A problematic estimate of warming from low-sulfur marine fuels

A new paper provides a reasonable forcing estimate but errs on the global temperature response

Some background on aerosols and shipping

Aerosols – small suspended particles of sulfur dioxide and other materials – are one of the trickiest parts of the climate system.

They tend to have a cooling influence on the Earth’s climate in two different ways – through reflecting incoming sunlight back to space (the direct aerosol effect) and through serving as cloud condensation nuclei (the indirect aerosol effect). While the direct aerosol effect is reasonably well constrained, the indirect aerosol effect is the single biggest uncertainty in our estimate of changes to the climate system since 1750, as shown in the figure below (from the recent IPCC AR6 report):

Estimates of changes in effective radiative forcing (and their uncertainties) since 1750 from different factors. Figure from the IPCC AR6 WG1.

The main contributor to aerosol cooling is our emissions of sulfur dioxide (SO2), primarily from burning sulfur-rich fossil fuels: coal, oil, and gas. Global SO2 emissions have fallen significantly in recent years as countries have increasingly prioritized reductions in outdoor air pollution – where SO2 is a major contributor to the millions of premature deaths that occur each year.

Overall global SO2 emissions have fallen by nearly 50% from their peak in the 1980s, with much of the decline occurring in the past 15 years. This has served to unmask a portion of the warming from greenhouse gases that the world would have otherwise experienced, and is contributing to a recent acceleration in the rate of warming.

Latest estimates of global total SO2 emissions (blue) and shipping sectors SO2 emissions (red) from the Community Emissions Data System (CEDS).

In 2020, the International Maritime Organization (IMO) adopted new regulations to reduce air pollution from shipping, imposing strict limits on the sulphur content of marine fuels. This quickly reduced the sulfur emissions from shipping more than 70% (red line in the figure above), contributing to the recent rapid decline in global SO2 emissions. Because these emissions are located over the oceans, there is reason to expect them to have a larger impact than reductions in SO2 emissions in regions with higher background SO2 concentrations (though this is somewhat counterbalanced by the concentration of ships in specific shipping lanes and the presence of natural dimethyl sulphide emissions from algae).

There has been a vigorous debate in the academic literature around the precise effect of these new IMO regulations on regional and global forcing and temperature response. Its in this context that a new paper by Tianle Yuan and colleagues was just released.

A new estimate of forcing over the ocean

The new paper by Yuan et al published in the journal Nature Communications: Earth & Environment provides an estimate of both the change in radiative forcing over the oceans associated with the IMO regulations and attempts to model the global surface temperature response. They frame this as an example of an “inadvertent geoengineering termination shock” that will inform our estimates of the effectiveness of future directed geoengineering. They also argue that the IMO regulations will result in a “more than doubling the long-term mean warming rate” this decade.

Yuan et al use a combination of satellite observations and a chemical transport model to quantify the radiative forcing from the IMO regulations. They find an increase in radiative forcing of 0.2 watts per meter squared (w/m^2) averaged over the entire ocean, with much higher values in specific regions of dense shipping traffic (e.g. 0.56 w/m^2 in the North Atlantic). This is about twice the value that Piers Forster and I initially estimated back in July 2023, but is in-line with the broader range of literature estimates.

Literature estimates of global mean radiative forcing from the IMO 2020 regulations, compiled in Hausfather and Forster 2023.

Unfortunately, while their radiative forcing estimate is well within the range of others in the literature, their calculation of the resulting warming relies on an overly simplified model that results in a substantial overstating of current warming impacts. There are two notable problems with their analysis:

First, their radiative forcing estimate of 0.2 w/m^2 is specifically over the global oceans (where the impact of shipping regulations would be primarily felt), but they proceed to use this value to calculate global temperature impacts of 0.16C at equilibrium (assuming a climate sensitivity of 3C per doubling CO2). However, the oceans only represent 71% of the surface of the planet, so an increase in radiative forcing of 0.2 w/m^2 over the oceans would imply a global average forcing increase of 0.14 w/m^2, resulting in 0.11C warming at equilibrium.

Second, and more importantly, the extremely simplified one-layer energy balance model they use to translate change in radiative forcing to global temperature response assumes a full equilibrium response to changing in forcing with a timescale 7 years (or, more precisely, an e-folding time of ~7 years). This is, to but it bluntly, simply wrong. Their simple model does not reflect real-world heat uptake by the ocean, and no actual climate model has equilibriation times anywhere near that fast.

Global surface temperature response to 0.2 w/m^2 forcing as calculated in the Yuan et al paper (blue line) compared to the same estimated using FaIR.

To provide a more realistic depiction of the warming expected using the forcing estimates in Yuan et al, I used 66 different runs of a climate model emulator (FaIR), with each run tuned to a different member of the latest generation of Earth System Models (the CMIP6 ensemble). Putting in a global average aerosol forcing change increase of 0.14 w/m^2 starting in 2020 gives us the figure below:

Global mean surface temperature warming associated with a 0.14 w/m^2 aerosol forcing across 66 different CMIP6 model emulations in FaIR.

Here we see a global average warming of ~0.08C (0.06C to 0.12C) by 2050 and ~0.05C (0.04C to 0.06C) in the year 2023. The eventual equilibrium response of 0.11C is not reached in the model for a few centuries, though the function approaches it logarithmically.

If we assume that Yuan et al’s 0.2 w/m^2 value over the oceans for the globe as a whole, we get 0.12C (0.08C to 0.17C) by 2050 and ~0.07C (0.05C to 0.09C) in 2023.

Global mean surface temperature warming associated with a 0.2 w/m^2 aerosol forcing across 66 different CMIP6 model emulations in FaIR.

While these values are not trivial, and certainly contributed to record warmth in 2023, its hard to explain more than 0.1C of the global mean temperature increase experienced to-date from the IMO regulations across the range of transient climate responses found in CMIP6 models. It remains a part of the explanation for the weirdness we experienced in the latter half of last year, but far from the full one.

There are a few more minor issues in the paper: taking a 7-year warming estimate and turning it into a decadal rate (whereas in their simple energy balance model the IMO regulations would contribute no additional warming after year seven), or the starting a new trend line at a local maximum in 2020 in their Figure 3. But the main takeaway here is that its hard to justify a global temperature effect of 0.16C that they report based on their estimate of the change in forcing over the ocean.