Theory¹ and numerical modelling² suggest that tropical cyclones (TCs) will strengthen with rising ocean temperatures. Even though models have reached broad agreement on projected TC intensification^3,4,5, observed trends in TC intensity remain inconclusive and under active debate^6,7,8,9,10 in all ocean basins except the North Atlantic, where aircraft reconnaissance data greatly reduce uncertainties¹¹. The conventional satellite-based estimates are not accurate enough to ascertain the trend in TC intensity^6,11, suffering from contamination by heavy rain, clouds, breaking waves and spray¹². Here we show that weak TCs (that is, tropical storms to category-1 TCs based on the Saffir–Simpson scale) have intensified in all ocean basins during the period 1991–2020, based on huge amounts of highly accurate ocean current data derived from surface drifters. These drifters have submerged ‘holy sock’ drogues at 15 m depth to reduce biases induced by processes at the air–sea interface and thereby accurately measure near-surface currents, even under the most destructive TCs. The ocean current speeds show a robust upward trend of ~4.0 cm s⁻¹ per decade globally, corresponding to a positive trend of 1.8 m s⁻¹ per decade in the TC intensity. Our analysis further indicates that globally TCs have strengthened across the entirety of the intensity distribution. These results serve as a historical baseline that is crucial for assessing model physics, simulations and projections given the failure of state-of-the-art climate models in fully replicating these trends¹³.

理论¹和数值模型²表明热带气旋 (TC) 将随着海洋温度上升而增强。尽管模型已就预计的 TC 强化^3,4,5达成广泛一致，但观察到的 TC 强度趋势仍然没有定论，并且在除北大西洋以外的所有海洋盆地中都存在激烈的争论^6,7,8,9,10 ，那里的飞机侦察数据大大减少不确定性¹¹．传统的基于卫星的估计不够准确，无法确定 TC 强度的趋势^6,11，受到大雨、云、碎波和喷雾的污染¹². 在这里，我们显示了 1991-2020 年期间所有海洋盆地的弱 TC（即热带风暴到基于 Saffir-Simpson 等级的 1 类 TC）根据从表面漂流物。这些漂流者在 15 米深处淹没了“神圣的袜子”锥套，以减少海气界面过程引起的偏差，从而准确测量近地表电流，即使在最具破坏性的 TC 下也是如此。全球洋流速度呈每十年约 4.0 cm s ^{-1的强劲上升趋势，对应于 1.8 m s -1的正趋势}每十年的 TC 强度。我们的分析进一步表明，全球 TC 在整个强度分布中都得到了加强。鉴于最先进的气候模型未能完全复制这些趋势¹³ ，这些结果作为历史基线对于评估模型物理、模拟和预测至关重要。