Shipping regulations lead to large reduction in cloud perturbations

Duncan Watson‐Parris, Matthew W. Christensen, Angus Laurenson, Daniel Clewley, Edward Gryspeerdt, Philip StierProceedings of the National Academy of Sciences 2022 · DOI

TL;DR

This paper leverages a novel machine learning approach to identify over a million ship tracks globally, creating the most comprehensive climatology to date. This dataset reveals a significant, non-linear reduction in ship tracks following the 2020 International Maritime Organization (IMO) fuel regulations, demonstrating a clear global cloud response to environmental policy.

What Problem Does This Paper Address?

Sulfur dioxide (SO₂) emissions from global shipping, a significant contributor to global SO₂ emissions, form sulfate aerosols. These aerosols have a cooling effect on the planet by scattering sunlight and influencing cloud properties (indirect effect). Quantifying this indirect effect is crucial for understanding Earth's radiative balance and is a major source of uncertainty in climate models. Ship tracks, visible linear features in clouds caused by these aerosols, offer a unique, localized "experiment" to study this effect. However, manually identifying and cataloging these tracks is incredibly time-consuming and has historically limited studies to specific regions and short time periods. This paper addresses the challenge of creating a large-scale, global dataset of ship tracks to enable a robust analysis of their response to regulatory changes.

Methods

The core of this research is a custom machine learning model based on a UNet architecture with a ResNet-152 backbone. This model was trained on MODIS Aqua satellite imagery, utilizing visible, near-infrared, and infrared channels to capture information about cloud properties relevant to ship track detection. The training data consisted of approximately 4,500 hand-logged ship tracks, meticulously marking their heads and turning points. The model was trained using Adam optimization and a binary cross-entropy Jaccard loss function, which proved effective in handling the class imbalance inherent in ship track detection. A key methodological advancement is the automation of ship track detection, enabling the cataloging of over 1 million tracks globally over a 20-year period. This automated approach significantly surpasses previous manual efforts in scale and scope.

Datasets

  • MODIS Aqua Satellite Imagery: Used for both training and inference. Specifically, "day microphysics" composites from channels 1, 20, and 32 were employed.
  • Hand-logged Ship Tracks: Approximately 4,500 manually identified ship tracks were used as ground truth for training the machine learning model.
  • CAMS-GLOB-SHIP v3.1 Product: This dataset provides monthly ship-borne SO₂ emissions data at 0.1° resolution, used for correlating emissions with ship track changes.
  • MODIS Level 3 Product (MYD08_L3): Monthly data on single-layer retrieved liquid cloud fraction was used to analyze environmental controls on ship track formation.

The paper introduces a new, large-scale benchmark dataset of over 1 million ship tracks, significantly expanding upon previous regional and shorter-term datasets.

Key Insights and Ideas

  • Machine Learning as a Game-Changer for Climate Science: The successful application of a deep learning model to automate ship track detection demonstrates the power of AI in tackling complex environmental monitoring challenges, enabling unprecedented global-scale analysis.
  • Non-linear Response of Clouds to Emissions: The finding that an 80% reduction in SO₂ emissions leads to only a 25% reduction in ship tracks highlights a non-linear relationship. This suggests that clouds are already significantly perturbed in shipping corridors, and further reductions in emissions yield diminishing returns in terms of ship track formation.
  • Clear Impact of Regulations: The study provides compelling evidence that stringent international regulations (IMO 2020) have a direct and measurable impact on cloud properties, even in the face of confounding factors like the COVID-19 pandemic.
  • Global vs. Regional Regulation Effectiveness: The analysis shows that while regional emission control areas (ECAs) had some impact, the global IMO regulations led to a more uniform and widespread reduction in ship tracks.

Key Results

  • Global Ship Track Climatology: Over 1 million ship tracks were cataloged globally over a 20-year period (2002-2021).
  • Significant Reduction Post-IMO 2020: Following the implementation of IMO 2020 regulations, the total number of ship tracks detected globally dropped by 25% in 2020 compared to previous years.
  • Non-linear Sensitivity: An 80% reduction in SO₂ emissions resulted in only a 25% decrease in the number of detected ship tracks, indicating a sublinear response.
  • Regional Differences: ECA regulations showed localized effects (e.g., off the coast of California), while global regulations had a more uniform impact across major shipping routes.
  • Cloud Fraction as a Key Factor: Ship track formation is strongly dependent on cloud fraction, with cleaner clouds being more susceptible to aerosol perturbations.

Limitations and Future Work

  • Sensitivity to Track Brightness and Linearity: The current model's detection is sensitive to the brightness and linearity of ship tracks. Older tracks with diminished perturbations might be missed. Future work could explore methods to better detect these less distinct tracks.
  • Confounding Factors: While the study accounts for COVID-19 disruptions, the interplay of various environmental factors and their influence on ship track detection could be further investigated.
  • Broader Cloud Perturbations: The paper acknowledges that ship tracks are only observable in homogenous cloud fields. Future research aims to combine ship track analysis with methods for detecting cloud perturbations in more complex cloud types (e.g., broken cumulus) to gain a more complete understanding of shipping's radiative forcing.
  • Saturation of Cloud Response: The non-linear response suggests that shipping corridors may be operating far from their preindustrial state, and further research could explore the degree to which cloud responses are saturated by current anthropogenic emissions.

Summary updated

Abstract

Global shipping accounts for 13% of global emissions of SO<sub>2</sub>, which, once oxidized to sulfate aerosol, acts to cool the planet both directly by scattering sunlight and indirectly by increasing the albedo of clouds. This cooling due to sulfate aerosol offsets some of the warming effect of greenhouse gasses and is the largest uncertainty in determining the change in the Earth's radiative balance by human activity. Ship tracks-the visible manifestation of the indirect of effect of ship emissions on clouds as quasi-linear features-have long provided an opportunity to quantify these effects. However, they have been arduous to catalog and typically studied only in particular regions for short periods of time. Using a machine-learning algorithm to automate their detection we catalog more than 1 million ship tracks to provide a global climatology. We use this to investigate the effect of stringent fuel regulations introduced by the International Maritime Organization in 2020 on their global prevalence since then, while accounting for the disruption in global commerce caused by COVID-19. We find a marked, but clearly nonlinear, decline in ship tracks globally: An 80% reduction in SO[Formula: see text] emissions causes only a 25% reduction in the number of tracks detected.
BibTeX
@inproceedings{watsonparris2022shipping,
  author = {Duncan Watson‐Parris and Matthew W. Christensen and Angus Laurenson and Daniel Clewley and Edward Gryspeerdt and Philip Stier},
  title = {{Shipping regulations lead to large reduction in cloud perturbations}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2022},
  doi = {10.1073/pnas.2206885119},
}
Figure