Assess severe fragmentation from occurrences and an isolation distance

Clusters occurrences into subpopulations - groups of occurrences no further apart than an isolation distance - and quantifies how much of the population lies in small subpopulations, to guide the "severely fragmented" judgement of IUCN Criterion B sub-criterion (a). This mirrors the fragmentation step of sRedList (Cazalis et al. 2024, after Santini et al. 2019) but from occurrence data alone, so no habitat model or density estimate is required.

Usage

assess_fragmentation(
  points,
  isolation_km,
  weights = NULL,
  small_size = NULL,
  n_thresholds = 50
)

Arguments

points: An sf of POINT geometries (one species). Use read_occurrences to produce it.
isolation_km: Isolation distance in km: occurrences further apart than this are treated as belonging to isolated subpopulations. Required.
weights: Optional numeric vector, one weight per occurrence in points (a relative subpopulation-size contribution, e.g. occupied area). NULL (default) weights every occurrence equally (1).
small_size: Optional single size the assessor considers "small". When supplied, prop_in_small (share of the population in subpopulations \(\le\) this size) and severe_suggested (TRUE when that share \(\ge\) 0.5) are returned. NULL (default) leaves the judgement to the curve.
n_thresholds: Number of points on the fragmentation curve (default 50).

Value

A list with n_subpop, isolation_km, sizes (a data.frame of cluster, size, n_points, ordered largest first), clusters (the subpopulation hulls/buffers as an sfc in WGS84), curve (a data.frame of size and prop_pop_le), median_subpop_size, largest_subpop_pct (percent of the population in the single largest subpopulation), prop_in_small / severe_suggested (when small_size is given, else NA) and crs_laea.

Details

Two occurrences are placed in the same subpopulation when they are within isolation_km of each other (a circular-buffer clustering with radius isolation_km / 2, unioned on a data-centred equal-area projection, as in calc_subpop). The relative size of each subpopulation is the sum of its occurrence weights (the number of occurrences by default; supply, for example, per-occurrence occupied area for an area-weighted size). The fragmentation curve then gives, for a range of candidate "small subpopulation" sizes, the proportion of the total population contained in subpopulations no larger than that size; the median_subpop_size is the size at which that proportion reaches one half (half the population lives in subpopulations smaller than this). If the assessor considers a subpopulation of median_subpop_size to be "small", the population can be considered severely fragmented.

This is a screening proxy: occurrence counts are sampling-biased and are not individuals. Use the result only when the mapped clusters make ecological sense as subpopulations, exactly as sRedList advises.

References

Santini L. et al. (2019) Applying habitat and population-density models to land-cover time series to inform IUCN Red List assessments. Conservation Biology 33:1084-1093. doi:10.1111/cobi.13279

Cazalis V. et al. (2024) Accelerating and standardising IUCN Red List assessments with sRedList. Biological Conservation 298:110761. doi:10.1016/j.biocon.2024.110761

Examples

f <- system.file("extdata", "example_occurrences.csv", package = "mappingAS")
occ <- read_occurrences(f)
sp1 <- occ[occ$species == occ$species[1], ]
fr <- assess_fragmentation(sp1, isolation_km = 20)
fr$n_subpop
#> [1] 4
fr$median_subpop_size
#> [1] 1