Warming-induced contraction of tropical convection delays and reduces tropical cyclone formation

Projected changes in global tropical cyclone (TC) frequency remain highly uncertain, with different modeling approaches and parameter choices yielding diverging outcomes. While many global climate models project a decrease in global TC frequency, some statistical–dynamical downscaling approaches suggest an increase when assuming unchanged statistics of convective disturbances. Several theoretical frameworks emphasize the role of atmospheric convection: (1) reduced time-mean upward mass flux and possible increases in upward mass flux per TC may hinder genesis; (2) increased tropospheric saturation deficit with warming dries the environment, slows moistening, and facilitates convective ventilation detrimental to TC formation; and (3) ocean warming may enhance convective activity and expand latitudes favorable for TCs, potentially increasing frequency. Another line of research uses Hadley circulation changes—its weakening and poleward expansion under warming—to qualitatively explain TC changes. However, prior work has paid limited attention to the seasonal migration and future changes of tropical convection (the ITCZ). This study investigates how a warming-induced contraction of tropical convection—stronger equatorial and weaker off-equatorial convection—affects the timing and frequency of TC genesis, aiming to reconcile projections by linking TC changes to robust, large-scale convection responses under anthropogenic warming.

The paper synthesizes multiple strands of prior research on TC–climate relationships centered on convection and large-scale circulation: (1) Hydrological-cycle and vertical mass flux theories propose that weaker tropical-mean ascent and altered mass flux per storm reduce TC genesis. (2) Moisture/ventilation perspectives emphasize increased saturation deficit and enhanced ventilation with warming, which dry and shear the environment, suppressing TC development. (3) Alternative views suggest warming increases convective activity, including TC-seeding disturbances, and broadens the TC-permitting latitude band. (4) Hadley circulation–based explanations focus on equatorial ascent and subtropical descent, with prior projections of Hadley weakening and poleward expansion used to qualitatively infer TC changes. Related work has also documented warming-induced narrowing (contraction) of tropical ascent and ITCZ changes in annual and seasonal means, potentially arising from enhanced equatorial warming and cloud-radiative feedbacks that sharpen meridional moist static energy gradients. Observational analyses have found signals of seasonal delays in tropical rainfall and ITCZ contraction in the Pacific. Yet, the connections between these large-scale changes and basin-scale TC seasonality and frequency have been underexplored, motivating the present study.

The study combines reanalysis, observed TC data, CMIP6 model output, and high-resolution large-ensemble simulations. Observations: ERA5 monthly fields at 0.25° resolution represent tropical convection; TC best tracks are from IBTRACS (US-sourced subset). Period emphasized is 1981–2010. CMIP6: A subset of ~1° models (e.g., BCC-CSM2-MR, CMCC-CM2-SR5, CNRM-CM6-1-HR, EC-Earth3, GFDL-CM4, HadGEM3-GC31-MM, MIROC6, MPI-ESM1-2-HR, MRI-ESM2-0, CESM2, TaiESM1) is analyzed using AMIP (reference) and SSP5-8.5 (future) to assess robustness of convection contraction. High-resolution large ensemble: Uses d4PDF with MRI-AGCM3.2H (~60 km). Historical simulation (1951–2010) is a 100-member ensemble driven by observed forcings with perturbed initial conditions and SSTs (random perturbations up to 30% of observed interannual SST variability). Future warming simulations apply greenhouse gas forcing corresponding to 2090 in RCP8.5, and SST patterns from six CMIP5 models (HadGEM2-AO, MPI-ESM-MR, MRI-CGCM3, NCAR-CCSM4, GFDL-CM3, MIROC5; denoted HA, MP, MR, CC, GF, MI) scaled to yield +4 K relative to preindustrial. Each pattern drives a 15-member ensemble (total 90). Analyses focus on 30-yr climatologies: historical (1981–2010) and warming (2081–2110). Significance is assessed via Student’s t-test using samples from the 100-member historical and 90-member warming ensembles. ITCZ latitude: Defined monthly as the centroid latitude of the top-quartile convection (within 25°S–25°N), weighting by intensity and grid-cell area to robustly track ITCZ location across seasons. TC detection/tracking: For the large ensemble, TC tracks from DIAS (Yoshida et al.) are used. Detection considers 850-hPa relative vorticity and wind maxima, warm-core structure aloft, vertical wind profile criteria, and special screening in the North Indian Ocean to exclude monsoon depressions; a minimum lifetime of 36 h is required. TC genesis is defined as the first point of each detected track. Seasonal cycle analytics: Genesis times are binned into 73 pentads (5-day bins) per year for each basin, normalized by basin total counts; TC-suppressed periods are bins with <1% of a basin’s TCs. Peak, quiet, and transition seasons are analyzed to quantify seasonal changes and season length (using a 0.01 probability threshold). Convection and environmental diagnostics: Tropical convection is represented by 500-hPa vertical motion (ω), with negative ω indicating ascent. The Seed Propensity Index (SPI) is used as a proxy for the frequency of rotating convective disturbances (“seeds”), defined as S = 1/(1 + Z^(−α)), with Z = (f + ζ)/(sqrt(β^2) + 0.7 ω), where f is the Coriolis parameter, β its meridional gradient, ζ the 850-hPa climatological relative vorticity; empirical parameters U = 20 m s⁻1 and α = 0.69 are used. For zonal-mean comparisons, zeroing SPI over subsidence regions is omitted to avoid artifacts from regional shifts. Environmental factors include 600-hPa relative humidity and vertical wind shear (200–850 hPa). Relationships between changes in equatorial (0–7.5° latitude) and off-equatorial (8–20°) convection and global TC frequency are assessed across ensemble members and six warming patterns for peak seasons in each hemisphere.

