Targeted marine cloud brightening weakens subsequent El Niño

Wildfires as a natural analog for MCB

To explore the viability of the cloud brightening from the 2019–2020 Australian wildfire event as a proxy for MCB, we compare the climate responses from the wildfires to simulations where instead MCB is deployed in the ocean regions with the strongest CDNC responses to the fires between December 2019 to February 2020 (fig. S1), which aligns with the peak season of the bushfire emissions (24). We find that MCB reproduces many of the key mechanisms identified from the wildfires including an immediate negative local shortwave forcing response (Fig. 1, A and D, and fig. S2A) and increase in cloud liquid water path (fig. S2B) in the SESP region. Cloud and radiative effects are followed by surface cooling (Fig. 1, C and F, and fig. S2D) and boundary layer drying (fig. S2C), which leads to a slight northward shift of the mean distribution of precipitation near the Intertropical Convergence Zone (ITCZ) across the equatorial Pacific (Fig. 1, B and E). The December 2019 to February 2020 mean shortwave cloud forcing in the cloud brightening region [gray contour in Fig. 1 (A and D)] is slightly higher in our MCB experiment (−7.7 W/m²) compared to the wildfires (−5.4 W/m²), but both experiments result in a similar global mean cooling during the subsequent La Niña peak (−0.1°C). Notably, MCB in the same regions with enhanced cloud albedo from the wildfires reproduces the strong tropical Pacific cooling pattern from Fasullo et al. (24). The precipitation response is also similar between the wildfires and MCB, particularly west Pacific equatorial drying, central/west Pacific northward shift in the ITCZ, and equatorial Indian Ocean poleward divergence in precipitation (Fig. 1, B and E). Thus, MCB can reproduce the simulated cloud brightening and La Niña–like responses from the 2019–2020 Australian wildfires. In addition, this supports the hypothesis from Fasullo et al. (24) that low cloud brightening from the wildfires was an important contributor to the 2020–2021 La Niña.

Deliberate MCB to modify El Niño

While implementing MCB in the ocean regions most influenced by the 2019–2020 Australian wildfires can effectively reproduce the 2020–2021 La Niña response, a targeted MCB strategy designed to deliberately modify El Niño would not necessarily replicate the wildfire response. Given the complex interactions between meteorological conditions, seasonality, and background aerosols, there is no reason to assume that the forcing produced by the fires is the limit of what could be achieved by MCB. For the El Niño modification MCB experiments, we use a higher CDNC perturbation of 500 particles/cm³ which is near the upper limit of what could be achieved with sea salt injection (34). We also intervene in a smaller, contiguous region of the SESP (7% global ocean area) with high brightening susceptibility (35, 36) rather than the larger patchwork of naturally brightened clouds associated with the fires (18.7% ocean area). Last, the timing of MCB to effectively modify El Niño would not correspond to the seasonal cycle of wildfires. Physical limitations in ENSO forecasting, in particular the spring predictability barrier (41), would be a determining constraint on MCB initiation because seasonal forecast models typically have low skill in predicting ENSO events before boreal summer, although some initial states may have predictability beyond a year (42). In addition, the duration of MCB deployment could vary depending on the desired climate outcome and available resources. We simulate a set of strategies to characterize the importance of timing in modifying El Niño (table S1), varying MCB initiation and termination preceding the two strongest El Niños of the past several decades: the 2015–2016 El Niño, a mixed regime event with central Pacific warming (43) and, for contrast, the 1997–1998 El Niño, an event characterized by strong eastern Pacific warming.

MCB deployed in the SESP after the spring predictability barrier weakens both El Niño events, and the responses depend on the timing of intervention (Fig. 2, A and B, and fig. S3). During the 2015–2016 El Niño peak (Fig. 2C; “ENSO peak” defined here as DJF), the earliest and longest “Full effort” MCB (Fig. 2D; June to February MCB) results in the strongest cooling in the SESP brightening region (−1.67°C) and reductions in Niño 3.4 SST (−1.88°C), virtually restoring neutral ENSO conditions by the end of the event peak. On the other end, the latest and shortest “11th hour” MCB (December to February MCB) leads to less cooling in the SESP (−0.52°C) and the smallest Niño 3.4 SST reductions (−0.31°C). “Early action” MCB (June to August MCB) has the same initiation month as the Full effort strategy and the same duration as the 11th hour strategy yet causes the least amount of cooling in the SESP (−0.31°C) and moderate Niño 3.4 SST cooling (−0.83°C).

