Decision-grade digital twins for in situ bioremediation of contaminated soil and groundwater: a critical review.

Field claims about in situ bioremediation often fail because monitoring data are not mapped to the causal process they are intended to prove. This critical review defines decision-grade digital twins as inferential systems, not as visualization tools, data lakes, or calibrated trend models. A decision-grade twin must connect a living conceptual site model with evidence streams that test competing hypotheses, prediction models with stated admissibility limits, uncertainty updates, and auditable trigger logic. The review evaluates how geochemistry, compound-specific isotope analysis, functional genes and transcripts, flux measurements, geophysical imaging, toxicity endpoints, and high-frequency sensing can support claims about pathway activation, treatment limitation, rebound risk, secondary impacts, and closure readiness. It also distinguishes the decision value of reactive transport models, hybrid process-data models, surrogate models, data-driven surveillance, and microbial process representations. Across hydrocarbons and polycyclic aromatic hydrocarbons, chlorinated solvents, pesticides, pharmaceuticals, and per- and polyfluoroalkyl substances, decision value depends on whether the evidence can distinguish among reaction, retention, redistribution, metabolite formation, and durable risk reduction. The review concludes that digital twins deserve decision-grade status only when they reduce ambiguity that can affect intervention choice, operating intensity, switching criteria, monitoring design, secondary-risk management, or closure judgment.