A deep-learning estimate of the decadal trends in the Southern Ocean carbon storage

Zemskova, Varvara; He, Tai‐Long; Wan, Zirui; Grisouard, Nicolas

doi:10.1038/s41467-022-31560-5

Cited by 14 publications

(7 citation statements)

References 73 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The deep learning model we use is adapted from the U-net model, which was originally proposed for biomedical segmentation problems ( 46 ). The U-net model has been recently widely applied in the field of earth science ( 47 – 50 ). The schematic diagram of the model architecture is shown in SI Appendix , Fig.…”

Section: Methodsmentioning

confidence: 99%

Increased methane emissions from oil and gas following the Soviet Union’s collapse

He,

Boyd,

Varon

et al. 2024

Proc. Natl. Acad. Sci. U.S.A.

Self Cite

View full text Add to dashboard Cite

Global atmospheric methane concentrations rose by 10 to 15 ppb/y in the 1980s before abruptly slowing to 2 to 8 ppb/y in the early 1990s. This period in the 1990s is known as the “methane slowdown” and has been attributed in part to the collapse of the former Soviet Union (USSR) in December 1991, which may have decreased the methane emissions from oil and gas operations. Here, we develop a methane plume detection system based on probabilistic deep learning and human-labeled training data. We use this method to detect methane plumes from Landsat 5 satellite observations over Turkmenistan from 1986 to 2011. We focus on Turkmenistan because economic data suggest it could account for half of the decline in oil and gas emissions from the former USSR. We find an increase in both the frequency of methane plume detections and the magnitude of methane emissions following the collapse of the USSR. We estimate a national loss rate from oil and gas infrastructure in Turkmenistan of more than 10% at times, which suggests the socioeconomic turmoil led to a lack of oversight and widespread infrastructure failure in the oil and gas sector. Our finding of increased oil and gas methane emissions from Turkmenistan following the USSR’s collapse casts doubt on the long-standing hypothesis regarding the methane slowdown, begging the question: “what drove the 1992 methane slowdown?”

show abstract

Section: Methodsmentioning

confidence: 99%

Increased methane emissions from oil and gas following the Soviet Union’s collapse

He,

Boyd,

Varon

et al. 2024

Proc. Natl. Acad. Sci. U.S.A.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Another approach, also using machine learning, by Zemskova et al (2022) extrapolated from satellite data by combining it with numerical model output. Unfortunately, these two estimates are limited, respectively, to the top 1500 m of the ocean, and to the Southern Ocean only.…”

Section: Future Applicationmentioning

confidence: 99%

An optimal transformation method applied to diagnosing the ocean carbon sink

Mackay,

Zika,

Sohail

et al. 2024

Preprint

View full text Add to dashboard Cite

Abstract. The ocean carbon sink plays a critical role in climate, absorbing anthropogenic carbon from the atmosphere and mitigating climate change. The sink shows significant variability on decadal timescales, but estimates from models and observations disagree with one another, raising uncertainty over the magnitude of the sink, its variability, and its driving mechanisms. There is a need to reconcile observationally-based estimates of air-sea CO2 fluxes with those of the changing ocean carbon inventory in order to improve our understanding of the sink, and doing so requires knowledge of how carbon is transported within the interior by the ocean circulation. Here we employ a recently developed Optimal Transformation Method (OTM) that uses water mass theory to relate interior changes in tracer distributions to transports and mixing and boundary forcings, and extend its application to include carbon using synthetic data. We validate the method using model outputs from a biogeochemical state estimate, and test its ability to recover boundary carbon fluxes and interior transports consistent with changes in heat, salt and carbon. Our results show that OTM effectively reconciles boundary carbon fluxes with interior carbon distributions when given a range of prior fluxes. OTM shows considerable skill in its reconstructions, reducing root-mean-squared errors from biased priors between model ‘truth’ and reconstructed boundary carbon fluxes by up to 71 %, with bias of the reconstructions consistently ≤ 0.06 mol-Cm−2 yr−1 globally. Inter-basin transports of carbon also compare well with the model truth, with residuals < 0.25 Pg C yr−1 for reconstructions produced using a range of priors. OTM has significant potential for application to reconciling observational estimates of air-sea CO2 fluxes with the interior accumulation of anthropogenic carbon.

show abstract

“…12 Based on seven sea surface features, deep-learning models were used to predict the increase in DIC concentration in the deep sea of the Southern Ocean. 18 The successful extrapolation of global ocean POC concentrations via an extreme gradient boosting (XGBoost) model based on in situ POC measurement data advanced dynamic research on global ocean POC. 36 The boosted regression tree model successfully predicted the GHG concentration in delta water bodies and emphasized the importance of small water bodies (<0.001 km 2 ) in regional carbon emissions.…”

Section: Ghg Emissions Recognition In Water Environmentsmentioning

confidence: 99%

“…Based on seven sea surface features, deep-learning models were used to predict the increase in DIC concentration in the deep sea of the Southern Ocean . The successful extrapolation of global ocean POC concentrations via an extreme gradient boosting (XGBoost) model based on in situ POC measurement data advanced dynamic research on global ocean POC .…”

Section: Machine Learning Promotes Large-scale Ghg Emissions Recognit...mentioning

confidence: 99%

“…The accumulated data are expected to provide effective solutions for accurately recognizing GHG emissions from water (Figure c). With a strong ability to learn complicated patterns, machine learning has been successfully applied to the large-scale prediction of DOC in inland waters based on satellite data and to the prediction of DIC in the Southern Ocean based on ocean surface data. , Machine learning analyzes the relationship between variables and labels in empirical data and makes predictions for unknown data. Given the experience of geophysical models and many experimental results, GHG emissions from water can be predicted based on complex high-dimensional environmental factors (e.g., geography, climate, human activities, and microbes).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning Provides Opportunities to Recognize Greenhouse Gas Emissions from Water at a Large Scale

Deng,

Hu,

2023

ACS EST Water

View full text Add to dashboard Cite

Water environments (e.g., oceans, lakes, and rivers) are important carbon sinks and sources and contribute to the carbon cycle of the earth's ecosystem. Machine learning provides a potential solution for recognizing greenhouse gas (GHG) emissions from water based on big data analysis. Data-driven machine learning can comprehensively recognize the key environmental drivers that affect GHG emissions. However, several urgent issues should be addressed to guarantee machine learning recognition of GHG emissions from water. For example, matching in situ data and databases is the greatest challenge in conducting large-scale machine learning research. It is imperative to unify in situ data collection methods and improve database quality (e.g., spatiotemporal matching and high resolution). Quantifying the contributions of human activity and climate change to GHG emissions from water is urgently needed to resolve future challenges. Beyond providing recognition and prediction, machine learning, due to its interpretability, can optimize the carbon cycle model; thus, empirical formulas for machine learning deserve attention. Overall, machine learning can manage large-scale and complicated data for climate change regarding GHG emissions from water.

show abstract

A deep-learning estimate of the decadal trends in the Southern Ocean carbon storage

Cited by 14 publications

References 73 publications

Increased methane emissions from oil and gas following the Soviet Union’s collapse

Increased methane emissions from oil and gas following the Soviet Union’s collapse

An optimal transformation method applied to diagnosing the ocean carbon sink

Machine Learning Provides Opportunities to Recognize Greenhouse Gas Emissions from Water at a Large Scale

Contact Info

Product

Resources

About