Solar radiation modification (SRM) climate interventions have been subject to increasing scientific interest in recent years owing to their potential to offset the impacts of greenhouse gas warming by reflecting a portion of incoming solar radiation. National and international assessments have made clear there is a need to better understand the feasibility, risks, benefits, and policy implications of SRM technologies . Marine cloud brightening (MCB) is an approach to increase sunlight reflection and cool climate by injecting sea salt aerosols (iSSA) into the lower marine atmosphere with the goal of making clouds more reflective. The idea is that the iSSA serve as cloud condensation nuclei (CCN), leading to clouds with smaller and more numerous cloud droplets. This change can increase the albedo of the clouds and reduce the amount of sunlight reaching Earth's surface, which we refer to as the cloud albedo effect and is also known as the Twomey effect . Increased cloud droplet number concentration (CDNC) further affects clouds by reducing the rate of precipitation formation, which leads to cloud adjustments via longer cloud lifetime the cloud lifetime effect – and by changing cloud and droplet dynamics, leading to cloud thickening or thinning, depending on cloud conditions e.g.,. Together these processes are referred to as aerosol-cloud interactions or indirect aerosol effects (Faci). Modeling and observational evidence from analogues of MCB, such as particle emissions from shipping and volcanic eruptions, suggest that MCB could produce substantial climate cooling , but the extent of the potential cooling and the impacts of increasing cloud reflectivity in different ocean regions are still uncertain.

Studies with Earth system models (ESMs) suggest that iSSA can achieve substantial cooling through its impact on clouds and, in some models, marine sky brightening (MSB) driven by the scattering of sunlight by the iSSA itself i.e., through aerosol–radiation interactions or direct aerosol forcing, Fari . Thus, iSSA injection could be a complement to emissions reductions that would offset some impacts of global climate change. ESMs suggest that the climate response to MCB depends strongly on the location(s) where MCB is applied hereafter H2026. Past modeling has largely focused on MCB implementation in the subtropical east Pacific and Atlantic oceans, due to the extensive coverage of marine stratocumulus clouds in these regions and the high susceptibility of those stratocumulus clouds to brightening by aerosol increases. However, ESMs consistently show strong La Niña-like climate responses to MCB interventions in these regions, and this response can result in highly inhomogeneous regional temperature responses and shifting precipitation patterns . Thus, while subtropical stratocumulus MCB may be suitable in terms of achieving efficient brightening and global-scale cooling, ESMs suggest it could also produce undesirable climate responses.

Other ESM studies indicate that shifting the implementation of MCB to the midlatitudes tends to produce more spatially uniform cooling patterns H2026, with few regional temperature anomalies and less dramatic precipitation shifts compared to subtropical MCB simulations. Furthermore, the ESM simulations presented in H2026 indicate that midlatitude MCB – in five midlatitude regions: North Pacific – NP [30N to 50° N; 170° E to 120° W], North Atlantic – NA [30 to 50° N; 70 to 0° W], South Pacific – SP [50 to 30° S; 170 to 90° W], South Atlantic – SA [50 to 30° S; 55° W to 15° E], and South Indian Ocean – SI [50 to 30° S; 30 to 100° E] – could produce both efficient cloud brightening and a stronger cooling response for a given amount of brightening than would MCB intervention in the subtropics, due to stronger positive cloud and albedo radiative feedbacks in these regions H2026. By combining iSSA emissions across Northern and Southern Hemisphere midlatitude ocean regions, we found that midlatitude MCB produces a cooling pattern that is anti-correlated with the greenhouse gas (GHG) warming pattern across three models (H2026). This suggests that midlatitude MCB could offset the GHG warming effectively without inducing undesirable regional climate impacts, in contrast to subtropical MCB .

In this study, we use the midlatitude MCB strategy identified in H2026 to design MCB intervention scenarios in which global mean surface temperature increases are held to approximately 1.5C of warming above preindustrial, defined as the 2020–2039 mean temperature following G6-1.5K-SAI protocol . These types of scenario simulations are vital for supporting policy-relevant climate impacts analyses of SRM . There has been extensive development and study of this type of scenario simulations for SAI, such as those developed for the Geoengineering Model Intercomparison Project GeoMIP –, the Geoengineering Large Ensemble GLENS –, and the ARISE-SAI protocol Assessing Responses and Impacts of Solar intervention on the Earth system with Stratospheric Aerosol Injection –. These efforts have provided vital information about the sensitivity of the climate response to the details of SAI deployment, such as the injection locations, target temperature, and start date of the intervention. Furthermore, they have enabled a rich literature on potential impacts of SAI across regions and climate system components.

