A Novel Atmospheric River Identification Algorithm

Abstract

In recent years, climatology, variability, hydrological impact, and climatic drivers of atmospheric rivers (ARs) are widely explored based on various AR identification algorithms. Different algorithms, varying in their tracing variables, thresholds, and geometric metrics criteria, will introduce uncertainty in further study of AR. Herein, a novel AR identification algorithm is proposed to address some current limitations. A coupled quantile and Gaussian kernel smoothing technique is proposed to make a balance in capturing the spatiotemporal variation of integrated water vapor transport climatology and avoiding largely biased estimation. In spite of variety of AR shape, orientation, and curvature, more reliable AR metrics (e.g., length and width) can be calculated based on the generated smooth AR trajectory, which is realized by modifying and integrating the concepts of local regression and K-nearest neighbors. An unprecedented and novel metric (i.e., turning angle series) is delivered to quantify AR curvature, serves as the key to distinguish tropical cyclone-like features, which often indicate occurrences of tropical cyclones. It also bridges ARs to their associated atmospheric circulation patterns. A pilot application of the algorithm is presented to identify persistent AR events related to flood triggering extreme precipitation sequences in the Yangtze River Basin (YRB). A dominating AR route, which connects Arabian Sea, Bay of Bengal, South China Sea, to Southeast China and YRB, terminates in the North Pacific, is found principal to the flood triggering extreme precipitation sequences in the YRB. In addition, this algorithm is extensible to other regions, even global domain.