What We Publish

Geoengineering as an optimization problem

George A. Ban-Weiss and Ken Caldeira · Environmental Research Letters 5, 034009 · 2010

Key finding. Seeking the latitudinal distribution of stratospheric aerosols that most closely reproduces a low-CO2 climate despite high CO2, the study finds that stratospheric aerosol loading higher in polar regions than in the tropics performs better than a uniform distribution.

Six panels in two columns, headed optimize to minimize temperature change and optimize to minimize change in precipitation minus evaporation. The top row shows the sulfate aerosol distribution against latitude, comparing a uniform loading with a parabolic one that is heaviest at the poles. The middle and bottom rows show the residual change in temperature and in precipitation minus evaporation by latitude, with the doubled-CO2 case as a dashed red line far from zero and the geoengineered cases close to it.
Optimising for temperature and optimising for the water cycle give different aerosol distributions. A polar-weighted (parabolic) loading flattens the residual temperature change across latitudes better than a uniform one, but no distribution zeroes both residuals at once. Figure 3 from Ban-Weiss and Caldeira (2010), Environmental Research Letters 5, 034009. Reproduced under CC BY 3.0. Extracted from the published PDF and resized for web display.

What question did this research address?

Earlier modelling had asked what would happen if aerosols were added to the stratosphere, and had established that doing so could reduce surface temperatures without recreating a low-CO2 climate.

This paper inverted the question. Rather than predicting the consequences of an arbitrary aerosol distribution, it asked which distribution would come closest to a chosen target — treating geoengineering as something to be optimized rather than merely simulated.

What did we find?

The optimization was carried out with the NCAR CAM3.1 general circulation model, searching over latitudinal aerosol distributions rather than testing a single prescribed case.

A polar-weighted distribution outperforms a tropical or uniform one. This follows from where greenhouse warming is concentrated — high latitudes warm most under increased CO2, so the compensating cooling should be concentrated there too.

Framing the problem as an optimization changes what the model is being asked. Instead of "what does this intervention do", the question becomes "what intervention best achieves this objective", which requires stating the objective explicitly.

No distribution recreates a low-CO2 climate exactly. The optimization finds the closest achievable approximation, not a solution.

Why does it matter?

The reframing is the durable contribution. Once geoengineering is posed as an optimization problem, the design space becomes something that can be searched, and subsequent work on tailoring interventions to objectives follows from that move.

Stating the objective explicitly also exposes a question that simulation alone conceals. Optimizing for global mean temperature, for regional temperature, or for precipitation gives different answers, so someone must decide what is being optimized for — and that is not a scientific question.

Citation

George A. Ban-Weiss and Ken Caldeira (2010). Geoengineering as an optimization problem. Environmental Research Letters 5, 034009.

Related