Global warming changes tropical cyclone translation speed

Tropical cyclones are among the most intense weather systems, and understanding how their characteristics change under global warming is crucial. Translation speed is a key metric because slower-moving TCs prolong hazardous conditions such as heavy rainfall and strong winds. With anticipated increases in TC-associated precipitation in a warmer climate, recent studies have debated historical and future changes in TC translation speed. Kossin (2018) reported a ~10% global slowdown of TC translation speed from 1949 to 2016 based on best-track observations, with the largest slowdown in the western North Pacific. Moon et al. (2019) challenged this, arguing that pre-satellite inhomogeneities (missing weaker, over-ocean, slower TCs) likely produced an artificial slowdown. Because observational homogeneity before the satellite era cannot be guaranteed, this study addresses two questions using numerical simulations: (1) whether there is a decreasing trend in TC translation speed during 1951–2011, and (2) how translation speed might change in a warmer future climate. The authors find no historical slowdown and project that global mean TC translation speed could increase under warming due to a poleward shift in TC occurrence, despite local slowdowns in the extratropics.

Prior literature presents conflicting views on TC translation speed trends and future changes. Kossin (2018) inferred a global slowdown in observed TC translation speed since mid-20th century, while Moon et al. (2019) attributed this to pre-satellite observational inhomogeneities that under-detected slow, weak, over-ocean TCs. Broader modeling studies show uncertainties in projections of TC characteristics including frequency, intensity, tracks, and latitude of occurrence. Several modeling efforts indicate decreased TC frequency near ~15° latitude and increases at higher latitudes, implying a poleward shift in relative TC frequency; others find basin-dependent or inconsistent track shifts, and some project increased global or regional TC frequency or intensity under warming. These uncertainties underline the need to use controlled, high-resolution ensemble simulations to assess translation speed changes and to consider both dynamical steering (e.g., changes in westerlies) and latitudinal distribution shifts when interpreting mean translation speed metrics.

The study uses high-resolution large-ensemble atmospheric general circulation model (AGCM) simulations with MRI-AGCM3.2 (60 km horizontal grid; same as MRI-AGCM3.2H in CMIP5). Historical simulations: forced with observed monthly SST and sea-ice from COBE-SST; period 1951–2010; 100-member ensemble using differing initial conditions and small SST perturbations based on analysis error; greenhouse gases set to observed annual global means. Future simulations: a pseudo–global warming experiment assuming a +4 K global mean surface air temperature relative to pre-industrial (consistent with late-21st-century RCP8.5). Six CMIP5 SST warming patterns were added to observed 1951–2010 SST, and for each pattern a 15-member ensemble was run (total 90 members), with the warming amplitude held constant in time. TC detection/tracking: applied to model output using criteria on sea level pressure, 850 hPa relative vorticity, 850 hPa and surface winds, vertical wind shear (850–300 hPa), warm-core temperature, and duration. Thresholds were tuned so the global number of TC geneses in current-climate simulations matches observations. The method shows comparable detection to other explicit schemes. Translation speed calculation: TC center positions at 6-hour intervals were used; great-circle distances between successive positions were computed, and per-member annual average translation speeds were derived. Ensemble means provided annual-mean translation speed. Frequencies: absolute frequency counted 6-hourly TC positions within specified latitude or speed bins; relative frequency computed against total counts. Observations: IBTRACS best-track data used for evaluation, focusing on 1982–2011 when global geostationary satellite coverage ensures higher homogeneity. Comparisons were performed globally, by hemisphere, and by basin; latitudinal distributions were assessed in 5° bins from 0° to 40°. Statistical significance assessed with two-tailed Student’s t tests.

