Offsetting the global warming caused by anthropogenic increases in atmospheric greenhouse gases by deliberate injection of aerosols into the stratosphere is the most studied of solar radiation management geoengineering schemes. The long-term success or failure of such schemes in achieving their stated goals is assessed by comparing simulated geoengineered temperature, precipitation and tropical cyclones metrics to equivalent fields in the simulated targeted climate simulations. Results using available data sets from three single model stabilized climate target experiments and three multi-model climate change reduction experiments are presented and compared against a measure of internal variability. While all but one experimental scheme is successful in achieving their targeted global mean annual surface temperature, their success at regional scales varies significantly and is often larger than the internal variability metric used here.
The human interference in the climate system due to a steady increase in heat trapping greenhouse gases has already altered the frequency and severity of extreme weather events and their impacts (Seneviratne et al., 2021). These changes are principally due to our consumption of fossil fuels for energy and it is not obvious that society has the political will to stabilize the climate anytime soon. The lower target of 1.5°C above preindustrial temperatures from the historic Paris Agreement is currently in serious jeopardy of exceedance and the upper target of 2.0°C above preindustrial temperatures will likely also be exceeded if emissions of greenhouse gases are not rapidly reduced to zero. The highly controversial geoengineering tactic of managing incoming surface solar radiation has long been offered to manage global temperatures if greenhouse gas (GHG) emissions cannot be eliminated. The concept is highly controversial from political, ethical and technical perspectives. Nonetheless, the US Congress has mandated further research in the topic (OSTP, 2023) following the recommendation of the US National Academy of Sciences, Engineering, and Medicine (NASEM, 2021). A variety of tactics have been proposed to reduce surface solar radiation and a series of numerical experiments using current climate models have been performed to investigate how such geoengineering might be implemented.
The climate is naturally a highly variable system across a wide range of temporal and spatial scales. The attribution of externally forced changes to the climate system must be interpreted in this context. By design, solar radiation management geoengineering schemes attempt to offset the warming of increased anthropogenic atmospheric GHG concentrations by artificially decreasing incoming surface solar radiation. The most commonly proposed scheme to accomplish such an increase in planetary albedo is the introduction of reflective sulfate aerosol precursors (i.e., sulfur dioxide, SO2) into the stratosphere although other concepts have been considered. In a perfectly geoengineered climate, the two external forcing agents would exactly offset each other. In practice, there are spatial and temporal differences between the responses of the two external forcing agents and this is reflected in simulations of geoengineered and target climates. An “ideal” geoengineered climate would minimize these differences in populated regions. What actually constitutes minimization is a value judgement. In this paper, differences between the geoengineered and target climates are evaluated in the context of natural variability. Other value judgements might consider the detectability of these differences (MacMartin et al., 2019; Barnes et al., 2022) or the effect of these differences on the climate impacts of human or ecological systems.
However, unlike attributing the relatively large changes in the climate observed since the pre-industrial period, these differences in response to forcing tend to be smaller presenting challenges in any attribution to a deliberate human intervention. Indeed, while the time to determining that a geoengineering scheme achieved its targeted global mean temperature can be as short as a decade or two, attributing success of achieving targeted regional temperatures would be much longer due to the higher internal climate variability at regional scales (MacMartin et al., 2019). Attributing success of achieving targeted mean precipitation rates could require over a century at regional scales because of yet larger relative internal variability (MacMartin et al., 2019).
In this paper, the success of the most recent geoengineering simulations is evaluated as to how well they achieve the intended goals of stabilizing mean and extreme metrics of surface air temperature, precipitation and tropical cyclone potential intensity in the context of natural internal variability. Extreme daily temperature and precipitation are defined here as 20-year return values. The six different experiments listed in Table 1 are designed to stabilize the climate to varying degrees under a range of GHG emission scenarios. As their targeted climate states are different, each geoengineering experiment must be evaluated separately.
Because the emissions scenarios, the targeted climates and the geoengineering analysis periods all differ in these experiments, a somewhat confusing intercomparison situation is presented. Table 1 compares their experimental protocols. The first three experiments in Table 1 are large ensemble but single climate model simulations. The large amount of available model output permits highly statistically significant statements about the results of that particular model. However, climate models differ greatly in their response to forcing agents and this source of structural uncertainty can be significant. The second group of three experiments are from the Geoengineering Model Intercomparison Project (GeoMIP) experiments, part of the 6th iteration of the internationally coordinated Coupled Model Intercomparison Project (CMIP6). While these experiments permit some assessment of model structural uncertainty, the ensemble sizes from some of the models are quite limited thereby introducing substantial statistical uncertainty.
In Table 1, the emissions scenarios of each experiment and the targets of stabilized global mean temperatures (labeled as “Target temperature”) are expressed as degrees Celsius (°C) above pre-industrial temperatures. As per the IPCC Working Group 1 definitions, the pre-industrial period is defined as an average over 1850 to 1900. The “Targeted periods” in Table 1 indicate the baseline period from the “Targeted emissions” simulations to compare the geoengineering simulation to. These baselines are obtained from the standard CMIP6 simulations (Eyring et al., 2016; O’Neill et al., 2016)
In the single model experiments, the target is to maintain global mean temperature at a constant 1.5°C above its preindustrial value. Hence, the geoengineered analysis period can be the entire duration of the geoengineering simulation. The GeoMIP experiments, on the other hand, aim to maintain global mean temperature from a very high emissions scenario (SSP585) to that expected from a lower emissions scenario (SSP245). Hence, temperatures continue to increase with time, albeit more slowly. Consistent with other GeoMIP analyses, we analyze the simulations toward their end in the late 21st century. Table S1 (see Supplementary Materials) lists the number of climate models used in each experiment and the number of realizations and years with appropriate daily model output for both the baseline and geoengineering simulations. Supplemental Table S2 lists individual model output details for the GeoMIP experiments. To increase data length sizes and increase statistical significance in both sets of experiments, each individual model’s realizations were concatenated. For the GeoMIP experiments, a regridded multi-model average was performed as a last step. Throughout this study, analysis of the temperature and precipitation metrics is confined to global land regions (excluding Antarctica), while analysis of the tropical cyclone metrics is confined to the tropical ocean (40°S–40°N).
2 A summary of the geoengineering schemes
ARISE-SAI-1.5: The ARISE-SAI-1.5 experiment begins in 2035 to inject stratospheric aerosols at an altitude of ∼21.5 km at four latitudes (30°S, 15°S, 15°N, 30°N) to maintain a targeted global mean surface air temperature near 1.5°C above its pre-industrial value (Richter et al., 2022). GHG emissions are as specified under the SSP2-4.5 protocols. SSP2-4.5 is considered a “medium emissions” future scenario (Riahi et al., 2017) with an anthropogenic forcing of ~4.5 W/m2 at the end of the 21st century (O’Neill et al., 2016). In the absence of geoengineering, global mean surface air temperature is projected to be 2.7C (likely range of 2.1–3.5°C) above preindustrial values at the end of the 21st century (IPCC AR6 WG1 SPM, 2021). Because SSP2-4.5 emissions continue to increase, the amount of injected aerosols must also continue to increase and is controlled by a feedback mechanism considering the global mean temperature, the north–south temperature gradient and the Equator-to-pole temperature gradient (Kravitz et al., 2017). This experiment uses a single model, the Community Earth System Model version 2 with the Whole Atmosphere Community Climate Model version 6 (CESM2-WACCM6) run with 10 realizations of the experiment. CESM2-WACCM6 is a so-called “high top” coupled model with a top of atmosphere at ∼140 km, well into the middle of the thermosphere (Gettelman et al., 2019). It also has a sophisticated stratospheric aerosol model (Tabazadeh et al., 1994). Both of these model properties are relevant to the physics of stratospheric aerosol injection and are more sophisticated than most CMIP6 models. As the target is to stabilize the climate at a fixed global warming level, quasi-stationarity in this analysis is assumed over the entire geoengineering simulation period and over the target simulation period (2020–2039). Hence, assessments of mean quantities are made by ensemble averaging over the entire periods. To calculate extreme values, annual and seasonal maxima are concatenated from each realization over the entire simulation periods to increase the data set size and reduce the fitting uncertainty of the Generalized Extreme Value (GEV) distribution coefficients (Wehner et al., 2024). A quasi-stationarity assumption permits the usage of the L-Moments method, a more stable method than Maximum Likelihood Estimators, to fit the GEV coefficients (Hosking et al., 1985; Hosking and Wallis, 1997). Twenty-year return values are simply then straightforwardly calculated from the fitted GEV parameters (Coles, 2001).
