Atmospheric Rivers in the Eastern and Midwestern United States Associated With Baroclinic Waves

Atmospheric rivers (ARs) significantly impact the hydrological cycle and associated extremes in western continental regions. Recent studies suggest ARs also influence water resources and extremes in continental interiors. AR detection tools indicate that AR conditions are relatively frequent in areas east of the Rocky Mountains. The origin of these ARs, whether from synoptic‐scale waves or mesoscale processes, is unclear. This study uses meteorological composite maps and transects of AR conditions during the four seasons. The analysis reveals that ARs east of the Rockies are associated with long‐wave, baroclinic Rossby waves. This result demonstrates that eastern North American ARs are dynamically similar to their western coastal counterparts, though mechanisms for vertical moisture flux differ between the two. These findings provide a foundation for understanding future climate change and ARs in this region and offer new methods for evaluating climate model simulations.

The vast majority of papers in the AR literature are focused on studies of western coastal zones, with most centered specifically on the United States (US) West Coast where much of the early research on ARs was directed.That said, there is an increasing recognition that ARs are also important in other regions, such as continental interiors and polar regions (Gorodetskaya et al., 2014;Nash et al., 2018;Wille et al., 2019), the interiors of Australia and China (Liang et al., 2020;Nash et al., 2021;Rauber et al., 2020;Reid et al., 2021;L. Xu et al., 2020;Y. Xu et al., 2020;H. Zhang et al., 2020), the Middle East and North Africa (Massoud et al., 2020), and the interior of the US east of the Rocky Mountains (Dirmeyer & Kinter, 2009, 2010;Moore et al., 2012;Slinskey et al., 2020).
For two specific examples, significant flooding events have occurred in the midwestern US in association with ARs: one in Nashville, Tennessee on 1-2 May 2010 (Moore et al., 2012) and one in Bloomington, Indiana on 18-19 June 2021.The Bloomington flood was a 100-year event in which multiple rain gauges recorded over 15 cm of rainfall in a 24-hr period.Analysis of the associated meteorology shows that the flood was associated with the combination of an AR, a cold frontal zone, and a mesoscale convective complex (see Figure S1 in Supporting Information S1).
Several studies (Lavers & Villarini, 2013;Mahoney et al., 2016;Nakamura et al., 2013;Nayak et al., 2016;Slinskey et al., 2020) demonstrate the importance of ARs for extreme precipitation in areas of the US east of the Rocky Mountains.However, the meteorological drivers of these ARs (hereafter "eastern" ARs for brevity) are unclear, as the literature on them is relatively sparse.Western US ARs are frequently associated with midlatitude cyclones; indeed, this frequent association appears in the definition of ARs (Ralph et al., 2018), and Lavers and Villarini (2013) show composites of mean sea-level pressure suggesting the influence of synoptic-scale dynamics on eastern ARs.In seeming contrast, some literature (Dirmeyer & Kinter, 2010;Gimeno et al., 2016) suggests that eastern ARs may be associated with the Great Plains Low Level Jet (GPLLJ).
This study aims to answer the question what is the primary meteorological driver of ARs that occur east of the Rocky Mountains?The literature suggests two hypotheses: (a) eastern US ARs, like their west coast counterparts, are driven by synoptic-scale eddies; and (b) eastern US ARs are associated with the GPLLJ.If baroclinic waves are the primary driver, then we would expect the signatures of these midlatitude systems to be evident in meteorological composites of times that satisfy AR conditions in the central US, and we would expect this signature to be present in all seasons.In contrast, the GPLLJ is primarily a warm-season phenomenon and is thought to be regulated by an inertial oscillation modulated by a consistent meridional buoyancy gradient (Gebauer & Shapiro, 2019), rather than synoptic-scale waves.
Using composites of reanalysis data, we find support for the baroclinic Rossby wave hypothesis.Our results show that eastern ARs are dynamically similar to their well-studied west coast counterparts in terms of their association with baroclinic waves.
