This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.
FluxFormer: Upscaled global carbon fluxes from eddy covariance data with multivariate timeseries Transformer
Downloads
Authors
Abstract
We provided a monthly global gross primary production (GPP) and ecosystem respiration (RECO) dataset from 1990 to 2019 at 0.25° × 0.25° spatial resolution named FluxFormer by utilizing the new plant function type dataset in combination with multivariate timeseries Transformer-based model. FluxFormer outperforms other satellite-derived upscaled products when comparing the correlation at site-level and seasonal pattern with FLUXNET 2015, especially in tropical regions. Additionally, our dataset shows the highest positive trend in GPP from 2001 to 2019, aligning with trends derived from dynamic global vegetation models that account for the CO2 fertilization effect. Notably, FluxFormer captures positive long-term trends that are not replicated by some existing products. FluxFormer could be used to validate terrestrial biosphere models and serve as
a tool for cross-checking other datasets. The FluxFormer GPP and RECO product is available at https://doi.org/10.5281/zenodo.10258644
DOI
https://doi.org/10.31223/X5BQ2H
Subjects
Earth Sciences, Physical Sciences and Mathematics
Keywords
TransformerGross primary production, Ecosystem respiration, Plant functional type, transformer, gross primary production, Ecosystem respiration, Plant functional type
Dates
Published: 2023-12-05 04:02
Last Updated: 2023-12-05 09:02
License
CC BY Attribution 4.0 International
Additional Metadata
Data Availability (Reason not available):
https://doi.org/10.5281/zenodo.10258644
There are no comments or no comments have been made public for this article.