Feedback: The “Feedback” experiment is an earlier controlled stratospheric aerosol injection scheme and is part of the Geoengineering Large Ensemble (GLENS). While using the same control mechanism as ARISE-SAI-1.5, it uses a similar but earlier version of the climate model, CESM1-WACCM6 (Tilmes et al., 2017, 2018) with 21 realizations of the experiment. Aerosol injection was maintained at the same latitude points but at higher altitudes of 5 km above the tropopause or about 23–25 km. The targeted global mean temperature is also 1.5°C above preindustrial temperatures but the GHG emissions scenario is SSP 5-8.5, a high scenario with extensive development of fossil fuels (Riahi et al., 2017). With an anthropogenic forcing of 8.5 W/m2 at the end of the 21st century (O’Neill et al., 2016), in the absence of geoengineering, global mean surface air temperatures are projected to be 4.4°C above preindustrial temperature with a likely range of 3.3°C to 5.7°C (IPCC AR6 WG1 SPM, 2021). Hence, the quantity of aerosols injected into the stratosphere is substantially higher than ARISE-SAI-1.5 causing the targeted climate more difficult to achieve. The methods used here to assess the ability to simulate the targeted climate are the same as for the ARISE-SAI-1.5 experiment.
Feedback_low: The “Feedback_low” experiment is essentially the same as the “Feedback” experiment except that the aerosol injection is at a lower altitude. While the experiment used the new model, CESM2-WACCM6, there are only 3 realizations of this experiment currently available.
3 Geoengineering Model Intercomparison Project Phase 6 (GeoMIP6)
The GeoMIP6 is a subproject of the 6th version of Coupled Model Intercomparison Project (Eyring et al., 2016), an internationally coordinated suite of multi-model simulations of a variety of geoengineering schemes (Kravitz et al., 2015).
G6Sulfur: As a coordinated multi-model project, the protocols for G6Sulfur experiment are more loosely defined than for the three single model experiments described above. Designed to be paired with the G6Solar experiment described below, aerosols are injected between 10°S and 10°N at altitudes between 18 and 20 km for climate models with interactive aerosol submodels. For those models without interactive aerosol components, an offline sulfate distribution is imposed (Kravitz et al., 2015; Visioni et al., 2021). In these experiments, the target is not to stabilize the climate to a fixed global mean surface air temperature, but to reduce the effect of the high emissions SSP 5-8.5 scenario to that of the medium emissions SSP 2-4.5 scenario. As in previous studies, here the geoengineering simulations are compared to the SSP 2-4.5 simulations at the end of the 21st century (Visioni et al., 2021; Tilmes et al., 2022). Results presented here for mean quantities are the multi-model means of each participating model’s ensemble mean. Return values are calculated for each model prior to the multi-model averaging by concatenating each realization over the 20-year period 2080–2099. Quasi-stationarity is assumed and the L-Moments methods is used to fit the GEV coefficients (Hosking and Wallis, 1997).
G6Solar: The protocols for the multi-model G6Solar experiment are designed to replicate the geoengineering forcings of the G6Sulfur experiment more simply by only altering the top of atmosphere incoming solar radiation (Kravitz et al., 2015; Visioni et al., 2021; Tilmes et al., 2022). While a much simpler simulation to perform, G6Solar simulations do not replicate the details of G6Sulfur, particularly in high latitude surface air temperatures and stratospheric temperatures (Tilmes et al., 2022). The methods used here to assess the ability to simulate the targeted climate are the same as for the G6sulfur experiment.
G7Cirrus: The G7Cirrus experiment considers a different approach to reduce incoming surface radiation. Not technically a solar radiation management scheme, G7Cirrus involves seeding cirrus clouds to increase ice crystal fall speed thus thinning the cirrus layer allowing more longwave radiation to escape (Kravitz et al., 2015). There are considerable uncertainties in this geoengineering scheme due to models’ questionable skill in reproducing cloud properties and their radiation feedbacks (Tselioudis et al., 2021; Zhang et al., 2023; Chao et al., 2024). The G7Cirrus protocols involve a constant amount of cloud seeding under the SSP 5-8.5 emissions scenario to reduce average global mean temperature in the decade 2020–2029 to that of the historical decade 1970–1979 (Kravitz et al., 2015). In practice, G7Cirrus has seen much less participation by both the modeling groups and analysts than G6sulfur and G6Solar.
Details of the geoengineering and target simulations for each experiment are listed in Table 1 and supplemental Tables S1 and S2. Data acquisition sources are listed below in the data section. For consistency in comparison of assessments of the mean and extreme, only those GeoMIP model simulations that provided the requisite daily data to calculate annual and seasonal maxima were retained for analysis of temperature and precipitation.
As targeted climates vary greatly between these experiments, the amount of intended geoengineered global cooling induced by geoengineering not only varies but increases as GHG increases, especially in the ARISE-SAI-1.5 and two Feedback experiments. The bottom row of Table 1 shows the global mean temperature reduction required by geoengineering to offset GHG warming averaged over the geoengineered periods for each experiment. Geoengineered global cooling required by the individual G6 models are shown in the bottom row of supplemental Table S2.
4 Natural variability: temperature
To assess how well the geoengineering experiments achieve their intended goals, it is important to place them in the context of natural variability. One method to do this is to compare the difference of the geoengineered climate and the targeted climate to some metric of natural variability (Morrison et al., 2024). If the difference between the geoengineered and targeted climates are less than these natural variability metrics, the external anthropogenic forcings of GHG and injected aerosol approximately cancel each other and the geoengineering could be assessed as “successful”. Natural variability metrics are highly dependent on the time scales of interest and methodology. Hence, the choice(s) of natural variability metrics are arbitrary and reflect a value judgement. Note that while the targeted climates vary across the geoengineering schemes considered here, the same measures of internal variability obtained from the Climate of the 20th Century (C20C+) project are applied them all. It is also important to note that measures from other models or metrics of natural variability would be different than these results. Smaller magnitude natural variability metrics would provide more stringent assessments of geoengineering success.
In this study, the targeted mean temperature and precipitation are ten to 20-year averages while the targeted temperature and precipitation extremes are 20-year return values over those periods. While there are many model data sets to quantify the long-term variability of 20-year mean values, the large ensemble of the Climate of the 20th Century (C20C+) project offers a unique data set to estimate natural variability of 20-year return values (Stone et al., 2019). This atmosphere-only experiment permits detailed analysis of the interannual variability in the probability of extreme events as there are many simulations performed under identical sea surface temperature conditions. There are 100 realizations of the lower resolution (1°) version of the Community Atmospheric Model version 5 (CAM5.1) over the period 1996 to 2017. This enables the 20-year return values of the annual maximum of the daily maximum temperature (TXx) to be calculated individually for each of these years (Wehner et al., 2018) by concatenating realizations for each single year. As these data sets for an individual year are considered to be a stationary, the L-moments method is appropriate to fit the parameters of the GEV distribution and estimate the 20-year return value for any single year (Hosking and Wallis, 1997; Wehner et al., 2024). Thus, the GEV parameters for each year are estimated independently of any other year permitting a straightforward estimate of interannual variability. The top panel of Fig. 1 shows the interannual standard deviation of the linearly detrended estimate of the 20-year return values of the annual maximum of the daily maximum temperature over land regions from the CAM5 C20C simulations over the period 1998–2017. This field will be used to assess the significance of the differences between geoengineered and target extreme temperatures.
As most high temperature extremes happen in the summer months, to facilitate comparison this study also examines the differences between geoengineered and target mean temperatures during the boreal summer and winter separately. To construct a natural variability metric, the 20-year mean is first calculated for each CAM5 C20C+ realization, followed by calculation of ensemble mean and variance across realizations. The middle panel of Fig. 1 shows the standard deviation across the 100 realizations of the 20-year mean (1998–2017) boreal summer surface air temperature from these simulations over land regions. Because of the reduction of snow cover and the resulting surface albedo increase, projected temperature increases in winter are generally larger than in summer in warmer futures. As the mechanisms of seasonal temperature changes in warmer futures are different, little relationship between winter and summer mean temperature changes should be expected. However, in a geoengineered climate, it is not as clear how the effect of engineering the two opposite seasons might be related. The lower panel of Fig. 1 shows a similar variability metric for boreal winter mean surface air temperature. The winter mean temperature variability is considerably larger than its summer equivalent but generally smaller than 1°C. Similar metrics for seasonal mean and extreme precipitation are presented below.
