Tuning seasonal and latitudinal variations of stratospheric aerosol optical depth in idealised geoengineering scenarios with very large ensembles of GCM runs

The climateprediction.net (CPDN) project can generate GCM ensembles with over 100,000 members, and explore a wider range of perturbed physics than any other GCM, providing unique opportunities both for studying the statistics of extreme events and for exploring high-dimensional and noisy optimization problems. We take advantage of this to attempt to optimize the seasonal and latitudinal variations of stratospheric aerosol optical depth (AOD) in a full GCM, assuming detailed control is possible. Large ensembles sizes provide a greater signal-to-noise ratio to quantify changes to the hydrological cycle, and allow for a large range of distributions to be tested in a single model.

Optimisation targets include cancellation of surface temperature anomalies, net top-of-atmosphere radiative forcing anomalies, and zonal precipitation - evaporation anomalies, within the model. In addition, idealised "basis vector" distributions based on stratospheric injections from the GLENS experiment and GeoMIP G4 aerosol optical depth distributions are analysed, to investigate the effects of AOD distributions sourced from injections simulated in different stratospheric models.

These experiments explore the possibility to "optimise" geoengineering, including trade-offs between local temperature changes and shifts in regional precipitation features, particularly monsoons and the ITCZ. We also hope to present results that investigate the impact of the timing of SRM interventions along an idealised emissions trajectory.