Search

The Geoengineering Model Intercomparison Project (GeoMIP) is a coordinating framework, started in 2010, that includes a series of standardized climate model experiments aimed at understanding the physical processes and projected impacts of solar geoengineering. Numerous experiments have been conducted, and numerous more have been proposed as "testbed' experiments, spanning a variety of geoengineering techniques aimed at modifying the planetary radiation budget: stratospheric aerosol injection, marine cloud brightening, surface albedo modification, cirrus cloud thinning and sunshade mirrors. To date, more than one hundred studies have been published that used results from GeoMIP simulations. Here we provide a critical assessment of GeoMIP and its experiments. We discuss its successes and missed opportunities, for instance in terms of which experiments elicited more interest from the scientific community and which didn't, and the potential reasons why that happened. We also discuss the knowledge that GeoMIP has contributed to the field of geoengineering research and climate science as a whole: what have we learned in terms of inter-model differences, robustness of the projected outcomes for specific geoengineering methods and future areas of models' development that would be necessary in the future. We also offer multiple examples of cases where GeoMIP experiments were fundamental for international assessments of climate change. Finally, we provide a series of recommendations, regarding both future experiments and more general activities, with the goal of continuously deepening our understanding of the effects of potential geoengineering approaches, as well as reducing uncertainties in climate outcomes, important for assessing wider impacts on societies and ecosystems. In doing so, we refine the purpose of GeoMIP and outline a series of criteria whereby GeoMIP can best serve its participants, stakeholders, and the broader science community. [ABSTRACT FROM AUTHOR]

There is now a substantial literature of climate model studies of equatorial or tropical stratospheric SO2 injections that aim to counteract the surface warming produced by rising concentrations of greenhouse gases. Here we present the results from the first systematic intercomparison of climate responses in three Earth System Models where the injection of SO2 occours at different latitudes in the lower stratosphere. Our aim is to determine commonalities and differences between the climate model responses in terms of the distribution of the optically reflective sulfate aerosols produced from the oxidation of SO2, and in terms of the surface response to the resulting reduction in solar radiation. A focus on understanding the contribution of characteristics of models transport alongside their microphysical and chemical schemes, and on evaluating the resulting stratospheric responses in different models is given in the companion paper (Bednarz et al., 2022). The goal of this exercise is not to evaluate these single point injection simulations as stand-alone proposed strategies to counteract global warming; instead we determine sources and areas of agreement and uncertainty in the simulated responses and, ultimately, the possibility of designing a comprehensive intervention strategy capable of managing multiple simultaneous climate goals through the combination of different injection locations. We find large disagreements between GISS-E2.1-G and the CESM2-WACCM6 and UKESM1.0 models regarding the magnitude of cooling per unit of aerosol optical depth (AOD) produced, from 4.7 K per unit of AOD in CESM2-WACCM6 to 16.7 K in the GISS-E2.1-G version with modal aerosol microphysics. By normalizing the results with the global mean response in each of the models, and thus assuming that the amount of SO2 injected is a free parameter that can be managed independently, we highlight some commonalities in the overall distributions of the aerosols, in the inter-hemispheric surface temperature response and in shifts to the Inter-Tropical Convergence Zone, and also some areas of disagreement, such as the aerosol confinement in the equatorial region and the transport to polar latitudes. [ABSTRACT FROM AUTHOR]

