Arctic near-surface air temperature_60° N_90° N_3h_5km

This study presents a continuous near-surface air temperature dataset over the Arctic Ocean, covering regions north of 60°N from May to August 2024, with a spatial resolution of 5 km and a temporal resolution of 3 hours. The dataset is developed to address the limitations of satellite-based Arctic temperature observations, including insufficient accuracy, poor temporal continuity, and incomplete spatial coverage. It is generated using a Physics-Guided Spatiotemporal Reconstruction and Correction (PGSRC) framework, which leverages the high-frequency revisit capability of polar-orbiting satellites in high-latitude regions to achieve seamless reconstruction. Validation against independent radiosonde data shows that the final product achieves a coefficient of determination (R²) of 0.83 and a root mean square error (RMSE) of 2.34 K, significantly outperforming conventional interpolation methods. In addition, the model demonstrates stable performance in cross-validation experiments based on temporal, spatial (station-based), and sample-based strategies, as well as in monthly-scale evaluations, with R² values consistently exceeding 0.87. The dataset also exhibits strong generalization capability under conditions of high cloud cover and severe data gaps.