MetaFlux: Meta-learning global carbon fluxes from sparse spatiotemporal observations. Nathaniel, J., Liu, J., & Gentine, P. Scientific Data, 10(1):440, July, 2023. ![MetaFlux: Meta-learning global carbon fluxes from sparse spatiotemporal observations [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Abstract We provide a global, long-term carbon flux dataset of gross primary production and ecosystem respiration generated using meta-learning, called MetaFlux . The idea behind meta-learning stems from the need to learn efficiently given sparse data by learning how to learn broad features across tasks to better infer other poorly sampled ones. Using meta-trained ensemble of deep models, we generate global carbon products on daily and monthly timescales at a 0.25-degree spatial resolution from 2001 to 2021, through a combination of reanalysis and remote-sensing products. Site-level validation finds that MetaFlux ensembles have lower validation error by 5–7% compared to their non-meta-trained counterparts. In addition, they are more robust to extreme observations, with 4–24% lower errors. We also checked for seasonality, interannual variability, and correlation to solar-induced fluorescence of the upscaled product and found that MetaFlux outperformed other machine-learning based carbon product, especially in the tropics and semi-arids by 10–40%. Overall, MetaFlux can be used to study a wide range of biogeochemical processes.
@article{nathaniel_metaflux_2023,
title = {{MetaFlux}: {Meta}-learning global carbon fluxes from sparse spatiotemporal observations},
volume = {10},
issn = {2052-4463},
shorttitle = {{MetaFlux}},
url = {https://www.nature.com/articles/s41597-023-02349-y},
doi = {10.1038/s41597-023-02349-y},
abstract = {Abstract
We provide a global, long-term carbon flux dataset of gross primary production and ecosystem respiration generated using meta-learning, called
MetaFlux
. The idea behind meta-learning stems from the need to learn efficiently given sparse data by learning how to learn broad features across tasks to better infer other poorly sampled ones. Using meta-trained ensemble of deep models, we generate global carbon products on daily and monthly timescales at a 0.25-degree spatial resolution from 2001 to 2021, through a combination of reanalysis and remote-sensing products. Site-level validation finds that MetaFlux ensembles have lower validation error by 5–7\% compared to their non-meta-trained counterparts. In addition, they are more robust to extreme observations, with 4–24\% lower errors. We also checked for seasonality, interannual variability, and correlation to solar-induced fluorescence of the upscaled product and found that MetaFlux outperformed other machine-learning based carbon product, especially in the tropics and semi-arids by 10–40\%. Overall, MetaFlux can be used to study a wide range of biogeochemical processes.},
language = {en},
number = {1},
urldate = {2024-11-15},
journal = {Scientific Data},
author = {Nathaniel, Juan and Liu, Jiangong and Gentine, Pierre},
month = jul,
year = {2023},
pages = {440},
}
Downloads: 0
{"_id":"RSWJKgAiLCBkdcs4J","bibbaseid":"nathaniel-liu-gentine-metafluxmetalearningglobalcarbonfluxesfromsparsespatiotemporalobservations-2023","author_short":["Nathaniel, J.","Liu, J.","Gentine, P."],"bibdata":{"bibtype":"article","type":"article","title":"MetaFlux: Meta-learning global carbon fluxes from sparse spatiotemporal observations","volume":"10","issn":"2052-4463","shorttitle":"MetaFlux","url":"https://www.nature.com/articles/s41597-023-02349-y","doi":"10.1038/s41597-023-02349-y","abstract":"Abstract We provide a global, long-term carbon flux dataset of gross primary production and ecosystem respiration generated using meta-learning, called MetaFlux . The idea behind meta-learning stems from the need to learn efficiently given sparse data by learning how to learn broad features across tasks to better infer other poorly sampled ones. Using meta-trained ensemble of deep models, we generate global carbon products on daily and monthly timescales at a 0.25-degree spatial resolution from 2001 to 2021, through a combination of reanalysis and remote-sensing products. Site-level validation finds that MetaFlux ensembles have lower validation error by 5–7% compared to their non-meta-trained counterparts. In addition, they are more robust to extreme observations, with 4–24% lower errors. We also checked for seasonality, interannual variability, and correlation to solar-induced fluorescence of the upscaled product and found that MetaFlux outperformed other machine-learning based carbon product, especially in the tropics and semi-arids by 10–40%. Overall, MetaFlux can be used to study a wide range of biogeochemical processes.","language":"en","number":"1","urldate":"2024-11-15","journal":"Scientific Data","author":[{"propositions":[],"lastnames":["Nathaniel"],"firstnames":["Juan"],"suffixes":[]},{"propositions":[],"lastnames":["Liu"],"firstnames":["Jiangong"],"suffixes":[]},{"propositions":[],"lastnames":["Gentine"],"firstnames":["Pierre"],"suffixes":[]}],"month":"July","year":"2023","pages":"440","bibtex":"@article{nathaniel_metaflux_2023,\n\ttitle = {{MetaFlux}: {Meta}-learning global carbon fluxes from sparse spatiotemporal observations},\n\tvolume = {10},\n\tissn = {2052-4463},\n\tshorttitle = {{MetaFlux}},\n\turl = {https://www.nature.com/articles/s41597-023-02349-y},\n\tdoi = {10.1038/s41597-023-02349-y},\n\tabstract = {Abstract\n \n We provide a global, long-term carbon flux dataset of gross primary production and ecosystem respiration generated using meta-learning, called\n MetaFlux\n . The idea behind meta-learning stems from the need to learn efficiently given sparse data by learning how to learn broad features across tasks to better infer other poorly sampled ones. Using meta-trained ensemble of deep models, we generate global carbon products on daily and monthly timescales at a 0.25-degree spatial resolution from 2001 to 2021, through a combination of reanalysis and remote-sensing products. Site-level validation finds that MetaFlux ensembles have lower validation error by 5–7\\% compared to their non-meta-trained counterparts. In addition, they are more robust to extreme observations, with 4–24\\% lower errors. We also checked for seasonality, interannual variability, and correlation to solar-induced fluorescence of the upscaled product and found that MetaFlux outperformed other machine-learning based carbon product, especially in the tropics and semi-arids by 10–40\\%. Overall, MetaFlux can be used to study a wide range of biogeochemical processes.},\n\tlanguage = {en},\n\tnumber = {1},\n\turldate = {2024-11-15},\n\tjournal = {Scientific Data},\n\tauthor = {Nathaniel, Juan and Liu, Jiangong and Gentine, Pierre},\n\tmonth = jul,\n\tyear = {2023},\n\tpages = {440},\n}\n\n\n\n\n\n\n\n","author_short":["Nathaniel, J.","Liu, J.","Gentine, P."],"key":"nathaniel_metaflux_2023","id":"nathaniel_metaflux_2023","bibbaseid":"nathaniel-liu-gentine-metafluxmetalearningglobalcarbonfluxesfromsparsespatiotemporalobservations-2023","role":"author","urls":{"Paper":"https://www.nature.com/articles/s41597-023-02349-y"},"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"https://bibbase.org/zotero/tereno","dataSources":["cq3J5xX6zmBvc2TQC"],"keywords":[],"search_terms":["metaflux","meta","learning","global","carbon","fluxes","sparse","spatiotemporal","observations","nathaniel","liu","gentine"],"title":"MetaFlux: Meta-learning global carbon fluxes from sparse spatiotemporal observations","year":2023}