Regional Modeling of Long-Term and Annual Flow Duration Curves: Reliability for Information Transfer with Evolutionary Polynomial Regression

Abstract

The estimation of long-term flow duration curves (FDC) and annual flow duration curves (AFDC) are frequently required for water resources planning and management at ungauged catchments. The index flow framework provides a simple mathematical model for linking both approaches in regionalization procedures. However, the reliable transfer of information may be unfeasible with established regression techniques due to the complex structure of variation between the statistical model parameters and catchments’ characteristics. This paper explores the evolutionary polynomial regression (EPR) technique for identifying regional equations for the parameters of the index flow model. Results showed that, at the expense of increased structural complexity, EPR might be an effective and robust tool for transferring information to ungauged sites. Long-term streamflow variability was relatively well captured under cross-validation, although the low flow regimes were misrepresented in most cases. As for the AFDCs, the model proved able to synthesize flow regimes in typical and wet years but failed to do so in dry conditions. Despite these limitations, the proposed approach may constitute a useful tool for supporting water resources management at the regional scale.