Intelligent sediment–groundwater digital twin: A systematic review, meta-analysis, and reference architecture for reliable metal pollution risk assessment

Sediment–groundwater interfaces regulate the mobilisation, transport, and attenuation of metal contaminants within highly heterogeneous and multi-scale subsurface systems. However, efforts to operationalise digital twins in this domain remain fragmented across modelling paradigms, disciplines, and scales of observation and prediction. This study provides a performance-centred, quantitatively harmonised meta-analysis of sediment–groundwater digital twin (SG-DT) implementations, explicitly addressing scale dependence, spatial heterogeneity, and process-level controls. A PRISMA 2020–compliant systematic review and meta-analysis of 175 studies (1997–2026; 27 countries) was conducted. SG-DT applications were stratified across a continuum from pore and laboratory scales to site, reach, and basin domains. Scaling effects were analysed via stratified synthesis and meta-regression, while heterogeneity was operationalised through facies-based zonation, parameter variability, and multi-source observational constraints. Predictive performance was quantified using a redesigned, baseline-consistent effect size (E), enabling harmonisation across heterogeneous metrics and validation protocols. Uncertainty was estimated using random-effects models (REML) with Hartung–Knapp adjustments, and heterogeneity was assessed via τ 2 and I 2 with scale-aware interpretation. Results indicate a robust positive pooled performance gain for surrogate-enabled SG-DTs relative to non-updating baselines under scale-consistent validation, while highlighting limitations related to data dependence and extrapolation. In a subset of comparable studies, SG-DTs achieved a pooled out-of-sample R 2 of 0.67 (95% CI: 0.62–0.72). Meta-regression identifies continuous data assimilation and explicit uncertainty quantification as key drivers of performance, whereas residual heterogeneity reflects unresolved scale mismatches and inconsistent representation of subsurface complexity. SG-DTs are defined as systems coupling a process-informed hydro(geo)chemical core capturing reaction kinetics and flow–transport–geochemical coupling, streaming observations, and an updating operator within a closed-loop framework. As secondary outputs, we provide a scale-aware reference architecture embedded in a continuous verification–validation–uncertainty quantification loop and the SED-GW-DT-REPORT v1.0 standard for reproducible, FAIR-aligned reporting. These findings establish an auditable, scale-consistent evidence base for advancing reliable digital twin development in sediment–groundwater systems.