We test the sensitivity of our results to choice of ARDT and to location by repeating the entire analysis with a more permissive ARDT, guan_waliser_v2 (Guan & Waliser, 2015;Guan et al., 2018), and by repeating the entire analysis with teca_bard_v1.0.1 in two different locations: Norman, OK and Washington, DC.These locations were chosen to represent three major regions east of the Rockies: the Great Plains, the Midwest, and the East Coast.Both the teca_bard_v1.0.1 and guan_waliser_v2 data come from the AR Tracking Method Intercomparison Project (ARTMIP) Tier 1 database (Shields et al., 2018), which spans the years 1980-2017.Details about each ARDT and results of these sensitivity studies are provided in Supporting Information S1 (Text S1 and Figures S2-S13).The guan_waliser_v2 ARDT detects nearly 10 times more timesteps with AR conditions occurring over Bloomington, IN because the requirement for "high-confidence" AR conditions with teca_bard_v1.0.1 is much more restrictive.The sensitivity study with teca_bard_v1.0.1 in Washington, DC yields similar numbers of ARs as Bloomington, IN, but the sensitivity study with teca_-bard_v1.0.1 in Norman, OK yields about a third fewer; see Table S1 in Supporting Information S1.
We generate composites of various meteorological quantities during the Bloomington, IN AR timesteps, as indicated above, within each season using the ERA5 reanalysis (European Centre for Medium-Range Weather Forecasts, 2019; Hersbach et al., 2020).Note that the AR timesteps come from MERRA-2 due to our use of the ARTMIP data set, but the meteorological composites come from ERA5.We utilize geopotential height, temperature, integrated vapor transport, integrated water vapor, winds, potential vorticity, vertical velocity, and mean sea-level pressure.Composites are generated using the teca_temporal_reduction application available within TECA (Loring et al., 2022;Prabhat et al., 2015).Additional methodological details, such as the objective method used to identify surface low and high pressure regions in the composite maps, are described in Text S2 in Supporting Information S1.

Results
Figure 1 shows composites of integrated vapor transport (IVT; vertically integrated horizontal moisture flux), total column water vapor (IWV), and the locations of surface lows and highs for all four seasons.The IVT and IWV fields show the distinctive signature of AR conditions, namely a long, narrow band of high water vapor transport co-located with high precipitable water content.In all four seasons, a surface low is present to the northwest of the central AR zone (southern Indiana), and a surface high is present over the Atlantic Ocean in all seasons except JJA which instead shows a broad ridge pattern over the region.The ARs occur within a region of high surface pressure gradient between these low and high-pressure regions.
To put the AR conditions in a synoptic context, Figure 2 shows composites of 850 mb potential temperature, 850 mb heights, 500 mb heights, and the same low/high-pressure regions shown in Figure 1.The upper-level heights show the clear presence of a longwave trough, with the mean trough axis 500-1,500 km to the west of the AR region in all four seasons and a ridge to the east, such that upper-level geostrophic winds are southwesterly over the AR region.The lower-level heights also show a clear longwave pattern, with a phase offset of several hundred kilometers to the east of the upper-level trough axis in all four seasons.The surface low sits within, or just to the east of, the low-level trough.The 850-mb potential temperature field also shows signs of a wave-like pattern, with a mean temperature gradient west of the AR region that would be associated with cold frontal zones, and signs of a warm frontal zone to the east of the AR region.Mean temperature features that could be correlated with fronts are much less well-defined in JJA, consistent with the weaker temperature gradients expected in Northern Hemisphere summer in midlatitudes.
In all four seasons, a mean upper-level trough exists west of the study region.If we were to treat each of the composite maps as representative of a typical event in that season, then this trough location indicates that the cyclonic vorticity associated with the trough is being advected eastward over the study region.The intensification of the winds with height (shown more clearly in Figure 3) indicates that the cyclonic vorticity advection increases with height.Such differential cyclonic vorticity advection is consistent with quasigeostrophic forcing favoring ascent over the region (Holton, 2004).
