GeoMIP Simulations: Testbed

The GeoMIP Testbed was introduced by Kravitz et al. (2015) as a way for other communities to participate in GeoMIP. The idea is that individuals can propose potential GeoMIP experiments to be performed by a small number of models. If these experiments turn out to be fruitful, they can then be considered for adoption by GeoMIP as official core experiments (via following this procedure).

All current testbed experiments will be listed here as they are added:

Overshoot Scenarios

GeoMIP6 includes an overshoot scenario, in which a temperature target is initially exceeded and then achieved late on in the simulation. This suite of three testbed experiments is designed to provide a set of experiments to explore this idea of an overshoot. Contact: Simone Tilmes (Simone Tilmes)

  1. G6sulfur_1.5 is designed to reduce surface temperatures in a high emission scenario (SSP5-8.5) to 1.5C above preindustrial conditions (1850-1900 average). This experiment requires sulfur injections at 4 locations in the stratosphere (15S, 15N, 30S, and 30N) or equivalent modifications of AOD using a feedback algorithm designed to maintain global mean temperature, the interhemispheric temperature gradient, and the equator-to-pole temperature gradient at the period corresponding to global mean temperature rise of 1.5C, as described by Kravitz et al., 2017. This experiment is designed to understand the effectiveness of sulfur injections with increasing injection amount between different models, comparisons of the required injection locations, and linearity of impacts with increasing injection amount.
  2. G8overshoot_1.5 is similar to G6sulfur_1.5 but uses the SSP5-34-OS future scenario as the baseline, which is the same as SSP5-8.5 until 2040. After 2040, this scenario assumes large amounts of decarbonization leading to an overshoot of CO2 mixing ratios above 550ppm around 2060 and reductions to below 500ppm by the end of the 21st century. G8overshoot_1.5 is equivalent to G6sulfur_1.5 experiment until 2040, after which modelers that use a feedback algorithm have to adjust the feedforward assumption to the SSP5-34-OS baseline scenario. This experiment is designed to produce a case where sulfur injections are gradually ramped up and ramped down for a limited time in order to keep surface temperatures at a temperature target of the control period. This experiment is described by Tilmes et al., 2019.
  3. G8overshoot_2.0 is similar to G8overshoot_1.5 but controls for a global temperature rise of 2.0C. Some models with low climate sensitivity may not be able to perform this experiment. This experiment, when compared to G8overshoot_1.5, will identify differences in the range of impacts if either the 1.5C or 2.0C temperature targets have been reached using SRM. This experiment is described by Tilmes et al., 2019.

Experiment G6sulfur is designed to reduce radiative forcing in a high emissions scenario to that of a moderate emissions scenario via simulation of stratospheric sulfate aerosol injection. This experiment would be useful in assessing the effectiveness of geoengineering as part of a portfolio of responses to climate change. However, this experiment only addresses one potential scenario, i.e., using geoengineering to achieve the forcing from a “medium” scenario. Increasing amounts of stratospheric SO2 injection would cause particles to coagulate and fall out more rapidly. Therefore, the relationship between the amount of injection and the resulting radiative forcing is projected to be sublinear. This problem prompts a natural question: how would the injection amount and the results from that injection differ if geoengineering were used to achieve a larger radiative forcing? This question is the first step in assessing any potential practical limits to stratospheric aerosol injection.

A natural first step in addressing this problem would in- volve a similar setup to that of G6sulfur. Against a background of the ScenarioMIP Tier 1 high forcing scenario, sulfate aerosol precursors would be injected into the stratosphere in sufficient amounts to reduce anthropogenic radiative forcing from the levels in the high forcing scenario to levels in the low forcing scenario.

GeoSulfur5, GeoSulfur20, GeoSulfur50

A different way of quantifying the effects of stratospheric aerosol geoengineering is to perform a series of experiments in which the hypothetical rate of injection of stratospheric sulfate aerosols is constrained. Such a simulation would be well suited to ascertain the range of model responses to a fixed amount of SO2 injection, highlighting model diversity. Against a background of the ScenarioMIP Tier 1 high forcing scenario, the modeling groups will inject 10, 20, or 50 Tg SO2 per year into the lower stratosphere, in a similar setup to experiment G4.


Experiment G1ocean-albedo has simulated the effects of marine cloud brightening by increasing ocean albedo by a constant multiplication factor. However, GeoMIP has not yet explored land-based approaches towards solar radiation management. Such approaches could readily be implemented on the regional scale, as human activities already control the albedo of a significant fraction of the land surface. We therefore propose an alternative experiment in which the land surface albedo is increased.

A protocol for this experiment has been designed and is available by clicking here. This experiment has specific model output requirements, which can be found by clicking here. For further details, please contact Annette Hirsch.


The G4Foam Experiment involves ocean albedo increases in several localized regions to take advantage of the power of climate system feedbacks in providing more broad-scale cooling. A formal protocol has not yet been established, but initial results from a single-model study are discussed by Gabriel et al. (2017).


Instead of using SO2 injection, a different injection strategy could be direct condensation of H2SO4. This method is hypothesized to have numerous advantages over SO2 injection, but it has not been tested in state-of-the-art Earth System Models, nor in a multi-model context. Details of the experimental protocol are provided here.