How climate change influences lake ice phenology is important in understanding the climate‐lake interactions. This study investigates the response of lake ice phenology to future climate scenarios. Thirty‐five lakes in the Tibetan Plateau were studied. Firstly, we applied a random forest (RF) model to simulate lake ice condition under different representative concentration pathways (RCPs). The results of the virtual experiments show that lake ice freeze‐up start date (FUSD) and break‐up end date (BUED) are well simulated by the RF model, with R²_FUSD > 0.91 under different RCPs, and R²_BUED > 0.84, except in RCP2.6 (R² = 0.74). Secondly, two non‐parametric methods (Mann‐Kendall and Sen's slope estimator) were used for analyzing the trends in lake ice phenology and its response to various emission scenarios. Lake ice phenology was largely affected by temperature changes under different RCPs, and it had a larger inter‐annual variability in the early period (2002–2050) than in the later period (2050–2099). FUSD of 35 lakes delayed by an average of 0.01, 0.04, and 0.04 days/yr from 2002 to 2098 under RCP4.5, RCP6.0, and RCP8.5, whereas BUED advanced by an average of 0.04, 0.11, and 0.21 days/yr from 2003 to 2099. Both delays in FUSD and advances in BUED contributed to shortened average ice duration of 35 lakes, which shortened by 0.05, 0.14, and 0.25 days/yr from 2002 to 2098 under RCP4.5, RCP6.0, and RCP8.5, respectively. These findings could help to tackle the impacts of climate change on the lake systems.

中文翻译：

---

青藏高原内陆未来气候情景下湖冰物候动态预测与分析

气候变化如何影响湖泊冰的物候学，对于理解气候与湖泊之间的相互作用非常重要。这项研究调查了湖冰物候对未来气候情景的响应。研究了青藏高原的35个湖泊。首先，我们应用随机森林（RF）模型来模拟不同代表性浓度路径（RCP）下的湖冰状况。虚拟实验的结果表明，RF模型很好地模拟了湖冰冻结开始日期（FUSD）和分解结束日期（BUED）， 在不同RCP下R ² _FUSD > 0.91，R ² _BUED  > 0.84，RCP2.6（R ² = 0.74）。其次，使用两种非参数方法（Mann-Kendall和Sen的斜率估计器）来分析湖泊冰物候趋势及其对各种排放情景的响应。在不同的RCPs下，湖冰的物候变化主要受温度变化的影响，其早期（2002-2050）的年际变化大于后期（2050-2099）的年际变化。在RCP4.5，RCP6.0和RCP8.5下，从2002年到2098年，35个湖泊的FUSD平均每年延迟0.01、0.04和0.04天/年，而BUED平均提前了0.04、0.11和0.21天从2003年到2099年，每年FUSD的延迟和BUED的增加都导致35个湖泊的平均冰期缩短，在RCP4.5，RCP6.0，和RCP8.5。