The transect composites (Figure 3; see Figure 2d for the trace of the transect) show the presence of an upper-level jet with a maximum to the northwest of the AR region (to the left of 0 in the transects) and just below the tropopause in all four seasons.The upper-level jet is strongest in DJF but weakest in JJA, and exhibits a westward tilt in all four seasons, with relatively strong winds from the upper levels down toward the surface.All four seasons also exhibit a relative maximum in wind speed near the surface approximately 200-300 km to the southeast (right of 0 in the transects), which indicates the presence of a low-level jet.These winds are thermallyforced, as indicated by composites generated using geostrophic winds instead of the full wind field; these composites (not shown) are essentially identical to those in Figure 2. The potential temperature field shows indications of a cold frontal region, with a dome of relatively cold air extending from the surface up to about 300 hPa to the northwest (left of 0).The actual values of potential temperature vary according to season, but the general structure of the frontal region is consistent.The maximum gradient in 1,000 mb temperatures is reached at or near the AR region, indicating that individual AR events may be associated with an impinging cold front.
Near-surface specific humidity (green dashed lines in Figure 3) reaches at least 10 g kg 1 in all seasons, with highest values primarily to the southeast of the AR region.The combination of high specific humidity, increased winds associated with the upper-level jet, and increased winds in the lower atmosphere result in high moisture transport directly over the AR region, consistent with the high IVT values shown in Figure 1.The region of high water vapor transport has a westward tilt, similar to the tilt in the tropospheric wind maximum, suggesting the importance of the upper level flow in generating the high IVT that defines the AR.

Discussion
Figures 1 and 3 bear a strong similarity to the map and transect plots shown by Ralph et al. (2018) in the American Meteorological Society Glossary definition of ARs: strong, filamentary moisture transport to the southeast of a surface low and cold frontal zone; and high moisture transport associated with high surface humidity and southwesterly winds from an upper-level jet and a pre-frontal low-level jet.Based on the qualitative definition given by Ralph et al. (2018), and based on the objective detection of AR conditions by TECA-BARD, it seems clear that the ARs discussed here are phenomenologically similar to their western coastal counterparts.
Likewise, Figures 2 and 3 show the distinctive characteristics of a longwave baroclinic Rossby wave: an upperlevel jet, presence of a frontal zone and a surface low, and westward tilting wind and potential temperature fields indicative of baroclinic waves.The westward tilt in the moisture flux suggests that the moisture flux is associated with the synoptic-scale, geostrophically-driven winds.This argues strongly against the hypothesis that central US ARs are primarily associated with the Great Plains low-level jet (GPLLJ).The clear signature of a baroclinic wave and upper-level dynamics (e.g., the tropopause folds in Figure 3) indicate that the moisture flux is associated with synoptic processes rather than the more mesoscale (and possibly boundary layer) scale processes associated with the Great Plains low-level jet.Note that this does not rule out the possibility that the GPLLJ is present during these AR conditions; notably, a masters thesis by Gyawali (2022) shows that most central Great Plains ARs also occur with a detected GPLLJ.But two factors suggest that synoptic-scale processes, rather than the GPLLJ, are the primary driver: (a) the similarity of the composites between seasons when the GPLLJ is not considered to be important (DJF) and seasons when it does have some influence (MAM and SON), and (b) Gyawali (2022) notes the similarity between mid-level height composites of AR + GPLLJ conditions and the dynamically-coupled GPLLJ composite conditions discussed by Burrows et al. (2019) in which the GPLLJ seems to be synoptically controlled.
Composites from all four seasons support the general idea that eastern US ARs are driven by longwave baroclinic Rossby waves, though there are some differences that are worth further investigation.The low amplitude of the upper-level wave in JJA (Figure 2c) is likely the result of averaging over multiple types of synoptic state such that the composite-mean pattern is weak, in contrast with the other seasons in which the composite-mean pattern is more representative of the typical synoptic state (see Figure S14 in Supporting Information S1).Additionally, DJF stands out from the other seasons in that the mean surface low is nearly co-located with the center of the AR (see Figure 2a) instead of being located well to the northwest of the AR.It is possible that surface convergence associated with lows in DJF may enable moisture-and resultant upper-level heating-from the AR to contribute to rapid deepening of these lows (Zhu & Newell, 1994;Z. Zhang et al., 2019).The use of simulation-based experiments and lagged composites may help clarify this.
