Vernal pools of the northeastern United States provide important breeding habitat for amphibians but may be sensitive to droughts and climate change. These seasonal wetlands typically fill by early spring and dry by mid-to-late summer. Because climate change may produce earlier and stronger growing-season evapotranspiration combined with increasing droughts and shifts in precipitation timing, management concerns include the possibility that some pools will increasingly become dry earlier in the year, potentially interfering with amphibian life-cycle completion. In this context, a subset of pools that continues to provide wetland habitat later into the year under relatively dry conditions might function as ecohydrologic refugia, potentially supporting species persistence even as summer conditions become warmer and droughts more frequent. We used approximately 3000 field observations of inundation from 449 pools to train machine-learning models that predict the likelihood of pool inundation based on pool size, day of the year, climate conditions, short-term weather patterns, and soil, geologic and landcover attributes. Models were then used to generate predictions of pool wetness across five seasonal time points, three short-term weather scenarios and four sets of downscaled climate projections. Model outputs are available through a website allowing users to choose the inundation thresholds, time points, weather scenarios and future climate projections most relevant to their management needs. Together with long-term monitoring of individual pools at the site scale, this regional-scale study can support amphibian conservation by helping to identify which pools may be most likely to function as ecohydrologic refugia from droughts and climate change.

中文翻译：

---

识别耐气候春池：干旱和气候变化下两栖动物繁殖的水文避难所

美国东北部的春季水池为两栖动物提供了重要的繁殖栖息地，但可能对干旱和气候变化很敏感。这些季节性湿地通常在早春填满，并在夏中至夏末干燥。由于气候变化可能会产生更早和更强的生长季节蒸散，再加上干旱增加和降水时间的变化，管理方面的担忧包括一些水池可能会在今年早些时候变得越来越干，这可能会干扰两栖动物生命周期的完成。在这种情况下，在相对干燥的条件下，在今年晚些时候继续提供湿地栖息地的水池子集可能起到生态水文避难所的作用，即使在夏季条件变得温暖和干旱更加频繁的情况下，也可能支持物种的持续存在。我们使用来自 449 个水池的大约 3000 次洪水现场观察来训练机器学习模型，该模型根据水池大小、一年中的某一天、气候条件、短期天气模式以及土壤、地质和土地覆盖属性来预测水池被淹的可能性. 然后使用模型对五个季节性时间点、三个短期天气情景和四组缩小的气候预测的水池湿度进行预测。模型输出可通过网站获得，允许用户选择与其管理需求最相关的洪水阈值、时间点、天气情景和未来气候预测。再加上在现场规模上对单个水池进行长期监测，