4.1 Temperature results
ARISE-SAI-1.5:Figure 2 shows the difference between the geoengineered and target temperatures for the ARISE-SAI-1.5 experiment. Mean boreal summer temperatures (middle panel) averaged over all land areas are 0.09°C warmer than the target with a root mean square error (RMSE) of 0.2°C. About 38% of the summer land area is warmer than the target and outside the range of the natural variability metric and about 10% cooler of the land area is than the target and outside the range of the natural variability metric of Fig. 1 (middle panel). However, the largest geoengineered boreal mean summer temperature differences are within 1°C of the target. Differences in 20-year return values (Fig. 2, upper panel) are more mixed in sign, warmer at high northern latitudes and cooler elsewhere, with an RMSE of 0.62°C. Despite the geographical variance in sign, the global mean land extreme temperature is 0.26°C warmer in the ARISE-SAI-1.5 experiment than its target. About 37% of the land area is warmer than the target and outside the range of the natural variability extreme metric (Fig. 1, upper panel) during summer but most of the cooler extreme temperature regions are within that metric. Mean boreal winter temperatures (lower panel) averaged over all land areas are 0.08°C warmer than the target with a root mean square error (RMSE) of 0.3°C. About 30% of the winter land area is warmer than the target and outside the range of the natural variability metric and about 11% of the land area is cooler than the target and outside the range of the natural variability metric of Fig. 1 (lower panel). Notably, most of the northern hemisphere winter warming is in current snow covered regions.
Mean and extreme temperatures in the ARISE-SAI-1.5 experiment are much closer to the intended targets than the other experiments considered here. However, this is a consequence of the fact that the greenhouse gas forcing in the ARISE-SAI-1.5 target is substantially less than the other experiments. As shown below, the ARISE-SAI-1.5 aerosol injection scheme is the least efficient of the geoengineering schemes considered here. Due to space limitations, the equivalent figures showing model performance for the other experiments can be found in the Supplementary Material. However, the temperature summary performance statistics are discussed individually for the remaining experiments and shown in Table 2.
Feedback: Figure S1 (see Supplementary Materials) shows the difference between the geoengineered and target temperatures for the Feedback experiment. This experiment is over geoengineered as global mean boreal summer land temperatures (middle panel) are 0.22°C cooler than the target with an RMSE of 0.55°C. About 18% of the land area is warmer than the target and outside the range of the natural variability metric, all at the high northern latitudes. About 53% of the land area is cooler than the target and outside the range of the natural variability metric. Twenty-year return values differences (Fig. S1, upper panel) that are cooler than the target and outside the range of the natural variability metric cover 31% of the land but are generally within 1.5°C of the target. Almost no extreme temperatures are warm enough to be outside the high end of the range of the natural variability metric. Mean boreal winter temperatures (Fig. S1, lower panel) averaged over all land areas are 0.27°C cooler than the target with a root mean square error (RMSE) of 0.8°C. About 18% of the winter land area is warmer than the target and outside the range of the natural variability metric and about 49% of the land area is cooler than the target and outside the range of the natural variability metric of Fig. 1 (lower panel). There is a strong contrast between the western and eastern portions of the Eurasian land mass in the winter temperature response.
Feedback_low: Figure S2 shows the difference between the geoengineered and target temperatures for the Feedback_low experiment. Patterns and magnitudes of these differences are very similar to the Feedback experiment. Comparison of Figs. S1 and S2 would suggest a slight response to a change in aerosol injection altitude, but these are mostly in places where the differences in geoengineered and target temperatures are less than the natural variability metrics. When averaged over global land areas, the temperature summary metrics are nearly the same in the two Feedback experiments as shown in Table 2.
G6solar: Figure S3 shows the multi-model ensemble mean differences between the geoengineered and target temperatures for the G6solar experiment at the end of the 21st century. The patterns of extreme and mean differences are similar with significantly larger geoengineered temperature than the targets at high northern latitudes. Mean boreal summer temperatures (Fig. S3, middle panel) averaged over all land areas are 0.15°C warmer than the target with an RMSE of 0.43°C (Table 2). About 48% of the land area is warmer than the target and outside the range of the natural variability metric and about 18% of the land area is cooler than the target and outside the range of the natural variability metric during summer. These summary statistics are similar for the 20-year temperature return values at 39% and 20%, respectively. However, these exceedances are generally less than 0.75°C for both the boreal summer mean and extreme temperatures. Mean boreal winter temperatures (Fig. S3, lower panel) averaged over all land areas are 0.28°C warmer than the target with a root mean square error (RMSE) of 0.6°C. About 68% of the winter land area is warmer than the target and outside the range of the natural variability metric and about 11% of the land area is cooler than the target and outside the range of the natural variability metric of Fig. 1 (lower panel). Warmer geoengineered temperatures in both seasons are mostly less than 1°C above the target.
G6sulfur: Figure S4 shows the multi-model ensemble mean differences between the geoengineered and target temperatures for the G6sulfur experiment at the end of the 21st century. Again, for this complementary experiment to G6solar, the patterns of extreme and mean differences are similar to each other but the geographical extent of higher geoengineered temperatures is considerably larger. Mean boreal summer temperatures (Fig. S4, middle panel) averaged over all land areas are 0.29°C warmer than the target with an RMSE of 0.7°C (Table 2). About 56% of the land area is warmer than the target and outside the range of the natural variability metric and about 21% of the land area is cooler than the target and outside the range of the natural variability metric during summer. Warmer geoengineered summer temperatures are mostly less than 1°C above the target. Cooler geoengineered summer temperatures are mostly in the tropics and Australia. These summary statistics for the 20-year temperature return values (Fig. S4, upper panel) are 22% and 29%, respectively. Mean boreal winter temperatures (Fig. S4, lower panel) averaged over all land areas are 0.3°C warmer than the target with a root mean square error (RMSE) of 1.1°C. About 50% of the winter land area is warmer than the target and outside the range of the natural variability metric and about 33% of the land area is cooler than the target and outside the range of the natural variability metric of Fig. 1 (lower panel). Warmer geoengineered winter temperatures often exceed 1°C above the target.
G7cirrus: Figure S5 shows the multi-model ensemble mean differences between the geoengineered and target boreal summer temperatures for the G7cirrus experiment over the period 2020–2029. Noting that only two modeling groups participated in this experiment, the experiment appears to fail as geoengineered temperatures are 3–5°C above the target. In fact, the effect of cirrus cloud modification appears to have little effect as the geoengineered temperature are not very different from the non-geoengineered temperatures at the same GHG emissions. Note that the temperature scale of Fig. S5 is five times that of the other temperature difference figures. Additional fields from the G7cirrus experiment are not further analyzed
5 Natural variability: precipitation
Similar to the mean and extreme temperatures metrics in Fig. 1, the C20C+ simulations can be used to construct natural variability metrics for precipitation. Because storm types vary significantly between seasons, it is useful to assess model precipitation performance seasonally. Part of the motivation behind a seasonal analysis is that the models’ abilities to simulate different storm types varies greatly, especially when cumulus convection is involved. In this paper, the maps of precipitation metrics and performance are shown by temporal seasons (DJF, MAM, JJA, SON) to avoid plotting discontinuities at the equator. However, the summary statistics are calculated on true boreal seasons. Precipitation analyses are performed here using percent differences. In the maps, this can highlight arid regions as small absolute differences can be large percent differences.
The top panels of Fig. 3 show the interannual standard deviation of the linearly detrended estimate of the 20-year return values of the seasonal maxima of the daily precipitation over land regions from the CAM5 C20C+ simulations. This field will be used to assess the significance of the differences between geoengineered and target extreme precipitation. Similar to the mean temperature metric, the 20-year seasonal mean precipitation is first calculated for each CAM5 C20C+ realization, followed by calculation of ensemble mean and variance across realizations. The bottom panels of Fig. 3 show the standard deviation across the 100 realizations of the 20-year seasonal mean (1998–2017) precipitation from these simulations over land regions.
5.1 Precipitation results
ARISE-SAI-1.5:Figure 4 shows the difference between the geoengineered and target mean and extreme seasonal precipitation for the ARISE-SAI-1.5 experiment. Global mean total precipitation is constrained by the global energy budget and the latent heat released upon condensation (Allen and Ingram, 2002). As a result, geoengineered global mean total precipitation is a complicated function of the alteration of long and shortwave radiative fluxes by the increasing GHG and aerosol concentrations (Ricke et al., 2023) and the differences between geoengineered and target climates would not be zero even if differences in global mean surface temperatures were. In the ARISE-SAI-1.5 experiment averaged over the entire period of 2036–2069, precipitation is about 2% higher than the target average over the entire globe. There is a slight negative trend in global mean precipitation (Richter et al., 2022) and by the end of the simulation it is lower than the targeted period. However, averaged over global land areas, the mean seasonal precipitation in the ARISE-SAI-1.5 simulation is less than 1% drier than the target simulation. But as the lower panels of Fig. 4 reveals, there are significant and potentially profound regional differences in seasonal mean precipitation. Of particular note is the substantial drying of the Amazon and southern Africa which could have impacts on regional vegetation and the global carbon budget. Drying in Australia during boreal summer and fall could lead to increased bushfire risk. Western Europe and central Eurasia would appear to be drier in most seasons. However, when considered in the context of the natural variability metrics in the lower panels of Fig. 3, only 5% of the global land area is drier and outside the natural variability seasonal mean metrics in boreal winter. Boreal summer and fall are drier than winter and fall with as much as 13% of the global land area drier and outside the natural variability metrics.
