Interactive comment on “ Tropical Moisture Exports , Extreme Precipitation and Floods in Northeast US ”

This study investigates the role of tropical moisture exports (TMEs) for extreme precipitation and floods in the northeastern U.S. The authors use an previously published TME climatology by Knippertz and Wernli, and quantify the statistical linkage between TMEs entering the northeastern U.S. and precipitation and floods in the same region. The objective of this study is fine, but the quality of the text and figures is insufficient. Important aspects of the study are not well described, the text is confusing in several places (and contains too many details), and again the figures provide too much information and lack clarity and compelling evidence. Since I think that a complete rewriting of the paper is required, including a redesign of the figures, I recommend very major revisions.


Introduction
Surface temperature gradients (Lorenz, 1984;Jain et al., 1999;Karamperidou et al., 2012) drive the large scale atmospheric circulation, and hence the meridional transport of moist air masses, primarily born in tropical oceanic areas, to higher latitudes, often manifested as localized threads of strong moist air fluxes.There are a number of studies focusing on 'Atmospheric Rivers' (ARs), a concept first introduced by (Zhu and Newell, 1994) who defined an atmospheric phenomenon that features a narrow corridor of concentrated moisture, enhanced and elated tropospheric water vapor transport that have a large hydrologic effect.Recent studies have linked Atmospheric Rivers (Zhu and Newell, 1998;Bao et al., 2006) to extreme precipitation and floods (Ralph et al., 2006;Leung and Qian, 2009;Lavers et al., 2011Lavers et al., , 2013;;Ralph and Dettinger, 2011;Lu et al., 2013;Lavers et al., 2016) as the major hydrometeorological contributor.ARs have been linked to extreme precipitation in the United Kingdom (Lavers et al., 2011(Lavers et al., , 2013;;Nakamura et al., 2013), Western France (Lu et al., 2013), Midwest United States (Nakamura et al., 2013) and the West coast of United States (Ralph et al., 2006;Dettinger, 2011;Ryoo et al., 2011;Lavers et al., 2016) which has been linked to the well-known "Pineapple express" (Higgins et al., 2000), that originates from oceanic areas adjacent to Hawaiian Islands.However, areas such as the Northeastern United States (NE-US) have not been studied as much in this context.In a previous study for Western France (Lu et al., 2013), we saw moisture tracks for some extreme storms in France originating in the Gulf of Mexico and passing through NE-US, with precipitation in the NE-US enroute to France, and we also observed seasonally changing spatial patterns of the TMEs tracks entered this region.Thus, the NE-US was chosen in this study as a pilot to develop the statistical and physical based framework for hydrometeorological extremes with a focus on diagnosis of associated atmospheric circulation and climate teleconnection.
The widely used definition of ARs as a narrow plume with at least 2cm of integrated water vapor (IWV), extending over thousands of kilometers long and hundreds of kilometers wide (Ralph et al. 2004;Neiman et al. 2008;Dacre et al. 2015;Jeon et al. 2015), has restricted the studies to popular AR-associated regions as indicated in the preceding paragraph.In contrast, Bao et al. (2006) and Knippertz and Martin (2007), suggested the term "moist conveyor belt" to broaden the scope of the studies to Tropical Moisture Exports (TMEs) that may have significant contributions to hydrometeorological extremes on a global scale (Knippertz and Wernli, 2010;Knippertz et al., 2013;Nakamura et al., 2013;Lu et al., 2013).Here, we explore how TMEs may be involved in moisture transport into the NE-US, and how their sources and trajectories vary by season and under different ENSO regimes.The linkage of TMEs to extreme precipitation and flooding in each of the seasons is also assessed.
Past studies on the nexus of moisture transport, extreme precipitation and floods, focused on the (1) link to selected historical floods events in various regions (Ralph et al., 2006;Lavers et al., 2011;Lu et al., 2013;Nakamura et al., 2013); (2) identification of moisture sources that contribute to extremes (Knippertz and Wernli, 2010;Ryoo et al., 2011;Knippertz et al., 2013;Lu et al., 2013;Nakamura et al., 2013); and (3) trajectory analysis of the air masses with investigation of the attendant atmospheric circulation (Wernli, 1997;Bao et al., 2006;Ryoo et al., 2011;Nakamura et al., 2013;Lu et al., 2013).Lavers et al. (2011) linked ARs with the top 10 largest winter floods in Britain since 1970; while Nakamura et al. (2013) identified strong association between anomalous atmospheric circulations that drive low-level flow of warm and moist air, and 21 extreme floods in Ohio River.Further, Knippertz and Wernli (2010), Lu et al. (2013) and Nakamura et al. (2013) noted that the poleward transport of tropical born moist air masses to the Northern Hemispheric extratropics provide the link between tropical moisture sources and extratropical extreme precipitation, occasionally with explosive cyclogenesis.Lu et al. (2013) associated TMEs from the Gulf of Mexico and tropical North Atlantic Ocean as the major moisture sources for the 1995 January flood in western France, and demonstrated the predictability of the extreme precipitation given only the mid-latitude sea level pressure (SLP) fields, suggesting that steering mechanisms were important.A similar study for the extreme floods in Ohio River Basin by Nakamura et al. (2013) identified a persistent dipole pattern in the SLP leading to the wave like transport of moisture from the Gulf of Mexico into the flooded area, every 4 to 7 days over the March-April-May season.Further, Cioffi et al. (2015) identified the influence of North Atlantic Oscillation and El Niño Southern Oscillation (ENSO) on the changes of space-time structure of precipitation extremes in Europe over the last century.

