The stratosphere, 10-50 km above the Earth’s surface, is home to the ozone layer that shields surface life from harmful ultraviolet radiation. It is also a region of the atmosphere that is extremely dry: stratospheric water vapour concentrations are measured in a few parts per million volume (ppmv), rather than the 2,000-20,000 ppmv at the Earth’s surface. The reason for this is due to how air typically enters the stratosphere. As air ascends in the tropics into the stratosphere, it is effectively freeze-dried as it passes through the region around the tropopause, 16 km above the tropical Earth surface, where temperatures are about -80°C.
Under climate change, it is expected that tropopause temperatures and the circulation of the stratosphere will be altered, leading to changes in stratospheric water vapour (SWV). These changes need to be understood as part of the whole picture of quantifying and reducing uncertainties in future projections.
Why is this important?
Firstly, SWV is a greenhouse gas that - if increasing - can amplify global warming (Figure 1). In addition, changes in SWV have been demonstrated to impact the tropospheric circulation and thus, potentially, the weather patterns we experience at Earth’s surface.
Secondly, changes in SWV affect the chemical mechanisms shaping the stratospheric ozone layer. In particular, large climate-driven increases in SWV, like those projected by several climate models, could substantially delay the recovery of the ozone layer and of the Antarctic ozone hole.
Figure 1. Sketch of the uncertain two-way coupling between global warming and SWV under increasing atmospheric CO2 concentrations. Our paper reduces (a) the uncertainties in the SWV response and (b) in how these feed back onto global warming. Adapted from Dessler et al. (2013).
The uncertainty in future changes in SWV is evident from projections made with the latest generations of global climate models. In Figure 2, we illustrate this issue for a standard climate modelling experiment, which aims at benchmarking responses of models that participate in major climate change assessments: following an abrupt quadrupling of atmospheric CO2 concentrations across a range of climate models, a few models show hardly any change in SWV, whereas others show increases greater than 200% compared to present-day levels. Importantly, this large modelling uncertainty has not reduced in the most recent generation of climate models (red lines) as compared to the previous - around 7 years older - generation of climate models (blue lines).
Figure 2. Large projection uncertainty in SWV. Shown are simulated changes in tropical lower SWV at 70 hPa (~18 km) and averaged within 30 °N – 30 °S (Δqstrat) after an abrupt quadrupling of atmospheric CO2 relative to pre-industrial levels. Clearly, the uncertainty has not reduced in the most recent set of 34 climate models (in red; the Coupled Model Intercomparison Project phase 6; CMIP6), as compared to an older set of 27 models from the previous CMIP5 (in blue). The dashed black lines mark the 0 to 4 ppmv interval, where 4 ppmv is broadly equivalent to present-day values.
How does our work tackle this uncertainty?
While it is relatively well understood why the stratosphere is so dry in the first place, the complexity of the underlying processes and the relatively short record of high-quality satellite observations has so far made constraining past and future changes in SWV challenging. Addressing uncertainty in SWV trends under global warming is indeed a longstanding research challenge.
To reduce this large uncertainty, we developed a statistical learning approach, which combines information from satellite data with large climate model data archives. The approach allows us to learn high-dimensional mathematical relationships from data, which are approximately climate-invariant (i.e., relationships that hold both for the present and in much warmer climates). With this new data-driven approach, which exploits machine learning ideas, we are now able to derive a first ‘observational constraint’ on the uncertainty. In essence, once the climate-invariance of the learned relationships has been validated across the climate models, we were then able to learn the equivalent relationships from Earth observations. These ‘observed relationships' were then used to constrain the uncertainty in the model projections under climate change.
Our observational constraint still implies that SWV is likely to increase with global warming (0.31 ± 0.39 ppmv per degree warming for a 90% confidence interval). It also allows us to say that those model projections with very large increases in SWV, which could for example substantially delay ozone recovery, are unlikely. However, changes in SWV will still likely amplify global warming in the future. Finally, we believe that the ideas behind this new observational constraint framework could help us improve future projections of other central changes in Earth’s climate, such as potential future changes in the ozone layer itself.
Written by Peer Nowack, Manoj Joshi, James Keeble, Sean Davis, Paulo Ceppi, Gabriel Chiodo, Mohamadou A. Diallo, and Birgit Hassler
References:
Dessler et al. (2013), https://csl.noaa.gov/news/2013/144_0930.html
Please sign in or register for FREE
If you are a registered user on Research Communities by Springer Nature, please sign in