Differences in geoengineered and target seasonal extreme precipitation (Fig. 4, upper panel) are spatially noisier than the seasonal mean precipitation differences. The root mean square differences are also significantly larger for extreme precipitation than for mean precipitation. However, since the CAM5.1 C20C+ metric of internal variability is large, the fraction of land regions with either larger or smaller extreme precipitation outside the range of internal variability is smaller than for mean precipitation in all seasons. Inspection of Table 3 reveals that this behavior holds for all of the geoengineering experiments.
Like for temperature, mean and extreme precipitation in the ARISE-SAI-1.5 experiment are much closer to the intended targets than the other experiments considered here. Again, this is a consequence of a lower GHG forcing in the targeted ARISE-SAI-1.5 climate.
Feedback: While the global land averaged seasonal mean precipitation is about the same for the Feedback and ARISE-SAI-1.5 experiments, the cooler Feedback experiment (Fig. S6, lower panel) shows larger regional differences from target than the ARISE-SAI-1.5 experiment (Table 3). Drying in parts of the Amazon and southern Africa occur in the same season in both experiments that are larger than the natural variability metrics. Australia is generally wetter although western Australia is drier in the winter consistent with the ARISE-SAI-1.5 experiment. However, the Feedback experiment also shows significant drying during the Indian monsoon season not seen in the ARISE-SAI-1.5 experiment. In Eurasia, a pattern of mid-latitude drying is persistent throughout all seasons and is often larger than the natural variability metric. Changes in North America are larger in the Feedback experiment, especially a springtime drying in the eastern and northern regions of the continent. The larger changes are reflected in the seasonal mean summary statistics of Table 3 with up to 21% of the land wetter and 24% drier than the target climate and ouside the range of the natural variability metrics.
Feedback Low: Figure S7 reveals that the pattern of percent differences in seasonal mean precipitation is not affected much by the altitude of aerosol injection. However, the magnitudes of both wetter and drier differences are considerably larger. Each of the seasonal mean summary statistics is larger with up to 26% of the land wetter and 31% drier than the target and outside the range the natural variability metrics. There is little significant difference in extreme precipitation between the two Feedback experiments.
G6solar: Despite very small global mean surface temperature differences between the G6solar experiment and its target, global average land average seasonal precipitation is up to 5% less. Large regions of drying are thus seen in the lower panels of Fig. S8. Similar to the Feedback experiments, this experiment dries the Amazon in 3 of the 4 seasons compared to its target. Australia is significantly drier in summer and fall but mean Indian monsoon rainfall is largely unaffected. Drying of the Northern Hemisphere is pervasive, except in the winter. The seasonal fractions of global land area drier than the target and outside the range of the natural variability metrics varies from 21% to 39%.
G6sulfur: The lower panels of Fig. S9 shows that the regional differences between the G6sulfur experiment and its target are both significantly wetter and drier than the companion G6solar experiment. As little as 25% of the geoengineered global land surface precipitation is within the natural variability limits of the targeted values. The Amazon is drier in every season as is much of southern Africa. However, the Indian Monsoon is wetter. Australia is also wetter in every season except for its western coast in summer. In the Northern hemisphere, the drying patterns of the companion G6solar experiment are substantially exaggerated in all seasons. Nearly the entire geoengineered Northern hemisphere land area is drier than the target in the summer. In winter and spring, midlatitude Eurasia is drier but the high latitudes are wetter than the target. Most of these differences are outside the range of the natural variability metric. Hence, the seasonal fractions of global land area drier than the target and outside the range of the natural variability metrics varies from 28% to 46%.
6 Tropical cyclone indices
The horizontal resolution of the models used on the current geoengineering simulations are not fine enough to realistically represent tropical cyclones. However, bulk indices of maximum potential wind speeds and of cyclogenesis can be calculated from the archived monthly output data. Emanuel’s Maximum Potential Intensity (MPI) is a well-known measure of the peak wind speeds possible in a perfect tropical cyclone (Emanuel, 1987). Assuming a Carnot cycle of transport of energy from the ocean surface to the top of the storm, MPI is a function of surface temperature and temperatures and humidity aloft. MPI has been demonstrated to be a credible upper bound on observed tropical cyclone wind speeds (Emanuel, 2000).
To assess how well the geoengineering simulations control the risk of intense tropical cyclones, MPI is first calculated from the monthly output, then the maximum monthly value is extracted for each year and compared between target and geoengineered simulations. Analysis is confined to ocean areas at latitudes between 40°S and 40°N.
A metric of natural variability for MPI can be obtained from the C20C+ CAM5.1 simulations as was done above for seasonal mean temperature and precipitation. The top panel of Fig. 5 shows the ensemble standard deviation of the 20-year mean (1998–2017) of the annual maximum MPI. The four lower panels of Fig. 5 show the MPI difference between the geoengineered simulations and their target simulations. Note that the color scale for the difference maps is five times larger than the natural variability metric revealing that these differences are larger over most of the tropics than the natural variability metric in all the geoengineering experiments. There is also little consistency between the geoengineering experiments. With slight decreases in intensity, the ARISE-SAI-1.5 experiment is the closest to its target in regions where tropical cyclones currently occur. The Feedback experiment shows somewhat larger decreases in MPI. The requisite data to calculate MPI from the Feedback_low experiment was not provided. The G6solar and G6sulfur experiments are completely different from each other with small changes in the former and large decreases in the latter.
The defining equation for MPI is
where C is a constant ratio of surface exchange coefficients of enthalpy and momentum, Δk the difference between saturation enthalpy of air at the surface and of boundary layer, Ts is the surface temperature and To is the temperature at the outflow (i.e., top) of the tropical storm. For intense tropical cyclones, the top of the storm can be near the tropopause. In the real world, GHG forcing has warmed the lower troposphere and cooled the stratosphere, thereby elevating tropopause height due to both thermal expansion at lower altitudes and contraction at high altitudes (Santer et al., 2003). Thus, attributable increases in current values of MPI (Gilford et al., 2024; Wehner and Kossin, 2024) are due to both surface warming and a higher and cooler tropopause. Changes in the vertical structure of tropical atmospheric temperature are more complex in the geoengineering experiments than observed in the real world.
All of the models in the injected sulfate aerosol experiments exhibit warming in the upper troposphere to varying degrees. But the temperature responses in the stratosphere and lower troposphere differ. The G6sulfur models generally exhibit a weak response in the stratosphere and a relatively strong cooling in the lower troposphere (Visioni et al., 2021) due to the injected aerosols (Kalidindi et al., 2015). Thus, the large MPI decreases in G6sulfur are likely a result of this complex tropospheric vertical temperature trend lowering the tropopause and warming To (Visioni et al., 2021) and thus reducing the efficiency of the Carnot engine in the MPI model. G6solar, on the other hand, exhibits slight temperature decreases throughout the tropical troposphere and little change in the stratosphere (Tilmes et al., 2022) affecting smaller MPI differences from the target. The Feedback experiment shows somewhat less cooling than the G6Sulfur models in tropical lower troposphere but larger cooling of the upper stratosphere (Tilmes et al., 2018) resulting in a smaller overall MPI decrease. MPI changes in the ARISE-SAI-1.5 experiment exhibit a more complex pattern than the Feedback experiment. While there are decreases in the regions where tropical cyclones occur, they are smaller. However, data to calculate temperature changes above 100 hPa was not provided for the ARISE-SAI-1.5 experiment, so its MPI response is unexplained.