Tropical Moisture Export Characterization
Tropical Moisture Exports (TMEs) were first documented in (Knippertz and Wernli, 2010) for Northern Hemisphere and later extended by Knippertz et al. (2013) to produce a global TMEs climatology.The TMEs tracks were calculated using 6-hourly ERA-Interim data (Dee et al., 2011) and range from 1989 to 2010, covering daily tracks born in the tropics [0° -20°N].One-day forward tracks are calculated from every 100 km × 100 km × 30 hPa box between 0 and 20°N, between 1000 hPa and 490 hPa.Specific humidity can be converted into water mass as each track represents the same atmospheric mass of ~ 3 × 10 12 kg (Knippertz and Wernli 2010).Tracks are calculated with the LAGRANTO Lagrangian analysis tool (Wernli, 1997;Wernli and Davies, 1997), by interpolating the relevant fields to the positions of the track at the 6-hr updating frequency.To ensure that the moisture characteristics of the tropical air parcels are maintained on their way across the subtropics, only tracks that reach 35°N within the next 5 to 6 days after crossing 20°N were retained; however, changes due to fluxes of heat and moisture from the underlying surface or mixing cannot be completely excluded.The water vapor fluxes of the retained tracks in the dataset must reach 100 g kg -1 m s -1 , a threshold chosen to represent 'fast' events and yet get meaningful statistics (Knippertz and Wernli, 2010).Knippertz and Wernli (2010) showed that TMEs contributes significantly (more than 60% of the average) to climatological precipitation in the mid-latitudes and identified four major source regions in the tropics: (1) "Pineapple Express' (PE) [170° -130°W].This source has a maximum activity in the December to Feburary (DJF) season and is almost absent in from June to Auguest (JJA).
(2) 'Great Plain' (GP) [100° -90°W].We extended this region to [100° -70°W] from the Knippertz and Wernli (2010) after initial analyses revealed that it was the dominant source region for the TMEs entering the NE-US.This region includes the Gulf of Mexico, and also parts of the continental region between the Rocky and the Appalachian Mountains.
(4) 'West Pacific' (WP) [120° -170°E].Although this source contributes the least, it is included in the analysis considering its correlation with the other three.Subsequently, we identify the TMEs tracks as PE, GP, GS or WP as associated with their birthplace.The source regions are marked in Figure 7 & 8.
The conceptual framework of the analysis presented in this paper is indicated in Figure 1.The causal structure illustrated considers the potential dependence of the TMEs birth process as a function of the source location, the season and ENSO state (this level of relationships is marked as red arrows in Figure 1).The birth process here refers to where and how many tracks are born and how these vary by season.The number of TMEs entering the NE-US on any given day depends on the associated birth process, the season, the source, the ENSO state, and the atmospheric circulation (this level is marked as blue arrows in Figure 1).The total water released (∆ ) by the TMEs in the NE-US on a given day is taken to depend on the number of TMEs' entering.The extreme precipitation amount, EP, is considered to depend on the ∆ (black arrows in Figure 1).We take the NE-US [39°N -48°N, 66°W -82°W] as the pilot study area to address the following questions and develop the framework that can be extended to other areas: (1) How were floods events in NE-US in different seasons related to TMEs in terms of the moisture sources, the trajectories and their seasonality?
(2) What is the connection between the birth process of TMEs and the role of large scale climate regulation, e.g.ENSO?
(3) What is the relationship between the entrance of TMEs to NE-US and the role of atmospheric circulation patterns?
(4) What is the link between TMEs and extreme precipitation events that may trigger floods?
The paper is organized in line with answering these research questions.The data used in this study is provided in Section 2. In Section 3, we illustrate the association of TMEs with examples of extreme floods in each season.
The second research question is addressed in Section 4. The entrance of TMEs to the NE-US and the associated atmospheric steering is discussed in Section 5.The link between moisture release and extreme precipitation that is closely related to floods is explored in Section 6.A discussion and summary of the key findings is provided in Section 7.

