Skip to contents

Summarise an ensemble of projection rasters

Source: R/spatial.R
ensemble_summary.Rd

Collapses a stack whose layers are ensemble members – competing models, emissions scenarios, bootstrap replicates – into one summary layer. The spread statistics are the point: a projection map without one says only what the ensemble guessed, not how much the members disagreed.

Usage

ensemble_summary(
  x,
  statistic = c("sd", "cv", "range", "iqr", "mean", "median"),
  na.rm = FALSE
)

Arguments

x

A SpatRaster whose layers are ensemble members.

statistic

What to compute across layers at each cell. "sd", "cv", "range" and "iqr" describe disagreement; "mean" and "median" describe the projection itself.

na.rm

Whether to ignore members that are missing at a cell. See Details – the default is deliberately FALSE.

Value

A single-layer SpatRaster.

Details

na.rm defaults to FALSE, which is the opposite of most R summaries and is deliberate. If one member is missing over part of the domain, taking the spread of the members that remain reports a narrower uncertainty exactly where the ensemble is least complete, and nothing on the resulting map says so. Leaving those cells NA makes the gap visible. Set na.rm = TRUE only once you know why the members differ in coverage.

"cv" is the standard deviation over the mean. It is the natural choice when members are on a count or density scale, where a spread of 5 means something very different at a mean of 10 than at a mean of 1000. It is a poor choice for probabilities, where the mean can approach zero and the ratio explodes; use "sd" there.

See also

plotUncertainty() to draw it, mess() for whether the projection is extrapolating in the first place.

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

Examples

if (requireNamespace("terra", quietly = TRUE)) {
  r <- terra::rast(nrows = 20, ncols = 20, vals = runif(400))
  ensemble <- c(r, r * 1.2, r * 0.7)
  names(ensemble) <- c("model1", "model2", "model3")

  ensemble_summary(ensemble, "sd")
}
#> class       : SpatRaster
#> size        : 20, 20, 1  (nrow, ncol, nlyr)
#> resolution  : 18, 9  (x, y)
#> extent      : -180, 180, -90, 90  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (CRS84) (OGC:CRS84)
#> source(s)   : memory
#> name        :       sd
#> min value   : 0.000152
#> max value   : 0.250746