Objective Aerodynamic roughness (z₀) reflects the extent to which surface roughness elements attenuate wind erosive forces and is a key parameter in soil wind erosion models. However, most existing remote sensing-based z₀ estimation models are developed using photosynthetic vegetation indices derived from the summer season, which limits their ability to capture variations in z₀ during spring. Therefore, this study focuses on the Xilingol steppe, a region highly susceptible to wind erosion in spring, to explore optimal remote sensing-based models for estimating spring z₀ and to analyze its influencing factors. The results are expected to improve the parameterization accuracy of z₀ and to contribute to the sustainable development of wind erosion-prone regions.

Methods Using measured z₀ values obtained through observation and lookup tables from the survey units of the 2011 National Soil and Water Conservation Census, a spring z₀ remote sensing estimation model based on MODIS (MCD43A4) data was developed using linear regression analysis. Furthermore, single-factor detection and interaction detection methods of geodetector were applied to analyze the influencing factors of the spatial differentiation of spring z₀ in the study area from 2010 to 2022.

Results 1) The simple tillage index (STI), calculated using the ratio of Band 6 to Band 7 in the short-wave infrared (SWIR) region of MODIS data, had the highest correlation with measured z₀. The correlations of the vegetation indices, in descending order, were as follows: STI > NDTI > DFI > LAI > NDVI. 2) The STI-z₀ remote sensing model (R² = 0.83) demonstrated the best fitting performance and the highest validation accuracy (R² = 0.53, RMSE = 0.1257), making it the optimal model for estimating spring z₀. 3) From 2010 to 2022, spring z₀ showed a fluctuating upward trend, with the highest annual mean z₀ occurring in 2013 and the lowest in 2011. The multi-year average z₀ was 0.55 cm. 4) The dominant factor influencing the spatial differentiation of spring z₀ in the study area from 2010 to 2022 was the dead fuel index (DFI), representing non-photosynthetic vegetation cover, followed by wind speed and vegetation type. Among topographic factors, slope aspect had the weakest explanatory power for z₀. Moreover, the interaction effects among influencing factors had greater explanatory power than any single factor.

Conclusions z₀ exhibits seasonal variability. The optimal remote sensing-based model for estimating springtime z₀ is established using the non-photosynthetic vegetation index STI rather than photosynthetic vegetation indices, and this difference primarily results from the seasonal effects on vegetation growth conditions. The findings of this study can provide data for improving the prediction accuracy of soil wind erosion models and offer a scientific basis for the long-term dynamic monitoring and prevention of soil wind erosion.