Atmospheric rivers (hereafter, ARs) are long and narrow structures in the lower troposphere that carry large amounts of water vapor . have estimated that ARs have a median length of about 3600 km, a median length / width ratio of about 7 and a mean integrated vapor transport (IVT) of 370 kg m-1 s-1. ARs are associated with the pre-cold frontal region and the warm conveyor belt (WCB) of extratropical cyclones . The maximum moisture flux often occurs within the low-level jet (LLJ) located at around 1 km height along the cold frontal boundary, and for this reason ARs and the LLJ are sometimes identified with each other . However, to account for most of the high IVT and moisture content defining an AR, usually a wider region encompassing the layers below 2.5 km ahead of the cold front must be considered . On occasion, the relationship between ARs and cyclones has been shown to be complex, with documented cases in which multiple cyclones are associated with a single AR .

ARs supply moisture to the WCB of the cyclones and are therefore considered as one of the potential precursors of extreme precipitation, particularly when landfall occurs e.g.,. The relationship between ARs and flood events has been extensively analyzed for the US West Coast region e.g.,, Europe and other regions worldwide e.g.,. It is important to better understand the physical mechanisms leading to extreme flooding associated with ARs, considering their impacts on human and natural systems and the mounting evidence that ARs are projected to become more frequent and intense in the future e.g.,.

Between three and five ARs can be found per hemisphere at any given time , accounting for approximately 84 % of the meridional IVT for the Northern Hemisphere and about 88 % in the Southern Hemisphere . Since these structures can transport an amount of precipitable water equivalent to several times the discharge of the Mississippi River , ARs have been identified as a primary feature of the global water cycle.

There are several proposed methods of AR detection, most of which are based on thresholds of integrated water vapor (IWV) and/or IVT, shape criteria, and from satellite or reanalysis data e.g.,. developed a global detection method using filters of intensity, direction, geometry and coherence of the structures. More recently, proposed a combined IVT and IWV variable-threshold detection algorithm, which operates both in summer and winter months. These objective detection criteria have shown that AR structures of IWV and IVT can extend from the tropics into midlatitudes; however, they do not provide information about the source and sink regions of AR water vapor.

Tropical moisture exports have been identified as a primary source of moisture for ARs in Europe and the US West Coast . AR structures link remote sources of moisture from the tropics to midlatitudes through long corridors of advection e.g.,. These studies have primarily used backward Lagrangian tools to evaluate the source–sink regions. However, moisture from midlatitudes (local sources) has also been identified as an important source of water vapor convergence in AR events . used the FLEXible PARTicle dispersion model (FLEXPART) to show that both tropical and local sources of moisture are present in AR landfall events for different European latitudes.

Some authors argue that local sources are primarily responsible for the high water vapor content within the AR core . By calculating the water vapor budget of 200 extratropical cyclones, conclude that tropical moisture reaching the extratropics only contributes to mid-level moisture, above the boundary layer. Following this perspective, ARs can be thought of as the footprints left behind the cyclone pathway, and not as a conduit for meridional transport of water vapor and latent heat. One possible explanation for the lack of agreement may be the sensitivity to the physics and parametrization schemes used in the latter analysis . The current understanding is that tropical moisture exports can provide a significant amount of moisture to ARs. Most of them also incorporate midlatitude sources of vapor along their path .

Considering that there is still an important discussion related to the origin of moisture in ARs, in this paper, we use a new forward moisture tracer tool coupled to the Weather Research and Forecast (WRF) model to evaluate the moisture sources of two particularly extreme AR case studies, as well as the transport mechanism of this humidity. The selected cases were associated with extreme precipitation and flooding that led to significant socioeconomic impacts. The first selected AR developed over the Pacific Ocean and affected the western coast of North America, whereas the second AR developed over the Atlantic Ocean and impacted the western coast of the Iberian Peninsula. The tracer tool allows us to track the tropical moisture associated with these two events and evaluate the relative contribution of this tropical moisture to total moisture and precipitation. In addition, the WRF tracer tool also provides information about the vertical distribution of tropical moisture, as well as the position of the maximum of moisture with regard to the LLJ.

