IPART: A Python Package for Image-Processing based Atmospheric River Tracking

Abstract

An atmospheric river (AR) in the field of meteorology/climatology refers to enhanced water vapor content in the lower troposphere. The term was coined as an analogy to terrestrial rivers in a sense that when viewed from satellite imagery or large scale atmospheric observation, they appear as narrow and elongated vapor filaments, representing transient intensified horizontal moisture fluxes (e.g., Gimeno et al., 2014; Dettinger, 2011). A typical atmospheric river can carry 7-15 times the water in the Mississippi River (Ralph et al., 2011), and at any time in winter, there are four to five such systems in the Northern Hemisphere alone (Zhu & Newell, 1998), accounting for 80-90% of the total north-south integrated vapor transport (Guan & Waliser, 2015; Zhu & Newell, 1998). Its dual hydrological role, both as a fresh water source for some water-stressed areas (Dettinger, 2011, 2013; Rutz & Steenburgh, 2012) and as a potential trigger for floods (Lavers et al., 2012; Lavers & Villarini, 2013; Moore et al., 2012; Neiman et al., 2008), has granted it increasing attention among the research community. Their long-term change in a warming climate also stands as a pressing research question. However, an important prerequisite to answer such questions is a robust and consistent detection method. As meteorologists and climatologists often deal with observational or simulation data in large sizes, an algorithmic method can ensure better efficiency, consistency and objectivity compared with human identification.