Accurate quantification of soil potassium (K) fractions is essential for sustainable nutrient management, but its spatial variability in arid ecosystems remains poorly quantified. This study presents the first high-resolution digital soil maps (DSM) of exchangeable, non-exchangeable, lattice, and total K fractions for the Thar Desert in India, one of the world's largest arid agro-ecosystems suffering chronic K deficiency. A total of 209 surface samples were analyzed, and predictive models were developed using Random Forest (RF) and Multiple Linear Regression (MLR) algorithms with terrain, climatic, and vegetation covariates. The RF model consistently outperformed MLR, achieving Nash-Sutcliffe efficiency (NSE) values of 0.15-0.42 across fractions. Elevation, mean annual temperature, and precipitation emerged as the dominant predictors of K distribution, driven by geomorphic and climatic processes. Uncertainty layers and management zones derived from ensemble dispersion highlight the spatial reliability of predictions and provide actionable insights for precision fertilization. In conclusion, this study demonstrates the potential of RF-based DSM to guide nutrient management in data-scarce, heterogeneous, and arid landscapes, such as the Thar Desert region. It supports targeted fertilizer use, long-term nutrient monitoring, and better soil policies in arid agricultural systems.Graphical abstractThis work illustrates the Thar Desert spatial variability of soil potassium fractions. Key datasets, including terrain attributes, bioclimatic variables, and legacy soil data, were represented to emphasize their role as environmental covariates. The analytical workflow is depicted through digital soil mapping approaches, particularly multiple linear regression (MLR) and random forest (RF) models, with data input, processing, and prediction steps. The model comparison is highlighted by presenting RF's superior performance over MLR, emphasizing higher predictive accuracy for exchangeable, non-exchangeable, lattice, and total potassium fractions. Results are represented through simplified thematic maps, showing the spatial distribution of K fractions across the study area, thereby connecting data, analyses, models, and outcomes in a logical sequence. Ultimately, the goal is to monitor the soil K in arid agroecosystems visually. The importance of this study lies in its ability to promote rapid comprehension by clearly communicating the implications for sustainable soil and nutrient management.