Anthropogenic forcings reverse a simulated multi-century naturally-forced Northern Hemisphere Hadley cell intensification

This study investigates how unprecedented the recent and projected weakening of the Northern Hemisphere (NH) Hadley circulation (HC) due to anthropogenic emissions is when compared to changes driven by natural forcings over the last millennium (850–1849). The HC is a key component of the tropical–subtropical climate system, governing heat and moisture transport and the spatial distribution of precipitation. While models project a robust anthropogenic weakening of the NH HC by the end of the 21st century, it remains unclear how these changes compare to naturally forced variations in earlier centuries, especially given the limited direct wind observations in the past. The research aims to quantify the historical context of NH HC changes and to elucidate the physical mechanisms—particularly the roles of static stability and latent heating—that underlie naturally forced HC variability. Understanding these contrasts is critical for assessing human influence on atmospheric dynamics and improving future projections that often omit natural forcing variability.

Previous work shows that climate models project a weakening of the NH HC under anthropogenic warming, primarily linked to changes in atmospheric temperature structure and static stability. Observation-based reanalyses have sometimes suggested recent HC intensification, but these findings may be affected by artifacts related to latent heating and observational uncertainties, especially over oceans. Recent studies using sea-level pressure as a proxy indicate a recent NH HC weakening consistent with model projections and attributable to anthropogenic emissions. For past centuries, proxy-based reconstructions (e.g., the “hockey stick” for NH temperatures) demonstrate significant natural variability, including the Medieval Climate Anomaly (MCA) and Little Ice Age (LIA), associated with changes in volcanic, solar, and other natural forcings. Prior modeling work highlights unprecedented zonal wind shifts in the 20th century and underscores the role of external forcings in last-millennium climate variability, but a quantitative comparison of NH HC strength changes across natural and anthropogenic forcing regimes has been lacking.

• Datasets and experiments: The analysis combines nine CMIP5 models participating in PMIP3 last-millennium runs (850–1849), the 12-member CESM Last Millennium Ensemble (CESM-LME, 850–1919), CMIP5 historical (1850–2005) and RCP8.5 (2006–2100) runs, and a 12-member subset of the CESM Large Ensemble (CESM-LE, 1920–2100). Where available, a milder forcing scenario (RCP4.5) is also examined.
• Metric of Hadley cell strength: The NH HC strength is defined as ψ_max, the maximum of the annual-mean meridional mass streamfunction at 500 mb, computed from zonal-mean meridional winds. A spatially averaged HC strength metric (ψ_avg) was also assessed for robustness.
• Forced response vs internal variability: Ensemble means (across CESM members or CMIP5 models) are used to isolate the forced response. Spread across CESM members characterizes internal variability; spread across CMIP5 models reflects both internal variability and inter-model differences. Long preindustrial control runs (CMIP5: total ~4900 years; CESM: 850- and 1850-control runs) quantify internal variability (“noise”).
• Trend analyses and detection: Linear regression is applied to estimate long-term trends (including over 850–1849). Robustness is assessed with the non-parametric Mann–Kendall test and by varying start/end years. Signal-to-noise ratio (SNR) analyses determine time of emergence (ToE) of forced signals: for anthropogenic weakening, signals are ψ_max trends starting in 1970 of various lengths; noise is the distribution of same-length trends in 850–1849 ensemble means. ToE is identified when the signal exceeds 2 standard deviations of the last-millennium forced variability. For naturally forced strengthening, signals are trends from 850 to each year (until 1849), and noise is internal variability from control runs.
• Attribution: Anthropogenic contribution to recent weakening is assessed using CMIP5/CMIP6 natural-only historical runs (hist-nat) vs full historical runs. The last-millennium strengthening is attributed via CESM-LME single-forcing ensembles (volcanic, solar, greenhouse gases, orbital, and anthropogenic land-use/land-cover). Bootstrapping harmonizes ensemble sizes when computing forced responses.
• Mechanism diagnosis: The Kuo–Eliassen (KE) equation, an elliptic linear diagnostic framework, decomposes HC changes between MCA and LIA into contributions from diabatic heating (latent and radiative), eddy heat and momentum fluxes, zonal friction, and static stability (S^2). Consistency is checked by comparing ΔΨ_KE with ΔΨ_max. A vertically integrated moisture budget relates precipitation (and latent heating) changes to mean meridional circulation, mean moisture, eddy moisture fluxes, and evaporation, clarifying drivers of latent heating gradients over the HC ascending branch.
• Model availability: KE terms and diabatic components are fully available for CESM last-millennium runs, enabling process attribution there; key results are cross-checked in CMIP5 where possible.