In contrast, there have been relatively few studies that have simulated hypothetical future scenarios in which MCB interventions are applied. Many MCB studies have used large step-change perturbations aimed at increasing the signal strength to understand the processes and dynamics underlying the response to MCB H2026. Fewer MCB studies have used simulation designs that are more suitable for impacts analysis in which MCB forcing is increased gradually to represent a hypothetical future deployment trajectory, which we refer to as ”scenario” simulations. , for example, evaluated a GeoMIP scenario (G3-SSCE) in 3 coupled climate models in which iSSA was emitted into the tropics between 30° S and 30° N and the emissions were scaled to maintain the top of atmosphere radiative forcing to RCP4.5 2020 levels. Two of the three ESMs in that study did not include important treatments of some aerosol-cloud interactions for the accumulation mode iSSA particles used for MCB, so it was necessary to use iSSA and CDNC fields from one model, NorESM, in the other two models resulting in some inconsistencies between the meteorology and iSSA/CDNC fields in those models. Furthermore, the tropical MCB strategy used in the study tended to under-cool higher latitude regions and over-cool the tropics . More recently, and simulated scenarios using variations of subtropical MCB intervention strategies. In , iSSA were added into four regions in the Eastern Pacific and increased over time to take the global mean temperature trajectory from SSP5-8.5 to SSP2-4.5. This resulted in a strong La Niña-like cooling in the Pacific which shifted circulation patterns, and drove regional sea level increases relative to SSP5-8.5. presented the first application of an algorithmic controller for an MCB scenario. Their scenario used the controller to vary the areal coverage of CDNC perturbations to maintain GMST at 1.5 °C over preindustrial. The strategy increased CDNC in the most susceptible clouds, primarily summertime subtropical stratocumulus clouds which tend to be most extensive in the subtropical east Pacific. As a result, this scenario exhibits many of the same climate responses as , including excess tropical Pacific cooling and exacerbating warming signals in parts of the midlatitude Pacific.

In this study, we seek to combine the knowledge gained from previous idealized and scenario simulations to define a new set of calibration simulations and a scenario simulation for MCB modeling. These simulations are intended to serve as a testbed to inform MCB simulations protocols that will be included in the next iteration of the Geoengineering Model Intercomparison Project GeoMIP –. GeoMIP is an international effort to coordinate ESM simulation design for a number of SRM methods to enable robust estimates of the potential SRM climate impacts, to quantify inter-model uncertainties, and to improve process understanding of SRM representation in ESMs . MCB simulations have been included in GeoMIP since the study by which defined the G1ocean-albedo, G4cdnc, and G4sea-salt simulations. These simulations used step perturbations to all ocean albedo, to CDNC over all oceans, and to iSSA flux between 30° N and 30° S, respectively, to represent MCB with varying levels of complexity. Analyses of these simulations revealed substantial inter-model spread in forcing due to CDNC and iSSA flux . These simulations, along with G3-SSCE , were designed based on equivalent SAI and solar constant decrease simulations. Specifically, G3-SSCE is based on the G3 simulation and G4cdnc and G4sea-salt are based on the G4 simulation .

Similarly, we seek to develop MCB protocols that can be compared to the SAI protocols developed in the recent G6-1.5K-SAI simulations included in GeoMIP and the CMIP7 “fast track” studies. The G6-1.5K-SAI is a protocol designed to balance simulation complexity and climate response targets by applying hemispherically symmetric stratospheric aerosol injections at 30° N and 30° S to maintain temperatures at 1.5 K above preindustrial. While SAI emissions could be varied across different locations to stabilise large-scale climate metrics e.g. or to mitigate changes in the ITCZ , injecting equal amounts of aerosol at 30° N and 30° S was found to lead to a similar climate response to algorithmically adjusting emissions across 15 and 30° N and S . It was also found that this distribution of emissions more effectively offsets the GHG warming pattern than equatorial SAI, which was used in previous GeoMIP SAI scenarios, while also only requiring adjustments to a single SAI emission rate, substantially reducing the difficulty of setting up, analysing, and comparing the scenario simulations across models.