The paper constitutes part 2 of a study performing a first systematic inter-model comparison of the atmospheric responses to stratospheric sulfate aerosol injections (SAI) at various latitudes as simulated by three state-of-the-art Earth System Models - CESM2(WACCM6), UKESM1.0, and GISS-E2.1-G. We use a set of five sensitivity experiments with constant annual injections of SO2 in the lower stratosphere at either 30° S, 15° S, 0°, 15° N or 30° N. We identify the similarities and differences in the simulated responses amongst the models as well as demonstrate the role of biases in the climatological circulation and specific aspects of the model microphysics and chemistry in driving the inter-model differences. Building on part 1 (Visioni et al., 2022), we explain the simulated differences in the aerosol spatial distribution between the models: CESM2 shows significantly faster shallow branches of the Brewer Dobson circulation facilitating transport of the relatively larger-sized aerosol to higher latitudes; UKESM shows a relatively isolated tropical pipe and older tropical age-of-air confining the relatively smaller-sized aerosols to the tropics; and the two GISS versions with either bulk or modal aerosol microphysics show elevated sulfate levels at higher latitudes as the result of smaller aerosol sizes and relatively stronger horizontal mixing (thus very young stratospheric age-of-air). We then elucidate the role of these factors in driving the stratospheric responses to SAI. We find a large spread in the magnitudes of the tropical lower stratospheric warming amongst the models, which can be partially attributed to the differences in aerosol distribution and sizes. Regarding the stratospheric ozone responses, we find a good agreement in the tropics between the models with modal microphysics, with lower stratospheric ozone changes consistent with the SAI-induced modulation of the large-scale circulation and the resulting changes in transport. In contrast to the relative agreement at low latitudes, we find a large inter-model spread in the Antarctic ozone responses that can largely be explained by the differences in the simulated latitudinal distributions of aerosols as well as the degree of implementation of heterogeneous halogen chemistry on sulfate. Finally, we also find large differences in stratospheric water vapour responses amongst the models, with CESM2 and GISS with modal microphysics both showing significant increases in stratospheric water vapour under SAI consistent with the increase in cold point temperatures that were largely not reproduced in UKESM. For the GISS runs with bulk microphysics, the SAI simulations show contrastingly different stratospheric responses to the models using the modal aerosol treatment, including the absence of lower stratospheric warming as well as significant reductions in stratospheric water vapour and ozone. The results point towards the importance of detailed treatment of aerosol processes, although some problems in halogen chemistry in GISS are identified that require further attention. Overall, our results contribute to an increased understanding of the underlying physical mechanisms as well as the sources of uncertainty in model projections of climate impacts from SAI. [ABSTRACT FROM AUTHOR]

As part of the Geoengineering Model Intercomparison Project a numerical experiment known as G6sulfur has been designed in which temperatures under a high-forcing future scenario (SSP5-8.5) are reduced to those under a medium-forcing scenario (SSP2-4.5) using the proposed geoengineering technique of stratospheric aerosol intervention (SAI). G6sulfur involves introducing sulphate aerosol into the tropical stratosphere where it reflects incoming sunlight back to space, thus cooling the planet. Here we compare the results from six Earth-system models which have performed the G6sulfur experiment and examine how SAI affects two important modes of natural variability, the northern wintertime North Atlantic Oscillation (NAO) and the Quasi-Biennial Oscillation (QBO). Although all models show that SAI is successful in reducing global-mean temperature as designed, they are also consistent in showing that it forces an increasingly positive phase of the NAO as the injection rate increases over the course of the 21st century, exacerbating precipitation reductions over parts of southern Europe compared with SSP5-8.5. In contrast to the robust result for the NAO there is less consistency for the impact on the QBO, but the results nevertheless indicate a risk that equatorial SAI could cause the QBO to stall and become locked in a phase with permanent westerly winds in the lower stratosphere. [ABSTRACT FROM AUTHOR]

Sulfate geoengineering (SG) methods based on lower stratospheric tropical injection of sulfur dioxide (SO2) have been widely discussed in recent years, focusing on the direct and indirect effects they would have on the climate system. Here a potential alternative method is discussed, where sulfur emissions are located at the surface in the form of carbonyl sulfide (COS) gas. A time-dependent chemistry-climate model experiment is designed from year 2021 to 2055, assuming a 40 Tg-S/yr artificial global flux of COS, geographically distributed following the present day anthropogenic COS surface emissions. The budget of COS and sulfur species is discussed, as well as the effects of SG-COS on the stratospheric sulfate aerosol optical depth (Δτ=0.080 in years 2046-2055), aerosol effective radius (0.46 µm), surface SOx deposition (+8.7 %) and tropopause radiative forcing (RF) (-2.0 W/m² for clear sky conditions and -1.5 W/m² including the cloud adjustment). Indirect effects on ozone, methane and stratospheric water vapor are also considered, along with the COS direct contribution (with an overall gas phase global radiative forcing of +0.23W/m²). According to our model results, the resulting net RF of this SG-COS experiment is -1.3 W/m² for the year 2050, and it is comparable to the corresponding RF of -1.7 W/m² obtained with a sustained injection of 4 Tg-S/yr in the tropical lower stratosphere in the form of SO2 (SG-SO2, able to produce a comparable increase of the sulfate aerosol optical depth). Significant changes of the stratospheric ozone response are found in SG-COS with respect to SG-SO2 (+4.9 DU versus +1.5 DU, globally). According to the model results, the resulting UVB perturbation at the surface accounts to -4.3% as a global-annual average (versus -2.4% in the SG-SO2 case), with a springtime Antarctic decrease of -2.7% (versus a +5.8% increase in the SG-SO2 experiment). Overall, we find that an increase in COS surface emission may be feasible, and produce a more latitudinally-uniform forcing without the need for the deployment of stratospheric aircrafts. [ABSTRACT FROM AUTHOR]

Simulating the complex aerosol microphysical processes in a comprehensive Earth System Model can be very computationally intensive and therefore many models utilize a modal approach, where aerosol size distributions are represented by observations-derived lognormal functions. This approach has been shown to yield satisfactory results in a large array of applications, but there may be cases where the simplification in this approach may produce some shortcomings. In this work we show specific conditions under which the current approximations used in modal approaches might yield some incorrect answers. Using results from the Community Earth System Model v1 (CESM1) Geoengineering Large Ensemble (GLENS) project, we analyze the effects in the troposphere of a continuous increasing load of sulfate aerosols in the stratosphere, with the aim of counteracting the surface warming produced by non-mitigated increasing greenhouse gases concentration between 2020-2100.We show that the simulated results pertaining to the evolution of sea salt and dust aerosols in the upper troposphere are not realistic due to internal mixing assumptions in the modal aerosol treatment, which in this case reduces the size, and thus the settling velocities, of those particles and ultimately changes their mixing ratio below the tropopause. The unnatural increase of these aerosol species affects, in turn, the simulation of upper tropospheric ice formation, resulting in an increase in ice clouds that is not due to any meaningful physical mechanisms. While we show that this does not significantly affect the overall results of the simulations, we point to some areas where results should be interpreted with care in modeling simulations using similar approximations: in particular, the evolution of upper tropospheric clouds when large amount of sulfate is present in the stratosphere, as after a large explosive volcanic eruption or in similar stratospheric aerosol injection cases. Finally, we suggest that this could be avoided if sulfate aerosols in the coarse mode, the predominant species in these situation, are treated separately from other aerosol species in the model. [ABSTRACT FROM AUTHOR]

We present here results from the Geoengineering Model Intercomparison Project (GeoMIP) simulations for the experiment G6sulfur and G6solar for six Earth System Models participating in the Climate Model Intercomparison Project (CMIP) Phase 6. The aim of the experiments is to reduce the warming from that resulting from a high-tier emission scenario (Shared Socioeconomic Pathways SSP5-8.5) to that resulting from a medium-tier emission scenario (SSP2-4.5). These simulations aim to analyze the response of climate models to a reduction in incoming surface radiation as a means to reduce global surface temperatures, and they do so either by simulating a stratospheric sulfate aerosol layer or, in a more idealized way, through a uniform reduction in the solar constant in the model. We find that, by the end of the century, there is a considerable inter-model spread in the needed injection of sulfate (29 ± 9 Tg-SO2/yr between 2081 and 2100), in how the aerosol cloud is distributed latitudinally, and in how stratospheric temperatures are influenced by the produced aerosol layer. Even in the simpler G6solar experiment, there is a spread in the needed solar dimming to achieve the same global temperature target (1.91 ± 0.44 %). The analyzed models already show significant differences in the response to the increasing CO2 concentrations for global mean temperatures and global mean precipitation (2.05 K ± 0.42 K and 2.28 ± 0.80 %, respectively, for the SSP5-8.5-SSP2-4.5 difference between 2081 and 2100): the differences in the simulated aerosol spread then change some of the underlying uncertainty, for example in terms of the global mean precipitation response (-3.79 ± 0.76 % for G6sulfur compared to -2.07 ± 0.40 % for G6solar against SSP2-4.5 between 2081 and 2100). These differences in the aerosols behavior also result in a larger inter-model spread in the regional response in the surface temperatures in the case of the G6sulfur simulations, suggesting the need to devise various, more specific experiments to single out and resolve particular sources of uncertainty. The spread in the modelled response suggests that a degree of caution is necessary when using these results for assessing specific impacts of geoengineering in various aspects of the Earth system: however, all models agree that, compared to a scenario with unmitigated warming, stratospheric aerosol geoengineering has the potential to both globally and locally reduce the increase in surface temperatures. [ABSTRACT FROM AUTHOR]

Solar geoengineering has been receiving increased attention in recent years as a potential temporary solution to offset global warming. One method of approximating global-scale solar geoengineering in climate models is via solar reduction experiments. Two generations of models in the Geoengineering Model Intercomparison Project (GeoMIP) have now simulated offsetting a quadrupling of the CO2 concentration with solar reduction. This simulation is artificial and designed to elicit large responses in the models. Here we show that energetics, temperature, and hydrological cycle changes in this experiment are statistically indistinguishable between the two ensembles. Of the variables analyzed here, the only major differences involve highly parameterized and uncertain processes, such as cloud forcing or terrestrial net primary productivity. We conclude that despite numerous structural differences and uncertainties in models over the past 20 years, including an increase in climate sensitivity in the latest generation of models, broad conclusions about the climate response to global solar dimming remain robust. [ABSTRACT FROM AUTHOR]

The realisation of the difficulty of limiting global mean temperatures to within 1.5 °C or 2.0 °C above pre-industrial levels stipulated by the 21st Conference of Parties in Paris has led to increased interest in solar radiation management (SRM) techniques. Proposed SRM schemes aim to increase planetary albedo to reflect more sunlight back to space and induce a cooling that acts to partially offset global warming. Under the auspices of the Geoengineering Model Intercomparion Project, we have performed model experiments whereby global temperature under the high forcing SSP5-8.5 scenario is reduced to follow that of the medium forcing SSP2-4.5 scenario. Two different mechanisms to achieve this are employed, the first via a reduction in the solar constant (experiment G6solar) and the second via modelling injections of sulfur dioxide (experiment G6sulfur) which forms sulfate aerosol in the stratosphere. Results from two state-of-the-art coupled Earth system models both show an impact on the North Atlantic Oscillation (NAO) in G6sulfur but not in G6solar. Both models show a persistent positive anomaly in the NAO during the Northern Hemisphere winter season in G6sulfur, suggesting an increase in zonal flow and an increase in North Atlantic storm track activity impacting the Eurasian continent leading to regional warming. These findings are broadly consistent with previous findings on the impact of stratospheric volcanic aerosol on the NAO and emphasise that detailed modelling of geoengineering processes is required if accurate impacts of SRM impacts are to be simulated. Differences remain between the two models in predicting regional changes over the continental USA and Africa, suggesting that more models need to perform such simulations before attempting to draw any conclusions regarding potential continental-scale climate change under SRM. [ABSTRACT FROM AUTHOR]

We investigate techniques for using deep neural networks to produce surrogate models for short term climate forecasts. A convolutional neural network is trained on 97 years of monthly precipitation output from the 1pctCO2 run (the CO2 concentration increases by 1% per year) simulated by the CanESM2 Earth System Model. The neural network clearly outperforms a persistence forecast and does not show substantially degraded performance even when the forecast length is extended to 120 months. The model is prone to underpredicting precipitation in areas characterized by intense precipitation events. Scheduled sampling (forcing the model to gradually use its own past predictions rather than ground truth) is essential for avoiding amplification of early forecasting errors. However, the use of scheduled sampling also necessitates preforecasting (generating forecasts prior to the first forecast date) to obtain adequate performance for the first few prediction time steps. We document the training procedures and hyperparameter optimization process for researchers who wish to extend the use of neural networks in developing surrogate models. [ABSTRACT FROM AUTHOR]

Climate sensitivity in Earth System Models (ESMs) is an emergent property that is affected by structural (missing or inaccurate model physics) and parametric (variations in model parameters) uncertainty. This work provides the first quantitative assessment of the role of compensation between uncertainties in aerosol forcing and atmospheric parameters, and their impact on the climate sensitivity of the Community Atmosphere Model, Version 4 (CAM4). Running the model with prescribed ocean and ice conditions, we perturb four parameters related to sulfate and black carbon aerosol radiative forcing and distribution, as well as five atmospheric parameters related to clouds, convection, and radiative flux. The atmospheric parameters explain more than 85 \% of the variance in climate sensitivity for the ranges of parameters explored here, with two parameters being the most important: one controlling low cloud amount, and one controlling the timescale for deep convection. Although the aerosol parameters strongly affect aerosol optical depth, the effects of these aerosol parameters on climate sensitivity are substantially weaker than the effects of the atmospheric parameters. Based on comparisons to inter-model spread of other ESMs, we conclude that structural uncertainties in this configuration of CAM4 likely contribute three times more to uncertainty in climate sensitivity than parametric uncertainty. We provide several parameter sets that could provide plausible (measured by a skill score) configurations of CAM4, but with different sulfate aerosol radiative forcing, black carbon radiative forcing, and climate sensitivity. [ABSTRACT FROM AUTHOR]

Marine cloud brightening has been proposed as a means of geoengineering/climate intervention, or deliberately altering the climate system to offset anthropogenic climate change. As an idealized representation of marine cloud brightening, this paper discusses experiment G1ocean-albedo of the Geoengineering Model Intercomparison Project (GeoMIP), involving an abrupt quadrupling of the CO2 concentration and an instantaneous increase in ocean albedo to maintain approximate net top-of-atmosphere radiative flux balance. Eleven Earth System Models are relatively consistent in their temperature, radiative flux, and hydrological cycle responses to this experiment. Due to the imposed forcing, air over the land surface warms by a model average of 1.14 K, while air over most of the ocean cools. Some parts of the near-surface air temperature over ocean warm due to heat transport from land to ocean. These changes generally resolve within a few years, indicating that changes in ocean heat content play at most a small role in the warming over the oceans. The hydrological cycle response is a general slowing down, with high heterogeneity in the response, particularly in the tropics. While idealized, these results have important implications for marine cloud brightening, or other methods of geoengineering involving spatially heterogeneous forcing, or other general forcings with a strong land/ocean contrast. It also reinforces previous findings that keeping top-of-atmosphere net radiative flux constant is not sufficient for preventing changes in global mean temperature. [ABSTRACT FROM AUTHOR]

We examine extreme temperature and precipitation under two potential geoengineering methods forming part of the Geoengineering Model Intercomparison Project (GeoMIP). The solar dimming experiment G1 is designed to completely offset the global mean radiative forcing due to a CO2-quadrupling experiment (abrupt 4 × CO2), while in GeoMIP experiment G4, the radiative forcing due to the representative concentration pathway 4.5 (RCP4.5) scenario is partly offset by a simulated layer of aerosols in the stratosphere. Both G1 and G4 geoengineering simulations lead to lower maximum temperatures at higher latitudes, and on land primarily through feedback effects involving high latitude processes such as snow cover, sea ice and soil moisture. Maximum 5-day precipitation increases over subtropical oceans, whereas warm spells decrease markedly in the tropics, and the number of consecutive dry days decreases in most deserts. The precipitation during the tropical cyclone (hurricane) seasons becomes less intense, whilst the remainder of the year becomes wetter. Aerosol injection is more effective than dimming in moderating extreme precipitation (and flooding), possibly due to stratospheric warming by aerosol injection working in tandem with sea surface temperature reductions to moderate extreme tropical storm cyclogenesis. The differences in the response of temperature extremes between the two types of geoengineering are relatively minor. Despite the magnitude of the radiative forcing applied in G1 being ~ 6.5 times larger than in G4, and differences in the aerosol chemistry and transport schemes amongst the models, one can discern clear differences in the precipitation extremes between the types of geoengineering probably due to the aerosol direct effect and related energetic changes. [ABSTRACT FROM AUTHOR]

Here we show results from Earth System Model simulations from the marine cloud brightening experiment G4cdnc of the Geoengineering Model Intercomparison Project (GeoMIP). The nine contributing models prescribe a 50 % increase in the cloud droplet number concentration (CDNC) of low clouds over the global oceans, with the purpose of counteracting the radiative forcing due to anthropogenic greenhouse gases under the RCP4.5 scenario. The model ensemble median effective radiative forcing (ERF) amounts to -1.9 Wm-2, with a substantial inter-model spread of -0.6 to -2.5 Wm-2. The large spread is partly related to the considerable differences in clouds and their representation between the models, with an underestimation of low clouds in several of the models. All models predict a statistically significant temperature decrease with a median of (for years 2020-2060) -0.95 [-0.18 to -1.19] K relative to the RCP4.5 scenario, with particularly strong cooling over low-latitude continentss. Globally averaged there is a weak but significant precipitation decrease of -2.24 [-0.49 to -2.90] % due to a colder climate, but at low latitudes there is a 1.20 % increase over land. This increase is part of a circulation change where a strong negative TOA short-wave forcing over subtropical oceans, caused by increased albedo associated with the increasing CDNC, is compensated by rising motion and positive TOA long-wave signals over adjacent land regions. [ABSTRACT FROM AUTHOR]