There are three forms of uncertainty that may impact the conclusions here: uncertainty in the detection of ARs, uncertainty associated with the choice of region over which to composite, and uncertainty in the reanalysis.Sensitivity tests using a different AR detection tool (guan_waliser_v2) and focusing on two different regions (Washington, DC and Norman, OK) show qualitatively identical results: Figures S2-S5 in Supporting Information S1 for the ARDT sensitivity test; and Figures S6-S13 in Supporting Information S1 for the region sensitivity test.Additionally, the use of the guan_waliser_v2 ARDT helps show that the coherent Rossby wave signatures in Figures 1-4 are not a trivial consequence of using an ARDT that detects synoptic-scale anomalies in vapor transport (see Text S1 in Supporting Information S1).Inda-Díaz et al. (2021) show that most ARDTs, and the guan_waliser_v2 ARDT in particular (see ARDT 17 in their Figure 1), detect ARs of a wide variety of sizes: from mesoscale up to synoptic scale.Finally, Collow et al. (2022) show that ARDTs can be sensitive to reanalysis.We compare ARs detected by teca_bard_v1.0.1 in MERRA-2 to ARs detected by teca_bard_v1.0.1 in ERA5 and find a high level of agreement (most times when AR probability is 100% in MERRA-2, AR probability is at least 95% in ERA5 and is 100% most of the time).All together, these analyses indicate that the composites presented here are robust to uncertainty in ARDT, geographic location, and reanalysis.
Taken together, Figures 1-3 provide strong evidence that eastern US ARs are dynamically similar to their wellstudied western US counterparts, though a key difference between the two is the mechanism for uplift and generation of precipitation.Orographic ascent in neutrally-stratified atmosphere provides an efficient mechanism for upward moisture flux (Cobb et al., 2021;Neiman et al., 2002;Neiman, Ralph, Wick, Kuo, et al., 2008;Ralph et al., 2005).The ubiquitous mountain ranges in the western US (e.g., the Coast Ranges, the Cascades, and the Sierra Nevadas) can provide this orographic forcing for ARs (B.L. Smith et al., 2010), though atmospheric stability and AR angle modulate the effectiveness of this orographic forcing (Hughes et al., 2014;Kingsmill et al., 2013;Neiman et al., 2002).In contrast, the relative dearth of topography in the area between the Rocky Mountains and the Appalachian mountains means that any upward moisture flux must come from dynamical and/ or convective processes, such as isentropic lift or convective instability.
Analysis of composite vertical velocities shows a broad area of low-level updraft across the majority of the AR region: Figure 4 shows composite vertical velocities at 700 hPa (in pressure coordinates: negative velocities indicate upward motion) over regions where IVT is greater than the 250 kg m 1 s 1 threshold that is often used as a baseline for AR presence (Rutz et al., 2014).These velocities reach up to 0.3 hPa s 1 in all seasons, though these values might be damped by the reanalysis resolution and the act of compositing.Considering that Figure 3 shows specific humidity values in the range 1-10 g kg 1 , this corresponds to a vertical moisture flux of O(0.1) mm d 1 or smaller.A moisture flux of this magnitude is too low to explain the extreme precipitation associated with AR conditions discussed by Slinskey et al. (2020) and shown in Figure S1 in Supporting Information S1.However, this broad region of synoptic-scale uplift may be enough to initiate convection.