Considering ENSO indicators beyond SST, the three scenarios with the longest durations have the greatest efficacy in reversing typical El Niño atmospheric conditions over the tropical Pacific, while the earliest initiating scenarios produce the largest reversal of typical El Niño state of the thermocline during the event peak (table S2). Perhaps intuitively, these results demonstrate that starting MCB early and leaving it on for the longest duration is more effective at weakening El Niño than implementing MCB only during the event peak.

We find similar relative responses between MCB strategies for the 1997–1998 El Niño (Fig. 2, B to F), compared to the 2015–2016 El Niño, but at a smaller magnitude, suggesting the exact responses are sensitive to different types of El Niño events and background conditions preceding El Niño within CESM2. Eastern Pacific El Niño events are typically associated with decreased low cloud cover off the coast of Peru and Chile. This is consistent in CESM2, where the 1997–1998 event has a stronger reduction low cloud cover (fig. S4F) and cloud liquid water path (fig. S4E) directly along the South American coast compared to the 2015–2016 event (fig. S4, B and C). However, because the MCB seeding region is farther west of the coastline and encompasses much of the central Pacific, we observe slightly diminished cloud fraction and enhanced cloud liquid water path preceding the 2015–2016 event compared to the cloudier and drier conditions preceding the 1997–1998 event. Thus, the background conditions preceding the 2015–2016 El Niño are easier to perturb with the same MCB interventions, which is supported by the larger increases in low cloud fraction and cloud liquid water path during the event peak leading to stronger decreases in specific humidity (boundary layer drying) and surface temperature in the SESP region for the 2015–2016 event (fig. S5) compared to the 1997–1998 event (fig. S6).

There are also notable side effects in the 2015–2016 event that do not appear in the 1997–1998 event, including significant warming over Europe and Asia (Fig. 2, D and F). The 2015–2016 Full effort MCB surface temperature response is similar to the negative South Pacific Meridional Mode (SPMM) (44) and positive North Pacific Meridional Mode (NPMM) (45) induced by the 2019–2020 Australian wildfires (25), which suggests that ENSO-targeted MCB could trigger other modes of variability and lead to unintended remote effects. Despite different spatial patterns in surface temperature between the 2015–2016 (Fig. 2D) and 1997–1998 events (Fig. 2F), the global mean temperature response to Full effort MCB is not statistically different from the control for both events (fig. S7).

MCB also changes the timing and magnitude of the ensuing La Niña event following El Niño. All MCB strategies result in colder Niño 3.4 SST anomalies relative to the control in boreal fall after the 2015–2016 El Niño peak, except Early action MCB which leads to warmer SSTs than the control through the next year (Fig. 2A). The delayed warming due to Early action MCB suggests that MCB preceding the El Niño growth phase damps the amplitude of the recharge oscillator (46) (e.g., smaller discharge of equatorial heat content during the warm phase leads to weaker recharge during the subsequent cold phase) while later MCB fundamentally disrupts the ENSO growth and decay dynamics. The three longest duration MCB strategies tend toward negative Niño 3.4 SST anomalies of −0.5°C or less (that is, a La Niña state) seasons earlier than in the control simulations and enhance the amplitude of the La Niña SSTs. Results from the 2015–16 simulations suggest that Early action MCB may provide a pathway for reducing the amplitude of El Niño without substantially influencing the onset of the subsequent La Niña, although further study is needed to understand the robustness of this result. Changes to the subsequent La Niña are less pronounced or insignificant following the 1997–1998 El Niño (Fig. 2B), suggesting that La Niña effects are also sensitive to the preceding El Niño conditions.

Early MCB disrupts ENSO growth feedbacks

If El Niño weakening was strictly caused by the cloud forcing imposed by MCB, we would expect the MCB strategies with the greatest radiative perturbations to produce the strongest reductions in peak Niño 3.4 SST anomalies. However, this relationship is only observed for Full effort MCB during the 2015–2016 El Niño, which has the largest mean shortwave forcing response from June to February in the seeding region (−27.1 W/m²; fig. S8) and DJF Niño 3.4 SST anomaly (−1.88°C; fig. S9). The outcomes of the other five strategies differ in ordering of their mean forcing and SST responses. For example, the 11th hour strategy has the smallest peak Niño 3.4 SST anomaly (−0.31°C) but only the third smallest mean forcing (−8.55 W/m²) likely due to the stronger incoming solar radiation during boreal winter in the Southern Hemisphere. Thus, the MCB’s effect on ENSO cannot be fully explained by the atmospheric thermodynamic response to radiative perturbations; it must also induce dynamical changes between the atmosphere and ocean that feed back onto the ENSO cycle.

