Tropical cyclones (TCs), especially those undergoing rapid intensification (RI), pose a significant threat due to their destructive potential and the inherent difficulty in forecasting them accurately. RI, defined as a rapid increase in TC intensity, transforms a relatively predictable hazard into an unpredictable one, severely limiting the lead time for effective warnings. While recent years have shown some improvement in RI forecasting in certain basins (East Pacific and North Atlantic), intensity forecast errors for RI events remain significantly higher compared to non-RI events. This forecasting challenge is further exacerbated by an observed upward trend in the proportion of storms experiencing RI in several basins. Future projections suggest an increased likelihood of RI, potentially making TC forecasting even more challenging. This study aims to address the knowledge gaps concerning the anthropogenic contribution to the observed changes in TC intensification rates and the influence of more favorable TC environments by explicitly comparing observational trends with climate model simulations. The challenge is disentangling the roles of multidecadal natural variability (e.g., Atlantic Multidecadal Variability (AMV) and Pacific Decadal Oscillation (PDO)) and anthropogenic climate change in driving these observed trends, particularly given the relatively short period with reliable TC intensity estimates. Previous work has attempted to isolate natural variability, but this study presents a more comprehensive analysis by employing a higher-resolution global climate model and a broader range of datasets. Understanding these driving forces is crucial for improving TC forecasting and developing effective mitigation strategies for coastal communities.

Previous research has documented upward trends in TC RI in various basins, using datasets such as IBTRACS and ADT-HURSAT. Studies have suggested links between rising sea surface temperatures (SSTs) and potential intensities (PIs), and more intense TCs and RI events. However, these studies lacked a comprehensive assessment of the anthropogenic contribution to these trends, and the role of changes in storm tracks and weather-scale variability. While some studies have examined basin-specific multidecadal variability, a robust global analysis incorporating both natural and anthropogenic forcing is lacking. The existing literature highlights the need for a more sophisticated investigation that can tease apart the relative contributions of natural variability and anthropogenic forcing in driving the observed increase in TC RI.

The study uses multiple datasets, including the International Best Track Archive for Climate Stewardship (IBTRACS) and the Advanced Dvorak Technique-Hurricane Satellite-B1 (ADT-HURSAT), spanning the period 1982-2017. ADT-HURSAT data, recently extended to cover a longer period, offers improved temporal consistency. The analysis focuses on the West Pacific, Atlantic, Australian, and global datasets due to consistent trends observed in these regions. To assess the role of natural climate variability, the authors use simulations from the high-resolution global climate model HiFLOR, specifically preindustrial control runs (1860CTL) with anthropogenic and natural forcing held constant at 1860 levels. Quantile delta mapping (QDM) is applied to correct for systematic errors in HiFLOR and to ensure realistic slopes in RI ratio. The observed RI ratio slopes (from IBTRACS and ADT-HURSAT) are then compared to the bias-corrected HiFLOR 1860CTL slopes over overlapping 36-year periods. Trends in the environmental favorability for TC intensification are examined using four environmental variables (vertically-averaged relative humidity (RH), vertical wind shear (SHR), SST, and PI) from ERA5 and MERRA-2 reanalyses. To isolate storm-ambient environments, the authors track storms, spectrally filter the environmental fields, and compute spatial averages within a 5° storm-centered radius. Critical environmental thresholds for RI are determined using a logit equation. The probability of RI is analyzed based on the number of critical thresholds met. Finally, the study examines the influence of anthropogenic forcing on tropical-mean environments by comparing multiple CMIP6 simulations with different forcing estimates (all-forcing, GHG-only, and natural-only). Linear trends in peak-season tropical-mean environmental variables are calculated and compared to observed trends from MERRA-2 and ERA5. Statistical tests (pairwise t-tests and Kolmogorov-Smirnov tests) are used to compare the slope distributions and assess the significance of the trends.

The study reveals a robust global increase in the proportion of RI events between 1982-2017 in the ADT-HURSAT dataset, significantly higher than the trends simulated by HiFLOR. This indicates that the observed increase is unlikely to be solely due to natural variability. In all analyzed basins, the IBTRACS RI ratio slopes were above the 99th percentile of the HiFLOR 1860CTL slopes. Further analysis of storm-local environments shows a significant increase in the proportion of cases meeting multiple critical environmental thresholds (favorable for RI) and a decrease in cases meeting few thresholds. The Atlantic basin exhibited the most substantial changes, with a more than doubling in the annual proportion of cases satisfying 3 or 4 thresholds. Analysis of tropical-mean environments in MERRA-2 and ERA5 reanalyses shows significant upward trends for PI and SST in both hemispheres, suggesting improvements in thermodynamic conditions conducive to RI. CMIP6 simulations support the observed trends, with significant differences between all-forcing and natural-only simulations, indicating the effect of anthropogenic forcing. However, there is a discrepancy between the PI trends in the CMIP6 AllForc ensemble and the reanalyses, potentially linked to differences in the vertical structure of temperature changes. The observed trend in PI is more than three times larger than the mean slope of the AllForc ensemble, suggesting a substantial but unexplained component in the observed trend.

The findings strongly suggest that anthropogenic forcing has contributed to the observed increase in TC RI. The consistent increase in the probability of RI across multiple basins and the global scale is compelling evidence. The observed changes in storm-local and tropical-mean environments (higher SSTs, PIs, and lower wind shears), coupled with the CMIP6 simulation results, point toward a substantial anthropogenic influence. The discrepancy between observed and modeled PI trends highlights the need for further investigation into model biases and the accurate representation of upper tropospheric warming. The results imply that track variability is unlikely to counteract the intensification effects of ongoing warming. This study underscores the importance of developing higher-resolution models capable of resolving mesoscale processes and accurately representing the complex interactions between TCs and their environment. The findings have important implications for improving TC forecasting and strengthening coastal resilience.

This study presents strong evidence that anthropogenic climate change has already contributed to the observed increase in rapidly intensifying hurricanes. The global increase in RI events, coupled with the observed shifts in favorable thermodynamic environments, and supported by climate model simulations, points to a substantial anthropogenic influence. The discrepancies between observed and modeled PI trends warrant further research to refine model representations of upper tropospheric warming. Improving coastal resilience is a crucial priority in light of the projected continued increase in hurricane intensity and RI frequency.

The study relies on a single bias-corrected climate model (HiFLOR) to estimate internal climate variability, although HiFLOR has demonstrated skill in simulating TCs. Furthermore, the model's horizontal resolution (0.25° × 0.25°) may not fully resolve small-scale processes associated with TC RI. The discrepancy between observed and modeled PI trends necessitates further investigation. Lastly, while the study focuses on a comprehensive analysis, the relatively short observational record for consistently measured TC intensity presents a limitation for long-term trend assessment.