From boundary delineation to causal interpretation in wetland mapping: HRU-scale wetland identification and evolution...

Wetland degradation and areal loss driven by global change and intensified human activities pose increasing challenges to wetland conservation and restoration. Reliable decision-making requires accurate identification of wetland extent and type, which is currently dominated by remote sensing-based mapping. However, gradual wetland boundaries, high intra-class heterogeneity, and spectral similarity among land-cover types often lead to classification uncertainty. Concealed wetlands are particularly prone to omission because of canopy obstruction, temporary absence of open water, and weak spectral separability. More importantly, remote sensing products alone have limited capacity to explain the hydrological mechanisms governing wetland expansion, contraction, and persistence under varying flow regimes. Hydrological connectivity provides a process-based perspective for linking wetland occurrence and function to runoff generation, lateral flow redistribution, and groundwater-surface water interactions within watersheds. To address these limitations, this study developed a SWAT+-based hydrological connectivity wetland classification system, termed SWAT+ HWC. Driven by physically based simulations at the hydrological response unit (HRU) scale, SWAT+ HWC provides a transferable framework for candidate wetland screening and functional wetland classification. Daily HRU-scale fluxes and state variables were used to derive ten process indicators characterizing dominant hydrological connectivity mechanisms, and candidate wetlands were further classified into three functional units: riparian wetlands, shallow depressional wetlands, and deeply connected wetlands. Classification thresholds were adaptively calibrated in quantile space using a particle swarm optimization-classification and regression tree algorithm combined with a K-out-of-N voting rule, thereby constraining the candidate wetland proportion while improving the spatial transferability of classification rules across watersheds with contrasting hydrogeomorphic settings. Model performance was evaluated through process-consistency analysis, quantitative comparison with the WetlandConnectivity v1.0 dataset, and spatiotemporal validation in the Altamaha River Basin. The identified wetland types showed stable correspondence with their expected dominant hydrological indicators. Quantitative evaluation demonstrated strong discriminatory performance across representative geomorphic units, with most ROC-AUC values ranging from 0.89 to 0.95, and seasonal wetland-area dynamics were highly synchronized with the reference dataset, with correlation coefficients exceeding 0.84. By linking wetland identification to testable hydrological process chains, SWAT+ HWC provides a physically traceable basis for wetland mapping, functional interpretation, and cross-period comparison. The resulting hydrological connectivity-based classification enables wetland mapping to move beyond identifying where wetlands are located toward explaining why wetlands occur and persist in specific landscape positions, thereby improving the interpretability and targeting of wetland conservation, ecological restoration, and watershed water resources management.