There is not yet a comprehensive theory about how climate change affects tropical cyclogenesis (Walsh et al., 2015; Sobel et al., 2021). While there is a consensus that the most intense tropical cyclones will increase in intensity with global warming (Cha et al., 2020; Seneviratne et al., 2021), opinions vary about how the total number of tropical storms might change (Chand et al., 2022, 2024; Emanuel, 2024). As a result, diagnostic indices reflecting tropical cyclogenesis are not without problems when applied to a changing climate (Wehner et al., 2015; Sobel et al., 2021). However, as the geoengineered climate is intended to be similar to the target climate, investigation via a genesis potential index (GPI) may be instructive. One commonly used variety of GPI is a function of Emanuel’s MPI, wind shear and low-level absolute vorticity (Emanuel and Nolan, 2004; Camargo et al., 2007):
where is η the absolute vorticity at 850 hPa (s−1), H is the percent relative humidity at 600 hPa, Vmax is the maximum potential intensity (m·s−1) and Vshear is the magnitude of the vertical wind shear between 850 hPa and 200 hPa (m·s−1).
The upper panel of Fig. 6 shows the natural variability metric for GPI from the C20C+ CAM5.1 simulations. Similar to the calculation of MPI, GPI is first calculated from monthly model output data followed by a determination of the annual maxima. The four lower panels of Fig. 6 show the GPI difference between the geoengineered simulations and their target simulations. GPI differences in ARISE-SAI-1.5 and Feedback experiments are more similar to each other than MPI differences. Generally, wind shear differences are the dominant factor in explaining GPI differences in these two experiments. While much of the geoengineered increases in GPI are not in regions currently experiencing significant tropical cyclogenesis, parts of the western Pacific, particularly near the Philippines could be adversely advected by these solar radiation management schemes. GPI differences in the G6solar and G6sulfur experiments are completely different. In the G6sulfur experiment, the very large reduction in MPI dominates geoengineered GPI leading to significant reduction in the potential for tropical cyclogenesis. Correlation between seasonal mean precipitation differences over ocean and GPI differences was found to be very low. However, it should be noted that at the horizontal resolutions of these climate models, they do not produce realistic tropical cyclones (Roberts et al., 2020).
7 Discussion and conclusions
In the absence of geoengineering, continuing to emit greenhouse gases into the atmosphere would undoubtedly result in extremely dangerous climates never experienced in the course of human existence (IPCC AR6 WG1 SPM, 2021). Solar radiation management schemes have been proposed as a method to mitigate some of these hazards but are extremely controversial for both technical and political reasons. It is not the purpose of this paper to advocate for or against this type of geoengineering but rather to attempt to provide some context of the results of existing simulations. To this end, it is proposed that geoengineering simulation results be considered in a background of natural variability. However, available simulations of geoengineering schemes vary widely in their targeted climates, thus complicating their intercomparison. The targeted climates in those experiments with less anthropogenic greenhouse gas forcing than the others should be easier to achieve as less geoengineered forcing would be required. As discussed below, normalizing the differences between geoengineered and targeted climates by the required global cooling reveals substantial variation in the effectiveness of the schemes considered here to offset the regional effects of anthropogenic greenhouse gas, especially for precipitation.
Here a convenient set of natural variability metrics are presented for temperature, precipitation and bulk tropical cyclone indices using a large ensemble of fixed sea surface temperature simulations from the Climate of the 20th Century (C20C+) experiment. This particular data set was chosen as it lends itself to quantifying the interannual variation in extreme temperature and precipitation via Generalized Extreme Value (GEV) statistical methods. While the metrics for mean quantities are also calculated from this experiment, there are many other possible choices. In particular, defining such natural variability metrics in terms of regional climate impacts could lead to more informed decisions about the success or failure of geoengineering.
Six different existing solar radiation management experiments are analyzed here. The ARISE-SAI-1.5, Feedback, and Feedback_low experiments involve a single model (CESM-WACCM) but containing moderately large ensembles of simulations. This particular model is a so-called “High Top” model with vertical coordinates throughout the entire stratosphere. This model feature is relevant to these geoengineering schemes that involve injection of sulfate aerosols into the stratosphere. The G6solar, G6sulfur, and G7cirrus experiments are part of the internationally coordinated Geoengineering Model Intercomparison (GeoMIP6) subproject of the Coupled Model Intercomparison Project (CMIP6). While these experiments involve multi-model simulations, ensemble size is considerably smaller than for the single model experiments. The participating GeoMIP models only partially encompass the full stratosphere.
Intercomparison of the six geoengineering schemes is complicated because each scheme is defined by different targeted climates and requires different amounts of geoengineering. The single model experiments, through a feedback scheme to control aerosol concentrations, aims to stabilize global mean surface temperature to fixed values. On the other hand, the GeoMIP6 schemes attempt to reduce the effect of a very high GHG emission scenario to that of a more moderate emission scenario. In these simulations, temperature continues to increase with time but at a lesser rate than in the absence of a solar radiation intervention. In all of these experiments, aerosol concentrations must continue to increase to offset the radiative effect of increasing GHG concentrations. Failure to maintain aerosol injections would result in a rapid adjustment of the climate system to the dangerous levels dictated by high GHG concentrations.
All the geoengineering experiments considered here, except G7cirrus, come close to meeting their targeted global mean surface temperature. Indeed, the differences between geoengineered and targeted temperatures are generally much less than current observed changes from the preindustrial period. However, geoengineered regional differences can be relatively large as seen in Fig. 2 and Figs. S1–S4. Differences between geoengineered and targeted extreme temperatures (20-year return values) over land are generally larger than mean temperature differences for all the experiments. But the natural variability of extreme temperatures, as defined here, is also larger leading to a reduction in the land area where extreme temperature differences are considered significant compared to mean summer temperature differences. As variability in winter temperatures is generally larger than summer temperatures, the winter metric for geoengineering success defined here is larger than the summer metric (Fig. 1). Winter mean temperature RMSE and global land temperature differences are larger than summer for the single model experiments but smaller for the multi-model experiments.
Due to its lower GHG forcing, the most recent experiment, ARISE-SAI-1.5, is superior to the others in achieving its targeted boreal winter and summer mean temperatures over land regions (Table 2). While the ARISE-SAI-1.5 geoengineered simulations are slightly warmer than its target, boreal mean winter and summer temperature differences are mostly insignificant over the populated tropical and midlatitude regions and less than 0.5°C except over the US Great Lakes region in the winter (Fig. 2). Extreme temperature differences are mixed but generally less than 1.0°C in these regions. In contrast, the two Feedback experiments are generally cooler than their targets over land (Figs. S1 and S2). Again, boreal mean summer temperature differences are mostly insignificant over the populated tropical and midlatitude regions and less than 0.5°C. Boreal winter temperatures are a mix of warmer and cooler temperatures than the target and differences exceed 1°C in several regions. Changing the altitude of aerosol injection has little effect on extreme and mean temperatures over land.
While both multi-model G6solar and G6sulfur experiments (Figs. S3 and S4) exhibit warmer winter and summer geoengineered surface air temperatures than the target at high latitudes, the differences are considerably larger over Northern Hemisphere land in the aerosol injection experiment. Visioni et al. (2021) showed this behavior was true for each of the individual models despite a larger cross model variance in G6sulfur due to differences in aerosol optical depths (see Fig. 7 of Visioni et al. (2021)). Kalidindi et al. (2015) also showed that top of atmosphere net radiative flux differences were of opposite signs for the G6solar and G6sulfur experiments over much of Northern Hemisphere land. It is quite likely that aerosol injection causes different changes to seasonal atmospheric circulation than simply changing the solar constant. Indeed, Banerjee et al. (2021) argued that aerosol injection leads to stratospheric circulation anomalies resembling the positive phase of the Northern Annular Mode in winter leading to warmer winter temperatures in Eurasia in the Feedback experiment. In both multi-model experiments, high northern latitude temperatures target exceedances are larger in winter than in summer. It should be clear that understanding how geoengineering strategies would change seasonal large-scale circulation is key to understanding its changes to temperature, precipitation and other high impact climate fields.
Projections of future changes in precipitation are considerably less confident than projections of future temperature changes (IPCC AR6 WG1 SPM, 2021). This is particularly true in regions where convective processes are an important contributor to seasonal rainfall due to known deficiencies in the cumulus parameterizations of CMIP class climate models (Rosa and Collins, 2013; Maher et al., 2018). While geoengineered precipitation differences from targets are generally smaller than observed and projected changes without geoengineering from pre-industrial values, some of these differences could be large enough to have impacts on agriculture and/or flood management. Previous analyses have highlighted circulation changes leading to increased drought risk, particularly in Africa (Abiodun et al., 2021; Alamou et al., 2022; de Perez et al., 2022). Controlling soil moisture by geoengineering temperature and thus potential evapotranspiration, would reduce the risk of agricultural drought in areas with small precipitation changes compared to scenarios without geoengineering (Odoulami et al., 2020; Liu et al., 2024). But in some of these geoengineered experiments, reduction in precipitation relative to target climates can be as large as the corresponding uncontrolled GHG emission scenario due to the complicated differences in temperatures aloft resulting from the complicated structural differences in long- and shortwave radiation fluxes.
