Polar bears (Ursus maritimus) require sea ice for capturing seals and are expected to decline range-wide as global warming and sea-ice loss continue^1,2. Estimating when different subpopulations will likely begin to decline has not been possible to date because data linking ice availability to demographic performance are unavailable for most subpopulations² and unobtainable a priori for the projected but yet-to-be-observed low ice extremes³. Here, we establish the likely nature, timing and order of future demographic impacts by estimating the threshold numbers of days that polar bears can fast before cub recruitment and/or adult survival are impacted and decline rapidly. Intersecting these fasting impact thresholds with projected numbers of ice-free days, estimated from a large ensemble of an Earth system model⁴, reveals when demographic impacts will likely occur in different subpopulations across the Arctic. Our model captures demographic trends observed during 1979–2016, showing that recruitment and survival impact thresholds may already have been exceeded in some subpopulations. It also suggests that, with high greenhouse gas emissions, steeply declining reproduction and survival will jeopardize the persistence of all but a few high-Arctic subpopulations by 2100. Moderate emissions mitigation prolongs persistence but is unlikely to prevent some subpopulation extirpations within this century.

北极熊（Ursus maritimus）需要海冰来捕获海豹，随着全球变暖和海冰损失持续^1,2，北极熊预计会在整个范围内下降。到目前为止，尚无法估计不同亚群何时可能开始下降，因为大多数亚群²都无法获得将冰的可利用性与人口统计性能相关联的数据，而预计的但尚待观察的低冰极值^3则无法获得先验值。。在这里，我们通过估算北极熊在幼崽募集和/或成年生存受到影响并迅速下降之前可以禁食的阈值天数，来确定未来人口影响的可能性质，时间和顺序。这些禁食影响阈值与预计的无冰天数相交，这是根据地球系统模型的大集合估算得出的⁴揭示了人口影响何时可能出现在整个北极的不同亚人群中。我们的模型捕获了1979-2016年间观察到的人口趋势，表明某些亚人群可能已经超过了招聘和生存影响阈值。这也表明，随着温室气体的高排放，到2100年，生殖和生存的急剧下降将危害除少数北极高种群以外的所有种群的持久性。适度的排放减轻将延长持久性，但不太可能阻止本世纪内某些种群的灭绝。