Per-variable BGC bias vs observations — diag_nutrient

Reports bias for whichever candidate variables this lake actually has usable data for.

Usage

diag_nutrient_budget(
  aeme,
  model,
  candidates = c("CHM_oxy", "NIT_amm", "NIT_nit", "NIT_tn", "PHS_frp", "PHS_tp",
    "PHY_tchla"),
  min_obs = 8,
  bias_ok_frac = 0.25
)

Arguments

aeme: Aeme object; already built and run for model.
model: character; single model code to diagnose (see check_model).
candidates: character vector; AEME variable names to check.
min_obs: integer; below this many in-window observations, the variable is skipped.
bias_ok_frac: numeric; a variable is classified "ok" when the absolute mean bias is below this fraction of the observed mean (a variable-appropriate relative threshold, since oxygen/nitrate/ chlorophyll live on very different absolute scales).

Value

data frame with columns var, mean_bias, mean_obs, rel_bias, n, direction ("model too high", "model too low" or "ok"). Zero rows if no candidate variable has enough usable data.