Sustainable soil and land management represents a critical challenge in the context of climate change, primarily due to the high spatial heterogeneity of landscapes and the complex temporal scales that govern soil functions and ecosystem dynamics. To address these challenges, integrated modelling approaches are essential to bridge scales, disciplines, and data sources, effectively linking mechanistic physical understanding with data-driven insights. Observations serve as the cornerstone of understanding pedo-hydrological processes, and while modern technological advancements provide a wealth of information, integrating these diverse measurement sources into data-driven and physics-informed models remains a significant hurdle in vadose zone hydrology and soil science in general. Recent breakthroughs in deep learning and AI have opened new pathways for modelling complex Earth system processes, offering cutting-edge applications to characterize soil biogeophysical and hydrothermal properties while predicting the transport of water, heat, and solutes. By assimilating data from field sensors to remote sensing platforms into physics-informed models and digital twin frameworks, soil processes can be simulated across multiple spatial scales. This integration enables more accurate and reliable predictions of critical issues such as climate change impacts, contamination, salinization, erosion, agricultural practices, and land-use change. This interdisciplinary approach-coupling models to represent interactions between soil, microbes, plants, and the atmosphere—not only highlights the promise and limits of integrated strategies but also provides a robust foundation for the resilient, sustainable management of soil and water resources across both agroecosystems and natural landscapes.