Abstract. Rain-on-snow (RoS) events in regions of ephemeral snowpack – such as the northeastern United States – can be key drivers of cool-season flooding. We describe an automated algorithm for detecting basin-scale RoS events in gridded climate data by generating an area-averaged time-series and then searching for periods of concurrent precipitation, surface runoff, and snowmelt exceeding pre-defined thresholds. When evaluated using historical data over the Susquehanna River Basin (SRB), the technique credibly finds RoS events in published literature and flags events that are followed by anomalously high streamflow as measured by gage data along the river. When comparing four different datasets representing the same 21-year period, we find large differences in RoS event magnitude and frequency, primarily driven by differences in estimated surface runoff and snowmelt. Using dataset-specific thresholds improves agreement between datasets but does not account for all discrepancies. We show that factors such as meteorological forcing and coupling frequency as well as choice of land surface model play roles in how data products capture these compound extremes and suggest care is to be taken when climate datasets are used by stakeholders for operational decision-making.

中文翻译：

---

不同气候数据集中算法检测的雨雪洪水事件：以萨斯奎哈纳河流域为例

摘要。短暂积雪地区（例如美国东北部）的雪雨（RoS）事件可能是冷季洪水的主要驱动因素。我们描述了一种自动算法，通过生成区域平均时间序列，然后搜索同时发生的降水、地表径流和融雪超过预定义阈值的时期，来检测网格气候数据中的流域规模的 RoS 事件。当使用萨斯奎哈纳河流域 (SRB) 的历史数据进行评估时，该技术可以可靠地发现已发表文献中的 RoS 事件，并标记事件，随后通过沿河的计量数据测量出异常高的水流。当比较代表同一 21 年期间的四个不同数据集时，我们发现 RoS 事件幅度和频率存在巨大差异，这主要是由估计的地表径流和融雪量的差异造成的。使用特定于数据集的阈值可以提高数据集之间的一致性，但不能解释所有差异。我们表明，气象强迫和耦合频率以及地表模型的选择等因素在数据产品如何捕获这些复合极端值方面发挥着重要作用，并建议利益相关者在使用气候数据集进行业务决策时要小心。