To diagnose the dynamical responses to MCB, we examine the progression of physical processes that drive El Niño growth and decay. In boreal spring, downwelling Kelvin waves deepen the thermocline, reduce upwelling, and lead to warmer SSTs in the eastern tropical Pacific, including in the Niño 3.4 region. This weakens the characteristic zonal SST gradient. Late boreal summer and fall are the typical growth phase of El Niño during which positive Bjerknes feedbacks (47) amplify the weakened zonal SST gradient and weakened (or reversed) easterly trade winds in the tropical Pacific (48).

By boreal winter, an unperturbed mature El Niño is typically associated with decreased Walker cell strength (see Materials and Methods), a more uniform equatorial SST gradient, and a shoaled thermocline slope index (see Materials and Methods). However, MCB results in a relatively strengthened Walker cell, increased zonal SST gradient, and a steeper thermocline slope compared to the unperturbed El Niño, effectively restoring neutral ENSO conditions for the atmosphere and ocean in the eastern Pacific while having a more moderate effect in the western Pacific (Fig. 3). While the strongest responses due to MCB tend to occur in the central and eastern Pacific (e.g., earlier increase in easterly trade winds and SST cooling relative to El Niño), Full effort MCB leads to a persistent deepening of the thermocline in the western equatorial Pacific during El Niño (fig. S10), which suggests that MCB is enhancing La Niña responses rather than merely masking surface warming in the Niño 3.4 region.

This change in typical atmospheric and oceanic conditions associated with a neutral ENSO state depends on the timing of MCB deployment. Both Walker cell strength and thermocline slope index (fig. S11) are statistically unchanged by 11th hour MCB. Because this strategy initiates MCB after the El Niño growth phase, the MCB predominantly imposes local radiative cooling in the SESP region rather than altering the dynamics of the ENSO cycle. All other MCB strategies significantly increase Walker cell strength, steepen thermocline slope, and reduce Niño 3.4 SST anomalies during the El Niño peak. For the June-initiated strategies, the magnitude of the changes during the event peak increases with duration. This suggests that MCB implemented in the SESP during the growth phase weakens El Niño most effectively because it disrupts the nonlinear Bjerknes feedbacks that would normally enhance El Niño conditions, but those effects weaken as time progresses after MCB is terminated. A simplified heat budget analysis (see Materials and Methods) comparing surface heat flux and upper ocean divergence shows that ocean divergence dominates the MCB cooling effect on the 2015–2016 El Niño event (fig. S12A), and the magnitude scales roughly with the Niño 3.4 SST anomalies for each strategy. The ocean divergence response is generally less during the 1997–1998 El Niño (fig. S12B), while the surface flux changes are similar to the 2015–2016 event (within 15 to 39% for the June-initiated strategies), further supporting that MCB is less effective at overriding the feedbacks that lead to Walker cell weakening in the 1997–1998 event due to the unfavorable eastern Pacific El Niño background cloud conditions.

While we focus our analysis on how MCB modifies feedbacks directly linked to ENSO, it is possible that MCB may trigger other modes of variability. Recent work on the 2019–2020 Australian wildfires shows that the state of the SPMM and NPMM influences whether a multiyear La Niña ensues (25). We find a similar westward propagating zonal equatorial SST gradient (Fig. 2, D and F, and fig. S10, B and E) and enhanced southeasterly cross-equatorial trade winds (figs. S10, A and D, S13, and S14), likely via wind-evaporation-SST feedback (49), due to Full effort MCB, suggesting that MCB excites a negative SPMM state which could further weaken El Niño (25, 44). On the other hand, if we remove the tropical Pacific ENSO variability [linearly regress out the first empirical orthogonal function (EOF) of detrended SSTs between 20°S to 20°N and 140°E to 90°W], then the resultant Pacific temperature response from Full effort MCB reveals a more prominent NPMM pattern and signal of the SESP MCB cooling, while the SPMM projection is no longer statistically significant (fig. S15) compared to Fig. 2 (D and F). Given the substantial overlap between the ENSO, SESP MCB, and NPMM/SPMM regions, it is challenging to establish a causal link between the direct effects of MCB and different modes of variability solely within a coupled modeling framework.