For our proposed protocol, which we label as “G6-1.5K-MCB”, we use a pattern of MCB iSSA emissions that produces a temperature response that more closely counters the GHG warming pattern than previous MCB strategies and removes many of the detrimental side effects on ENSO. This protocol is the product of the road map proposed in which set out a three stage procedure: (1) quantify forcing due to iSSA emissions, (2) quantify circulation and radiative feedback response to iSSA emissions in different regions, and (3) design simulations to achieve specific climate objectives. Based on analysis of fourteen regional iSSA perturbations, H2026 identified five midlatitude regions that produce a temperature signal that offsets the GHG warming pattern: the South Atlantic (SA), South Pacific (SP), South Indian (SI), North Atlantic (NA), and North Pacific (NP) oceans. Building from these studies, we propose a set of simulations guided by a number of goals:

A clear and detailed specification for benchmarking simulations. Stage 1: fixed sea surface temperature (SST) simulations needed to enable assessment of the inter-model differences in MCB aerosol-cloud interaction forcing (Faci) and aerosol-radiation interaction forcing (Fari) and the resulting effective radiative forcing (Ftot) from iSSA emissions. Stage 2: followup coupled simulations with idealized constant iSSA emission perturbations that are used to determine the ESM's climate sensitivity to MCB. Stage 3: the G6-1.5K-MCB scenario simulations, which use the sensitivity (response per emission rate) from stage 2 to inform the MCB iSSA emission rates required to offset the SSP2-4.5 warming. The uncertainties in MCB Faci requires a strong focus on process understanding, motivating the inclusion of the stage 1 idealized fixed SST simulations. The stage 2 coupled benchmarking simulations follow a similar methodology to G4seasalt, which was included in a previous GeoMIP iteration . The similar simulation design enables a degree of comparison across generations of models and MCB studies, though the differences in iSSA emission location mean one-to-one comparisons are not possible. Finally, the information from stage 2 helps to identify the iSSA emissions required for stage 3.

Owing to the large uncertainty in Faci due to iSSA emissions, benchmarking simulations will be required to characterize a given ESM for the G6-1.5K-MCB simulations . We encourage participants to conduct two benchmark simulations (Stage 1 and 2) following the descriptions below to enable the community to use them to assess the inter-model uncertainties in the forcing and response. These simulations are based on the stage 1 and stage 2 simulations of the MCB-REG protocol described by .

Inj-seasalt-midlat-SST (Stage 1) Fixed sea surface temperature (SST) simulation with time-constant 100 Tg yr⁻¹ iSSA emissions in the five midlatitude regions (NP, NA, SP, SA, SI – see Sect. 2.2) to estimate the ERF from midlatitude MCB (ERF_100 Tg). iSSA is emitted where ocean fraction > 0.5 such that areal-sum emissions are 25 Tg yr⁻¹ in each of the northern hemisphere (NH) regions and 16.67 Tg yr⁻¹ in each of the southern hemisphere (SH) regions. We recommend that the simulations be run for at least 5 years to ensure there is sufficient signal in the radiative forcing.

The temperature response estimated from Stage 2 can then be used to estimate the required iSSA time series for the G6-1.5K-MCB simulations (Sect. ). We note that for the G6-1.5K-MCB simulations shown here, we have not conducted the stage 1 and 2 simulations as outlined above. Instead, we use existing simulations from R2024 and H2026 wherein the benchmark emission rates used in Stage 2 are calculated as the emission rate to achieve -1.8 W m⁻² in fixed SST simulations with MCB iSSA emitted into three subtropical regions (R2024). This benchmark emission rate was then used in coupled simulations with MCB in the five midlatitude regions (NP, NA, SP, SA, SI) in H2026. As a result, the ERF in the coupled calibration simulations is larger than the -2 W m⁻² target we outline above (Fig. c, f and i). The initial emission amount of 100 Tg yr⁻¹ and the Stage 2 target ERF of -2 W m⁻² are intended to provide a guide for modelers, based on our experience designing these simulations (see Sect. 2.2).

The third stage of the protocol is the G6-1.5K-MCB scenario which follows the structure of the G6-1.5K-SAI scenario , with MCB in the midlatitude boxes as described above rather than SAI:

Background Scenario: SSP2-4.5.

