Answers are grounded in this paper and cite their sources.
Alpine ecosystems remain underrepresented in global carbon flux studies due to sparse ground-based observations, likely reducing the reliability of global carbon flux products in these regions. MODIS gross primary productivity (GPP) and many machine-learning-based gridded flux products show increased uncertainty in alpine ecosystems due to limited observational constraints and challenges in representing fine-scale spatial heterogeneity. Here we developed a footprint-aware environmental response function (ERF) based approach to map carbon flux at half-hourly temporal resolution, and 30 m grid spacing over an alpine steppe landscape at Nam Co on the Tibetan Plateau (TP). The framework combined eddy covariance (EC) observations with meteorological variables, land surface temperature, and Sentinel-1 SAR backscatter, and used XGBoost to learn the relationship between fluxes and environmental conditions. The ERF model showed good performance and low bias across all flux components on held-out test data, with R² values of 0.64 for net ecosystem exchange (NEE), 0.78 for GPP, and 0.87 for ecosystem respiration. Independent validation of mapped half-hourly NEE against the NEE measured at Nam Co showed an overall RMSE of 1.34 g C m⁻² d⁻¹, a mean bias of −0.76 g C m⁻² d⁻¹ and R 2 of 0.23. Bias was near zero (−0.1 g C m⁻² d⁻¹) under well-mixed turbulent conditions (u* > 0.51 m s⁻¹), but ranged from -0.87 to -1.21 g C m⁻² d⁻¹ under low to moderate turbulence. Comparison of ERF-modelled GPP with widely used MODIS GPP indicated substantial underestimation of ecosystem carbon uptake by MODIS at the landscape scale (≈ 75 % relative bias). These results show that high-resolution flux mapping can help improve carbon flux representation over heterogeneous alpine terrain, although underestimation under low-turbulence conditions remains a limitation in data-scarce alpine ecosystems.
No takes yet. Share an insight, caveat, or question.
Pillai et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: