Expand SA quantile samples into one parameter set per run

get_parameter_samples returns sensitivity analysis samples as one table per PFT, quantiles down the rows and traits across the columns. The config writers want the opposite shape: one row per model run, holding the parameter values that run uses. This converts between the two, using the labels the design carries.

Usage

sa_run_samples(sa_samples, design_matrix)

Arguments

sa_samples

Sensitivity analysis samples, a named list with one data.frame per PFT, quantiles as rownames and traits as columns, as returned by get_parameter_samples.

design_matrix

The design from generate_OAT_SA_design, carrying sa_pft, sa_trait and sa_quantile.

Value

A named list with one data.frame per entry of sa_samples, each with one row per design row and one column per trait. This is the shape write.ensemble.configs takes as ensemble.samples.

Details

Every run starts from each PFT's median values. A run that the design labels as moving a trait then has that one value replaced with the trait's value at the labelled quantile, which is what makes a one-at-a-time design separable. The first row, the median run, keeps every value at its median.

Entries that are not PFTs, such as env, are carried at their medians for every run, matching what write.sa.configs passes today.

See also

generate_OAT_SA_design for the labels this reads.