Figure  shows that midlatitude MCB is able to maintain GMST to the target values, within the 0.2 K tolerance in all three models (and for CESM2-WACCM6 in Fig. b). Figure  displays the emission required to maintain GMST to the target temperature in each model (Fig. a–c), the resulting temperature timeseries (Fig. d–f), and the rolling mean GMST anomaly from SSP2-4.5 divided by the emission rate (Fig. g–i). We see an order of magnitude difference in the iSSA emission mass required between the models, with CESM2.1-CAM6 reaching ∼5 Tg yr⁻¹, E3SMv2.0 reaching ∼30 Tg yr⁻¹, and UKESM reaching ∼100 Tg yr⁻¹ by 2085. We note that due to the difference in emitted iSSA size distributions between UKESM1.1 and E3SMv2.0/CESM2.1, the global iSSA number flux requirements are more similar between the models with CESM2.1-CAM6 at ∼8.2×1019 s⁻¹, E3SMv2.0 at ∼4.9×1020 s⁻¹, and UKESM1.1 at ∼5.4×1020 s⁻¹, though there is still over a factor of 6 difference in mass emission rates across the models, even after accounting for the size differences. The CESM2-WACCM6 emission rates are similar to CESM2.1-CAM6 up to 2070 (Fig. a). Owing to the idealized and differing iSSA emission distributions in the models and the difficulty that global models have simulating cloud responses to aerosols, these simulations are not suitable for estimating the iSSA mass or number emission required for actual deployment but are instead aimed at identifying and understanding the sources of inter-model differences in cloud and climate response to MCB.

For CESM2.1-CAM6, E3SMv2.0, and CESM2-WACCM6, the emission rates estimated by the algorithmic controller (blue lines) are lower than the estimate derived using a simple linear assumption (green lines in Figs. a, b and a). This implies the MCB impact on GMST is greater than expected later in the century (Fig. d and e). Correspondingly, we see that the feedforward response results in excess cooling after 2050 in CESM2.1-CAM6/WACCM6 and after 2060 in E3SMv2.0 (Fig. d and e). Similarly, for UKESM1.1 the feedforward emission estimate results in excess cooling after 2060, so in the final manually adjusted simulation the emissions are held at 100 Tg yr⁻¹ to meet the GMST target (Fig. c). Thus, across all three models there is increasing cooling efficiency from midlatitude MCB after the mid 21^st century (Fig. g–i). Considering the non-linearity in the Twomey effect means brightening efficiency will decrease as iSSA emissions increase, this suggests other nonlinearities drive this change in efficiency. Decreases in other aerosol emissions and increasing liquid water content due to precipitation suppression in the models may increase Twomey effect brightening, though in reality the magnitude and even sign of the latter effect is uncertain due to buffering by other processes such as enhanced entrainment . The MCB forcing and climate response have complex and uncertain interactions with changing temperatures, cloud fields, sea ice, ocean circulation etc. as GHG forcing increases . These results also differ from those of , who performed model iSSA experiments seeding relatively pristine stratocumulus cloud regions of the East Pacific and found a decreasing cooling per unit mass injection.

This is similar to findings from early SAI controller simulations , where it is attributed to decadal and longer timescale climate responses to SAI not captured by the short 10-year system identification simulations used in that study. However, in the MCB simulations shown here, the feedforward estimate is computed using the last 30 years of a 50-year simulation, meaning it is less likely that the overestimate in the iSSA in the feedforward is due to the climate not being sufficiently equilibrated in the system identification simulations. However, we note that there are slower multi-decadal to centennial timescale responses that will not be captured without even longer simulations . This suggests that non-linearities other than the timescale of response may be influencing the midlatitude MCB simulations. In contrast, more recent SAI controller simulations ARISE-SAI – find that the linear assumption underestimates the emissions required throughout the 21st century. Recent MCB controller work, which modified the areal extent of CDNC perturbations , found that the linear assumption failed early in the simulation but was a reasonable estimate in the late 21st century. These diverging behaviours demonstrate the need for algorithmic controllers or manual re-adjustments, as it can be difficult to anticipate the direction and magnitude of non-linearities in the response prior to conducting the simulations.

The MCB cooling results in global mean precipitation reductions which exceed the increase from GHG, resulting in a net decrease across all three models (Fig. a–c, CESM2-WACCM6 in Fig. c). This corresponds to similar findings for anthropogenic sulphate aerosols and SAI, and is related to the weaker global mean precipitation increase due to longwave forcing from GHG versus precipitation decrease due to shortwave forcing from aerosols . The drying is largest in UKESM1, which has an anomaly of -0.05 mm d⁻¹ relative to the reference period while CESM2.1-CAM6 and E3SMv2.0 have anomalies of -0.03 and -0.02 mm d⁻¹ respectively. This may be because UKESM1.1 cools slightly below the target later in the century, resulting in a greater net global mean precipitation change.

Despite the net decrease in global precipitation, CESM2.1-CAM6, CESM2-WACCM6, and E3SMv2.0 show no change in land mean precipitation minus evaporation (P-E) (Fig. d and e CESM2-WACCM6 in Fig. d) for G6-1.5K-MCB relative to the reference period. UKESM1.1 shows a modest decrease in land P-E, though it is a smaller decrease than the global mean (Fig. f). Previous analysis of ocean albedo change simulations suggest that shortwave cooling over oceans results in differences in ocean versus land energy budgets that may induce circulation changes which shift precipitation from oceans onto land . This appears to be the case for our MCB simulations as well, and the global mean decrease is a combination of this neutral land precipitation with a larger ocean precipitation decrease. This may be a point of divergence from SAI, which causes forcing over both land and ocean. When we compute the land mean precipitation alone, we see that simulations in all three models indicate there is little change.

Differences in forcing across hemispheres tend to induce shifts in tropical precipitation towards the hemisphere with greater net radiative flux due to the tendency of the intertropical convergence zone (ITCZ) to follow the energy flux equator, the latitude at which column-integrated energy fluxes go to zero . We compute the ITCZ latitude as the latitude of the precipitation median ϕmedian between ϕ1=20° S and ϕ2=20° N , which is the latitude ϕmedian such that 4∫ϕ1ϕmediancos⁡(ϕ)Pdϕ=∫ϕmedianϕ2cos⁡(ϕ)Pdϕ for precipitation P. In CESM2.1-CAM6, SSP2-4.5 results in a southward shift of the ITCZ (Fig. g) while in E3SMv2.0 there is a northward shift (Fig. h), and in UKESM1.1 the ITCZ remains relatively stationary (Fig. i). The mass-balanced MCB case used in G6-1.5K-MCB has relatively little effect on the ITCZ position, relative to SSP2-4.5 in all three models. However, we do see that the ITCZ is shifted slightly northward in CESM2.1-CAM6 and southward in E3SMv2.0, offsetting the SSP2-4.5 effect. Interestingly, we see that the distribution of MCB emissions changes the ITCZ response substantially. In CESM2.1-CAM6, the ERF-balanced case shows a stronger southward shift of the ITCZ and the feedforward equal-emission simulation shows a northward shift, owing to the stronger forcing in the NH for the ERF-balanced case and in the SH for the equal emission case. This demonstrates the challenge of anticipating the tropical circulation change in the models. Not only is the response sensitive to the MCB strategy, but the models do not agree on the baseline GHG response either.

We evaluate the pattern of climate response by comparing anomalies of the SSP2-4.5 and G6-1.5K-MCB simulations in 2065–2084 (last 20 years of the simulations) relative to the reference period (2020–2039) for each model. We can then compute the net effect of MCB as the difference between the 2065–2084 averages for G6-1.5K-MCB minus SSP2-4.5. Annual mean temperature anomaly patterns are similar across the models for SSP2-4.5 (Fig. a–c), with Arctic and land amplification signals characteristic of the transient climate response to GHG. However, the models differ in the North Atlantic, where CESM2.1-CAM6 sees a notable “warming hole”, UKESM1.1 sees a minimum in the warming, and E3SMv2.0 sees strong warming. This is consistent with a decline in the Atlantic Meridional Overturning Circulation in both CESM2.1-CAM6 and UKESM1, and no decline in E3SMv2.0 . Furthermore, E3SMv2.0 and UKESM1 warm by ≈1.7  K, compared to 1.4 K in CESM2.1-CAM6 between 2020–2039 and 2065–2084. CESM2-WACCM6 shows anomaly patterns that are very similar to CESM2.1-CAM6 and we do not discuss it further here (Fig. ).

The G6-1.5K-MCB simulations show substantially reduced temperature anomalies (Fig. d–f). All three models show small (≈0.1 K) net cooling relative to the reference period, indicating that MCB emissions have successfully brought the global-mean temperature to within 0.2 K of the target. Temperatures are reduced relative to the SSP2-4.5 case in all regions and in all models (Fig. g–i), indicating that midlatitude MCB would not exacerbate warming in any region. There are large regions of the globe wherein the G6-1.5K-MCB temperatures are not significantly different from the reference period, especially over land regions. In CESM2.1-CAM6, 50 % of global and 52 % of land area see insignificant temperature anomalies (Fig. a). In E3SMv2.0, it is 74 % of global and 82 % of land area (Fig. b) and in UKESM1.1 it is 39 % of global and 62 % of land area (Fig. c). However, we do see significant residual temperature changes in many regions. In all three models, we see over cooling in the midlatitude oceans, such as in the North Pacific and SH midlatitudes, corresponding to the regions where the MCB is applied, and under cooling in certain land regions, such as over North America.

