Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Abstract. Accurate estimates of long-term surface temperature (ST) changes are fundamental not only for assessing observed warming, but also for improving the reliability of future climate projections. However, substantial missing information in global ST datasets, remains a major source of uncertainty in estimating global or regional temperature changes. Recent advances in artificial intelligence (AI) have promoted the effective application of deep learning approaches, such as image inpainting and transfer learning, in reconstructing incomplete geophysical datasets. This study develops a historical climate field reconstruction framework based on a Partial Convolutional (PConv) neural network, using climate model simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6) as training samples. Global surface temperature observations were constructed by merging the China Global Land Surface Air Temperature dataset (C-LSAT2.1) with the 200-member sea surface temperature ensemble from HadSST4.2.0.0, provided by the UK Met Office Hadley Centre, thereby forming an input ensemble. Model uncertainty was further quantified using the Deep Ensemble approach, generating multiple reconstruction members. The final monthly global surface temperature reconstruction was obtained by averaging all reconstruction ensemble members, while the corresponding uncertainty was estimated from the ensemble. Furthermore, validation against station observations indicates that the reconstructions perform well over Antarctica after 1961, where observational coverage is extremely sparse. Based on this framework, we developed the China global Artificial Intelligence Reconstructed Surface TemperatureCMIP6 (C-AIRSTM) dataset, providing spatially complete global monthly ST anomaly reconstructions since 1850 with a spatial resolution of 5°×2.5°, with anomalies referenced to the 1961–1990 climatology. This dataset provides improved support for extending long-term climate records and for applications in polar climate assessment, as well as in climate monitoring, detection, and attribution studies. The C-AIRSTM datasets can be downloaded at https://doi.org/10.6084/m9.figshare.33122444 (Ouyang et al., 2026). It is also available from http://www.gwpu.net/en/h-col-103.html (last access: 30 July 2026).