This study’s field observations of Light Source Efficiency (LSE) in the Taklamakan Desert have unveiled significant daily average variations across different wavelengths, with LSE values ranging from 0.827 at 9.1 μm to a peak of 0.969 at 12.1 μm, and notably, a substantial daily variation (DV) of Δε = 0.080 in the 14.3 μm band. These findings underscore the necessity for wavelength-specific analysis in LSE research, which is crucial for enhancing the precision of remote sensing applications and climate models. This study’s high-temporal-resolution FTIR field observations systematically reveal the diurnal dynamics of infrared surface emissivity in the desert for the first time, challenging existing satellite-based inversion products and highlighting the limitations of traditional temperature–emissivity separation algorithms in arid regions. The diurnal fluctuations are governed by three primary mechanisms: the amplification of lattice vibrations in quartz minerals under high daytime temperatures, changes in the surface topography due to thermal expansion and contraction, and nocturnal radiative cooling effects. The lack of a significant correlation between environmental parameters and the emissivity change rate suggests that microclimate factors play a dominant indirect regulatory role. Model comparisons indicate that sinusoidal functions outperform polynomial fits across most wavelengths, especially at 12.1 μm, confirming the significant influence of diurnal forcing. The high sensitivity of the 14.3 μm band makes it an ideal indicator for monitoring desert surface–atmosphere interactions. This study provides three key insights for remote sensing applications: the development of dynamic emissivity correction schemes, the prioritization of the long-wave infrared band for surface temperature inversion in arid regions, and the integration of ground-based observations with geostationary high-spectral data to construct spatiotemporally continuous emissivity models. Future research should focus on multi-angle observation experiments and the exploration of machine learning’s potential in cross-scale emissivity modeling.

Representative severe-weather events and high-reflectivity radar echo samples are relatively scarce in inland Western China, limiting the local training of deep learning-based extrapolation models. To examine the applicability and limitations of cross-regional transfer learning under limited target-domain data availability, this study adopts a source-domain pretraining and target-domain fine-tuning strategy. Using East China as the source domain and northwestern Xinjiang as the target domain, we conduct a unified comparison across three representative architectures: PredRNN, SimVP, and Earthformer. SimVP is further employed to examine the effects of target-domain data volume and module-wise transferability. In the evaluated experiments, transfer learning improved the Critical Success Index, Probability of Detection, and Fractions Skill Score at both 60- and 120-min lead times across the selected reflectivity thresholds, although changes in the False Alarm Ratio varied by architecture and threshold. The three case studies provided complementary evidence of improved echo-structure preservation. For SimVP, transfer gains are more pronounced when target-domain data are limited and generally diminish as more local training data become available. This finding suggests that pretraining can improve the use of limited target-domain data, while the continued accumulation of representative local severe-weather observations remains necessary. Within the tested SimVP freezing configurations, module-wise analysis further shows that the Encoder parameters responsible for spatial feature extraction exhibit relatively strong cross-regional reusability, whereas updating the Translator responsible for latent-space spatiotemporal evolution while keeping the Encoder frozen improves forecast performance. These module-level findings are specific to the evaluated SimVP setting. Overall, the results support the practical applicability of transfer learning in the studied East China-to-Xinjiang scenario. The transfer gains are jointly influenced by differences in meteorology and radar observations, and their broader applicability remains to be tested in future studies across different regions and model architectures.