Other residual signals are more model dependent. In CESM2.1-CAM6, there is large cooling over the North Atlantic and Europe because midlatitude MCB does not counteract the GHG “warming hole”, but instead induces more cooling (Fig. d). This may be because MCB is not increasing the strength of the Atlantic Meridional Overturning circulation (AMOC) enough to offset the decrease due to GHG (not shown). CESM2.1-CAM6 also shows residual warming over Antarctica and the surrounding ocean and over South America, Southern Africa, South Asia, and Australia. This is partly related to the design of the simulation, as the excessive North Atlantic cooling will lead the controller algorithm to reduce MCB emissions and undercool other regions of the globe. In E3SMv2.0, the anomalies over the North Atlantic are consistent with the reference period and overcooling is instead highest in the South Atlantic and Indian oceans (Fig. e). E3SMv2.0 shows residual warming over central Asia, extending from the Middle East to Western China. In UKESM1, the simulations show excess cooling in the North Atlantic. However, as the UKESM1.1 SSP2-4.5 simulation does not see a notable warming hole, the GHG signal is not exacerbated in this case. There is a substantial residual Arctic and Siberian warming in UKESM1, indicating midlatitude MCB does not fully counteract the Arctic amplification signal in this model. However, it is worth noting that in UKESM1.1 Arctic warming is 4.3±0.2 K in 2065-84 under SSP2-4.5 and 1.4±0.3 K under G6-1.5K-MCB, showing there is substantial Arctic cooling.

The simulations show there are notable inter-model differences in the G6-1.5K-MCB temperature anomalies in part because they are residuals of two fields. The effect of midlatitude MCB alone, calculated as the difference between G6-1.5K-MCB minus SSP2-4.5 (Fig. g–i), has an overall similar pattern across the three models, with global cooling, some Arctic Amplification, and cooling of land relative to oceans. Together this shows a MCB cooling pattern that broadly resembles the opposite of the SSP2-4.5 warming pattern, except in the southern ocean. Nevertheless, it is important to note that the models show qualitatively different regional temperature change patterns under G6-1.5K-MCB (Fig. d–f). This demonstrates the importance of understanding the underlying climate dynamics of the responses to SRM, how they might differ across models, and how representative these models are of the real world.

For annual mean precipitation, all three models show the characteristic GHG-driven precipitation anomaly pattern under SSP2-4.5, with increases over the tropical Pacific and Atlantic and in the mid and high latitudes along with decreases in subtropical regions (wet-get-wetter and dry-get-dryer) (Fig. a–c). The point-wise signal to noise is much smaller for precipitation, so differences between the models are difficult to detect at a statistically significant level. In the G6-1.5K-MCB simulations, there is a net decrease in global mean precipitation across all three models (Fig. d–f). However, we see that there are very few land regions with statistically significant precipitation anomalies. The main notable signal is a positive anomaly in Central Africa for E3SMv2.0. Precipitation decreases mainly occur over oceans, with significant drying anomalies in regions of the Atlantic and Pacific for all models. This means that the precipitation response to MCB is close to opposite that of SSP2-4.5 (Fig. g–i), characterized by decreases over the tropics, particularly the tropical Pacific, and mid/high latitudes along with increase in subtropical regions.

The annual mean anomalies may obscure important seasonal signals. We see that for G6-1.5K-MCB, the residual temperature anomalies differ in DJF versus JJA (Fig. ). In all three models, there is significant residual JJA warming in Northern Eurasia and in CESM2.1-CAM6, the region shows an overcooling signal in DJF. The models also show that the under cooling over North America occurs primarily in boreal summer (JJA). The residual Arctic warming in UKESM1.1 mainly occurs in the wintertime (DJF), corresponding to the season in which Arctic amplification temperature anomalies are greatest. Seasonal precipitation maps show few significant signals over land (Fig. ) except for E3SMv2.0 showing significant precipitation increases in central Africa in DJF and West Africa in JJA.

