Skip to contents

Computes availability: the probability that an animal is at the surface, and so can be seen, while it is within the observers' view. Use it to build the availability component of a g(0) correction, one row per key (month, platform, year), and pass the result to g0().

Usage

availability(
  surface,
  dive,
  window,
  se_surface = NULL,
  se_dive = NULL,
  cov_surface_dive = 0,
  key = NULL
)

Arguments

surface

Numeric. Mean surfacing interval, in seconds. Must be positive. A vector gives one result per element.

dive

Numeric. Mean diving interval, in seconds. Must be positive. Length 1 or the length of the longest input.

window

Numeric. Time the animal is in view, in seconds. Must not be negative. Derive it with view_window_aerial(), view_window_vessel() or view_window(). Length 1 or the length of the longest input.

se_surface, se_dive

Numeric or NULL (default). Standard errors of surface and dive, in seconds. Give both or neither: propagating one and dropping the other would understate the total variance.

cov_surface_dive

Numeric. Covariance between the two interval means, in seconds squared. Default 0, which is right when they come from separate follows and optimistic otherwise. Used only with standard errors.

key

Optional vector of labels (months, platforms, years), one per result. Must have the same length as the result. Stored as character.

Value

A tibble with one row per element of the longest input, and columns:

key

The label from key, or NA.

component

Always "availability".

value

Availability, between 0 and 1.

se

Standard error of value, or NA without se_surface and se_dive.

This is the component shape g0() accepts, so availability and perception rows can be stacked.

Why availability is computed, not estimated

An animal that is submerged while the platform passes is missed at every perpendicular distance equally. Availability therefore scales the detection function \(g(x)\) without changing its shape, and the distances carry no information about it. The inputs come from outside the survey: surface and dive from focal follows, tags or drones, and window from the platform's geometry and speed.

How it works

Following Laake et al. (1997), for a mean surfacing interval \(E(s)\), a mean diving interval \(E(d)\), and a window \(w\) (the time the animal is in view):

$$a = \frac{E(s) + E(d)\left(1 - e^{-w/E(d)}\right)}{E(s) + E(d)}$$

At \(w = 0\), \(a\) is \(E(s)/(E(s) + E(d))\), the proportion of time at the surface. As \(w\) grows, \(a\) approaches 1. The middle term is the chance that an animal that was submerged when the window opened surfaces before it closes.

  1. The inputs are checked and recycled to a common length. Each must have length 1 or the common length.

  2. surface and dive must be positive and window must not be negative.

  3. se_surface and se_dive must be given together or not at all.

  4. \(a\) is computed for each element.

  5. If standard errors were given, the standard error of \(a\) is computed by the delta method: the gradient of \(a\) with respect to \((E(s), E(d))\) is taken numerically and combined with their covariance matrix, as in effect_estimates_ddf().

Without se_surface and se_dive the standard error is NA. The function does not invent a precision, because the CV of availability often dominates the CV of abundance. If the raw focal follows are available, bootstrapping them (as Ganley et al. 2019 did) avoids assuming the interval means are jointly normal; the result can be passed as se_surface and se_dive, or the component rows can be built directly.

Use one value per key, not one constant

Availability varies. Ganley et al. (2019) found right whale availability in Cape Cod Bay ranging from 0.27 in January to 0.91 in April, following the depth of the copepods the whales fed on (see ganley_availability). A single constant suited to one month can be wrong by a factor of three for another. Because it scales every candidate model equally, model selection cannot detect the error. Compute one value per key and pass a vector.

This function gives availability only. Perception (whether an animal at the surface and in view is seen) is a separate component and needs a second observer team to estimate. g0() warns if it is missing.

References

Laake, J.L., Calambokidis, J., Osmek, S.D. and Rugh, D.J. (1997) Probability of detecting harbor porpoise from aerial surveys: estimating g(0). The Journal of Wildlife Management 61:63-75. doi:10.2307/3802415 Source of the formula.

Laake, J.L. and Borchers, D.L. (2004) Methods for incomplete detection at distance zero. In Advanced Distance Sampling, pp. 108-189. Oxford University Press. Why perception and availability need separate data.

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 Focal follows and monthly availability for right whales.

Roberts, J.J., Yack, T.M., Fujioka, E., Halpin, P.N., Baumgartner, M.F. and others (2024) North Atlantic right whale density surface model for the US Atlantic evaluated with passive acoustic monitoring. Marine Ecology Progress Series 732:167-192. doi:10.3354/meps14547 Corrections for perception and availability by platform, team and conditions.

See also

view_window_aerial(), view_window_vessel() and view_window() for the window; g0() for combining components; sweep_models() for the detection function fit, which this does not enter.

Examples

# With a zero window, availability is the proportion of time at the surface
availability(surface = 60, dive = 240, window = 0)
#> # A tibble: 1 × 4
#>   key   component    value    se
#>   <chr> <chr>        <dbl> <dbl>
#> 1 NA    availability   0.2    NA

# A longer window raises it: some submerged animals surface in time
availability(surface = 60, dive = 240, window = 30)
#> # A tibble: 1 × 4
#>   key   component    value    se
#>   <chr> <chr>        <dbl> <dbl>
#> 1 NA    availability 0.294    NA

# One row per month
availability(
  surface = c(45, 60, 90),
  dive    = c(300, 240, 150),
  window  = 24,
  key     = c("Feb", "Mar", "Apr")
)
#> # A tibble: 3 × 4
#>   key   component    value    se
#>   <chr> <chr>        <dbl> <dbl>
#> 1 Feb   availability 0.197    NA
#> 2 Mar   availability 0.276    NA
#> 3 Apr   availability 0.467    NA

# With standard errors on the intervals
availability(surface = 60, dive = 240, window = 24,
             se_surface = 8, se_dive = 25)
#> # A tibble: 1 × 4
#>   key   component    value     se
#>   <chr> <chr>        <dbl>  <dbl>
#> 1 NA    availability 0.276 0.0297