Figure  shows the comparison between WRF-simulated and observed precipitation. Observations are from the dataset for the Pacific simulation and from the Iberia02 precipitation dataset in the case of the Atlantic simulation. The latter dataset is a combination of Spain02 and Portugal02 , both of which include a high density of good-quality stations . Further comparison of the simulations with observations, against IWV (IVT, Eq. 1) from NASA's Modern-Era Retrospective Analysis for Research and Applications (MERRA) is provided in Fig. .

Whereas the simulated IVT field is realistic when compared to observations, WRF tends to overestimate precipitation. The overestimation is particularly high in the mountains of Oregon and Washington for the 2007 event. However, despite the fact that precipitation is arguably the most difficult variable to simulate in a numerical model e.g.,, the spatial pattern of precipitation is realistically represented.

Figure shows the three-dimensional distribution of water vapor mixing ratio (panel a), and tracer water vapor mixing ratio (panel b) for the event in the Pacific that made landfall along the US West Coast. While the former accounts for the total amount of moisture, the latter shows only the moisture originating from tropical latitudes, labeled with the 3-D mask depicted in Fig. . The simulation was started 8 days before the time shown in Fig. . Panels (c) and (d) show a snapshot of the water vapor mixing ratio and the tracer water vapor mixing ratio in the form of a series of cross-section slices that allow the visualization of the vertical distribution of moisture. The images suggest that the vast majority of the moisture contained in the pre-frontal region has its origin in the tropical regions. This is especially true at lower latitudes. The maximum content of tropical moisture remains mostly in the lower levels. The high moisture values behind the front and at the leading edge of the AR structure, where the WCB is located, are not related to tropical advection and are thus generated by convergence of moisture from local sources occurring along the frontal region. Slightly different conclusions are obtained for the Atlantic case study shown in Fig. . For the sake of simplicity, only the eastern longitudes of the domain of simulation are shown. Even though the tropical moisture still remains in the lowest levels of the troposphere, its contribution to the total is less significant than for the Pacific case study. The reason is shown in Fig. e–h, indicating that most of the connection of the WCB with tropical regions is through a much longer path, due to the blocking position of the Azores High.

Figure a shows the percentage of IWV that comes from the tropics (IWVTR/IWV) for the Pacific event. In a region extending from the tropics to the coast of Washington state, tropical moisture accounts for about 80–90 % of the precipitable water and locally exceeding this contribution. These high percentages extend inland along the northwestern US coastal regions. Likewise, Fig. b shows the 24 h accumulated percentage of precipitation that is composed of condensed tropical water vapor (PrecTR/Prec). For clarity, we only plot the region where precipitation exceeded 3 mm day-1 . Precipitation of tropical origin accounts for 70 to 90 % of total precipitation from northern California to southern Oregon, and the ratio decreases at higher latitudes. Interestingly, tropical moisture is funneled by local topography, and it contributes to about 70 to 80 % of precipitation west of the Cascade Range. In the Atlantic case, we also see a clear plume where tropical water vapor accounts for more than 80 % of precipitable water; however, the percentage decreases to around 70 % close to the center of the system and just before arriving on the Iberian coast (Fig. c–d). In this case, cyclogenesis occurs just off the coast of Galicia, on the northwestern tip of the Iberian Peninsula, and thus the enhanced convergence of the existent local moisture feeds the AR and is involved in the heavy precipitation, which consequently is only between 60 and 80 % of tropical origin. The high ratios observed in mountainous ranges such as the Pyrenees and the Rockies, far ahead of the cold front and the AR associated with the systems, are very likely due to the slantwise lift of tropical moisture in mid-levels along the warm frontal boundary.

Figure plots a range of transversal cross sections showing the vertical distribution of the tracer water vapor mixing ratio through the central axis of the AR, as well as wind speed. From these results, there is evidence that the maximum of tropical moisture does not necessarily coincide with the LLJ, which is the maximum in wind speed at lower levels. More precisely, at the root of the AR, at subtropical latitudes (Fig. d), most of the tropical moisture remains close to the surface and below the LLJ, which can be identified at a height of 1 km. As the central axis of the LLJ goes upward with latitude, tropical moisture tends to ascend in the vertical column, but the maximum of moisture can be located in front of or behind the LLJ, as well as remaining near the surface levels. In the leading part of the AR, the interaction of the humidity with the topography of the Pacific coast of North America makes the situation more difficult to analyze. Very likely, the complex formation process of the cyclone, from the interaction of two preexistent frontal systems, loaded with tropical moisture, adds complication to the thermodynamic structure and moisture distribution of the resulting front. Notwithstanding, this AR event is a particularly well defined case from the perspective of vertically integrated quantities such as IVT and IWV.

An analogous plot for the Atlantic case is presented in Appendix A (Fig. A2). The results are similar to the Pacific case, with no clear one-to-one connection between the LLJ and the maximum in tropical humidity. We note that no general conclusion can be obtained from particular case studies, but these results suggest that the perception that ARs are clearly associated with the LLJ of extratropical cyclones should be reviewed.

Figure a shows the area-averaged total precipitation (black circles) and the ratio between tracer precipitation and total precipitation (red crosses) throughout the region highlighted in Fig. b for the Pacific event. As expected, the plot shows that the maximum in tropical precipitation is observed during the landfall of the cold front and the AR, which closely coincides with the maximum in total precipitation. The secondary maximum in the total precipitation observed 1 day before the landfall of the AR event is due to the landfall of the warm front associated with the cyclone (see Fig. A3). The convergence of local moisture would be the dynamical source for precipitation in the warm front.

A new 3-D Eulerian forward water vapor tracer tool implemented in the WRF model has been used to analyze two important AR events. The first event developed over the Pacific Ocean and corresponds to the Great Coastal Gale of 2007 on the US Pacific West Coast, which resulted in an estimated USD 678 million in direct economic damages . The Atlantic event corresponds to an atmospheric river event in October 1987 that resulted in record winds of 100 km h-1 and daily precipitation of over 100 mm day-1 in Galicia (in the northwest of Spain) and Portugal. In an effort to understand the origin of moisture for these two AR events, we use 3-D water vapor tracers to quantify the percentage of total precipitable water and precipitation that originates from the tropics.

Results show that most of the moisture within and surrounding the two atmospheric rivers had its origin in the tropical regions that we labeled with the 3-D tracer mask. Consequently, most of the precipitation that fell during these two events was composed of condensed tropical water vapor. The Pacific event shows a more intense connection with tropical regions than the Atlantic case. As a result, the percentage of tropical precipitation for this event over North America is higher and peaks around 85 %. Nevertheless, for the Atlantic event, still more than 60 % of the resulting precipitation is of tropical origin.

The two selected case studies have been chosen due to the associated severity of flooding and socioeconomic damages. Both correspond to a strong AR feeding the system of a very intense extratropical storm. The conclusions drawn from these two AR events are thus not necessarily representative of the bulk of Pacific or Atlantic AR events. However, the results highlight the importance of tropical moisture for the two case studies. We also find evidence that convergence of local moisture also contributes to total precipitable water, especially in the post-frontal region, the leading edge of the AR and in the far northern latitudes where the tropical link has weakened. It is well known that in a mature system, when baroclinic structures are well differentiated, the stored water vapor tends to be constant (e.g., ), and since the fate of tropical moisture is to precipitate, local convergence should keep the balance by lateral inflow.

Based on these results, we hypothesize that the highest amounts of precipitable water can only be attained in a system in which a clear tropical source of moisture is sustained until landfall. Strong ARs with a direct link to tropical latitudes should be expected to result in more precipitation than those with local convergence as a primary feeding mechanism. It is our aim for the future to extend this work by including more cases.

Finally, our findings suggest that the maximum of tropical moisture does not necessarily coincide with the LLJ of either extratropical cyclone analyzed. Instead, this maximum is located near surface levels at lower latitudes to gradually ascend in northern latitudes, but still remaining below 2 km, mostly within the boundary layer, in contrast with findings in other studies . The maximum of tropical moisture may be situated below and toward the back, or ahead of the LLJ, which is located along the cold front. Both events are clear examples of ARs from the point of view of vertically integrated variables, such as IWV and IVT used in most detection algorithms; however, the vertical distribution of moisture of tropical origin reflects the complex processes leading to precipitation. The new 3-D tracer tool will allow us to delve into these processes and explore the role of tropical moisture exports in the initiation and intensification of AR events.