Irrespective of which mode is dominant or whether the mode is triggered by MCB at all, the initial state of meridional mode may modulate the ENSO response. An approach that considers the state of other modes of variability, while beyond the scope of the present study, may provide additional predictive skill in achieving a particular MCB response and a worthy topic for investigation in follow-up studies.

Regional climate impacts of modifying El Niño with MCB

While the societal impacts of El Niño events are heterogeneous, El Niño has been linked to net reduced economic growth (7). El Niño years are also warmer at the global mean level (50–52) so, on average, will boost the predominant effects of anthropogenic warming (2). While our analysis illustrates that MCB can weaken El Niño as measured by the Niño 3.4 index (Fig. 2, A and B), Southern Oscillation Index (table S2), and some of its physical manifestations across the tropical Pacific (Fig. 3), whether MCB is ultimately attractive will most likely depend on whether harmful regional impacts associated with El Niño are reduced.

To identify the regions that typically experience substantial temperature and precipitation changes during El Niño in CESM2, we cluster geographic regions with temperature and precipitation anomalies preceding and during the 2015–2016 (Fig. 4, A and B, and fig. S16) and 1997–1998 (fig. S17) El Niño events (June to February mean) greater than twice the SD after normalizing by the monthly SDs from the CESM2 Large Ensemble (LENS2) (53) output for 50 ensemble members. By averaging the anomalies before and during the El Niño peak, we account for the possibility that certain regions will experience the effects of El Niño in the growth phase leading up to DJF, but we exclude the time after the event peak to avoid characterizing La Niña effects that may have emerged earlier due to MCB. Last, within each of the El Niño impact regions, we calculate the change in temperature (Fig. 4, C and E) and precipitation (Fig. 4, D and F) anomalies due to Full effort, Early action, and 11th hour MCB as illustrative cases.

During the 2015–2016 El Niño, the three MCB strategies generally reduce the temperature and precipitation effects that would occur under El Niño for most regions (e.g., places that would get warmer under El Niño get cooler due to MCB, places that get wetter under El Niño get drier due to MCB, etc.). Full effort MCB results in the largest reduction of warm (Fig. 4C) and dry (Fig. 4D) El Niño effects for all regions except Western Africa where drying is insignificantly increased due to MCB. Cool (Fig. 4C) and wet (Fig. 4D) El Niño effects are weakened under Full effort MCB, with the exceptions of Eastern Africa and Southeastern Asia which undergo enhanced insignificant and significant cooling, respectively, due to Full effort MCB on top of the El Niño cooling effects. The effects for Early action and 11th hour MCB are generally smaller and statistically insignificant compared to Full effort MCB, but they still reduce the El Niño impacts for all regions except a slight increase of warming in South America due to 11th hour MCB. While the results are consistent with our physical expectations overall, targeted MCB interventions do not perfectly reverse the effects of El Niño. In general, MCB appears more effective at significantly weakening regional warm anomalies and precipitation changes than at reducing El Niño induced regional cooling. Cases in which MCB enhances rather than reduces El Niño impacts, while atypical, present ethical risks that must be carefully considered before any future implementation. While we focus here on the contiguous regions which have statistically significant temperature and precipitation changes during El Niño events, the teleconnected temperature (figs. S9 and S18) and precipitation (figs. S13 and S14) responses to MCB are globally distributed. There are regions not significantly affected during El Niño that are substantially modified by MCB. For example, the significant warming over Europe and Asia due to MCB during the 2015–2016 event (Fig. 2D) occurs in an area that experienced statistically insignificant temperature changes during the 2015–2016 El Niño (Fig. 4A). Nonetheless, these results demonstrate that timely strategic MCB deployment could reduce some of the remote teleconnections associated with El Niño.

The regional effects of MCB on the 1997–1998 El Niño are generally smaller and less consistent than the 2015–2016 El Niño effects (Fig. 4, E and F). While most warm and wet effects are weakened due to the three MCB strategies (~70%), around 41% of the cool and dry effects are increased across the different strategies. This underscores the point that responses to ENSO-targeted MCB are likely to be dependent on the background conditions, type of El Niño event, and MCB strategy. In particular, these results suggest that MCB may be more effective at weakening central Pacific El Niño than eastern Pacific El Niño events.