Data
The results of this study mainly use the TMEs dataset documented in (Knippertz and Wernli, 2010).The dataset covers from 1989 to 2010, recording daily tracks born in the tropics that meet the following criteria: (1) they reach 35°N within the next 5 to 6 days after crossing 20°N, and (2) water vapor flux of any track is not less than 100 g kg -1 m s -1 .For all TME trajectories, position information (longitude and latitude) is available every 6 hours, and up to 7 days.
We calculate the change of moisture (recharge or release) along each track as where j is the time point on the track (the j th position along the track), ( ) is the k th TMEs track's specific humidity in g kg -1 (units) at time point j and Δ ( ) is the change of specific humidity.Each of the tracks has its positions recorded every 6 hours, up to 7 days, with a total of 29 time points.Thus, j ranges from 0 to 28, with the birth location at j = 0 and the point before death location at j = 28.
The total release of water from the air parcel in the study area was calculated as the change of total specific humidify (including liquid phase) integrated over all the tracks for a given date starting from their entrance as follows: where ∆ ( ) is the total change of specific humidity of all the tracks that are active in the NE-US on day t, N t is the total number of tracks active in the NE-US on day t.Note that we consider the tracks leaving/exiting at the different hours of the day, i.e. 0 o 'clock, 6 o'clock, 12 o'clock, 18 o'clock, 24 o'clock.By examining the selected tracks and their release of moisture to the study area, the associated moisture birth location and their trajectories to the NE-US can be identified and the precipitation resulted from the release of water vapor from the moist air parcel can also be computed.
For the analysis of atmospheric circulation patterns, we choose the sea level pressure (SLP) data from the NCEP/NCAR Reanalysis project (Kalnay et al., 1996) provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA at http://www.esrl.noaa.gov/psd/.We use the daily SLP data, with a resolution of 2.5° latitude× 2.5° longitude, covering the same period as the TMEs dataset, 1989 -2010.We further derived the daily SLP anomalies against calendar day climatology as calculated as follows.
where i is the i th day in year j, i = 1, …, 365 (366 for leap years) and j = 1989, …, 2010; is the SLP anomaly on i th day in year j, is the calendar day climatology of SLP on the i th day of a year.
The Oceanic Niño Index (ONI) is provided by NOAA/National Weather Service, NOAA Center for Weather and Climate Prediction, Climate Prediction Center at http://www.cpc.ncep.noaa.gov/.The ONI has become the de-facto standard that NOAA uses for identifying El Niño (warm) and La Niña (cool) events in the tropical Pacific.The warm and cold episodes are based on a threshold of ± 0.5 o C of the running 3-month mean of ERSSTv3b data (Smith et al., 2008) SST anomaly for the Niño 3.4 region (5°N -5°S, 120° -170°W).Cold and warm episodes are defined when the threshold is met for a minimum of 5 consecutive over-lapping seasons.
The flood events in NE-US we discuss in the following section are recorded by the Dartmouth Flood Observatory (DFO, at their website: http://floodobservatory.colorado.edu/).The Dartmouth Flood Observatory Achieve contains global floods events from 1985 to present based on data acquired by NASA, the Japanese Space Agency and the European Space Agency.

TMEs and Floods in the NE-US
For each of the four seasons, i.e., Dec -Jan -Feb (DJF), Mar -Apr -May (MAM), Jun -Jul -Aug (JJA) and Sep -Oct -Nov (SON), major flood events in the NE-US as recorded in the DFO data base were identified.For each event, TMEs tracks that were born in any of the four major source regions, listed in Section 1.1, and within 7 days of the onset of heavy precipitation in the flooded region were identified.Of these, those that entered the NE-US, including those that continued out of the region, were further identified.This two-step identification procedure yields a list of TMEs tracks that both were from the major source regions and contributed to the precipitation leading to the target flood events: a representative flood event is identified for each season and the TMEs tracks associated with it are shown in Figure 2. The colors indicate the changes of specific humidity along the tracks, calculated as in Eq. 1. Moisture release is recorded as blue or light blue dots in Figure 2, while moisture recharge magnitude is shown as red or yellow dots.Basic attributes of the exemplified flood events are summarized in Table 1.Enlarged panels of Figure 2 are attached in Appendix A for reference.There are strong spatial patterns of the trajectories given the TMEs tracks' sources: tracks from PE follow typical northeastwardly 'Pineapple Express' sine-wave like trajectory showed in the AR literature focusing on the west coast [e.g., (Dettinger, 2011;Dettinger et al., 2011;Ralph and Dettinger, 2011)]; tracks from GP and GS penetrated to the north until reaching 50°N, where the jet stream locates, and then turned to the east.Lu et al. (2013) showed remarkably similar trajectories of the two major sources, i.e.GP and GS, associated with the more than 100-yr heavy precipitation induced flood event in Western France in January 1995.It suggested that moist air parcels originated from the tropics can have far impacts on extremes in the higher latitudes; and floods occurred at different time and over different regions may share the same moisture sources in terms of the TMEs.
The major moisture sources identified for the April 2005 event (Figure 2 (g) -(l)) are GP and GS.There were a few TMEs tracks from PE, which contributed only to the early stage of the heavy rainfall, indicated by both the number of tracks and moisture release.It took averagely 6 -7 days, depending on the carrying wind speed and season, for TMEs tracks born in PE to reach the NE-US.The major contributors are thus GP and GS, which are both very active in MAM (Knippertz and Wernli, 2010).The PE TMEs (Figure 2 (g) -(h)) propagated further north reaching 50°N, different from the winter event shown in Figure 2 (a) -(f).The same is the case for the TMEs born in GP and GS (Figure 2 (g) -(l)).This may be associated with the beginning of the seasonal northward shift of the jet stream (Laing and Michael Fritsch, 1997).The moisture released from the TMEs was a significant fraction of the moisture carried by the tracks, and Figure 2   The four flood event examples show that in the NE-US, TMEs may be closely related to floods year-around with varying major moisture sources and trajectories.The main year-around moisture sources shown in these four events are GP and GS with some contributions from PE in DJF.

