Computes the TRMSE between simulated and observed values using a Box-Cox
transformation, reducing the influence of extreme values.
Usage
trmse(sim, obs, lambda = 0.3)
Arguments
- sim
numeric. Simulated values.
- obs
numeric. Observed values.
- lambda
numeric. Box-Cox transformation parameter. Default: 0.3.
Value
numeric. The TRMSE value.
Details
TRMSE applies a Box-Cox transformation to both observed and simulated
values before computing RMSE:
$$
TRMSE =
\sqrt{\frac{1}{n} \sum_{i=1}^{n} \bigg(
\frac{(S_i + 1)^{\lambda} - 1}{\lambda} -
\frac{(O_i + 1)^{\lambda} - 1}{\lambda}
\bigg)^2}
$$
where \(S_i\) and \(O_i\) are simulated and observed values,
respectively. NA values are removed before computation.
References
van Werkhoven, K., Wagener, T., Reed, P., & Tang, Y. (2009).
Sensitivity-guided reduction of parametric dimensionality for multi-objective
calibration of watershed models.
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 TRMSE
trmse(sim, obs)