Effects of climate change on the movement of future landfalling Texas tropical cyclones

Texas has experienced several devastating tropical cyclones (TCs) since 2000 (e.g., Allison 2001, Ike 2008, Harvey 2017), with differing damage mechanisms (extreme rainfall and flooding versus wind and storm surge). TC impacts depend on multiple factors, including intensity, size, landfall angle, translation speed, and sea level; climate change influences these through warmer SSTs, increased atmospheric moisture, and altered large-scale circulation. While TC frequency and intensity have been extensively studied, recent investigations into historical and future changes in translation speed have yielded inconclusive results. This study asks: How will anthropogenic climate change influence the movement—especially translation speed—of future landfalling Texas TCs? Using large-ensemble and multi-model datasets, clustering of daily steering winds, and statistical–dynamical downscaling, the study evaluates changes between 1979–2005 and 2074–2100 (RCP8.5), focusing on June–September conditions and landfalling TCs near Texas. The authors find a robust increase in northward steering winds over Texas and corresponding increases in the likelihood of faster-moving landfalling TCs by late 21st century.

Prior work has documented changes in TC frequency and intensity and explored translation speed trends, with mixed conclusions. Kossin (2018) suggested a global slowdown (including over North Atlantic land areas), but subsequent studies raised concerns about data inhomogeneities and did not confirm a hemispheric slowdown. High-resolution AGCM studies (Yamaguchi et al. 2020; Zhang et al. 2020) project reduced translation speeds at higher latitudes associated with weakened westerlies, alongside shifts in TC occurrence. Regional pseudo-global warming simulations (Gutmann et al. 2018) reported mixed storm-specific changes (e.g., a slower Ike), but differ in scope and methodology from the present regional, multi-model, statistical approach centered on Texas landfalls. Broader literature on circulation responses highlights intensified and westward-shifted subtropical highs and a weakened North American monsoon in warming climates, which can modify steering flows over Texas.

- Periods and scenario: Current climate 1979–2005 and future climate 2074–2100 under RCP8.5. - Steering wind definition: Steering winds computed as 0.8×(850 mb winds) + 0.2×(200 mb winds) for both zonal (U) and meridional (V) components. - Datasets for large-scale circulation: Three large-ensemble fully coupled climate datasets: NCAR CESM Large Ensemble (LENS; ~1°; 40 members), MPI Grand Ensemble (MPI-GE; ~1.8°; 100 members), and GFDL-CM3 Large Ensemble (GFDL-LE; ~2°; 20 members). Additionally, 14 CMIP5 models (historical and RCP8.5) were analyzed; fields interpolated to 0.75° for multi-model mean. Variables: monthly U, V, Z500, sea-level pressure (monthly; daily U,V for some analyses). - Analysis of mean changes: June–September averages of steering winds and Z500; future minus current differences; statistical significance assessed (two-tailed t test at 95% for LENS; model agreement threshold for CMIP5). - Stationary-wave mechanism analysis: Decomposition of low- and upper-level stationary streamfunctions and sea-level pressure to interpret responses, referencing prior findings on subtropical high intensification and monsoon weakening. - Daily steering wind regime clustering: Self-organizing map (SOM) clustering applied to daily steering wind vectors (U,V) over a domain surrounding Texas, using LENS daily data for current and future periods combined. SOM configuration: MATLAB selforgmap, hextop topology, linkdist distance, 10 clusters (2×5 grid), 1000 ordering steps. Robustness: 50 repetitions with one-fifth subsampling; uncertainties in frequencies reported as standard errors. Changes decomposed into frequency and pattern components using defined equations. - Downscaling of TCs (CHAZ): The Columbia TC HAZard model generates synthetic TC genesis, tracks, and intensity using environmental inputs (PI, vertical wind shear, moisture metric CRH or SD, 850 mb vorticity, steering flow). Downscaled from six CMIP5 models for HIST (1981–2005) and RCP8.5 (2071–2099). Analysis subset: storms passing within 300 km of Houston. CHAZ track model advects seeds with steering flow plus beta drift; intensity via empirical regression plus stochastic term; landfall intensity via separate regression. Bias in HIST forward speed distributions addressed via two bias-correction methods (Gaussian parameter correction and quantile matching); corrections derived from HIST and applied to RCP8.5; conclusions tested for robustness. - Observational and reanalysis reference: IBTrACS v4 for observed tracks (1981–2018); NOAA CPC precipitation (0.5°) for rainfall composites; NCEP–DOE Reanalysis 2 for winds and Z500 (1979–2018).