TMEs Birth and ENSO
In each season, TMEs from different moisture sources were seen to be associated with floods in the NE-US.Cioffi et al. (2015) also linked NAO and ENSO to the space-time structure of precipitation extremes in Europe and its changes, based on the data over the past century.Cioffi et al. (2015) also encouraged a further study on the influence of ENSO on extremes, together with its modulation of the atmospheric circulation.Thus, the seasonality of and interannual variations in TMEs birth and its association with ENSO are analyzed in this section.Figure 3 presents the seasonality of and the interannual variations in the TMEs born in the four major sources under different ENSO phases.All four sources show a strong seasonality of the TMEs birth process.

TMEs entrance and Atmospheric Circulation Patterns
The origins of the TMEs entering NE-US vary seasonally, due in part to the seasonality of the TMEs birth, and in part to the seasonal and interannual variations in atmospheric circulation patterns.The conditional probability P(Source|NE) that the TMEs that entered the NE in a given month comes from a particular source, and its interannual variation are illustrated in Figure 4.The calculation of conditional probability in this study followed the classic ratio analysis.The corresponding data for the total number of tracks entering by calendar month, the total number of tracks from the four sources considered here, and the annual total for each source are supplemented and presented in the Appendix B Table B.2. First, we note that the four sources considered account for 85% of all tracks entering the NE-US on an annual average basis, varying cyclically from a minimum of 70% in December to a maximum of 98% in July (Table B.2).The number of TMEs tracks entering peaks in the winter with a secondary maximum in June.On an annual basis, approximately 73% of the tracks come from GP, 14% from GS, 11% from PE and 3% from WP. Consequently, it is not a surprise that on average, 45% (Nov) to 79% (June) of the tracks on a monthly basis come from GP.The fraction coming from GS increases from June to October and peaks in August (28%).PE's contributions are important in November to February (~20% of the tracks), and WP is a weak spring/fall contributor accounting for a maximum of 4% of the tracks in December and February.In summary, GP is important year around, but particularly in April -July, GS has increased contributions in June -October, and PE in November -February, with WP a possible contributor in winter.
From Figure 4, we see that GP accounts for a relatively stable seasonality of tracks entering across the year, with a consistent peak in June (~80%), and minimum in Nov -Feb (~45%) and high interannual variability in August, September and November.GS is an important source in July -September with high interannual variability.PE is primarily a contributor in Nov -Feb, while WP has a low contribution throughout the year with relatively high variability across years in fall and spring.
Tracks entering from GP into the NE-US are positively correlated at a significance level of 0.05, with those entering from GS in January (correlation=0.62),February (0.79), March (0.71), September (0.37), November (0.51) and December (0.68); with PE tracks entering in November (0.37) and December (0.18), and with WP tracks in January (0.52) and February (0.39).Tracks entering from GS and PE are negatively correlated in January (-0.32) and October (-0.38).Tracks entering from PE and WP are positively correlated in December (0.26).
The seasonal and interannual variability in the relative contributions of the tracks from different sources may be due to changes in the TMEs birth or TMEs steering characteristics.To develop some understanding of these issues, the statistics of TMEs birth and the conditional probability of tracks born in a region entering the North East, P(NE|Source) are illustrated in Figure 5 and supplemented in Appendix B Table B.2.  B.2).In terms of the proportion of tracks born in the four source regions that essentially entered NE-US, on average only 13% of these tracks enter the NE-US, with a minimum of 8% in August, a maximum of 20% in November and December and an active 18% in JFM.Consequently, the seasonal cycle of the tracks entering the NE-US has a minimum in August, with peaks in December to March and in June.
On average, 42% of the GP TMEs enter NE-US with small variation from month to month, slightly lower in May-August, and higher in March-April and September to December (Figure 5(b) & Table B.3).The GP birth, i.e.P(GP) (Figure 3 (a)), varies across the year with a peak in June, while its entrance, i.e.P(NE|GP) (Figure 6 (a)), undergoes small changes through the year but has strong interannual variations in January and October.A t-test for the difference in mean GP TMEs entrance counts considering unequal variance in each phase, i.e., El Niño and La Niña phases, yields no difference.Therefore, the variations in the birth process of GP tracks over the year dominate the contributions to the seasonality of the TMEs that enter the NE-US.(1989 -2010).TMEs active entering days are categorized as the days which have the number of total TMEs tracks entering NE-US exceeding its monthly 90% percentile; the composite is done for each month, with 90% percentiles calculated for each month over the 22 years; TMEs Inactive entering days are categorized as the days which have the number of total TMEs tracks entering NE-US below its monthly 10% percentile; the composite is done for each month, with 10% percentiles calculated for each month over the 22 years; the four major TMEs sources in the tropics are marked in blue rectangular boxes with their names: PE (Pineapple Express), GP(Great Plains), GS(Gulf Stream) and WP (West Pacific).Unit of the color bar is hPa.(1989 -2010).TMEs active entering days are categorized as the days which have the number of total TMEs tracks entering NE-US exceeding its monthly 90% percentile; the composite is done for each month, with 90% percentiles calculated for each month over the 22 years; TMEs Inactive entering days are categorized as the days which have the number of total TMEs tracks entering NE-US below its monthly 10% percentile; the composite is done for each month, with 10% percentiles calculated for each month over the 22 years; the four major TMEs sources in the tropics are marked in blue rectangular boxes with their names: PE (Pineapple Express), GP(Great Plains), GS(Gulf Stream) and WP (West Pacific).Unit of the color bar is hPa.