The simulations proposed here have two primary purposes: first, to elucidate the differences across models in how clouds respond to iSSA emissions and, second, to assess the climate impacts of midlatitude MCB implementation. In addition to the core baseline CMIP7 variables , we request that participating models store variables for assessing aerosol-cloud interactions. These include monthly variables requested by CFMIP, including COSP satellite simulator output and aerosol fields, and AerChemMIP, including sea salt aerosol emissions, deposition, and concentrations and aerosol number concentration full lists in the aerchemmip_2d_monthly, aerchemmip_3d_monthly, CFMIP-monthly, and CFMIP-aero variable groups in the CMIP7 data request v1.2.2,). If possible, performing additional radiation calls to compute clean-sky and clean-clear-sky fluxes following is recommended, as it enables clearer decomposition of the SW forcing into Fari and Faci. Furthermore, we request that modelers document information such as emitted iSSA size distribution and flux specifications and, if possible, number concentrations for sea salt aerosol and other atmospheric aerosol. This information will ensure the community can make clear comparisons between models, considering the large dependence of the MCB forcing on the size distribution of the emitted iSSA. These detailed aerosol and cloud variable outputs are highest priority for Inj-seasalt-midlat-SST, but would be of interest in the coupled G6-1.5K-MCB and Inj-seasalt-midlat simulations as well.

Priority variables for impact analysis are summarized in , most of which are included in the baseline CMIP7 variable lists. GeoMIP SAI data has been requested for marine and land biogeochemistry, fisheries, and agricultural assessment and modeling, so similar output would be recommended for MCB as well. Impact analysis variables are highest priority for the G6-1.5K-MCB simulations, as we anticipate this simulation will be of greatest interest for impact assessments.

Recent studies have significantly advanced the understanding of how MCB is represented in ESMs and the potential climate impacts of MCB H2026. Assessment of the climate response in ESMs to different regional MCB perturbations has indicated that midlatitude MCB intervention can produce a cooling pattern that resembles and opposes the GHG warming pattern (H2026), in contrast to the strongly patterned cooling in response to subtropical or tropical MCB interventions that have been used in previous studies. It may therefore be thought of as a “cooperative” MCB scenario in which MCB is deployed to meet a global target, as it broadly reduces temperatures globally and produces fewer unintended regional climate changes compared to subtropical MCB. To facilitate evaluation of midlatitude MCB, and comparison of MCB versus SAI, we propose here a protocol named G6-1.5K-MCB, that uses midlatitude iSSA emissions, but is in other respects quite similar to G6-1.5K-SAI, which has already been accepted as a GeoMIP scenario.

Owing to the substantial uncertainty in the representation of aerosol-cloud interactions in ESMs and consequently the wide range in ESM radiative responses to MCB, we further define a two-stage set of benchmarking simulations (Inj-seasalt-midlat-SST and Inj-seasalt-midlat) with the aim of establishing the iSSA emission strength information needed to set up the stage 3 G6-1.5K-MCB simulations for each model. Furthermore, we request aerosol and cloud information from the fixed SST simulation to enable clear comparisons of the MCB ERF across models. These simulations use a protocol similar to past GeoMIP MCB simulations, G4cdnc and G4seasalt, which applied an idealized constant MCB perturbation in the tropics between 30° S and 30° N starting in 2015 of an RCP4.5 simulation. We include suggestions for analogous CDNC perturbation simulations, as some ESMs may not represent iSSA-cloud interactions with sufficient fidelity to conduct useful iSSA emission simulations. Though CDNC and iSSA perturbations differ, such as due to the advection of aerosol when using iSSA emissions and the partitioning between Faci and Fari, CESM2.1-CAM6 and E3SMv2.0 simulation results suggest that the climate response will be relatively insensitive to the perturbation method – if similar radiative forcings are used . Nevertheless, modelers should use iSSA perturbations when possible.

Using three ESMs, we have conducted a set of testbed simulations in which MCB is applied in five midlatitude ocean regions (North Atlantic and Pacific; South Atlantic, Pacific, and South Indian oceans). Simulation results in all three indicate that midlatitude MCB can successfully maintain global mean temperatures close to the 2020–2039 reference period climate. In two of the three models (CESM2.1-CAM6 and E3SMv2.0), the majority of the forcing occurs via Faci while UKESM1.1 shows similar contributions from Fari and Faci for a 200 Tg yr⁻¹ emission rate simulation. This is in contrast to previous work that found a predominant role for Fari . Ongoing CMIP6-generation model intercomparison work suggests that the MCB ERF is mainly Faci in other current-generation ESMs as well. The models require substantially different iSSA emission rates in the G6-1.5K-MCB simulations, with a factor of 20 difference in the iSSA mass flux and a factor of 7 difference in number flux between CESM2.1-CAM6 and UKESM1.1 to maintain the temperature target. The mass emission rates used in CESM2.1-CAM6 and E3SMv2.0 simulations are lower than many previous estimates . For the 2075–2084 average, CESM2.1-CAM6 and E3SMv2.0 are lower by 53× and 8.9× respectively compared to the equivalent period in NorESM1 (266 Tg yr⁻¹) which had the lowest emission rate in . Thus, the results presented here indicate that the emissions required to produce significant cooling with MCB may be much lower than indicated in previous studies.

