Recent years have witnessed the rise of big data and artificial intelligence (AI) as transformative forces in multiple scientific domains, and digital soil mapping (DSM) is no exception. Based on a review of recent literature, we summarize key characteristics and research focuses of recent DSM research and highlight the role of big data and AI in them. As soil sample data remain fundamental to DSM, much efforts are put into sampling design optimization, and legacy data have been widely used. Big Earth data create great opportunities for the development of innovative environmental covariates, such as novel covariates generated using time series remote sensing data. While supervised-machine learning is dominant in soil–environment relationship modelling, deep learning approaches adopting semi-supervised and self-supervised learning are emerging. The objective of DSM has extended from solely spatial to spatiotemporal mapping. To better harness the power of AI and big data to achieve more accurate and efficient soil mapping, we suggest three directions for future DSM research: advancing spatiotemporal soil sampling to decipher spatial and temporal variation patterns of soil, integrating DSM methodologies (especially machine learning) with pedological knowledge, and developing foundation models for soil mapping.
Lin et al. (2026) studied this question.