Infrared land surface emissivity dynamics in the Taklimakan desert from 2001 to 2023

Land surface emissivity (LSE) regulates long-wave energy partitioning in hyper-arid deserts, yet its multi-decadal behaviour and wavelength-specific controls on multi-decadal scales remain poorly resolved. Here we integrate 23-year satellite observations (CAMEL ESDR V003, 2001–2023) with ERA5 reanalysis to quantify spatiotemporal LSE dynamics at 8.3, 8.6 and 9.1 μm across the Taklimakan Desert and to disentangle thermal versus hydrological drivers. Using Mann-Kendall trend analysis, partial correlation, and Random Forest (RF, n = 1000), we identify three key findings. First, LSE increased paradoxically by 0.053 to 0.062 per decade (2001–2023) despite regional drying (-0.15 g kg⁻¹ decade⁻¹) and warming. Thermal-aeolian processes explained 68 ± 7% of this variance. Second, surface temperature independently reduced LSE by 0.0029 ± 0.0012 per °C, with maximum effect at 9.1 μm (-0.0035 ± 0.0015 per °C). Third, distinct wavelength-dependent characteristics, where the 9.1 μm band exhibits the highest interannual stability (no significant trend, p > 0.05) while the 8.3 μm band shows the greatest spatial heterogeneity (ΔLSE > 0.07 in central dunes).Thus, thermal-aeolian coupling, rather than soil moisture, dominates LSE variability in hyper-arid systems. Our findings provide critical constraints for desert–climate feedback models and offer wavelength-specific coefficients (-0.0035 LSE °C⁻¹ at 9.1 μm) to improve future space-borne land-surface temperature retrievals (e.g., ESA LSTM).