Translating these regional climate responses into human welfare-relevant impacts is not straightforward. Recent literature identifies a robust link between El Niño economic damages and the E-index (7), a common ENSO metric (54) based on the two leading EOF modes of SST anomalies in the tropical Pacific (see Materials and Methods). A strong positive E-index is typically associated with eastern Pacific El Niño events and high global income losses on the order of trillions of dollars (7). To orient our findings in the context of this recent literature, we calculate the E-index changes associated with the simulated MCB strategies (fig. S19 and table S2) The reduction in E-index due to our MCB interventions suggests the potential for large global economic benefits, especially under events like the 2015–2016 El Niño, but any decoupling of traditional ENSO indices and typical remote climate effects means that innovations in El Niño impact assessment will be needed to more fully understand the potential consequences of an MCB-modified El Niño.

Simulating MCB with the SMYLE using CESM2

We use the SMYLE (31) using the CESM2 (23) to simulate the climate responses of MCB. SMYLE is an initialized prediction system designed to test prediction skill for lead times ranging from 1 month to 2 years (31). SMYLE has similar ENSO prediction skill comparable to other operational seasonal forecast systems including the European Centre for Medium-Range Weather Forecasts seasonal forecast system 5. Building off these advances in seasonal initialized prediction modeling, we follow a similar model setup as used in (24). We run ensembles of 2-year-long MCB simulations using SMYLE preceding three historical ENSO events: (i) the 2020–2021 La Niña to approximate the forcing and subsequent La Niña–like response from the fires presented in (24) and illustrate the connection to MCB (20 members), (ii) the 2015–2016 El Niño to test MCB’s potential efficacy in weakening the most extreme El Niño of the 21st century (43) (10 members), and (iii) the 1997–1998 El Niño to check the robustness of (ii) (10 members). We use the corresponding month initialized hindcast from the SMYLE experiment with no MCB as our control to compute the responses to MCB and the full 20-member SMYLE experiment with no MCB for our reference historical monthly climatology (1970–2014).

MCB perturbations are represented in CESM2 following the same methodology as in (20) by nudging the CDNC within the target region. For the 2020–2021 La Niña MCB experiment (i), we nudge the CDNC to a constant of 100 particles/cm³ in the ocean grid cells with CDNC perturbations greater than the 85th percentile from December 2019 to February 2020 (fig. S1, D to F). For both the (ii) 2015–2016 and (iii) 1997–1998 El Niño MCB experiments, we increase the CDNC to 500 particles/cm³ over the SESP region (Latitude: 30°S to 0°; Longitude: 150°W to 85°W). We chose this higher perturbation because it is near the upper limit of what might be physically achievable from work that explicitly modeled sea salt aerosol treatments (34). While the wildfires brightened a larger cloud deck to contribute to the 2020–2021 La Niña, the SESP region was chosen for the El Niño targeted experiments due to its robust La Niña–like response across ESMs, including ones that test multiple locations in CESM2 (15, 59). In our first set of experiments for (i), we initialize 20 ensemble members in August 2019 and apply MCB in the ocean grid cells with the highest CDNC perturbation due to the Australian fires from December 2019 through February 2020 and then shut MCB off for the remainder of the simulation to allow for the development of any ENSO conditions. In (ii) and (iii), we run a set of six May-initialized experiments (10 members each) with varying MCB initiation months and durations. Boreal spring initialization was chosen to reduce the simulation runtime before MCB initiation. The six MCB strategies are implemented as varied combinations of seasonal [June, July, August (JJA); September, October, November (SON); December, January, February (DJF)] deployments with the earliest experiments initiating MCB in June and the latest experiments ending MCB in February (table S1).

Walker circulation strength index

We use Walker cell strength as a proxy for the atmospheric component of the Bjerknes feedback (47) due to the tight coupling between the Walker circulation and equatorial Pacific SSTs. The Walker circulation strength index (60, 61) is defined as the difference in sea level pressure between the Indian Ocean/west Pacific (80°E to 160°E, 5°S to 5°N) and central/eastern Pacific (160°W to 80°W, 5°S to 5°N) regions (Fig. 3 and fig. S11A). A high Walker circulation strength index amplifies positive Bjerknes feedbacks, while a low Walker strength index weakens Bjerknes feedbacks.

Thermocline slope index

We approximate the oceanic component of the Bjerknes feedback using the thermocline depth approximated as the depth of the ensemble mean DJF 20°C isotherm (Z20) over the equatorial Pacific (120°E to 80°W, 2°S to 2°N) following (46) (Fig. 3). We calculate the thermocline slope index (fig. S11B) as the difference in Z20 anomalies (relative to the reference SMYLE climatology) between the western (160°E to 150°W, 2°S to 2°N) and eastern Pacific (90°W to 140°W, 2°S to 2°N) (62). A steeper thermocline slope is associated with neutral ENSO/La Niña conditions, while a shoaled thermocline is indicative of El Niño conditions.