TMEs and Extreme Precipitation
The moisture release from the tracks in NE-US is highly correlated with the total number of TMEs entering NE-US (concurrent correlation between number of TMEs entering and ∆Q in NE-US is 0.88 (p-value < 10-4).We estimated the conditional density of daily precipitation given total daily change of specific humidity (∆Q in Eq. 2) of TMEs entering the NE-US using the local polynomial density estimation with the R package 'hdrcde' (Kim et al., 2011).Figure 9 shows that as the daily moisture release (∆Q) by the TMEs increases, the daily precipitation increases with a shift in the conditional distribution that is marked beyond a threshold of ∆Q of about 3500 g kg-1.This observation based on data pooled over the whole year motivates a seasonal analysis of the association between extreme TMEs and extreme precipitation for different seasons, which is illustrated in Figure 10.The boxplot of the moisture releases (∆Q) from the TMEs to NE-US are significantly different (p-value 0.01 to 0.001) with two-sample Kolmogorov-Smirnov test) given extreme or non-extreme rainfall states for all the four seasons.An extreme rainfall event is defined as one exceeding the 99th percentile of daily rainfall (including days with no or trace rain) in that season, e.g., an extreme rainfall event in DJF must exceed the 99th percentile of daily rainfall amounts in all the Dec, Jan and Feb over the 22 years.The seasonal 99th percentile thresholds are 12.7cm (DJF), 12.2cm (MAM), 11.4cm (JJA) and 15.0cm (SON).For all the four seasons, non-extreme rainfall days have their average moisture releases from TMEs around 0 g kg-1 with thin, long tails.We further examined the ENSO influence on the extreme rainfall events occurrence.The number of extreme events fall into each ENSO episode is tabulated in Table 2. Summarizing on a monthly basis for the 22 years, there are 36 extreme events in La Niña years, with 16 events in El Niño years out of 83 total events at the 99 th of daily rainfall by season.Extremes in June have 5 out of total 7 in La Niña years, against 0 in El Niño years.March, April and September also have similar observations.However, due to the limitation of the sample size, the statistical significance is weak (e.g., p=0.18 for June, even though we have 0 El Niño and 5 La Niña cases).
On a monthly basis, the numbers of TMEs entering the NE-US for extreme precipitation events is consistently larger than those for non-extreme events.The definition of monthly extreme rainfall events is the same as that for seasonal extreme events except that the 99 th percentile thresholds are taken for each month.Table 3 provides the ratio of the average number of TMEs tracks entering NE-US from each source in every month.The ratio is the average over all extreme events divided by the average over all non-extreme events.As the year-around major source, GP shows consistent intensification of TMEs entrance on extreme rainy days.The average TMEs from GP on extreme rainy days are 4 to 7 times of the average on non-extreme days.The difference of TMEs counts for days in the above and below 99 th percentile rainfall categories is statistically significant with a p-value of less than 0.001 for the null hypothesis of no difference.The second major source, GS shows year-around intensification of TMEs entrance except for July when the ratio is close to 1.The intensification is stronger in GS than GP due to the fact that the average TMEs from GS on non-extreme rainfall days are less than those from GP, but their average number of TMEs tracks on extreme rainy days are comparable.This suggests that for extreme rainfall events, GP and GS are both important.Both PE and WP are active contributors to NE-US from Oct to Apr (Figure 6 (c) & (d)) and contribute very little from May to Sep.The average TMEs entrance from PE and WP are less than 3 for non-extreme rainy days from May to Sep/Aug, and hence the corresponding entries in Table 3 are left blank.Although the ratios for PE in Mar and for WP in April are large, their contributions to extreme rainfall events are less than those from GP and GS.Table 3.The ratio of average TMEs tracks entering NE-US from each source: extreme rainfall events over non-extreme rainfall events; extreme rainfall events are defined as the ones exceeding the 99 th percentile of daily rainfall for each month; some entries are blank because the average number of tracks are less than 3 for non-extreme events; the differences between the two samples (i.e.

