Response of stratospheric water vapour to warming constrained by satellite observations

Peer Nowack, Paulo Ceppi, Sean Davis, Gabriel Chiodo, William T. Ball, Mohamadou Diallo, Birgit Haßler, Yue Jia, James Keeble, Manoj JoshiNature Geoscience 2023 · DOI

This study, published in Nature Geoscience, addresses a critical "missing link" in climate modeling: how much stratospheric water vapour (SWV) will increase as the planet warms. By applying a statistical-learning framework to satellite observations and historical temperature data, the researchers found that many current climate models significantly overestimate this response. Their findings provide a new, tighter constraint that reduces the uncertainty of future projections by half, with major implications for our understanding of climate sensitivity and ozone recovery.

TL;DR

Researchers used machine learning to discover a robust relationship between atmospheric temperature patterns and stratospheric water vapour. Applying this "fingerprint" to satellite data reveals that many climate models are over-sensitive, leading to an observationally constrained estimate that slashes the uncertainty of future water vapour increases by 50%.

What Problem Does This Paper Address?

The stratosphere is bone-dry, but even tiny changes in its water vapour content have outsized effects. Stratospheric water vapour (SWV) is a potent greenhouse gas; it also influences the chemical reactions that govern the recovery of the ozone layer.

The amount of water that enters the stratosphere is regulated by a "cold trap" in the tropical upper troposphere. As air rises, most water freezes out. However, climate models struggle to simulate the complex, small-scale processes (like convection and cloud microphysics) that happen in this region. Consequently, state-of-the-art models (CMIP5 and CMIP6) disagree wildly on how much SWV will increase under global warming, with projections ranging from negligible increases to massive surges. This "longstanding uncertainty" makes it difficult to predict exactly how fast the surface will warm or when the ozone hole will fully heal.

Methods

The authors developed a statistical-learning framework based on ridge regression.

  • The Predictor: They used high-dimensional 3D temperature structures (from 70 to 250 hPa) across the tropics and mid-latitudes.
  • The Target: Tropical lower stratospheric water vapour concentrations ($q_{strat}$).
  • The Innovation: Instead of relying on simple climate indices (like ENSO), the model "learns" the complex spatial and temporal co-variations between temperature and SWV from historical data.

Crucially, they validated this approach using a "perfect-model" setup. They showed that if they trained their regression on a climate model's historical data, it could accurately predict that same model's future response to a quadrupling of $CO_2$. This proved the relationship was "climate-invariant," allowing them to safely apply the functions learned from real-world satellite observations to future warming scenarios.

Datasets

  • SWOOSH (Stratospheric Water and OzOne Satellite Homogenized): A multi-satellite dataset providing the primary observational record of SWV from 1984 to 2020.
  • Reanalysis Data: ERA5, MERRA-2, and JRA-55 were used for historical atmospheric temperature fields.
  • CMIP5 & CMIP6: Data from 61 different global climate models were used to evaluate the spread of current projections and validate the machine learning framework.

Key Insights and Ideas

  • Model Over-sensitivity: The most striking insight is that many climate models are "too sensitive." They project SWV increases that are physically inconsistent with the historical relationships seen in satellite data.
  • The "Tape Recorder" Signal: The machine learning model successfully identified physical realities without being "told" about them. For instance, the regression coefficients (the weights the model gives to different data points) clearly showed the upward propagation of information over time, reflecting the slow vertical ascent of air through the tropical stratosphere (often called the "atmospheric tape recorder").
  • Narrowing the Gap: By discarding model projections that didn't match the observational constraint, the researchers reduced the 95th percentile of the uncertainty distribution by 50%. This suggests that the "extreme" warming scenarios driven by massive SWV increases are highly unlikely.

Key Results

  • The Constraint: The study establishes an observationally constrained range for SWV changes of $0.31 \pm 0.39$ ppmv K⁻¹ (parts per million volume per degree of global warming).
  • Model Comparison: Out of 61 models, about one-quarter projected increases greater than 1 ppmv K⁻¹, which this study deems "highly unlikely."
  • Radiative Impact: The researchers translated this SWV constraint into a radiative feedback value of $0.14$ W m⁻² K⁻¹ (median). This represents a significant reduction in the predicted "boost" that SWV gives to global warming compared to previous unconstrained model averages.

Limitations and Future Work

  • Methane Oxidation: The study focuses on the lower stratosphere (70 hPa). At higher altitudes, the breakdown of methane becomes a major source of water vapour. This process was not the focus of the constraint but remains a factor for the middle and upper stratosphere.
  • Vertical Resolution: The authors speculate that the "over-sensitivity" in models might be due to low vertical resolution around the tropopause, which fails to capture the sharp temperature gradients of the cold trap.
  • Expanding the Scope: Future work could apply this statistical-learning approach to other regions, such as the extratropical lowermost stratosphere, or to other chemically active species like ozone to further refine climate sensitivity estimates.

Summary updated

Abstract

Future increases in stratospheric water vapour risk amplifying climate change and slowing down the recovery of the ozone layer. However, state-of-the-art climate models strongly disagree on the magnitude of these increases under global warming. Uncertainty primarily arises from the complex processes leading to dehydration of air during its tropical ascent into the stratosphere. Here we derive an observational constraint on this longstanding uncertainty. We use a statistical-learning approach to infer historical co-variations between the atmospheric temperature structure and tropical lower stratospheric water vapour concentrations. For climate models, we demonstrate that these historically constrained relationships are highly predictive of the water vapour response to increased atmospheric carbon dioxide. We obtain an observationally constrained range for stratospheric water vapour changes per degree of global warming of 0.31 ± 0.39 ppmv K<sup>-1</sup>. Across 61 climate models, we find that a large fraction of future model projections are inconsistent with observational evidence. In particular, frequently projected strong increases (>1 ppmv K<sup>-1</sup>) are highly unlikely. Our constraint represents a 50% decrease in the 95th percentile of the climate model uncertainty distribution, which has implications for surface warming, ozone recovery and the tropospheric circulation response under climate change.
BibTeX
@inproceedings{nowack2023response,
  author = {Peer Nowack and Paulo Ceppi and Sean Davis and Gabriel Chiodo and William T. Ball and Mohamadou Diallo and Birgit Haßler and Yue Jia and James Keeble and Manoj Joshi},
  title = {{Response of stratospheric water vapour to warming constrained by satellite observations}},
  booktitle = {Nature Geoscience},
  year = {2023},
  doi = {10.1038/s41561-023-01183-6},
}

Loading related content...

Figure