Sustainable soil–water management under climate and socio-economic pressures requires predictive capability that is both mechanistic and continuously corrected by observations. Data assimilation (DA) provides the formal machinery to merge models with heterogeneous measurements—from satellite evapotranspiration and soil moisture to cosmic-ray neutron sensing, proximal geophysics, lysimeters, and groundwater hydrographs—while propagating uncertainty. This review (based on 90 references) synthesizes frontiers in DA and modeling for soil–water systems across scales, emphasizing (i) multi-source observation operators and scaling; (ii) coupled crop–vadose–groundwater modeling frameworks and their structural hypotheses; (iii) modern DA methods (ensemble, variational, particle-based, and hybrid physics–ML) for joint estimation of states, parameters, and biases; and (iv) emerging digital twins that enable predict-then-verify management loops for irrigation, recharge enhancement, and drought risk reduction. We highlight how tracer-aided and isotope-informed components can improve evapotranspiration partitioning and recharge threshold detection, and how agent-based or socio-hydrological coupling can represent human decision feedback. Finally, we outline research gaps in uncertainty quantification, benchmarking, reproducibility, and governance needed to operationalize trustworthy soil–water digital twins for resilient food and water systems.
Ying Zhao (Sat,) studied this question.