Summary and Discussion
The key findings of the paper are summarized as follows: 1.The NE-US floods in the four seasons are closely related to TMEs as evidenced by the historical flood events (Figure 2).
2. The four major moisture sources of TMEs account for approximately 85% of all the TMEs entering the NE-US, by calculating the ratio of TMEs born in these four major moisture sources over all TMEs that entering the study region (given in Table B.1 that 36694/43256 ≈ 85%).The birth processes in the four source regions are relatively independent, except for moderate association between GP and GS in some months.They all have strong seasonality and interannual variation, which determine their contributions to the NE-US.Their overall contributions can be ordered as GP>GS>PE>WP (by comparing the last row in Table B.1), with GP and GS as the year-around sources, and PE active in winter, and WP as the smallest contributor.
3. Depending on the month, some of the interannual variations of TMEs birth are associated with ENSO phases (Figure 3).Since GP is the dominant contributor and year-around source, the influence of ENSO on GP-TMEs' birth affects the TMEs' entrance to NE-US.The intensification of TMEs born between October and January during El Niño suggests that more TMEs could potentially enter NE-US under the influence of ENSO warm phase, which may result in more moisture release and precipitation.Since the ENSO warm phase also leads to an intensification of the GS birth process, the two major sources would contribute more TMEs potentially, when the steering mechanisms cooperated in the convergence of tracks to the NE-US.
The GP and GS TMEs track entrances are highly correlated over several months, and this may reflect the common influences on the birth and steering processes.).However different from the ENSO effect on the birth process, no significant difference was observed among different ENSO episodes for the entrance given a birth source, suggesting the ENSO has more influence on birth than steering mechanism.The composite SLPa of active TMEs days and inactive TMEs days suggests that low wave number patterns of atmospheric circulation mark the anomalous transport (Figure 7 and Figure 8).Depending on the calendar month, these patterns are not always symmetric for active vs inactive TMEs periods suggesting that the circulation processes driving the TMEs to the NE-US are nonlinear.
5. The number of tracks entering and the associated moisture release are highly correlated, with correlation 0.88 (p-value < 10 -4 ).This translates into a strong influence of active TMEs periods on the occurrence of extreme rainfall.The distribution of TMEs for extreme rainfall events (>the 99 th percentile of daily rainfall) is significantly different from the one given non-extreme events (Figure 10), suggesting the important role of TMEs in determining extreme precipitation.This argument carries forward into shifts in extreme rainfall event occurrence across different ENSO phases as they influence the TMEs birth and steering.
The study of atmospheric rivers as an influence on floods induced by extreme precipitation has significantly enhanced the interaction between hydrologists and climate scientists towards an improved understanding of the synoptic and climatological factors that govern such phenomena.The NE-US has not been the target of many of these investigations, partly because of the way atmospheric rivers have been defined, although some debate on the definition still remains.The Knippertz and Wernli work established the broader role of tropical moisture exports in the climatology of mid-latitude precipitation, and provided a data base that allows a consistent exploration of the Lagrangian transport of moisture from the tropics to the mid-latitudes.Our initial work on a flood over Western France in 1995 revealed that at synoptic scales systematic organization of moisture as reflected in the data not including 1995 could be identified by a simple statistical model and used effectively for an out of sample prediction of the extreme precipitation in January 1995 that led to the flood event.Several of the moisture tracks associated with that flood event also had a moisture release over the NE-US, stimulating the work reported in this paper.Here, we explore climatological aspects of the links between TMEs birth, steering, moisture release and extreme precipitation, providing the first such chronology of the year-round links between these factors, as well as the potential links to ENSO.
Given the short record of TMEs available to us, a structured exploration of interannual variability and an association with ENSO and lower frequency phenomenon was not possible.However, we expect to extend the work reported here in multiple directions.First, we expect that a national analysis will be much more informative as to the spatial and temporal shifts in TMEs' influence on extreme precipitation by season and as driven by identified low frequency climate mechanisms.Developing a formal spatiotemporal model that considers the connectivity across such a source-destination network and its modulation by atmospheric and ocean circulation precursors is a challenge worthy of pursuit.As we develop such a model, we expect to utilize longer atmospheric-reanalysis data and to develop a longer TMEs record from it using LAGRANTO as was done by Knippertz and Wernli using the ERA-Interim data.Such a development may provide a much better empirical understanding of extreme precipitation dynamics across the USA, and thus provide an important diagnostic tool for the performance of climate models for seasonal forecasting or for climate change simulations.The nationwide study will also consider the recycling of moisture suggested by Dirmeyer et al. (2009).

Figure 1 .
Figure 1.Influence diagram of the factors considered and their proposed relationship (as the arrows directed) investigated in this paper.EP is extreme precipitation, ∆ represents moisture release from the TMEs tracks

Figure 2 .
Figure 2. (a) -(f): DJF Flood: TMEs tracks that born between the 13 th and the 18 th Jan in 1996, when two successive flood events occurred, one started on the 15 th in New York, Pennsylvania, Ohio, West Virginia, New Jersey; the other started 4 days later in Chateauguay, Quebec.(g) -(l): MAM flood: TMEs tracks that born between the 24 th and the 29 th Mar in 2005; broad area including New York, New Jersey, Delaware had flood started on the 1 st of April.(m) -(r): JJA flood: TMEs tracks that born between the 17 th and the 22 nd Jul in 1998; heavy rainfall induced flood event in Green Mountain to Bradford, Vermont and New Hampshire.(s) -(x) SON flood: TMEs tracks that born between the 3 rd and 8 th Oct in 2005 associated with 10 days flooding from the 8 th to the 17 th of Oct in Southwest New Hampshire, Vermont, New Jersey and Connecticut.The red box highlights the NE-US with color indicating the changes of moisture from the tracks: red = recharge, blue= release; each dot corresponds to a 6-hour update of the air parcel position, i.e., trajectory of the tracks.Enlarged panels are provided in the Appendix A. The unit of the color bar is g kg -1 .(Dartmouth Flood Observatory Archive http://floodobservatory.colorado.edu/)

Figure 2
Figure2(a) -(f) shows that the associated TMEs for the January 1996 flood (Table1) were born in (1) 'Pineapple Express' (PE), i.e.Niño 3.4 region near Hawaii; (2) 'Great Plain' (GP), i.e.Gulf of Mexico and (3) 'Gulf Stream,' (GS), i.e. tropical North Atlantic ocean east to the Bahamas.The beginning of the heavy rainfall appears related to tracks born in all the three regions as shown in Figure2(a) -(c).Later in the event (Figure2(d) -(f)) TMEs that entered the NE-US were all born in GP and GS.The change of moisture content of the tracks indicates that for tracks born in PE, the relatively longer travel time and distance resulted in more recharging and releasing on their way to the NE-US, compared to GP and GS, which featured less changes of moisture content before reaching the flooded area, and relatively more release in the NE-US (more dark blue dots of tracks in Figure2 (d) -(f) in the red box -NE-US).There are strong spatial patterns of the trajectories given the TMEs tracks' sources: tracks from PE follow typical northeastwardly 'Pineapple Express' sine-wave like trajectory showed in the AR literature focusing on the west coast [e.g.,(Dettinger, 2011;Dettinger et al., 2011;Ralph and Dettinger, 2011)]; tracks from GP and GS penetrated to the north until reaching 50°N, where the jet stream locates, and then turned to the east.Lu et al. (2013) showed remarkably similar trajectories of the two major sources, i.e.GP and GS, associated with the more than 100-yr heavy precipitation induced flood event in Western France in January 1995.It suggested that moist air parcels originated from the tropics can have far impacts on extremes in the higher latitudes; and floods occurred at different time and over different regions may share the same moisture sources in terms of the TMEs.
(h), (j) -(l) show that the most extensive releases were occurred in the NE-US.The major moisture sources associated with exemplified summer flood in June 1998 are GP and GS, with some from the tropical East Pacific.The most notable feature in Figure2(m) -(r) is the well-organized spatial trajectory of all TMEs tracks that bring moisture from the warmer tropical ocean to the Great Plains, starting to release moisture in the Mississippi river basin before reaching NE-US.A flood event in Southwest Iowa occurred right before this NE-US flood, which agrees with the observation that the TMEs (Figure2(m) -(r)) followed the path passing both of the flooded regions in a consistent order.Note that only TMEs tracks that finally entered NE-US have been retained in Figure2(m) -(r), and there were additional TMEs tracks that were responsible for the Southwest Iowa floods but they did not eventually reach the NE-US.Due to the longer distance of these trajectories and the weaker westerly in summer, it took longer for the TMEs to reach NE-US after they were born.The trajectories of TMEs exiting the NE-US were different in Figure2(m) & (n) and Figure2(q) & (r).The TMEs in (Figure2(m) & (n)) went further north to Quebec, Canada; while the ones born on the 21 st and 22 nd (Figure2(q) & (r)) followed a wavelike trajectory starting from the Great Lakes.The changing trajectories were likely associated with synoptic scale transients(Lu et al., 2013).

Figure 2
Figure 2 (s) -(x) shows the TMEs tracks that contributed to the October 2005 flood event (Table1).The major moisture sources identified are GP and GS.The most notable feature in Figure2 (s) -(x) is the similarity to the trajectories of those born in GS in DJF (Figure2 (a) -(f)).TMEs born in GS remained in the oceanic sector before reaching the NE-US, which resulted in a continuous recharging of moisture to the tracks.At the same time in October 2005, it was reported that remnants of Tropical Storm Tammy and Subtropical Depression Twenty-two merged with incoming continental cold fronts to produce torrential rains over the NE-US.The trajectories of the GS TMEs are consistent with these storms.Such a feature is not typically consistent with ARs.The release of the moisture was concentrated in the NE-US.

Figure 3 .
Figure 3. Seasonality and interannual variations of the monthly total number of TMEs born in the four major source regions (a) GP, (b) GS, (c) PE and (d) WP and the influence of ENSO scenarios on their monthly average amounts: blue is the composite of the El Niño years; red is the composite of the La Niña years; black is the composite of the neutral years

Figure 4 .
Figure 4. Boxplots of the conditional probabilities of TMEs birthplace (sources regions), given TMEs tracks entered NE-US showing the seasonal and interannual variability of contributions of TMEs tracks from (a) GP, (b) GS, (c) PE and (d) WP

Figure 5 .
Figure 5. (a) Seasonal variation of the TMEs born in the four major sources and the seasonally varying portions of these TMEs entering NE-US, note that the TMEs born refer to the y-axis on the left and entrance refer to the y-axis on the right, whose scale is 1/10 of the left; (b) seasonal variation of the probability of TMEs entering NE-US given TMEs born in four major source regions

Figure 6 .
Figure 6.Seasonality and interannual variations of the conditional probabilities of TMEs Entrance to NE-US for each month, given TMEs born in the four major source regions in the tropics: (a) GP, (b) GS, (c) PE and (d) WP; and the influence of ENSO episodes: blue is the composite of the El Niño years; red is the composite of the La Niña years; black is the composite of the neutral years

Figure 7 .
Figure 7. Composite of daily Sea Level Pressure anomalies of the top 10% TMEs active entering NE-US days in each month from January to June ((a) -(f)), over the 22 years(1989 -2010); and composite of daily Sea Level Pressure anomalies of the top 10% TMEs inactive days in each month from January to June, (g) -(l), over the 22 years(1989 -2010).TMEs active entering days are categorized as the days which have the number of total TMEs tracks entering NE-US exceeding its monthly 90% percentile; the composite is done for each month, with 90% percentiles calculated for each month over the 22 years; TMEs Inactive entering days are categorized as the days which have the number of total TMEs tracks entering NE-US below its monthly 10% percentile; the composite is done for each month, with 10% percentiles calculated for each month over the 22 years; the four major TMEs sources in the tropics are marked in blue rectangular boxes with their names: PE (Pineapple Express), GP(Great Plains), GS(Gulf Stream) and WP (West Pacific).Unit of the color bar is hPa.

Figure 8 .
Figure 8. Composite of daily Sea Level Pressure anomalies of the top 10% TMEs active entering NE-US days in each month from July to December ((a) -(f)), over the 22 years (1989 -2010); and composite of daily Sea Level Pressure anomalies of the top 10% TMEs inactive days in each month from July to December, (g) -(l), over the 22 years(1989 -2010).TMEs active entering days are categorized as the days which have the number of total TMEs tracks entering NE-US exceeding its monthly 90% percentile; the composite is done for each month, with 90% percentiles calculated for each month over the 22 years; TMEs Inactive entering days are categorized as the days which have the number of total TMEs tracks entering NE-US below its monthly 10% percentile; the composite is done for each month, with 10% percentiles calculated for each month over the 22 years; the four major TMEs sources in the tropics are marked in blue rectangular boxes with their names: PE (Pineapple Express), GP(Great Plains), GS(Gulf Stream) and WP (West Pacific).Unit of the color bar is hPa.

Figure 9 .Figure 10 .
Figure 9. Kernel conditional density of daily precipitation given total changes of specific humidity of TMEs entered NE-US from 1989 -2010, estimated using local polynomial.∆Q is calculated by Eq.1

4.
The seasonal and interannual variations in atmospheric circulation patterns also play an important role in determining the TMEs' entrance to NE-US.Aggregating cross the four major sources, the annual maximum of TMEs entrance occurs in June and the minimum occurs in winter (Figure5 (a)).The order of importance of the four sources is consistent with observations in the floods examples in the beginning: GP>GS>PE>WP(Figure 5 (b)

Figure A. 2 .
Figure A.2. Enlarged figure for Figure 2 (g) -(l): MAM flood: TMEs tracks that born between the 24 th and the 29 th Mar in 2005; broad area including New York, New Jersey, Delaware had flood starting on the 1 st of April

Figure A. 3 .Figure
Figure A.3.Enlarged figure for Figure 2 (m) -(r): JJA flood: TMEs tracks that born between the 17 th and the 22 nd Jul in 1998; heavy rainfall induced flood event in Green Mountain to Bradford, Vermont and New Hampshire

Table 1 .
Facts of the example flood events in the each of the four seasons, records are from Dartmouth Flood Observatory Archive

Table 2 .
ENSO influence on the number extreme rainfall events, i.e. exceeding the 99 th percentile threshold of that calendar month over the 22 years' data period