Object-based algorithm provides additional spatiotemporal information of precipitation, besides traditional aspects such as amount and intensity. Using the Method for Object-based Diagnostic Evaluation with Time Dimension (MODE-TD, or MTD), precipitation features in western Canada have been analyzed comprehensively based on the Canadian Precipitation Analysis, North American Regional Reanalysis, Multi-Source Weighted-Ensemble Precipitation, and a convection-permitting climate model. We found light precipitation occurs frequently in the interior valleys of western Canada while moderate to heavy precipitation is rare there. The size of maritime precipitation system near the coast is similar to the continental precipitation system on the Prairies for moderate to heavy precipitation while light precipitation on the Prairies is larger in size than that occurs near the coast. For temporal features, moderate to heavy precipitation lasts longer than light precipitation over the Pacific coast, and precipitation systems on the Prairies generally move faster than the coastal precipitation. For annual cycle, the west coast has more precipitation events in cold seasons while more precipitation events are identified in warm seasons on the Prairies due to vigorous convection activities. Using two control experiments, the way how the spatiotemporal resolution of source data influences the MTD results has been examined. Overall, the spatial resolution of source data has little influence on MTD results. However, MTD driven by dataset with coarse temporal resolution tend to identify precipitation systems with relatively large size and slow propagation speed. This kind of precipitation systems normally have short track length and relatively long lifetime. For a typical precipitation system (0.7 \(\sim \) 2 \(\times \) 10\(^{4}\) km\(^{2}\) in size) in western Canada, the maximum propagation speed that can be identified by 6-h data is approximately 25 km h\(^{-1}\), 33 km h\(^{-1}\) for 3-h, and 100 km h\(^{-1}\) for hourly dataset. Since the propagation speed of precipitation systems in North America is basically between 0 and 80 km h\(^{-1}\), we argue that precipitation features can be identified properly by MTD only when dataset with hourly or higher temporal resolution is used.

基于对象的算法除了提供诸如数量和强度之类的传统方面之外，还提供了额外的降水时空信息。使用基于时间维度的基于对象的诊断评估方法（MODE-TD或MTD），基于加拿大降水分析，北美区域再分析，多源加权集合降水，和允许对流的气候模型。我们发现加拿大西部的内陆山谷经常出现轻度降水，而那里很少有中度至重度降水。沿海附近的海洋降水系统的规模类似于大草原上的中度到重度降水的大陆性降水系统，而草原上的轻度降水的规模大于沿海地区。就时间特征而言，太平洋海岸中度至重度降水的持续时间比轻度降水的持续时间长，大草原上的降水系统通常比沿海降水的移动速度快。在年度周期中，由于对流活动剧烈，大草原上的西海岸在寒冷季节会有更多的降水事件，而在暖季则有更多的降水事件。使用两个控制实验，已经研究了源数据的时空分辨率如何影响MTD结果的方式。总体，源数据的空间分辨率对MTD结果影响不大。但是，由时间分辨率较粗糙的数据集驱动的MTD倾向于识别规模相对较大且传播速度较慢的降水系统。这种降水系统通常轨道长度短且寿命相对较长。对于典型的降水系统（0.7在加拿大西部，\（\ sim \） 2 \（\ times \） 10 \（^ {4} \） km \（^ {2} \）大小），最大传播速度可以通过6小时来确定数据大约每小时25 km h \（^ {-1} \），3小时33 km h \（^ {-1} \）和每小时数据集100 km h \ {^ {-1} \）。由于北美降水系统的传播速度基本上在0至80 km h \（^ {-1} \）之间，因此我们认为只有在使用时分分辨率更高或时间分辨率更高的数据集时，MTD才能正确识别降水特征。