Projected changes in extreme winds over Japan from a large-ensemble climate experiment

Extreme winds are a major climate hazard in Japan, yet their response to different levels of global warming is still not well quantified. This study investigates projected changes in daily maximum 10 m wind speeds using the d4PDF large-ensemble, which provides thousands of simulated years for the present climate and for +1.5 K, +2 K, and +4 K global warming levels. We focus on changes in hazard-relevant return levels at 311 land grid points across Japan. To represent physically distinct storm regimes, we adopt a seasonal piecewise extreme value framework. Block maxima are extracted separately for DJF, MAM, JJA, and SON and fitted with generalized extreme value (GEV) distributions using a probability integral transform estimator. The seasonal GEV models are then combined to derive annual 5-, 10-, 50-, and 100-year return levels for each grid and climate experiment, allowing us to quantify how contributions from typhoon-season and winter-season extremes change with warming. Present-day return levels show strong spatial and seasonal contrasts, with the highest values in summer and autumn over southern Japan, Pacific-facing coasts, and the Okinawa/Ryukyu region. Future changes are summarized as future-to-present ratios of annual return levels. For the 5-year return level, the signal is weak and mixed, with mean ratios close to 1.0 in all warming levels. In contrast, the 50- and 100-year return levels increase more clearly, with mean ratios of about 1.02-1.03 in +1.5 K, 1.04-1.05 in +2 K, and 1.08-1.09 in +4 K. The largest increases occur in already wind-exposed southern and coastal regions, highlighting a potential amplification of high-impact wind risk in a warmer climate.