From national to global: Mapping global center-pivot irrigated cropland with knowledge-guided data augmentation

Center-pivot irrigated cropland (CPIC) is a large-scale, intensive agricultural irrigation practice with high water-use efficiency. This characteristic enables CPIC to increase crop yields and promote agricultural development, while simultaneously causing groundwater overexploitation and subsequent environmental issues. Mapping the global distribution of CPIC is critical for enhancing water resource management, guiding agricultural policies, and assessing ecological impacts. This study developed a novel methodological framework integrating CPIC prior knowledge into the generation of a global training dataset through designed algorithms, ultimately producing a 2020 global CPIC map (GCMap) based on Landsat imagery. Specifically, we first proposed a shape-prior guided Projection Simulation (PS) to enhance the diversity of national-scale CPIC samples, enabling global CPIC pseudo-label generation through a teacher model. Subsequently, a stratified sampling strategy was implemented to improve sample representativeness across different regions using the pseudo-labels of global CPIC. Furthermore, a dual-temporal correction mechanism was proposed to clean sample labels, ultimately generating high-confidence global training samples. Finally, based on this dataset, a student model was trained using a noise-resistant loss function and deployed on cloud computing platforms to generate the GCMap rapidly. Following Olofsson's design-based stratified sampling and error-matrix framework, we derived statistically rigorous area estimates with explicit 95% confidence intervals and assessed mapping accuracy using the corresponding user's accuracy (UA), producer's accuracy (PA), and overall accuracy (OA) metrics. In comparison with current CPIC products, our GCMap attained a PA of 0.86 ± 0.02, representing a marked reduction in omission errors. Our results show that the global error-adjusted area of CPIC in 2020 was estimated at 18.70 ± 0.42 Mha, representing a 63% increase relative to the previous estimate of 11.5 Mha and revealing substantial underestimation in previous estimates. Geographically, the semi-arid zone has the largest share of CPIC, covering 9.27 ± 0.29 Mha (49.28% of the total), followed by the arid zone (4.69 ± 0.20 Mha, 24.95%) and humid zone (2.99 ± 0.22 Mha, 15.92%). Further analysis indicates that most CPIC is distributed in aquifers with limited underwater recharge. Approximately 15.25 ± 0.87 Mha (82.04%) of CPIC areas are located in regions with recharge rates below 100 mm/year, of which 9.97 ± 0.48 Mha (53.62% of total CPIC) fall within extremely low recharge zones (<20 mm/year). These findings provide an essential foundation for further research on irrigation-groundwater interactions and sustainable policymaking about agricultural development and water usage. Our GCMap inventory is publicly available in an online repository at: https://doi.org/10.12436/GCMap.62.20250711.