Satellite precipitation products are not free of errors. These errors may show specific temporal patterns related to the life cycle of precipitation events. Understanding such patterns is key to uncertainty quantification, product improvements, and hydrologic applications. Here we investigate satellite error patterns during the life cycle of precipitation events over the contiguous United States (CONUS), using the Global Precipitation Measurement (GPM) Integrated Multi-satellite Retrievals for GPM (IMERG) as the satellite product and the Ground Validation Multi-Radar/Multi-Sensor (GV-MRMS) as reference to define “events” in a Eulerian perspective, both at the 30 min and 0.1°×0.1° native resolution of IMERG. We reveal significant variation in IMERG’s biases (both detectability and intensity bias) before/during/after the events, with a marked temporal asymmetry with respect to the mid-point of the event duration. Overall, the miss/false proportions of precipitation occurrence peak near the event temporal boundaries, with miss proportion higher near event ends, and false detection proportion higher near event beginnings. Precipitation intensity tends to be overestimated near the boundaries as well, while it is underestimated during the early- to mid-stages of the event. Diagnostic analysis controlling for data source inhomogeneity in IMERG and intensity variations throughout events traces back this stage-dependent performance to the Passive Microwave (PMW) retrieval algorithm, possibly due to the variation in cloud physical properties during event life cycles. Further investigations over different seasons/regions/times of day reveal distinct event-stage-dependent error patterns, which are likely linked to the different convective precipitation proportions. Consequently, a conditional analysis of error patterns on storm-type-related environmental variables, i.e., Convective Available Potential Energy (CAPE) and dewpoint is performed revealing significant relationships that underscore their prognostic value for characterizing the stage-dependent error curves. This study underscores the robust dependency of satellite errors on event stages and conversely, indicates the possible accuracy improvement of satellite precipitation products by integrating event stage information, as well as comprehensively leveraging environmental variables in future algorithms.

中文翻译：

---

了解多卫星降水产品在降水事件生命周期中的误差模式，用于诊断和算法改进


卫星降水产品并非没有错误。这些误差可能显示与降水事件生命周期相关的特定时间模式。了解这些模式是不确定性量化、产品改进和水文应用的关键。在这里，我们研究了美国本土降水事件生命周期中的卫星误差模式，使用全球降水测量 （GPM） GPM 综合多卫星检索 （IMERG） 作为卫星产品，并以地面验证多雷达/多传感器 （GV-MRMS） 为参考，以欧拉视角定义“事件”，在 IMERG 的 30 分钟和 0.1°×0.1° 原始分辨率下。我们揭示了事件之前/期间/之后 IMERG 偏差 （可检测性和强度偏差） 的显着变化，相对于事件持续时间的中点存在明显的时间不对称性。总体而言，降水发生的漏/假比例在事件时间边界附近达到峰值，事件结束时的缺失比例较高，事件开始时的误检测比例较高。在边界附近，降水强度也往往被高估，而在事件的早期到中期，降水强度被低估。控制 IMERG 中数据源不均匀性和整个事件中强度变化的诊断分析将这种阶段依赖性性能追溯到被动微波 （PMW） 检索算法，这可能是由于事件生命周期中云物理特性的变化。对不同季节/地区/一天中时间的进一步调查揭示了不同的事件阶段依赖性误差模式，这可能与不同的对流降水比例有关。 因此，对风暴类型相关环境变量（即对流可用势能 （CAPE） 和露点）的误差模式进行了条件分析，揭示了强调它们在表征阶段相关误差曲线方面的预后价值的重要关系。本研究强调了卫星误差对事件阶段的稳健依赖性，反过来，表明通过整合事件阶段信息以及在未来算法中综合利用环境变量，可以提高卫星降水产品的准确性。