All of geoengineering experiments considered here exhibit somewhere an increased risk of precipitation drought to varying degrees. While detailed differences are discussed above, even though precipitation differences are smallest in the ARISE-SAI-1.5 experiment (Fig. 4), the risk of African drought is increased. Drying of the Amazon would also be of concern due to its enormous reserve of biological carbon. Areas of increased risk of drought are further enlarged in the Feedback experiment extending into the northern midlatitudes in all seasons (Figs. S6 and S7). Changing the altitude of aerosol injection again has little effect. Differences between G6solar and G6sulfur are pronounced. In particular, a profound change in the hydrological cycle in G6sulfur leads to large scale reductions in seasonal mean precipitation. The dissimilarity of temperature differences between G6solar and G6sulfur suggests that simply dimming the solar constant uniformly is a poor predictor of the effect of sulfate aerosol injection in the stratosphere.
In these CMIP-class models, changes in extreme precipitation are constrained by the Clausius-Clapeyron (C-C) relationship leading to changes of about 6%/°C–7%/°C (Kharin et al., 2013; Li et al., 2021). However, evidence from high resolution event attribution studies suggest that changes in storm structure can lead to precipitation changes far exceeding C-C (Reed et al., 2021; Otto et al., 2023; Reed and Wehner, 2023; Tradowsky et al., 2023). Because of the C-C constraint and relatively small differences in geoengineered and target temperatures, differences in extreme precipitation (20-year return values) in the ARISE-SAI-1.5 are spatially noisy due to sampling limitations. In the other experiments with large seasonal mean precipitation differences, there is some degree of correlation between extreme and mean precipitation differences, due to the circulation differences that lead to storm track differences. This is particularly evident in the G6sulfur experiment (Fig. S9). However, because of the models’ limitation to realistically reproduce extreme storms (especially in convection prone regions and seasons), these extreme precipitation differences should be interpreted with caution.
While surface temperature differences between the geoengineering experiments and their targets are generally small, differences in the vertical temperature structure can be large. Geoengineered changes in temperatures aloft are more complex in the geoengineered simulations than in both current observations and projected model changes without geoengineering. Observed changes are characterized by warming from the surface throughout the troposphere and a pronounced cooling in the stratosphere. In all of the injected aerosol geoengineering experiments, the upper troposphere warms but changes in the lower troposphere and stratosphere vary. Understanding geoengineered changes in the vertical temperature structure is key to understanding how intense tropical cyclones and other strong storms would be modified. Generally, in the regions where tropical storms presently occur, Emanuel’s Maximum Potential Intensity, a well-established measure of a tropical storm’s lifetime maximum surface wind speeds, is generally less in the injected aerosol geoengineered simulations than in their targets in all of the experiments considered here.
However, there is little consistency in the differences in an oft used Genesis Potential Index to estimate the frequency of tropical cyclogenesis. With no established theory connecting climate change to tropical storm frequency, confidence in storm frequency differences would be low even if the experiments were more consistent. To better address this issue, geoengineering experiments with high resolution models may be informative (Wehner et al., 2014; Roberts et al., 2020).
While the ARISE-SAI-1.5 experiment is most effective at achieving its targeted climate, it requires the least amount of injected aerosol cooling in the experiments considered here (Table 1). Thus, its injection protocol is not the most effective in reducing the effect of a specified GHG forcing. Tables 4 and 5 show the RMSE from Tables 2 and 3 normalized by the intended geoengineered cooling. In this respect, the effectiveness of the five aerosol injections experiments are more similar to each other in offsetting the effect of increased GHG on summer mean land surface air temperature than Table 2 and Figs. 2, S1–S4 would indicate. Winter mean temperatures are less effectively controlled than summer mean temperatures in all experiments. There are substantial differences between the experiments in the seasonal mean precipitation normalized RMSE. The Feedback experiments are considerably more efficient than the other experiments in offsetting the effect of increased GHG on precipitation.
Visioni et al. (2023) found that global mean annual precipitation scaled linearly with injected stratospheric aerosol cooling using the CESM2-WACCM6 model and variants of the ARISE-SAI-1.5 experiment. However, in the multimodel/multi-experiment simulations analyzed here, there is no clear scaling relationship in global mean seasonal or annual precipitation over land. Supplemental Figs. S10 to S12 show the global mean RMSE for surface air temperatures and annual mean precipitation over land as a function of the intended geoengineered global cooling. Summer land surface air temperature RMSE appears to be independent of this cooling for the G6Solar and G6Sulfur experiments. There may be a multi- model linear relationship in the RMSE of annual precipitation over land with global cooling for the G6Sulfur experiment but not for the G6Solar experiment. For all fields, the RMSE for MPI-ESM2-LR model is an outlier, possibly due to an in-house aerosol data set (Visioni et al., 2021).
Consideration of the success or failure of a geoengineering scheme has been posed as a detection and attribution problem (MacMartin et al., 2019; Barnes et al., 2022). This permits framing two different questions. First, is the geoengineered climate significantly different than the non-geoengineered climate? Second, is the geoengineered climate significantly different than the targeted climate? Clearly if the answer to the first question is no, so is the answer to the second question and this is indeed the situation for the available G7cirrus simulations. Focused on the first question, Barnes et al. (2022) showed that the time to detecting that geoengineering altered extreme temperature and precipitation can be quite short, even at regional scales in the GLENS (feedback) experiment. This is a consequence of the large future changes from the high GHG emissions scenario and is similar to attributing current anthropogenic changes. Such detection times would be longer in more moderate GHG emissions scenarios. MacMartin et al. (2019) addressed the second question using the same experiment for mean temperature and precipitation finding that detection times at regional scales can be many decades. Their result is a consequence of the small differences, by design, between geoengineered and target climates (Figs. S1 and S6 here). Here, the difference between the geoengineered and target climates can be interpreted as the signals in MacMartin et al. (2019) while the natural variability metrics presented in this study can be interpreted as the noise in that study. While a regional failure of a geoengineering scheme may be indicated by the results presented here (Figs. 2, 4–6, S1–S9), determination of such failure may require many decades following the logic of MacMartin et al. (2019).
While this study has focused on assessing geoengineered changes to the climate system relevant to impacts on people, a more complete understanding of the processes driving these changes would require further analyses of the effect of geoengineering on large scale circulation and the vertical structure of the atmosphere. The ability to achieve the desired targeted climate is not independent of the effectiveness of a scheme to offset GHG forcing. Design of future numerical geoengineering experiments would be well served by coordinating targeted climates. Indeed, the effectiveness of different schemes could be non-stationary. The current mix of experiments does not easily facilitate such intercomparison.