Simplified heat budget analysis

While a formal heat budget analysis is beyond the scope of this proof-of-concept study, we conduct a simplified analysis comparing the relative contribution of net surface heat flux and upper ocean divergence to the MCB cooling effect on ENSO. We calculate the surface heat flux response by integrating the surface heat flux anomaly over time from June to February and space within the Niño 3.4 region (120°W to 170°W, 5°S to 5°N). We calculate the upper ocean divergence by first multiplying potential temperature by the specific heat capacity (3990 J/kg/°C) and density of seawater (1026 kg/m³) to obtain total heat. We then integrate the total heat over depth in the upper 100 m and area within the Niño 3.4 region and taking the mean from June to February to get upper ocean divergence. Comparing these two quantities shows the relative contribution of surface heat flux changes and ocean dynamics to the Niño 3.4 cooling response to MCB.

E-index and C-index ENSO metrics

The E-index and C-index are common ENSO metrics (54) based on the two leading modes of SST anomalies in the tropical Pacific. A positive E-index is typically associated with eastern Pacific El Niño events, while a negative C-index indicates a central Pacific La Niña event.

We calculate the E-index (E = [PC1 − PC2]/√2) using linearly detrended SST anomalies in the tropical Pacific (20°S to 20°N and 140°E to 80°W) referenced to the 1970–2014 historical climatological monthly means from the SMYLE Forced Ocean-Sea Ice (FOSI) configuration of CESM2 (31). We use SSTs from SMYLE-FOSI as the control simulations for this analysis instead of the SMYLE hindcast output to capture the “true” internal variability of tropical Pacific SSTs for the EOF analysis because the SMYLE hindcast artificially deflates variability near the initialization points. The MCB experiments are bias-corrected to align with SMYLE-FOSI by calculating the absolute difference between the MCB and SMYLE control SSTs and adding the response to SMYLE-FOSI. The detrended SST anomaly time series for SMYLE-FOSI and the corrected MCB simulations are then averaged over DJF and concatenated to calculate the DJF EOFs and E-index for each year (fig. S19). The C-index (C = [PC1 + PC2]/ √2) is calculated in a similar fashion, but as it pertains to La Niña events, we do not include it further in the analysis to estimate economic changes during El Niño.

Funding:

This work was supported by the NSF NCAR Early Career Faculty Innovator grant cooperative agreement no. 1755088 (J.S.W. and K.R.); NSF NCAR grant cooperative agreement no. 1852977 (J.T.F., N.R., and C.-C.C.); National Defense Science and Engineering Graduate Fellowship (J.S.W.); Achievement Rewards for College Scientists Foundation Scholarship (J.S.W.); National Aeronautics and Space Administration Award 80NSSC21K1191 (J.T.F.); National Aeronautics and Space Administration Award 80NSSC17K0565 (J.T.F.); National Aeronautics and Space Administration Award 80NSSC22K0046 (J.T.F.); National Science Foundation Award 2103843 (J.T.F.); US Department of Energy, Office of Science, Office of Biological and Environmental Research, Regional and Global Model Analysis component of the Earth and Environmental System Modeling Program under award number DE-SC0022070 (N.R.); and National Oceanic and Atmospheric Administration Earth’s Radiation Budget Grant NA22OAR4310481 (C.-C.C.).

Author contributions:

Conceptualization: J.T.F. and K.R. Data curation: N.R. Methodology: J.S.W., J.T.F., N.R., C.-C.C., and K.R. Investigation: J.S.W., N.R., and K.R. Formal analysis: J.S.W. Funding acquisition: K.R. Project administration: K.R. Resources: N.R. and K.R. Software: J.S.W., N.R., and C.-C.C. Supervision: K.R. Validation: N.R. Visualization: J.S.W. Writing—original draft: J.S.W. and K.R. Writing—review and editing: J.S.W., J.T.F., N.R., and K.R.

Competing interests:

The authors declare that they have no competing interests.

Data, code, and materials availability:

All data and code needed to evaluate and reproduce the results in the paper are present in the paper and/or the Supplementary Materials and are available at https://doi.org/10.5281/zenodo.20712074. The CESM2 SMYLE MCB simulations generated for this study are available at https://doi.org/10.5065/DGNA-PJ49. This study did not generate new materials.

This PDF file includes: