Skip to content

Metrics

Goodness of fit metrics for comparing routed discharge against observations.

river_route.metrics

mean_error(y_true, y_pred)

The mean of the simulated minus the observed values: positive when the simulation is too high on average.

mean_absolute_error(y_true, y_pred)

The mean of the absolute differences between the simulated and the observed values.

mean_square_error(y_true, y_pred)

The mean of the squared differences between the simulated and the observed values.

pearson_r(y_true, y_pred)

The Pearson correlation coefficient of the observed and the simulated values.

kling_gupta_efficiency_2012(y_true, y_pred)

Kling-Gupta efficiency per Kling et al. (2012): 1 - sqrt((r - 1)^2 + (beta - 1)^2 + (gamma - 1)^2), with r the correlation, beta the ratio of the simulated to the observed mean, and gamma the ratio of the simulated to the observed coefficient of variation. 1 is a perfect simulation. NaN when either series is constant or has a mean of zero, where a coefficient of variation is undefined.