Marine cloud brightening through sea spray injection has been proposed as a climate engineering method for avoiding the most severe consequences of global warming. A limitation of most of the previous modelling studies on marine cloud brightening is that they have either considered individual models, or only investigated the effects of a specific increase in the number of cloud droplets. Here we present results from coordinated simulations with three Earth system models (ESMs) participating in the Geoengineering Model Intercomparison Project (GeoMIP) G4sea-salt experiment. Injection rates of accumulation mode sea spray aerosol particles over ocean between 30° N and 30° S are set in each model to generate a global-mean effective radiative forcing (ERF) of -2.0 W m-2 at the top of atmosphere. We find that the injection increases the cloud droplet number concentration in lower layers, reduces the cloud-top effective droplet radius, and increases the cloud optical depth over the injection area. We also find, however, that the global-mean clear-sky ERF by the injected particles is as large as the corresponding total ERF in all three ESMs, indicating a large potential of the aerosol direct effect in regions of low cloudiness. The largest enhancement in ERF due to the presence of clouds occur as expected in the subtropical stratocumulus regions off the west coasts of the American and African continents. However, outside these regions, the ERF is in general equally large in cloudy and clear-sky conditions. These findings suggest a more important role of the aerosol direct effect in sea spray climate engineering than previously thought. [ABSTRACT FROM AUTHOR]

This study evaluates the forcing and feedback of net shortwave radiation at the surface in the G4 experiment of the Geoengineering Model Intercomparison Project by analysing outputs from six participating models. G4 involves injection of 5 Tg yr-1 of SO2, a sulphate aerosol precursor, into the lower stratosphere from year 2020 to 2070 against a background scenario of RCP4.5. A single layer atmospheric model for shortwave radiative transfer is used to estimate the direct forcing of solar radiation management (SRM) and feedback effects from changes in the water vapour amount, cloud amount, and surface albedo (compared with RCP4.5). The analysis shows that the globally and temporally averaged SRM forcing ranges from -3.6 to -1.6 W m-2, depending on the model. The feedback effects due to changes in the water vapour and cloud amounts on net shortwave radiation have heating effects ranging from approximately 0.4 to 1.5 W m-2 and weaken the effect of SRM by around 50%. The surface albedo changes have a cooling effect, which is locally strong (~ 4 W m-2) in snow and sea ice melting regions, but minor for the global average. The analyses show that the results of the G4 experiment, which simulates sulphate geoengineering, include large inter-model variability both in the direct SRM forcing and the feedback from changes in the cloud amount, and imply a high uncertainty in modelled processes of sulphate aerosols and clouds. [ABSTRACT FROM AUTHOR]

We introduce system identification techniques to climate science wherein multiple dynamic input-output relationships can be simultaneously characterized in a single simulation. This method, involving multiple small perturbations (in space and time) of an input field while monitoring output fields to quantify responses, allows for identification of different timescales of climate response to forcing without substantially pushing the climate far away from a steady state. We use this technique to determine the steady state responses of low cloud fraction and latent heat flux to heating perturbations over 22 regions spanning Earth's oceans. We show that the response characteristics are similar to those of step-change simulations, but in this new method, the responses for 22 regions can be characterized simultaneously. Furthermore, we can estimate the timescale over which the steady state response emerges. The proposed methodology could be useful for a wide variety of purposes in climate science, including characterization of teleconnections and uncertainty quantification to identify the effects of climate model tuning parameters. [ABSTRACT FROM AUTHOR]

Climate emulators trained on existing simulations can be used to project the climate effects that would result from different possible future pathways of anthropogenic forcing, without relying on general circulation model (GCM) simulations for every possible pathway. We extend this idea to include different amounts of solar geoengineering in addition to different pathways of green-house gas concentrations by training emulators from a multi-model ensemble of simulations from the Geoengineering Model Intercomparison Project (GeoMIP). The emulator is trained on the abrupt 4 x CO2 and a compensating solar reduction simulation (G1), and evaluated by comparing predictions against a simulated 1% per year CO2 increase and a similarly smaller solar reduction (G2). We find reasonable agreement in most models for predicting changes in temperature and precipitation (including regional effects), and annual-mean Northern hemisphere sea ice extent, with the difference between simulation and prediction typically smaller than natural variability. This verifies that the linearity assumption used in constructing the emulator is sufficient for these variables over the range of forcing considered. Annual-minimum Northern hemisphere sea ice extent is less-well predicted, indicating the limits of the linearity assumption. For future pathways involving relatively small forcing from solar geoengineering, the errors introduced from nonlinear effects may be smaller than the uncertainty due to natural variability, and the emulator prediction may be a more accurate estimate of the forced component of the models' response than an actual simulation would be. [ABSTRACT FROM AUTHOR]