Skip to contents

Goodness-of-fit (GOF) metrics and FDC-based signatures

Source: R/performance-02-metrics.R
gof_metrics.Rd

Computes a set of goodness-of-fit (GOF) metrics and flow duration curve (FDC) signatures to evaluate model performance by comparing simulated vs. observed values.

Usage

gof_metrics(
  obs,
  sim,
  metric,
  minimize = FALSE,
  all_metrics = FALSE,
  digits = 4
)

Arguments

obs

numeric. Observed values.

sim

numeric. Simulated values.

metric

character. Name of the metric to compute. See Details for the list of available metrics.

minimize

logical. Whether to transform the metric so that it can be used as a minimization objective in optimization-based model calibration workflows. Default: FALSE.

all_metrics

logical. Whether to compute all available metrics. If TRUE, a tibble with all metrics is returned. Default: FALSE.

digits

integer. Number of decimal places for rounding. Default: 4.

Value

If all_metrics = FALSE, a numeric value for the requested metric. If all_metrics = TRUE, a tibble with two columns:

  • gof: Metric name.

  • value: Metric value (rounded).

Details

The following metrics are available. Parentheses indicate value domains (e.g., -Inf to Inf), and brackets indicate exceedance probability segments of FDC (e.g., [-EP: 0–2%]):

  • NSE: Nash-Sutcliffe efficiency (-Inf to 1).

  • log_NSE: Log-transformed NSE (-Inf to 1).

  • mNSE: Modified NSE (-Inf to 1).

  • rNSE: Relative NSE (-Inf to 1).

  • KGE: Kling-Gupta efficiency (-Inf to 1).

  • KGElf: KGE for low values (-Inf to 1).

  • KGEnp: Non-parametric KGE (-Inf to 1).

  • KGEkm: Knowable Moments KGE (-Inf to 1).

  • RMSE: Root mean squared error (>= 0).

  • TRMSE: Box-Cox transformed RMSE (>= 0).

  • MAE: Mean absolute error (>= 0).

  • PBIAS: Percent bias (-Inf to Inf).

  • VE: Volumetric efficiency (-Inf to 1).

  • RSR: Ratio of RMSE to SD of observations (0 to +Inf).

  • rPearson: Pearson correlation (-1 to 1).

  • rSpearman: Spearman correlation (-1 to 1).

  • R2: Coefficient of determination (0 to 1).

  • FDCsign: Aggregated FDC signature (>= 0).

  • FDC_Speak: Percent bias in peak segment volume [-EP: 0–2%] (-Inf to Inf).

  • FDC_Shigh: Percent bias in high segment volume [-EP: 2–20%] (-Inf to Inf).

  • FDC_Smid: Percent bias in midsegment slope [-EP: 20–70%] (-Inf to Inf).

  • FDC_Slow: Percent bias in low segment volume [-EP: 70–100%] (-Inf to Inf).

Metrics TRMSE and all FDC* signatures are computed using functions from this package (trmse and fdc_signature), while the remaining metrics are computed with the hydroGOF package.

See also

Other Performance evaluation: evaluate_swat(), fdc_signature(), pbias_fdc_midsegment(), pbias_fdc_target_segment(), plot_fdc(), plot_mean_annual_cycle(), plot_timeseries()

Examples

# Synthetic daily flow data
set.seed(123)
obs <- abs(rnorm(730, mean = 50, sd = 20))
sim <- obs * runif(730, min = 0.8, max = 1.5)

# Compute a single metric (NSE)
gof_metrics(obs, sim, "NSE")

# Compute all metrics
gof_metrics(obs, sim, all_metrics = TRUE)