Predict at new locations via universal kriging

Given the fitted stage 2 model, predict the full parameter vector and its covariance at new spatial locations. Then extract the target components (default: first source's params, e.g. 1-3 for NOAA) and their marginal covariance.

Usage

predict_krig(model, new_sites, predict_params = NULL)

Arguments

model

A fitted evfuse_model from fit_spatial_model.

new_sites

Data frame with lon and lat columns for prediction locations.

predict_params

Integer vector of parameter indices to extract for predictions (default: first source's params, i.e. 1:3).

Value

A list with components:

pred_mean

Matrix (n_new x p) of predicted parameter means.

pred_cov

List of n_new pxp predictive covariance matrices.

noaa_mean

Matrix (n_new x length(predict_params)) of predicted target GEV parameters (mu, log_sigma, xi).

noaa_cov

List of n_new marginal covariances for target params.