Integrating Tracers and Soft Data Into Multi-Criteria Calibration: Implications From Distributed Modeling in a Riparian Wetland

Calibrating distributed hydrological models often leads to equifinality due to complex model structures, which can be further exacerbated in wetlands due to spatio-temporal heterogeneity in ecohydrological processes. Here, step-wise calibrations of the physically-based distributed model EcH₂O-iso was conducted in a data-rich wetland by minimizing a weighted average of the errors on discharge, stream isotopes, groundwater (GW) isotopes, and soil moisture. Results showed multi-criteria calibration outperformed single-criterion calibration as it strongly increased the overall performance, yet only marginally degraded performance of each calibration target. Isotopes were highlighted as appropriate auxiliary data as they effectively constrained the model with relatively small weights (0.1). However, those parameter sets that minimize the errors could still lead to physically implausible simulations of uncalibrated internal states or fluxes. This was further demonstrated by an approach developed to check internal fluxes based on soft data (transpiration and lateral flow), suggesting 54% of optimized models gave “right answers for the wrong reasons.” By excluding those models against soft data, such an approach further constrained equifinality, and unraveled potential inconsistencies between observations and calibration. Modeling represented the wetland as a slow-draining system mainly fed by GW, but also influenced by near-surface flow during winter or summer convectional events. Further, heterogeneity in hydrological functioning was partly attributed to distinct evapotranspiration patterns between contrasting vegetation communities. Therefore, this study not only provided insights into wetland functioning, but also revealed potential equifinality even with abundant data for calibration, and potential solutions based on the integration of isotopes and soft data.