GIS‐Based Soil Fertility Mapping in Agricultural Landscapes: A Systematic Review of Methods and Applications

Spatial heterogeneity in soil‐fertility constitutes a fundamental barrier to sustainable crop production management; yet no rigorous synthesis of geographic information system (GIS)‐based mapping methodologies currently exists in the published literature. This systematic review consolidates current evidence on methods, performance benchmarks, and emerging directions in GIS‐based soil fertility‐mapping across global agricultural systems. Adhering to PRISMA 2020 guidelines, four major databases, Web of Science, Scopus, PubMed, and Google Scholar, were searched for peer‐reviewed publications from 2000 to 2024. Of 3847 records identified, 93 studies fulfilled inclusion criteria for qualitative synthesis, while 47 contributed to a quantitative random‐effects metaanalysis. Reviewed studies encompassed 41 countries across six agroecological regions; East and South Asia (33%) and Sub‐Saharan Africa (20%) were most represented. Each study was assigned a single, dominant interpolation method to avoid double‐counting; on this basis, geostatistical approaches, chiefly ordinary and regression kriging, predominated (60%), followed by machine‐learning (ML) techniques (28%), deterministic algorithms such as inverse distance weighting (IDW; 11%), and spline‐based methods (1%). Total nitrogen, exchangeable calcium, and electrical conductivity (ECa) were the most frequently mapped individual parameters, while micronutrients (Zn, Fe, Mn, and Cu), taken together as a class, were reported most often overall. Incorporating remote sensing and digital elevation model (DEM) covariates reduced prediction error (root mean‐square error [RMSE]) by 12%–23%, while GIS‐informed variable‐rate fertilizer prescriptions achieved a mean 22% reduction in fertilizer use without yield penalties in the subset of studies reporting such outcomes (Section 4.3). GIS‐based soil‐fertility mapping has transitioned into an operational precision agriculture decision‐support framework. ML delivers the highest predictive accuracy when training data are adequate; geostatistical approaches retain value where samples are sparse. Key research gaps include standardized sampling protocols, cross‐zone model transferability, temporal monitoring frameworks, and open‐data infrastructure for smallholder systems.