This has several possible explanations. First, the size distributions used here are smaller than many previous studies , except for where iSSA with a dry diameter of 100 nm shows 2.1 W m⁻² with 20 Tg yr⁻¹. Second, we use a lower target than which cooled from SSP5-8.5 to SSP2-4.5, requiring over 3 °C of cooling versus up to 2 °C in this scenario. Third, we use midlatitude emissions, which the three models here suggest produce stronger MCB forcing (H2026). Fourth, there have been updates in the representation of aerosol-cloud interactions, such as prognostic sea salt aerosol interactions with clouds (which was not available in two of the three models in ) and influence of CDNC on autoconversion rates, which is the source of the cloud lifetime effect. We emphasize that considering the idealized iSSA emission distributions used and the uncertainty in Faci across ESMs, these emission rates should not be treated as estimates of the iSSA mass or number emission required in an actual deployment.

In contrast to MCB strategies focused on subtropical regions, which tend to produce regional temperature changes that exacerbate GHG warming , the simulation results indicate midlatitude MCB produces more uniform cooling and does not exacerbate GHG warming in any regions. Midlatitude MCB also produces fewer large precipitation shifts compared to subtropical MCB and the spatial extent of statistically significant annual mean precipitation anomalies is much reduced compared to SSP2-4.5. Studies of the large scale response to other midlatitude perturbations, such as ocean heat flux anomalies and regional sulphate emissions , suggest that the more uniform cooling seen for midlatitude MCB may also occur in other ESMs. Notable inter-model differences in the temperature response pattern can be seen in the Arctic, North Atlantic, and Southern ocean. These appear to be primarily related to differences in the circulation response to MCB, as they largely occur away from the midlatitude forcing regions. Future assessments of potential MCB deployments therefore depend critically on the continued advancement of climate dynamics research. Additional differences may arise from model variations in the susceptibility of NH versus SH regions to MCB, which could give rise to differing inter-hemispheric asymmetries in the forcing and response.

Like G6-1.5K-SAI, the MCB scenario protocol described here attempts to strike a balance between simplicity to enable standardized, inter-comparable simulations and sufficient complexity to be relevant for assessing MCB as a climate intervention strategy. This mid-latitude implementation of MCB is able to simultaneously maintain annual-average land average temperatures, precipitation, and P-E at close to baseline values. In many SAI scenarios explored to date, temperatures, and precipitation cannot simultaneously be returned to baseline values , though a more thorough comparison is required to understand any potential differences between the intervention methods. The scenario choices here do not represent the most optimal or realistic MCB strategy, nor should it be interpreted as being endorsed by the modeling community as a template for real-world deployment. Depending on the desired climate outcomes, further optimization of the MCB emission pattern can likely be achieved. Here, we have used global annual average temperature as a target metric in our controller and feed-forward runs. This results in regions of over- and under-cooling in the scenario simulation with midlatitude MCB and SSP2-4.5. Furthermore, there are significant inter-model differences in the pattern of temperature response to the SSP2-4.5 scenario and midlatitude MCB. Differences in the climate response to MCB in different regions suggest it may be possible to use a more complex set of targets and controllable parameters in a controller simulation that would address the regional inhomogeneity in climate warming to some extent, for example by adjusting iSSA emissions independently in different regions. Additionally, it is important to note that a significant component of the uncertainty in the climate response to MCB stems from uncertainties in the underlying climate dynamics, which also drive significant uncertainties in future climate absent MCB. On the other hand, MCB has the potential to be deployed regionally H2026 or may be deployed in a manner closer to weather modification to reduce local heat impacts, as is proposed in the Great Barrier Reef project , which will have distinct regional and global climate impacts. Both more refined controller simulations and regional MCB scenarios are important areas of research and a multi-model G6-1.5K-MCB ensemble could serve as a common reference point for other, more exploratory analyses. As the ESM community looks towards the seventh iteration of CMIP, there will be substantial updates to ESMs and changes to the emission trajectories used by the models. G6-1.5K-MCB will provide a valuable bridge between generations; providing knowledge and expertise that can be applied to define the next generation G7-1.5K-MCB scenario.