Seasonal advance of intense tropical cyclones in a warming climate

The study investigates whether and how the seasonal cycle of intense tropical cyclones (TCs; lifetime maximum winds >110 kt) has changed in recent decades under anthropogenic warming. TCs are among the most damaging natural hazards. Although trends in total global TC numbers are debated, increases in intense TCs and poleward migrations of TC activity have been reported. Intense TCs typically peak in autumn due to ocean heat availability, while other high-impact weather (e.g., monsoon-driven extreme rainfall) often peaks in summer. Determining if intense TCs are occurring earlier in the year is important for hazard assessment and potential compounding with other extremes. The authors hypothesize that intense TCs have undergone a seasonal advance linked to earlier rapid intensification (RI) facilitated by earlier-onset favourable oceanic conditions in a warming climate.

Prior work has documented: (1) global warming of about 1.0 °C above pre-industrial levels, mostly since mid-20th century; (2) mixed evidence on annual global TC counts but increases in intense TCs; (3) poleward migration of TC activity, especially in the western North Pacific, with implications for seasonality; (4) challenges for high-resolution models to represent intense TCs, yet statistical evidence suggests greater sensitivity of intense TCs to anthropogenic warming; and (5) earlier onset of Atlantic TC formation dates without shifts in accumulated cyclone energy milestones. The bimodal distribution of TC lifetime maximum intensity is strongly tied to RI; most intense TCs undergo at least one RI. This literature motivates focusing on seasonality of intense TCs and RI under warming.

Data and definitions: Intense TC occurrence time is defined as the date when a TC first reaches its lifetime maximum intensity. Primary TC data are from ADT–HURSAT (1981–2017). Best-track datasets (1981–2021) are also analyzed for robustness. Seasons are defined as June–November (NH active season) and December–April (SH active season). Early season is June–August (NH) and December–February (SH); late season is September–November (NH) and March–April (SH). Trend estimation and spatial analysis: Monthly counts of intense TCs and their linear trends (per decade) are computed. Interseasonal difference trends (early minus late season) are mapped on 5°×5° grids to reveal spatial patterns. Median occurrence date time series during the active season are constructed, and linear trends (days per decade) are estimated with 95% confidence intervals. Statistical significance is assessed primarily via non-parametric tests (e.g., Mann–Kendall), with additional tests for robustness. Intensity stratification: Shifting rates of occurrence dates are computed for TCs stratified by peak intensity thresholds to compare intense vs less-intense storms. Rapid intensification (RI) analysis: RI events are defined as ≥35 kt increase in 24 hours (~95th percentile of 24-h intensity change). Seasonal cycles and trends of RI occurrence dates are computed, and their relationship to intense TC occurrence dates is quantified via correlation. Environmental factors: Oceanic and atmospheric conditions favoring RI are analyzed: potential intensity (PI) and ocean heat content (OHC) as oceanic drivers; relative humidity (RH) and vertical wind shear (VWS) as atmospheric drivers. PI is computed from oceanic variables (ECMWF) and atmospheric variables (ERA5). OHC is from ECMWF. The fractional tropical area covered by high PI and high OHC (upper percentiles indicative of favorability for RI) is calculated monthly, and linear trends in these fractions are assessed during 1981–2017. RH and VWS seasonal cycles and trends are similarly examined across three atmospheric reanalyses. Climate model detection and attribution: CMIP6 historical simulations (multimodel mean) are used to detect changes in seasonal cycles of favorable oceanic conditions (fractional area with high PI and OHC) over 1981–2014. CESM2 large ensemble is used to suppress internal variability. DAMIP single-forcing experiments are analyzed to attribute changes to greenhouse gases (GHG), natural forcing, and anthropogenic aerosols, isolating their contributions to seasonal shifts. Projections under high-emission scenarios are examined for end-of-century amplification of seasonal advances. Impact assessment: To explore implications for compounding hazards, extreme rainfall events and those attributable to intense TCs are analyzed for South China and the Gulf of Mexico, focusing on changes between July–September relative to the typical bimodal precipitation climatology in South China and TC-affected rainfall in the Gulf of Mexico. Persistent rainfall event counts are also examined.

