Mapping spatial heterogeneity of non-structural carbohydrates in Haloxylon ammodendron using remote sensing in extreme desert environments

Abstract

Arid regions occupy 40 % of the Earth's land surface and play a significant role in the global carbon budget. Dominant desert shrub Haloxylon ammodendron, which thrives under extreme climatic conditions, contributes substantially to the stability of desert ecosystems. However, its species distribution patterns and carbon spatial dynamics in deserts remain unknown. Based on our newly development “Satellite-Airborne-Field” Ensemble learning platform for Systematic plant CArbohydrate estimation in Nature (SAFESCAN), we accurately estimated the spatial patterns of non-structural carbohydrate (NSC) content in leaves (LNSC) and branches (BNSC), with R² values of 0.83 and 0.72, respectively. This analysis incorporated 51 environmental variables and 564 samples collected from 81 subplots across multiple temperate deserts between 75°E and 115°E. The results identified the northern Gurbantunggut Desert and western edges of the Taklamakan Desert in Xinjiang as primary distribution areas for H. ammodendron, with northern Inner Mongolia and northern Qinghai also suitable for its survival. LNSC showed a west-to-east decreasing trend, while BNSC exhibited an increasing trend. The SAFESCAN platform demonstrated strong predictive capacity for shrub-dominated arid regions but may overestimate the relationship between NSC content and certain environmental variables. By integrating macro-scale species distribution patterns and local individual patterns of H. ammodendron, SAFESCAN provides NSC content maps, laying the foundation for future efforts in desert plant conservation and cultivation.

Keywords

• Desert shrub
• Ensemble machine learning
• Spatial pattern
• Environmental drivers