Machine learning is reshaping remote sensing from a primarily observational science into an operational source of spatial intelligence for decision-making. This editorial introduces the Research Topic “Machine learning for advanced remote sensing: from theory to applications and societal impact”, which brings together studies connecting methodological advances in machine learning with spatially explicit applications in agriculture, infrastructure, maritime monitoring, satellite autonomy, image enhancement, and ecosystem assessment. A central theme emerging from the collected studies is that remote-sensing machine learning remains most powerful when it is coupled with domain knowledge and spatial-temporal reasoning. The editorial further outlines priorities for future research, including geographically and physically aware foundation models, rigorous validation that accounts for spatial and temporal autocorrelation, efficiency- and privacy-aware operational deployment, and explicit assessment of the societal value of remote-sensing machine learning.