Upscaling Tower-Based Net Ecosystem Productivity to 250 m Resolution with Flux Site Distribution Considerations
Qizhi Han, Liangyun Liu, Xinjie Liu
Abstract
Net ecosystem productivity (NEP) is an extremely important flux for terrestrial ecosystems, indicating the value of net ecosystem exchange (NEE) between terrestrial ecosystems and the atmosphere, excluding carbon fluxes from disturbances. Leveraging flux network NEE annual measurements, this study focuses on upscaling the tower-based NEP to a global 250 m resolution dataset with flux site distribution considerations. Firstly, the data augmentation method was presented to address issues related to the uneven spatial distribution of flux sites. Secondly, a random forest model was developed for NEP estimation using the optimized tower-based NEP and remotely sensed and meteorological gridded sample sets, giving an R2 value of 0.73 and an RMSE value of 149.83 gC m−2 yr−1. Finally, a global NEP product at a 250 m resolution was generated (2001–2022, average 13.79 PgC yr−1) and evaluated. In summary, we present a solution to the overestimation of global NEP by data-driven methods, producing a long-time-series, high-resolution NEP dataset that is more comparable to atmospheric inversion results. This dataset enhances comparability with atmospheric inversion results, thereby boosting our confidence in conducting a consistency analysis of terrestrial carbon sinks across different methods within the framework.