How much do two predicted distributions overlap?

Compares two suitability surfaces cell by cell and reports how similar they are. Used to ask whether two species, two seasons, or the same species under two climate scenarios occupy the same space.

Usage

niche_overlap(x, y, statistic = c("both", "D", "I"), na.rm = FALSE)

Arguments

x, y

Two SpatRaster layers of predicted suitability, or two numeric vectors of equal length. Must line up cell for cell.

statistic

"D" for Schoener's D, "I" for Warren's I, or "both".

na.rm

Whether to drop cells missing from either surface. See Details.

Value

A named numeric vector.

Details

Both statistics run from 0 (no overlap) to 1 (identical), and both begin by rescaling each surface to sum to 1, so what is compared is the shape of each distribution rather than its level. A model predicting uniformly higher suitability than another can still overlap it perfectly.

Schoener's D is one minus half the summed absolute difference. Warren's I works on square roots, which makes it less sensitive to a handful of cells where the two surfaces disagree sharply. They usually agree; where they do not, D is being moved by a few strong disagreements and I by the broad pattern.

na.rm defaults to FALSE for the same reason ensemble_summary() does. Dropping cells missing from one surface compares the two over a domain neither was asked about, and nothing in the resulting number says how much was discarded. Make the coverage match first, deliberately.

Neither statistic is a test. A D of 0.7 is not evidence of anything on its own – two surfaces built from the same covariates over the same domain will overlap substantially whatever the species do. niche_equivalency() is the randomisation test that gives it a reference distribution.

References

Warren, D. L., Glor, R. E., & Turelli, M. (2008). Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution. Evolution, 62(11), 2868-2883. doi:10.1111/j.1558-5646.2008.00482.x

See also

niche_equivalency() to test it against a null, plotUncertainty() for ensemble spread.

Other spatial plots: ensemble_summary(), hex_bin(), mess(), niche_equivalency(), plot.fancyfx_equivalency(), plotExtrapolation(), plotHexbin(), plotUncertainty(), thin_points()

Examples

set.seed(1)
a <- runif(400)
b <- a * 0.8 + runif(400) * 0.2

niche_overlap(a, b)
#>         D         I 
#> 0.9334834 0.9948262 
niche_overlap(a, a)          # identical surfaces
#> D I 
#> 1 1