A global, daily, and spatially seamless daytime and nighttime FY-3D LST dataset based on OESTR model

Land surface temperature (LST) is one of the key parameters of surface physical processes and is essential for many applications. FY-3D LST derived from passive microwave sensors can provide all-weather observations, but mainly due to discontinuous satellite revisit orbits, there are significant spatiotemporal gaps, which greatly limit its applications. In this study, a joint training strategy for daytime and nighttime data and an orbit-encoding-based spatiotemporal reconstruction (OESTR) model were proposed to address the sample imbalance between daytime and nighttime observations and the insufficient exploitation of temporal information. Additionally, based on the OESTR model, a global, daily, and spatially seamless daytime and nighttime FY-3D LST dataset from 30 April 2019 to 31 December 2025 was generated. The proposed method provides an effective solution for LST reconstruction, and the generated spatially seamless FY-3D LST dataset offers valuable support for various scientific applications.