• Warming-induced contraction of tropical convection is robust in high-resolution large-ensemble simulations and CMIP6: equatorial ascent strengthens while off-equatorial ascent on ITCZ flanks weakens, especially in the Northern Hemisphere. The contraction is most evident during the transition from boreal spring to summer and delays the poleward migration of the ITCZ.
• TC genesis frequency decreases with warming, with a strong seasonal dependence. Decreases are relatively larger in fractional terms during transition months (e.g., May–June in the Northern Hemisphere), indicating season shortening, and large in magnitude near peak seasons close to the ITCZ, linking TC decreases to convection contraction.
• Shortened and delayed TC seasons across basins under +4 K warming: defining season length with a 0.01 probability threshold, seasons shorten by about 5–10%. Early-season delays are widespread (around June NH; December SH), generally under 15 days in all basins except the Northeastern/East Pacific. Transition periods account for a large share of projected TC decreases: from −13% (North Atlantic) to −200% (East Pacific) across basins, with a six-basin average of −75% of the total decrease.
• Suppressed peak-season activity near the ITCZ: SPI (proxy for rotating convective disturbances) decreases near the NH ITCZ and on the poleward flank of the SH ITCZ. Weaker off-equatorial convection reduces low-level vorticity generation via column stretching. Contraction also induces TC-unfavorable environments off the ITCZ: mid-tropospheric drying and enhanced vertical wind shear.
• Sensitivity to warming patterns and convection changes at seasonal peaks: Across six SST-warming patterns, the sign of global TC frequency change is robustly negative, but magnitude correlates with convection changes. Correlations between TC frequency changes and convection changes are: NH equatorial (0–7.5°N) r = 0.35; NH off-equatorial (8–20°N) r = 0.30; SH equatorial (0–7.5°S) r = −0.24; SH off-equatorial (8–20°S) r = 0.71. The SH result is strongly influenced by South Indian and South Pacific basins, including an equatorward shift of the South Pacific Convergence Zone.
• The contraction framework helps explain ensemble spread in projected global TC frequency by linking variations in equatorial/off-equatorial convection (and associated warming patterns, e.g., El Niño-like SST changes) to TC responses.

The study addresses the central question of how anthropogenic warming affects TC frequency and seasonality by identifying a robust, large-scale mechanism: contraction of tropical convection (stronger equatorial, weaker off-equatorial ascent). This contraction delays the seasonal ITCZ migration into the TC-favoring regime, shortening and delaying TC seasons, and suppresses peak-season TC formation by reducing seed propensity and making the environment drier with stronger shear in off-equatorial latitudes where TCs preferentially form. By partitioning convection changes into equatorial and off-equatorial components, the analysis shows that both components contribute to reduced TC activity, with statistically significant correlations across diverse warming patterns. The results synthesize prior hypotheses (weaker mass flux, increased saturation deficit/ventilation, fewer seeds) into a unified framework centered on tropical convection. Patterns of surface warming (e.g., El Niño-like SST anomalies) modulate the magnitude of contraction and, consequently, TC frequency changes, helping to explain inter-scenario ensemble spread. While the framework does not explicitly resolve TCs at the tropical–extratropical interface, it implies indirect influences via changes in upper-level outflow and subtropical westerlies. Overall, the contraction mechanism provides a physically grounded explanation for model-projected TC frequency declines and seasonal delays, clarifying how large-scale convection changes translate to TC climatology.

The paper demonstrates that anthropogenic warming induces a contraction of tropical convection—strengthening near the equator and weakening off-equator—that delays ITCZ poleward migration, shortens TC seasons by about 5–10% under +4 K, and reduces TC genesis, particularly near seasonal peaks and close to the ITCZ. Transition periods disproportionately contribute to the overall decrease in TC frequency (average −75% across six basins). Environmental changes accompanying contraction (reduced seed propensity, mid-tropospheric drying, enhanced shear) further hinder genesis. The study unifies disparate hypotheses about future TC changes under a single framework centered on tropical convection and shows that convection contraction and warming pattern differences explain a substantial portion of ensemble spread in projected TC frequency. Future work should refine theoretical understanding, improve climate-model fidelity (especially tropical Pacific warming patterns and Southern Hemisphere convection), and consider additional processes (planetary waves, midlatitude circulations) and tropical–extratropical interactions to enhance confidence in TC risk projections.

• Model dependence and data availability: Conclusions rely largely on the d4PDF high-resolution large ensemble and available CMIP6 models; attempts to include more high-resolution models were limited by missing variables.
• Southern Hemisphere uncertainty: Large-ensemble projections show muted SH convection changes relative to CMIP6, partly due to specific warming patterns (CC, GF) with weaker El Niño-like warming; SH results thus have greater uncertainty.
• Seasonality and tracking biases: Differences between automated model tracking and human-curated observational tracks complicate direct comparison of genesis locations and season metrics. The large-ensemble seasonal cycle has known biases in some basins.
• Forcing magnitude and other anthropogenic/natural factors: The primary analysis uses +4 K warming scenarios; present-day observed changes (e.g., Atlantic season lengthening) may be influenced by aerosols or natural variability, and may not align with +4 K projections.
• Proxy measures and computational compromises: SPI is used as a proxy for rotating convective disturbances; simplifications (e.g., omitting zeroing in subsidence regions; using simple shear and mid-level humidity instead of a full ventilation index) may affect quantitative details though not the qualitative conclusions.
• Scope: The framework does not explicitly address TCs at the tropical–extratropical interface, which may change under warming and influence high-latitude TC risk.