- Observational detection of seasonal advance: The median occurrence date of intense TCs shifted earlier by 3.7 days per decade in the Northern Hemisphere and 3.2 days per decade in the Southern Hemisphere (ADT–HURSAT, 1981–2017), significant at the 95% level. Best-track analyses corroborate earlier shifts (NH ~2.3–2.7 days/decade; SH ~1.8–4.1 days/decade). - Spatial pattern: Early-season increases and late-season decreases in intense TC occurrence are found across most tropical oceans, with strong earlier-shifting trends in the western North Pacific, eastern North Pacific east of 130°W, Gulf of Mexico, western North Atlantic, South Indian (west of 80°E), near northern Australia, and South Pacific east of 180°. Some regions (central Pacific, NE coast of Australia) show later shifts. - Basin details: The western North Pacific shows the largest advance (~8.1 days per decade). The North Atlantic exhibits large advances in the Gulf of Mexico/western basin and little change in the eastern basin. - Intensity dependence: Earlier seasonal shifts are significant for intense TCs but not for less-intense TCs or when all TCs are pooled, indicating the phenomenon is concentrated in the most intense storms. - Rapid intensification linkage: RI events also shift earlier (NH 3.6 days/decade; SH 4.1 days/decade), with high correlation between annual median occurrence dates of intense TCs and RI (r=0.88 NH; r=0.79 SH), implicating RI timing as a key driver of intense TC seasonality changes. - Environmental drivers: The fractional area of the tropics with high PI and high OHC increases in the early season and decreases in the late season, mirroring the RI and intense TC shifts. Atmospheric RH and VWS show no significant seasonal-cycle trend changes across reanalyses, suggesting oceanic factors dominate. - Model detection: CMIP6 historical simulations and the CESM2 large ensemble detect an earlier onset of favorable oceanic conditions (high PI/OHC), consistent with observations. - Attribution: DAMIP experiments indicate GHG forcing produces the early-season increases and late-season decreases in favorable oceanic conditions. Natural forcing shows no significant effect; anthropogenic aerosols modestly offset GHG-induced earlier shifts in specific late-season months/hemispheres. Overall, GHGs are the primary driver. - Future projection: The earlier seasonal shift of favorable oceanic conditions is projected to amplify by late century under high emissions. - Impacts/overlap with rainfall extremes: In South China, extreme rainfall increases during July–September (the inter-peak gap between June monsoon and October TC peaks), with intense TC-induced events accounting for almost all of the July increase and about half of August’s increase. In the Gulf of Mexico, July–September extreme rainfall increases are also tied to intense TCs. Persistent rainfall events increase alongside earlier TC-induced extremes, implying heightened compound risk.

The findings demonstrate that the seasonal cycle of intense TCs has advanced since the 1980s, largely due to an earlier onset of RI events. This addresses the research question by linking observed changes in intense TC timing to physical mechanisms—seasonally dependent ocean warming that boosts PI and OHC earlier in the year. The lack of significant changes in RH and VWS underscores the dominant role of oceanic preconditioning. Attribution analyses identify GHG forcing as the primary driver, strengthening the case for anthropogenic influence. The significance lies in risk management: as intense TCs occur earlier, their temporal separation from other summer-peaking hazards (e.g., monsoonal extreme rainfall) diminishes, elevating the probability of compound events. The western North Pacific shows the largest advances, with concrete manifestations in increased mid-summer extreme rainfall in South China, much of it intense-TC-driven, and similar signals in the Gulf of Mexico. These shifts challenge existing preparedness, resource allocation, and infrastructure planning predicated on historical seasonality.

This study provides robust detection and attribution of a seasonal advance in intense tropical cyclones across most basins since the 1980s, quantified at roughly 3–4 days per decade hemispherically and up to ~8 days per decade in the western North Pacific. The advance is tightly coupled to earlier rapid intensification, driven by earlier-onset favorable oceanic conditions (higher PI and OHC), which are primarily forced by greenhouse gas increases. Atmospheric factors (RH, VWS) show little contribution to the seasonal shift. The earlier timing heightens overlaps with summer-peaking extreme rainfall, increasing compound risk and persistent rainfall events. Future research should: (1) quantitatively assess joint probabilities and impacts of compound events involving intense TCs and other extremes; (2) improve high-resolution modelling of intense TCs and RI to refine projections of seasonal shifts; (3) evaluate regional socio-economic vulnerabilities to earlier intense TC occurrences; and (4) explore adaptation strategies and early-warning systems that account for changing seasonality.

- Data and detection: ADT–HURSAT and best-track intensity estimates carry uncertainties; some hemisphere/basin trends have wide confidence intervals (e.g., SH). The eastern North Pacific shows non-significant shifts, indicating regional variability. - Process representation: Atmospheric reanalysis products may have limitations detecting subtle seasonal changes in RH and VWS. - Modelling: Coarse-resolution climate models have difficulty explicitly resolving intense TCs and RI; inferences rely on environmental proxies (PI, OHC) rather than simulated storm statistics. - Attribution nuances: Aerosol effects and internal variability can modulate seasonal signals regionally; while large ensembles mitigate variability, residual uncertainties remain.