|BARNHART, BRAD - Environmental Protection Agency (EPA)|
|SAWICZ, KEITH - Environmental Protection Agency (EPA)|
|FICKLIN, DARREN - Indiana University|
Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/20/2017
Publication Date: 8/31/2017
Citation: Barnhart, B.L., Sawicz, K.A., Ficklin, D.L., Whittaker, G.W. 2017. MOESHA: A genetic algorithm for automatic calibration and estimation of parameter uncertainty and sensitivity of hydrologic models. Transactions of the ASABE. 60(4):1259-1269. https://doi.org/10.13031/trans.12179.
Interpretive Summary: Characterization of uncertainty and sensitivity of model parameters is an essential and often overlooked facet of hydrological modeling. This paper introduces an algorithm called MOESHA that combines input parameter sensitivity with model calibration. Model calibration usually has the sole objective of minimizing model error. Our algorithm extends previous work by adding minimization of uncertainty and robustness to model error as objectives. Application of the algorithm to the Dee River catchment in Wales estimates the trade-offs among model error, uncertainty and robustness in model calibration. The results show that the additional objectives using uncertainty and robustness improve calibration and identify sensitive variables.
Technical Abstract: Characterization of uncertainty and sensitivity of model parameters is an essential and often overlooked facet of hydrological modeling. This paper introduces an algorithm called MOESHA that combines input parameter sensitivity analyses with a genetic algorithm calibration routine to dynamically sample parameter space as an alternative to traditional static space-sampling methods, such as stratified sampling or Latin hypercube sampling. In addition to calibrating input parameters to a hydrologic model, MOESHA determines the optimal distribution of input parameters that maximizes model robustness and minimizes error. Subsequently, the variance of the input parameter distributions are used to differentiate between impactful and non-impactful input parameters. In this way, an optimally calibrated model is produced, and estimations of model uncertainty as well as the relative impact of input parameters on model output (i.e., sensitivity) are determined. A case study using a single-cell hydrological model (EXP-HYDRO) is used to test the method using river discharge data from the Dee River catchment in Wales. We compare the results of MOESHA with Sobol’s global sensitivity analysis method and demonstrate that the algorithm is able to pinpoint non-impactful parameters while achieving excellent calibration results.