Abiodun B J, Odoulami R C, Sawadogo W, Oloniyo O A, Abatan A A, New M, Lennard C, Izidine P, Egbebiyi T S, MacMartin D G (2021). Potential impacts of stratospheric aerosol injection on drought risk managements over major river basins in Africa.Clim Change, 169(3−4): 31
[2]
Alamou A E, Obada E, Biao E I, Zandagba E B J, Da-Allada C Y, Bonou F K, Baloïtcha E, Tilmes S, Irvine P J (2022). Impact of stratospheric aerosol geoengineering on meteorological droughts in West Africa.Atmosphere (Basel), 13(2): 234
[3]
Allen M R, Ingram W J (2002). Constraints on future changes in climate and the hydrologic cycle.Nature, 419(6903): 224–232
[4]
Banerjee A, Butler A H, Polvani L M, Robock A, Simpson I R, Sun L (2021). Robust winter warming over Eurasia under stratospheric sulfate geoengineering – the role of stratospheric dynamics.Atmos Chem Phys, 21(9): 6985–6997
[5]
Barnes E A, Hurrell J W, Sun L (2022). Detecting changes in global extremes under the GLENS-SAI climate intervention strategy.Geophys Res Lett, 49: e2022GL100198
[6]
Camargo S J, Emanuel K A, Sobel A H (2007). Use of a genesis potential index to diagnose ENSO effects on tropical cyclone genesis.J Clim, 20(19): 4819–4834
[7]
Cha E J, Knutson T R, Lee T C, Ying M, Nakaegawa T (2020). Third assessment on impacts of climate change on tropical cyclones in the Typhoon Committee Region – Part II: future projections.Trop Cyclone Res Rev, 9(2): 75–86
[8]
Chand S S, Walsh K J E, Camargo S J, Kossin J P, Tory K J, Wehner M F, Chan J C L, Klotzbach P J, Dowdy A J, Bell S S, Ramsay H A, Murakami H (2024). Reply to: limitations of reanalyses for detecting tropical cyclone trends.Nat Clim Chang, 14(2): 146–147
[9]
Chand S S, Walsh K J E, Camargo S J, Kossin J P, Tory K J, Wehner M F, Chan J C L, Klotzbach P J, Dowdy A J, Bell S S, Ramsay H A, Murakami H (2022). Declining tropical cyclone frequency under global warming.Nat Clim Chang, 12(7): 655–661
[10]
Chao L W, Zelinka M D, Dessler A E (2024). Evaluating cloud feedback components in observations and their representation in climate models.Journal of Geophysical Research: Atmospheres, 129: e2023JD039427
[11]
ColesS (20012001. An Introduction to Statistical Modeling of Extreme Values Springer Series in Statistics. Springer
[12]
de Perez E, Fuentes I, Jack C, Kruczkiewicz A, Pinto I, Stephens E (2022). Different types of drought under climate change or geoengineering: systematic review of societal implications.Front Clim, 4: 959519
[13]
Emanuel K (1987). The dependence of hurricane intensity on climate.Nature, 326(6112): 483–485
[14]
Emanuel K (2000). A statistical analysis of tropical cyclone intensity.Mon Weather Rev, 128(4): 1139–1152
[15]
Emanuel K (2024). Limitations of reanalyses for detecting tropical cyclone trends.Nat Clim Chang, 14(2): 143–145
[16]
Emanuel K A, Nolan D S (2004). Tropical cyclone activity and global climate.Bull Am Meteorol Soc, 85: 666–667
[17]
Eyring V, Bony S, Meehl G A, Senior C A, Stevens B, Stouffer R J, Taylor K E (2016). Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization.Geosci Model Dev, 9(5): 1937–1958
[18]
Gettelman A, Mills M J, Kinnison D E, Garcia R R, Smith A K, Marsh D R, Tilmes S, Vitt F, Bardeen C G, McInerny J, Liu H L, Solomon S C, Polvani L M, Emmons L K, Lamarque J F, Richter J H, Glanville A S, Bacmeister J T, Phillips A S, Neale R B, Simpson I R, DuVivier A K, Hodzic A, Randel W J (2019). The Whole Atmosphere Community Climate Model Version 6 (WACCM6).J Geophys Res Atmos, 124(23): 12380–12403
[19]
Gilford D M, Giguere J, Pershing A J (2024). Human-caused ocean warming has intensified recent hurricanes.Environmental Research: Climate, 3(4): 045019
[20]
Hosking J R M, Wallis J R, Wood E F (1985). Estimation of the generalized extreme-value distribution by the method of probability-weighted moments.Technometrics, 27(3): 251–261
[21]
HoskingJ, WallisJ R (19971997. Regional Frequency Analysis. Cambridge University Press
[22]
IPCC AR6 WG1 SPM (20212021. IPCC AR6 WG1 Summary for Policymakers. In: Masson-Delmotte V, Zhai P, Pirani A, Connors S L, Péan C, Berger S, Caud N, Chen Y, Goldfarb L, Gomis M I, Huang M, Leitzell K, E, Lonnoy J, Matthews B R, Maycock T K, Waterfield T, Yelekçi O, Yu R, Zhou B, eds. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge: Cambridge University Press
[23]
Kalidindi S, Bala G, Modak A, Caldeira K (2015). Modeling of solar radiation management: a comparison of simulations using reduced solar constant and stratospheric sulphate aerosols.Clim Dyn, 44(9−10): 2909–2925
[24]
Kharin V V, Zwiers F W, Zhang X, Wehner M (2013). Changes in temperature and precipitation extremes in the CMIP5 ensemble.Clim Change, 119(2): 345–357
[25]
Kravitz B, MacMartin D G, Mills M J, Richter J H, Tilmes S, Lamarque J F, Tribbia J J, Vitt F (2017). First simulations of designing stratospheric sulfate aerosol geoengineering to meet multiple simultaneous climate objectives.J Geophys Res Atmos, 122(23): 12616–12634
[26]
Kravitz B, Robock A, Tilmes S, Boucher O, English J M, Irvine P J, Jones A, Lawrence M G, MacCracken M, Muri H, Moore J C, Niemeier U, Phipps S J, Sillmann J, Storelvmo T, Wang H, Watanabe S (2015). The Geoengineering Model Intercomparison Project Phase 6 (GeoMIP6): simulation design and preliminary results.Geosci Model Dev, 8(10): 3379–3392
[27]
Li C, Zwiers F, Zhang X, Li G, Sun Y, Wehner M (2021). Changes in annual extremes of daily temperature and precipitation in CMIP6 Models.J Clim, 34(9): 3441–3460
[28]
Liu Z, Lang X, Jiang D (2024). Stratospheric aerosol injection geoengineering would mitigate greenhouse gas-induced drying and affect global drought patterns.Journal of Geophysical Research: Atmospheres, 129: e2023JD039988
[29]
MacMartin D G, Wang W, Kravitz B, Tilmes S, Richter J H, Mills M J (2019). Timescale for detecting the climate response to stratospheric aerosol geoengineering.J Geophys Res Atmos, 124(3): 1233–1247
[30]
Maher P, Vallis G K, Sherwood S C, Webb M J, Sansom P G (2018). The impact of parameterized convection on climatological precipitation in atmospheric global climate models.Geophys Res Lett, 45(8): 3728–3736
[31]
Morrison A L, Barnes E A, Hurrell J W (2024). Natural variability can mask forced permafrost response to stratospheric aerosol injection in the ARISE-SAI-1.5 simulations.Earths Future, 12: e2023EF004191
[32]
NASEM (20212021. Reflecting Sunlight: Recommendations for Solar Geoengineering Research and Research Governance. National Academies of Sciences, Engineering, and Medicine. Washington, D.C.: The National Academies Press
[33]
O’Neill B C, Tebaldi C, van Vuuren D P, Eyring V, Friedlingstein P, Hurtt G, Knutti R, Kriegler E, Lamarque J F, Lowe J, Meehl G A, Moss R, Riahi K, Sanderson B M (2016). The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6.Geosci Model Dev, 9(9): 3461–3482
[34]
Odoulami R C, New M, Wolski P, Guillemet G, Pinto I, Lennard C, Muri H, Tilmes S (2020). Stratospheric aerosol geoengineering could lower future risk of ‘Day Zero’ level droughts in Cape Town.Environ Res Lett, 15(12): 124007
[35]
OSTP (20232023. Congressionally Mandated Research Plan and an Initial Research Governance Framework Related to Solar Radiation Modification. Office of Science and Technology Policy, Washinton, D.C.
