Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning

Abstract Global measurements of ocean p C O 2 are critical to monitor and understand changes in the global carbon cycle. However, p C O 2 observations remain sparse as they are mostly collected on opportunistic ship tracks. Several approaches, especially based on direct learning, have been used to u...

Full description

Saved in:
Bibliographic Details
Main Authors: Siyeon Kim, Juan Nathaniel, Zhewen Hou, Tian Zheng, Pierre Gentine
Format: Article
Language:English
Published: Nature Portfolio 2024-10-01
Series:Scientific Data
Online Access:https://doi.org/10.1038/s41597-024-03959-w
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850100129741144064
author Siyeon Kim
Juan Nathaniel
Zhewen Hou
Tian Zheng
Pierre Gentine
author_facet Siyeon Kim
Juan Nathaniel
Zhewen Hou
Tian Zheng
Pierre Gentine
author_sort Siyeon Kim
collection DOAJ
description Abstract Global measurements of ocean p C O 2 are critical to monitor and understand changes in the global carbon cycle. However, p C O 2 observations remain sparse as they are mostly collected on opportunistic ship tracks. Several approaches, especially based on direct learning, have been used to upscale and extrapolate sparse point data to dense estimates using globally available input features. However, these estimates tend to exhibit spatially heterogeneous performance. As a result, we propose a physics-informed transfer learning workflow to generate dense p C O 2 estimates that are grounded in real-world measurements and remain physically consistent. The models are initially trained on dense input predictors against p C O 2 estimates from Earth system model simulation, and then fine-tuned to sparse SOCAT observational data. Compared to the benchmark direct learning approach, our transfer learning framework shows major improvements of up to 56-92%. Furthermore, we demonstrate that using models that explicitly account for spatiotemporal structures in the data yield better validation performances by 50-68%. Our strategy thus presents a new monthly global p C O 2 estimate that spans for 35 years between 1982-2017.
format Article
id doaj-art-beae03fcefaf43baa25dd0e5fecfbfea
institution DOAJ
issn 2052-4463
language English
publishDate 2024-10-01
publisher Nature Portfolio
record_format Article
series Scientific Data
spelling doaj-art-beae03fcefaf43baa25dd0e5fecfbfea2025-08-20T02:40:21ZengNature PortfolioScientific Data2052-44632024-10-0111111310.1038/s41597-024-03959-wSpatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learningSiyeon Kim0Juan Nathaniel1Zhewen Hou2Tian Zheng3Pierre Gentine4Department of Statistics, Columbia UniversityDepartment of Earth and Environmental Engineering, Columbia UniversityDepartment of Statistics, Columbia UniversityDepartment of Statistics, Columbia UniversityDepartment of Earth and Environmental Engineering, Columbia UniversityAbstract Global measurements of ocean p C O 2 are critical to monitor and understand changes in the global carbon cycle. However, p C O 2 observations remain sparse as they are mostly collected on opportunistic ship tracks. Several approaches, especially based on direct learning, have been used to upscale and extrapolate sparse point data to dense estimates using globally available input features. However, these estimates tend to exhibit spatially heterogeneous performance. As a result, we propose a physics-informed transfer learning workflow to generate dense p C O 2 estimates that are grounded in real-world measurements and remain physically consistent. The models are initially trained on dense input predictors against p C O 2 estimates from Earth system model simulation, and then fine-tuned to sparse SOCAT observational data. Compared to the benchmark direct learning approach, our transfer learning framework shows major improvements of up to 56-92%. Furthermore, we demonstrate that using models that explicitly account for spatiotemporal structures in the data yield better validation performances by 50-68%. Our strategy thus presents a new monthly global p C O 2 estimate that spans for 35 years between 1982-2017.https://doi.org/10.1038/s41597-024-03959-w
spellingShingle Siyeon Kim
Juan Nathaniel
Zhewen Hou
Tian Zheng
Pierre Gentine
Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
Scientific Data
title Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
title_full Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
title_fullStr Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
title_full_unstemmed Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
title_short Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning
title_sort spatiotemporal upscaling of sparse air sea pco2 data via physics informed transfer learning
url https://doi.org/10.1038/s41597-024-03959-w
work_keys_str_mv AT siyeonkim spatiotemporalupscalingofsparseairseapco2dataviaphysicsinformedtransferlearning
AT juannathaniel spatiotemporalupscalingofsparseairseapco2dataviaphysicsinformedtransferlearning
AT zhewenhou spatiotemporalupscalingofsparseairseapco2dataviaphysicsinformedtransferlearning
AT tianzheng spatiotemporalupscalingofsparseairseapco2dataviaphysicsinformedtransferlearning
AT pierregentine spatiotemporalupscalingofsparseairseapco2dataviaphysicsinformedtransferlearning