• Unprecedented anthropogenic weakening: Ensemble means of CESM and CMIP5 show that recent and projected NH HC weakening under anthropogenic forcing is unprecedented relative to forced changes over the last millennium. Using SNR, the anthropogenically forced weakening signal emerges from last-millennium forced variability by the early 2010s.
• Reversal of a multi-century trend: Over 850–1849, the NH HC exhibits a statistically significant intensification: about 1.3 × 10^6 kg−1 yr−1 in CESM mean and 0.9 × 10^6 kg−1 yr−1 in CMIP5 mean. Nearly all individual CESM members and CMIP5 models show strengthening.
• Natural-forcing attribution and detection: Single-forcing CESM-LME analyses indicate that natural forcings (volcanic, solar, greenhouse gases in preindustrial context, orbital) collectively account for most of the last-millennium strengthening; anthropogenic land-use alone is insufficient. The naturally forced strengthening emerges from internal variability around 1600 in CESM mean and around 1400 in CMIP5 mean; emergence windows span roughly 1500–1700 (CESM members) and 1350–1750 (most CMIP5 models).
• Physical mechanisms: KE diagnostics identify two dominant contributors to the last-millennium HC intensification: (1) reduced static stability (S^2) due to tropospheric cooling with greater upper- than lower-tropospheric cooling in the tropics, and (2) an increased meridional gradient of latent heating over the NH HC ascending branch. Other terms (radiative heating, eddy heat/momentum fluxes, friction) have smaller and generally weakening effects.
• Moisture and latent heating budget: The enhanced meridional gradient of latent heating is linked to increased precipitation in the ascending branch from dynamical changes. Moisture budget decomposition shows that changes in mean meridional circulation are the primary contributor to precipitation increases; changes in mean moisture, eddy moisture fluxes, and evaporation play smaller or opposing roles.
• Consistency with future projections: Despite opposite signs of temperature change (cooling during MCA→LIA vs warming in the 21st century), latent heating gradients tend to intensify the NH HC in both periods, though via different mechanisms (dynamics dominate in the last millennium; thermodynamics dominate under warming).
• Minor ITCZ and SH impacts: The ITCZ position/width changes are small over the last millennium, and both hemispheric low latitudes cool similarly; natural forcings have minor effects on SH HC strength, consistent with small anthropogenic effects there in the 20th–21st centuries.
• Anthropogenic attribution for recent trend: Natural-only (hist-nat) simulations exhibit much smaller recent trends than full historical runs, supporting attribution of recent NH HC weakening to anthropogenic emissions.

The analysis directly addresses the question of how the recent and projected NH HC weakening compares to past, naturally forced variability. It demonstrates that anthropogenic emissions have produced a weakening that not only exceeds naturally forced variations of the last millennium but also reverses a multi-century naturally forced strengthening. The detection and attribution framework shows that natural forcings drove a robust intensification detectable above internal variability, while recent weakening is attributable to anthropogenic forcing. Mechanistic diagnostics connect last-millennium intensification to decreased static stability under tropospheric cooling and to dynamical increases in the meridional gradient of latent heating over the ascending branch. These findings underscore a profound anthropogenic influence on large-scale atmospheric dynamics. Given the HC’s role in shaping the tropical–subtropical hydrological cycle, these unprecedented dynamical changes likely propagate to significant regional climate impacts. The results also highlight that projections may be biased if future natural forcing variability (e.g., volcanic) is inadequately represented, emphasizing the need to integrate natural forcings into projection frameworks.

This work quantitatively situates the anthropogenic NH Hadley cell weakening within the context of last-millennium variability, showing it to be unprecedented and a reversal of a multi-century naturally forced strengthening. It attributes most of the historical strengthening to natural forcings and elucidates the governing processes—reduced static stability under cooling and enhanced meridional latent heating gradients driven by circulation changes. The study advances understanding of how external forcings shape large-scale tropical circulation and stresses that accurate projections require proper accounting of natural forcing variability. Future research should (1) improve representation and scenario coverage of natural forcings (especially volcanic) in projections, (2) further investigate causal feedbacks between HC intensity and latent heating gradients, (3) expand multi-model diagnostics with comprehensive output (e.g., KE terms) across ensembles, and (4) better reconcile observational constraints and proxies on historical circulation changes.

• Model dependence and output availability: Full KE diagnostics and diabatic components are available primarily in CESM for last-millennium runs; CMIP5 cross-checks are more limited. Inter-model configuration differences contribute to spread in magnitude and time of emergence.
• Causality: The KE framework is diagnostic; causal relationships and feedbacks (e.g., between HC strength and latent heating gradients) cannot be definitively inferred from these equations.
• Forcing uncertainties: Reconstructions of last-millennium forcings and future projections’ limited inclusion of natural forcings (notably volcanic) introduce uncertainty in both past attribution and future emergence times.
• Observational constraints: Direct historical wind observations are lacking; some observational precipitation datasets over oceans are uncertain, complicating validation of long-term circulation changes.
• Units/metrics: While ψ_max is a standard metric, different HC strength metrics can yield nuances; however, robustness tests indicate consistent main conclusions.