Estimating Global GPP From the Plant Functional Type Perspective Using a Machine Learning Approach

Abstract

Gross primary production (GPP) is the total amount of carbon sequestered by plants in an ecosystem through photosynthesis (Beer et al., 2010;Turner et al., 2006). GPP is the largest carbon flux of the terrestrial carbon cycle, which has significant impacts on atmospheric CO 2 concentration and terrestrial ecosystem carbon cycle (

FluxFormer: Upscaled global carbon fluxes from eddy covariance data with multivariate timeseries Transformer

“…We also investigated the long-term trends of GPP and RECO from 2001 to 2019 and observed that our data exhibits the highest positive trend in GPP during this period, with a growth rate of 0.45 PgC per year. This finding aligns with studies such as [Piao et al, 2020, Guo et al, 2023, Yang et al, 2022, supporting the assumption that the CO 2 fertilization effect should increase GPP over time. In contrast, MetaFlux and the widely used product FLUXCOM fail to replicate the long-term trend of GPP, contradicting the currently recognized significant greening observed from regional to global scales [Piao et al, 2020].…”

“…The results are illustrated in Figure 5a and Figure 5b. We observed that our data exhibits lower correlation with CSIF and TROPOMI SIF in tropical regions (Central and South America, West and Central Africa, and Southeast Asia) and Our long-term GPP trend aligns with the expected increase due to the CO 2 fertilization effect, anticipated to enhance the land carbon sink [Piao et al, 2020, Guo et al, 2023, Yang et al, 2022.…”

“…Estimating GPP involves various methods, such as simulating dynamic global vegetation models (DGVMs) like those employed in the TRENDY project [Sitch et al, 2015, Le Quéré et al, 2018, upscaling from measurements obtained through eddy covariance (EC) flux tower and satellite observations [Jung et al, 2019, Zeng et al, 2020. However, all these approaches rely on plant functional types (PFTs) to estimate ecosystem productivity [Poulter et al, 2011, Lin et al, 2021, Guo et al, 2023. Inconsistencies in PFT maps can significantly contribute to uncertainties in GPP estimations, as well as other climate-relevant variables, at both regional and global scales [Poulter et al, 2011].…”

Exploring the Potential of Long Short‐Term Memory Networks for Predicting Net CO2 Exchange Across Various Ecosystems With Multi‐Source Data

“…The observation stations affiliated with both networks utilize high‐precision instruments and equipment to record meteorological and ecosystem data (Baldocchi, 2020; Novick et al., 2018). Researchers use data from the flux networks to analyze and comprehend factors related to climate change and energy and material exchange processes in terrestrial ecosystems, particularly NEE and GPP (Guo et al., 2023; Xu et al., 2019). These measurements are reliable, allowing for robust analysis of daily, monthly, and IAVs in the North American region.…”

“…Data‐driven ML methods are simple and effective in evaluating NEE, as they are entirely adaptable to the data and do not rely on assumptions about terrestrial ecosystem patterns (Peylin et al., 2013). Various ML algorithms have been used to estimate ecosystem carbon fluxes, including Artificial Neural Networks (Papale & Valentini, 2003), Model Tree Ensemble (Liang et al., 2020), and Random Forest (RF; Guo et al., 2023).…”