- Robust northward steering wind response over Texas: All model groups project stronger June–September-averaged northward (meridional) steering winds over Texas by 2074–2100 relative to 1979–2005. The magnitude near Houston (~95°W, 30°N) is about 33% of the historical climatology. Eastward (zonal) steering winds weaken between 30°N–50°N, consistent with a poleward shift/weakening of midlatitude westerlies. - Mechanism: The enhanced northward steering over Texas aligns with an intensifying and westward-shifting North Atlantic subtropical high (low-level) and a weakening North American monsoon (upper-level), which constructively increase meridional steering across levels. - Daily regime changes: SOM clustering shows a ~7% increase in the frequency of regimes with northward steering (e.g., clusters C1, C4, C5) and a ~7% decrease in regimes with southward steering (C6, C9), with additional weakening of southward patterns. The aggregate across clusters reproduces the mean northward wind response. - Downscaled TC translation speeds near Houston (within 300 km): Probability density shifts toward faster motion under RCP8.5. Fast-moving TCs (≥20 km h⁻¹) increase from 31.4% to 37.6% (+6.2 percentage points), while slow-moving TCs (≤5 km h⁻¹) decrease from 20.4% to 16.6% (−3.8 percentage points), indicating a 10-percentage-point shift from slow to fast categories. The meridional component drives most of the shift: −9.8% from slow (|v| ≤5 km h⁻¹) toward fast northward (v ≥15 km h⁻¹); the zonal component shift is −2.9% from slow (|u| ≤5 km h⁻¹) toward fast westward (u ≥15 km h⁻¹). Results are consistent after bias correction. - No evidence of increased likelihood of slow-moving landfalling Texas TCs by late 21st century; instead, a higher probability of faster-moving landfalling TCs is projected.

The projected strengthening of northward steering winds over Texas, arising from changes in the subtropical high and weakened North American monsoon, implies faster northward translation of landfalling TCs near Texas. This regional response complements basin- or hemispheric-scale findings that emphasize slower motion at higher latitudes due to weakened westerlies; those do not contradict a Texas-specific increase in meridional steering at lower latitudes. Comparisons with prior literature indicate that reported historical global slowdowns (e.g., Kossin 2018) may be affected by data inhomogeneities and are not directly comparable in scope or region. High-latitude decreases in translation speed (Yamaguchi et al. 2020; Zhang et al. 2020) are consistent with weakened westerlies but focus on broader scales. Regional pseudo-global warming case studies (Gutmann et al. 2018) differ in approach and storm sampling. Importantly, faster-moving TCs do not inherently reduce risk; damage mechanisms can shift (e.g., surge and wind for faster-moving storms versus extreme rainfall for stalled ones). The findings underscore the need for region-specific assessments of TC movement under climate change and consideration of multiple hazard drivers for adaptation planning.

This study integrates large-ensemble and multi-model analyses, clustering of daily steering winds, and statistical–dynamical downscaling to assess future changes in the movement of landfalling Texas TCs. It finds a robust increase in June–September northward steering winds over Texas by late 21st century, driven by an intensified, westward-expanded Atlantic subtropical high and a weakened North American monsoon. Consistent with these circulation changes, downscaled simulations project a 10-percentage-point shift from slow-moving to fast-moving TCs near Houston, with increased frequency of fast (≥20 km h⁻¹) and decreased frequency of slow (≤5 km h⁻¹) storms. The results do not support an increased likelihood of slow-moving landfalling Texas TCs under RCP8.5 but instead indicate higher odds of faster-moving events. Future work should extend analyses to mid-century and alternative emissions scenarios (e.g., RCP4.5), further disentangle dynamic versus thermodynamic contributions, and incorporate other risk factors (intensity, size, sea level, moisture) to comprehensively evaluate regional TC hazards.

- Scenario and period focus: Results pertain to late 21st century under RCP8.5, potentially differing under other scenarios/timelines. - Model and internal variability: Regional circulation projections carry uncertainties from model biases and internal variability despite large ensembles. - Mechanistic diversity across models: Differences in Z500 responses among ensembles (e.g., presence/absence of regional lows) suggest model-dependent details. - Downscaling biases: CHAZ underestimates slow-moving and overestimates fast-moving storms in HIST; addressed via bias correction, but residual uncertainties remain. - TC frequency uncertainty: Future TC frequency in CHAZ depends on moisture metrics (CRH vs SD); study emphasizes speed distributions rather than absolute counts. - Regional and process scope: Focus on Texas landfalls and movement near landfall; does not directly assess changes in TC intensity, size, or full-life-cycle speeds. - Hazard translation: Faster motion does not straightforwardly map to reduced risk; other factors (e.g., surge, wind, rainfall, sea level, urbanization) are not fully integrated into risk estimates.