Quantify model-conditional or resampling surface variability

Usage

ps_surface_uncertainty(
  x = NULL,
  points = NULL,
  method = NULL,
  approach = c("kriging_variance", "conditional_simulation", "tps_standard_error",
    "resampling_sensitivity"),
  nsim = 100,
  probabilities = c(0.05, 0.5, 0.95),
  resampling_design = NULL,
  template = NULL,
  mask = NULL,
  keep_realizations = FALSE,
  output_directory = NULL,
  seed = 1,
  progress = NULL,
  exceedance_levels = NULL
)

Arguments

x: A structured interpolation result.
points: Points used for resampling sensitivity when x is absent.
method: Interpolation method for resampling.
approach: Uncertainty or sensitivity approach.
nsim: Number of simulations or resamples.
probabilities: Pointwise quantile probabilities.
resampling_design: "case", "jackknife", or a list with a type and spatial group vector.
template, mask: Mapping geometry controls.
keep_realizations: Retain realization rasters in memory.
output_directory: Optional realization directory.
seed: Deterministic seed.
progress: Optional callback.
exceedance_levels: Optional head levels for exceedance probability.

Value

A potentiomap_uncertainty object. Resampling products are sensitivity summaries, not formal confidence intervals.

Examples

data("synthetic_wells")
p <- ps_make_points(synthetic_wells[1:14, ], "x", "y", "gw_elevation",
                    "well_id", "EPSG:26916")
fit <- suppressWarnings(ps_interpolate(p, methods = "OK", grid_res = 250,
                                       return = "result"))
uncertainty <- ps_surface_uncertainty(fit, approach = "kriging_variance")
uncertainty$method_manifest
#>   method         approach
#> 1     OK kriging_variance
#>                                                                                                             assumptions
#> 1 Model-conditional kriging variance under the fitted trend and variogram; this is not total hydrogeologic uncertainty.
#>   simulation_count seed
#> 1                0    1
# Kriging variance is conditional on the retained covariance model.