• Model–observation agreement (1982–2011): Simulations reproduce the global annual-mean TC translation speed with both observations and simulations averaging 17.6 km/h in the post-geostationary era. Latitudinal distributions of translation speed are also reasonably reproduced globally and in the western North Pacific, though some basin-specific discrepancies exist (e.g., eastern North Pacific) and model year-to-year variability is smoother.
• No historical slowdown (1951–2011): Simulations show no decreasing trend in global or hemispheric TC translation speed, nor in individual basins. Observed speeds exceed simulated ones in the pre-satellite era, and latitudinal distributions differ, consistent with observational inhomogeneities driving the slowdown reported by Kossin (2018).
• Future global mean increase: Comparing current (1951–2011) to future (2051–2110, +4 K) climates, the global annual-mean translation speed increases from 17.5 to 18.0 km/h (p < 0.001). Hemispheric and basin means: Northern Hemisphere 18.3→18.6 (p < 0.001); Southern Hemisphere 16.0→16.3 (p < 0.001); North Atlantic 22.1→22.6 (p < 0.001); Western North Pacific 18.2→18.4 (p = 0.017); Eastern North Pacific 18.5→18.7 (p = 0.305, not significant); Northern Indian 13.8→13.7 (p = 0.106, not significant); Southern Indian 15.9→15.8 (p = 0.140, not significant); South Pacific 16.1→17.0 (p < 0.001).
• Extratropical slowdown but poleward shift in occurrence: In a warmer climate, weakened midlatitude westerlies reduce steering flow, decreasing average translation speed in the extratropics. Simultaneously, absolute TC frequency decreases in the tropics/subtropics and relative frequency increases at higher latitudes. Since extratropical TCs typically move faster, this poleward shift in relative frequency compensates for local slowdowns, yielding a higher global mean speed.
• Speed histograms by latitude: At 0–10° latitude, relative frequency of speeds ≤20 km/h increases in the future but absolute counts do not necessarily rise. At 30–40°, both absolute and relative frequencies of speeds ≤30 km/h increase and >30 km/h decrease, indicating a slowdown at higher latitudes across many basins.

The findings address the study’s core questions. First, the absence of a simulated slowdown in 1951–2011, along with discrepancies between observations and simulations in the pre-satellite era, supports the view that previously reported historical slowdowns likely reflect observational inhomogeneities rather than a robust climate signal. Second, projections under strong warming indicate that although extratropical TC translation speeds decrease due to weakened westerlies, the relative poleward shift in TC occurrence increases the contribution of faster-moving extratropical TCs to the global mean, producing a net global mean increase. This emphasizes that mean translation speed is sensitive to latitudinal TC distribution. The results are broadly consistent with other modeling studies that show decreased tropical/subtropical TC frequency and increased higher-latitude relative frequency, though literature also documents significant inter-model spread in track changes, frequency, and intensity responses. Consequently, while the global mean may increase, regional impacts could differ: translation speeds at higher latitudes likely decrease, potentially lengthening local hazard durations despite global mean increases. Further, model resolution and lack of ocean coupling may influence intensity distributions and track behaviors, highlighting the need for continued investigation with coupled, higher-resolution models to refine projections and reduce uncertainties.

Two key conclusions emerge. (1) Historical large-ensemble simulations show no significant decrease in TC translation speed over 1951–2011, suggesting that reported observational slowdowns likely stem from data inhomogeneities in the pre-satellite era. (2) Under a +4 K warming scenario, the global annual-mean TC translation speed is projected to increase modestly, driven by a poleward shift in TC relative frequency that offsets local extratropical slowdowns due to weakened westerlies. For risk assessment, mean translation speed is a limited metric; local latitudinal changes and full speed distributions matter more for impacts. Future research should employ finer-resolution, ocean-coupled models, examine basin-specific track changes, and better constrain uncertainties in TC frequency and poleward shifts to improve confidence in regional translation speed projections.

• Observational inhomogeneity prior to the satellite era complicates trend detection; comparisons rely mainly on 1982–2011 for robust evaluation.
• Model resolution (60 km) may not fully capture intensity distributions (e.g., bimodality) and inner-core dynamics that could influence tracks and speeds.
• The AGCM is atmosphere-only; absence of air–sea coupling may affect TC intensity evolution and tracks, potentially altering relative frequencies by latitude.
• Ensemble averaging smooths interannual variability, reducing standard deviation relative to observations in some basins.
• Basin-specific discrepancies remain (e.g., eastern North Pacific translation speeds), and multi-model studies show substantial spread in projected track shifts and frequencies, limiting generalizability of some conclusions.
• Future experiment uses a fixed +4 K warming amplitude and idealized SST warming patterns, which may not capture transient evolution or full coupled feedbacks.