hydroeval is an open-source evaluator of goodness of fit between simulated and observed streamflow time series in Python. It is licensed under GNU GPL-3.0. The package provides a bundle of the most commonly used objective functions in hydrological science. The package is designed to calculate all objective functions in a vectorised manner (using numpy, and therefore C code in the background) which makes for very efficient computation of the objective functions.
If you are using hydroeval, please consider citing the software as follows (click on the link to get the DOI of a specific version):
Hallouin, T. (XXXX). HydroEval: Streamflow Simulations Evaluator (Version X.X.X). Zenodo. https://doi.org/10.5281/zenodo.2591217
Brief overview of the API
import hydroeval as he
simulations = [5.3, 4.2, 5.7, 2.3]
evaluations = [4.7, 4.3, 5.5, 2.7]
nse = he.evaluator(he.nse, simulations, evaluations)
kge, r, alpha, beta = he.evaluator(he.kge, simulations, evaluations)
Objective functions available
The objective functions currently available in hydroeval to evaluate the fit between observed and simulated streamflow time series are as follows:
- Nash-Sutcliffe Efficiency (nse)
- Original Kling-Gupta Efficiency (kge) and its three components (r, α, β)
- Modified Kling-Gupta Efficiency (kgeprime) and its three components (r, γ, β)
- Non-Parametric Kling-Gupta Efficiency (kgenp) and its three components (r, α, β)
- Root Mean Square Error (rmse)
- Mean Absolute Relative Error (mare)
- Percent Bias (pbias)
Moreover, some objective functions can be calculated in a bounded version following Mathevet et al. (2006):
- Bounded Nash-Sutcliffe Efficiency (nse_c2m)
- Bounded Original Kling-Gupta Efficiency (kge_c2m)
- Bounded Modified Kling-Gupta Efficiency (kgeprime_c2m)
- Bounded Non-Parametric Kling-Gupta Efficiency (kgenp_c2m)
Finally, the evaluator can take an optional argument transform. This argument allows to apply a transformation on both the observed and the simulated streamflow time series prior the calculation of the objective function. The possible transformations are as follows:
- Inverted flows (using transform='inv')
- Square Root-transformed flows (using transform='sqrt')
- Natural Logarithm-transformed flows (using transform='log')
Acknowledgement
Early versions of this tool were developed with the financial support of Ireland's Environmental Protection Agency (Grant Number 2014-W-LS-5).