[36]
Otto F E L, Zachariah M, Saeed F, Siddiqi A, Kamil S, Mushtaq H, Arulalan T, AchutaRao K, Chaithra S T, Barnes C, Philip S, Kew S, Vautard R, Koren G, Pinto I, Wolski P, Vahlberg M, Singh R, Arrighi J, van Aalst M, Thalheimer L, Raju E, Li S, Yang W, Harrington L J, Clarke B (2023). Climate change increased extreme monsoon rainfall, flooding highly vulnerable communities in Pakistan.Environmental Research: Climate, 2(2): 025001
[37]
Reed K A, Wehner M F (2023). Real-time attribution of the influence of climate change on extreme weather events: a storyline case study of Hurricane Ian rainfall.Environmental Research: Climate, 2(4): 043001
[38]
Reed K, Wehner M F, Stansfield A M, Zarzycki C M (2021). Anthropogenic influence on Hurricane Dorian’s extreme rainfall.Bull Am Meteorol Soc, 102(1): S9–S15
[39]
Riahi K, van Vuuren D P, Kriegler E, Edmonds J, O’Neill B C, Fujimori S, Bauer N, Calvin K, Dellink R, Fricko O, Lutz W, Popp A, Cuaresma J C, Samir K C, Leimbach M, Jiang L, Kram , T , Rao S, Emmerling J, Ebi K, Hasegawa T, Havlik P, Humpenöder F, Da Silva L A, Smith S, Stehfest E, Bosetti V, Eom J, Gernaat D, Masui T, Rogelj J, Strefler J, Drouet L, Krey V, Luderer G, Harmsen M, Takahashi K, Baumstark L, Doelman J C, Kainuma M, Klimont Z, Marangoni G, Lotze-Campen H, Obersteiner M, Tabeau A, Tavoni M (2017). The shared socioeconomic pathways and their energy, land use, and greenhouse gas emissions implications: an overview.Global Environmental Change, 42: 153–168
[40]
Richter J H, Visioni D, MacMartin D G, Bailey D A, Rosenbloom N, Dobbins B, Lee W R, Tye M, Lamarque J F (2022). Assessing responses and impacts of solar climate intervention on the earth system with stratospheric aerosol injection (ARISE-SAI): protocol and initial results from the first simulations.Geosci Model Dev, 15(22): 8221–8243
[41]
Ricke K, Wan J S, Saenger M, Lutsko N J (2023). Hydrological consequences of solar geoengineering.Annu Rev Earth Planet Sci, 51(1): 447–470
[42]
Roberts M J, Camp J, Seddon J, Vidale P L, Hodges K, Vanniere B, Mecking J, Haarsma R, Bellucci A, Scoccimarro E, Caron L P, Chauvin F, Terray L, Valcke S, Moine M P, Putrasahan D, Roberts C, Senan R, Zarzycki C, Ullrich P (2020). Impact of model resolution on tropical cyclone simulation using the HighResMIP–PRIMAVERA multimodel ensemble.J Clim, 33(7): 2557–2583
[43]
Rosa D, Collins W D (2013). A case study of subdaily simulated and observed continental convective precipitation: CMIP5 and multiscale global climate models comparison.Geophys Res Lett, 40(22): 5999–6003
[44]
Santer B D, Wehner M F, Wigley T M L, Sausen R, Meehl G A, Taylor K E, Ammann C, Arblaster J, Washington W M, Boyle J S, Brüggemann W (2003). Contributions of anthropogenic and natural forcing to recent tropopause height changes.Science, 301(5632): 479–483
[45]
SeneviratneS I, ZhangX, AdnanM, BadiW, DereczynskiC, Di LucaA, GhoshS, IskandarI, KossinJ, LewisS, OttoF, PintoI, SatohM, Vicente-SerranoS M, WehnerM, ZhouB (20212021. Weather and climate extreme events in a changing climate. In: Masson-Delmotte V, Zhai P, Pirani A, Connors S L, Pé C, Berger S, Caud N, Chen Y, Goldfarb L, Gomis M I, Huang M, Leitzell K, Lonnoy E, Matthews J B R, Maycock T K, Waterfield T, Yelekçi O, Yu R, Zhou B, eds. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK and New York, NY, USA
[46]
Sobel A H, Wing A A, Camargo S J, Patricola C M, Vecchi G A, Lee C Y, Tippett M K (2021). Tropical cyclone frequency.Earths Future, 9: e2021EF002275
[47]
Stone D A, Christidis N, Folland C, Perkins-Kirkpatrick S, Perlwitz J, Shiogama H, Wehner M F, Wolski P, Cholia S, Krishnan H, Murray D, Angélil O, Beyerle U, Ciavarella A, Dittus A, Quan X W, Tadross M (2019). Experiment design of the international CLIVAR C20C+ detection and attribution project.Weather Clim Extrem, 24: 100206
[48]
Tabazadeh A, Turco R P, Jacobson M Z (1994). A model for studying the composition and chemical effects of stratospheric aerosols.J Geophys Res, 99(D6): 12897–12914
[49]
Tilmes S, Richter J H, Kravitz B, MacMartin D G, Mills M J, Simpson I R, Glanville A S, Fasullo J T, Phillips A S, Lamarque J F, Tribbia J, Edwards J, Mickelson S, Ghosh S (2018). CESM1(WACCM) Stratospheric Aerosol Geoengineering Large Ensemble Project.Bull Am Meteorol Soc, 99(11): 2361–2371
[50]
Tilmes S, Richter J H, Mills M J, Kravitz B, MacMartin D G, Vitt F, Tribbia J J, Lamarque J F (2017). Sensitivity of aerosol distribution and climate response to stratospheric SO2 injection locations.J Geophys Res Atmos, 122(23): 12591–12615
[51]
Tilmes S, Visioni D, Jones A, Haywood J, Séférian R, Nabat P, Boucher O, Bednarz E M, Niemeier U (2022). Stratospheric ozone response to sulfate aerosol and solar dimming climate interventions based on the G6 Geoengineering Model Intercomparison Project (GeoMIP) simulations.Atmos Chem Phys, 22(7): 4557–4579
[52]
Tradowsky J S, Philip S Y, Kreienkamp F, Kew S F, Lorenz P, Arrighi J, Bettmann T, Caluwaerts S, Chan S C, De Cruz L, de Vries H, Demuth N, Ferrone A, Fischer E M, Fowler H J, Goergen K, Heinrich D, Henrichs Y, Kaspar F, Lenderink G, Nilson E, Otto F E L, Ragone F, Seneviratne S I, Singh R K, Skålevåg A, Termonia P, Thalheimer L, van Aalst M, Van den Bergh J, Van de Vyver H, Vannitsem S, van Oldenborgh G J, Van Schaeybroeck B, Vautard R, Vonk D, Wanders N (2023). Attribution of the heavy rainfall events leading to severe flooding in Western Europe during July 2021.Clim Change, 176(7): 90
[53]
Tselioudis G, Rossow W B, Jakob C, Remillard J, Tropf D, Zhang Y (2021). Evaluation of clouds, radiation, and precipitation in CMIP6 models using global weather states derived from ISCCP-H cloud property data.J Clim, 34: 7311–7324
[54]
Visioni D, Bednarz E M, MacMartin D G, Kravitz B, Goddard P B (2023). The choice of baseline period influences the assessments of the outcomes of stratospheric aerosol injection.Earths Future, 11: e2023EF003851
[55]
Visioni D, MacMartin D G, Kravitz B, Boucher O, Jones A, Lurton T, Martine M, Mills M J, Nabat P, Niemeier U, Séférian R, Tilmes S (2021). Identifying the sources of uncertainty in climate model simulations of solar radiation modification with the G6sulfur and G6solar Geoengineering Model Intercomparison Project (GeoMIP) simulations.Atmos Chem Phys, 21(13): 10039–10063
[56]
Walsh K J E E, Camargo S J, Vecchi G A, Daloz A S, Elsner J, Emanuel K, Horn M, Lim Y K, Roberts M, Patricola C, Scoccimarro E, Sobel A H, Strazzo S, Villarini G, Wehner M, Zhao M, Kossin J P, La Row T, Oouchi K, Schubert S, Wang H, Bacmeister J, Chang P, Chauvin F, Jablonowski C, Kumar A, Murakami H, Ose T, Reed K A, Saravanan R, Yamada Y, Zarzycki C M, Luigi Vidale P, Jonas J A, Henderson N (2015). Hurricanes and climate: the U. S. Clivar working group on hurricanes.Bull Am Meteorol Soc, 96(6): 997–1017
[57]
Wehner M F, Duffy M L, Risser M, Paciorek C J, Stone D A, Pall P (2024). On the uncertainty of long-period return values of extreme daily precipitation.Front Clim, 6: 1343072
[58]
Wehner M F, Kossin J P (2024). The growing inadequacy of an open-ended Saffir–Simpson hurricane wind scale in a warming world.Proc Natl Acad Sci USA, 121(7): e2308901121
[59]
Wehner M F, Reed KA, Li F, Prabhat , Bacmeister J, Chen C-T, Paciorek C, Gleckler P J, Sperber K R, Collins W D, Gettelman A, Jablonowski J (2014). The effect of horizontal resolution on simulation quality in the Community Atmospheric Model, CAM5.1.J Adv Model Earth Syst, 6(4): 980–997
[60]
Wehner M, Prabhat , Reed K A, Stone D, Collins W D, Bacmeister J (2015). Resolution dependence of future tropical cyclone projections of CAM5.1 in the U. S. CLIVAR Hurricane Working Group idealized configurations.J Clim, 28(10): 3905–3925
[61]
Wehner M, Stone D, Shiogama H, Wolski P, Ciavarella A, Christidis N, Krishnan H (2018). Early 21st century anthropogenic changes in extremely hot days as simulated by the C20C+ detection and attribution multi-model ensemble.Weather Clim Extrem, 20: 1–8
[62]
Zhang Y, Jin Z, Ottaviani M (2023). Comparison of clouds and cloud feedback between AMIP5 and AMIP6.Atmosphere (Basel), 14(6): 978
RIGHTS & PERMISSIONS
The Author(s). This article is published with open access at link.springer.com and journal.hep.com.cn
AI Summary 中Eng×
Note: Please be aware that the following content is generated by artificial intelligence. This website is not responsible for any consequences arising from the use of this content.
PDF
(9185KB)
