a common window. Use the values as measurements; do not recompute them.
- Do not pass
availability_variancetog0()without deciding what it is It is labelled a variance and ranges from 0.04 to 0.09. Read as a variance, January's standard error is \(\sqrt{0.04} = 0.2\) against an estimate of 0.27, a CV of 74%, which would dominate any correction. Read as a standard error, the CV is 15%. Neither matches the very small error bars in the paper's Fig. 4A. The column keeps the paper's label and is not converted.
The survey platform
A Cessna 336/337 Skymaster at 185 km/h and 228 or 304 m altitude, surveying
January to May. Time in view at the effective trackline was 51.22 s, rising
with perpendicular distance (see view_window_aerial()). The effective
trackline is at 100 m, not 0 m, because the aircraft's flat windows leave a
blind spot beneath it, and the surveys were left-truncated at 100 m. At the
fitting end the same blind spot is handled by sweep_models(left = ) or the
gamma key.
Two inconsistencies in the source
These are in the paper, not transcription errors.
February's sample size is 9 in Table S1 (summing to the table's total of 86) but 10 in the Fig. 2 caption (summing to the 87 stated in the Methods).
April's availability is 0.91 in Table S1 and Section 3.1, but the abstract gives the seasonal range as 0.27-0.85.
Table S1 is followed here, as the tabulated source.
Examples
ganley_availability
#> # A tibble: 4 × 8
#> month percent_surface_time mean_dive mean_surface availability
#> <fct> <dbl> <dbl> <dbl> <dbl>
#> 1 Jan 16 533 48 0.27
#> 2 Feb 34 256 226 0.52
#> 3 Mar 31 219 67 0.52
#> 4 Apr 55 88 801 0.91
#> # ℹ 3 more variables: availability_variance <dbl>, n_follows <int>,
#> # hours_followed <dbl>
# The monthly values as a g(0) component. A standard error must be chosen
# first (see the variance column above). This example reads the column as a
# variance; decide which reading is right before relying on it.
avail <- data.frame(
key = as.character(ganley_availability$month),
component = "availability",
value = ganley_availability$availability,
se = sqrt(ganley_availability$availability_variance)
)
suppressWarnings(g0(avail))
#> <dsfit_g0>
#> assumed 1: perception
#> components:
#> availability source not recorded
#>
#> key g0 se cv
#> Jan 0.27 0.2000 0.741
#> Feb 0.52 0.3000 0.577
#> Mar 0.52 0.2449 0.471
#> Apr 0.91 0.2236 0.246
#>
#> Divide abundance by g0, and propagate cv. This package does not
#> apply it: correction happens at the abundance step.
# The pooled row
attr(ganley_availability, "all_sightings")$availability
#> [1] 0.63