Compute nonstationary return levels with covariate — compute_return_levels

Takes kriging predictions of 4 GEV parameters (mu0, mu1, log_sigma, xi) and a future covariate value, and computes r-year return levels with confidence intervals via the delta method and/or simulation.

Usage

compute_return_levels_ns(
  predictions,
  covariate_value,
  r = 100,
  alpha = 0.05,
  method = "both",
  n_sim = 2500,
  seed = NULL
)

Arguments

predictions: An evfuse_predictions object from predict_krig with 4-column noaa_mean and 4x4 covariance matrices.
covariate_value: Scalar or vector (length n_new) of centered covariate values at prediction sites.
r: Return period in years (default 100).
alpha: Confidence level for CIs (default 0.05 for 95% CIs).
method: One of "delta", "simulation", or "both" (default).
n_sim: Number of simulations (default 2500).
seed: Random seed for simulation method.

Value

A data frame with columns: lon, lat, return_level, se_delta, ci_lower_delta, ci_upper_delta, se_sim, ci_lower_sim, ci_upper_sim.

Details

The effective location is mu = mu0 + mu1 * covariate_value, where covariate_value should be centered relative to the same reference used in fit_gev_ns.