Among the four seasons, all but DJF have appreciable mean convectively available potential energy (CAPE; see Figure 4 and Figures S5, S9, and S13 in Supporting Information S1) over the study region, and even DJF shows hints of elevated CAPE extending from the Gulf of Mexico.This suggests that ARs in this region fuel convection through providing: (a) an adequate supply of high moisture content, (b) a source of unstable air, and (c) a broad region of upward motion.Even absent an orographic source of uplift, these three factors combine to provide an efficient mechanism for translating horizontal moisture flux into intense vertical moisture flux within convective regions.These three ingredients, in combination with the wind shear (Figure 2) associated with the growing baroclinic wave that drives the AR, are well-known ingredients for severe convective environments.One therefore might expect a strong association between mesoscale convective systems (MCS) and ARs in this region, and this warrants further study.This association between ARs and environments favorable for MCS development may also open new opportunities for using paleoclimate proxies to study ARs and climate change.For the western US, the presence of terrigenous sediment layers can provide a proxy of AR-driven activity, since terrestrial flood events tend to be primarily associated with ARs in the region (Du et al., 2018;Hendy et al., 2015).Such a proxy is inapplicable in the continental interior, but recent work by Sun et al. (2021) shows that the hydrogen isotopic composition of leaf wax preserves a signal associated with MCS.The authors primarily associate this proxy with changes in the GPLLJ, but analysis of paleoclimate simulations suggest that ARs-and changes therein-may have played a major factor in the hydroclimate of the continental interior since the Last Glacial Maximum (Lora et al., 2023;Skinner et al., 2020Skinner et al., , 2023)).Taken together, the analysis here suggests that ARs may be a factor in modulating MCS activity in the region.Further analysis of the proxy developed by Sun et al. (2021) may provide a novel way to study paleoclimate changes in ARs in the continental interior.

Conclusions
This analysis provides clear evidence that ARs in the eastern US are driven by synoptic-scale processes, and in particular that ARs seem to be associated with longwave baroclinic Rossby waves.This does not preclude the idea that the GPLLJ can sometimes play a role in these ARs, but the evidence presented here suggests that the primary means of generating strong, and southwesterly, horizontal moisture flux is through geostrophic forcing of winds from a synoptic-scale wave.This horizontal moisture flux-and associated unstable air-then drives vertical moisture flux (and precipitation) through convective processes rather than orographic processes as in the western US.
As Slinskey et al. (2020) report, a high proportion of central and eastern US extreme precipitation is associated with ARs, but it is not known whether this extreme precipitation results from ARs alone.Figure S1 in Supporting Information S1 indicates that some extreme precipitation events are associated with more than one meteorological phenomenon (e.g., a front, an AR, and a mesoscale convective system as in that case), and analysis of Figures 2-4 suggest that these ARs occur in an environment favorable for mesoscale convection.It is not clear how frequently such co-occurrences happen or whether they systematically intensify precipitation.We are currently working on follow-up studies to assess this.
Given that eastern US ARs are synoptically forced, it seems reasonable to expect that climate models should be able to resolve this association between midlatitude cyclones and ARs in this region.A recent intercomparison of simulations and AR detection tools shows that most climate models simulate a relative maximum in AR frequency in the midwestern and eastern US (O'Brien et al., 2021), suggesting that this may be the case.Building composites, like the ones shown here but for historical climate model simulations, could provide a way to directly evaluate the dynamics of simulated ARs.In contrast, the mechanisms for vertical moisture flux-which appear to be convective in nature-could be quite challenging for models to adequately simulate.Such a phenomenonfocused perspective could provide a way to elucidate specific model deficiencies as well as possible indications for how to fix them.A recent workshop has advocated for such an approach as a promising way to rapidly improve the simulation of precipitation in climate models (Pendergrass et al., 2020).This work helps pave the way for advancing a theory-based understanding of ARs and climate change in the eastern US that builds on the well-established thermodynamic scaling of moisture (i.e., Clauius-Clapeyron scaling) in ARs (Payne et al., 2020).The results here show that eastern US ARs are strongly associated with midlatitude cyclones, and there is an increasing body of literature about the theoretical effects of climate change on the location and frequency of these storms (Feldl et al., 2017;Shaw, 2019;Shaw et al., 2016).Overall, it could be beneficial to extend this work further to assess the degree to which different areas of high AR frequencyparticularly the inland ones-seem to be associated with midlatitude cyclones.

Figure 1 .
Figure 1.Composite maps of integrated vapor transport (shading), total column water vapor (red contours), and surface low and high pressures (L and H symbols) for atmospheric river conditions over Bloomington, IN in (a) DJF, (b) MAM, (c) JJA, and (d) SON.The dashed black curve in (d) shows the location of the transect in Figure 3, and the yellow star shows the location of Bloomington, IN.

Figure 2 .
Figure 2. Composite maps of 850 mb potential temperature (shading), 850 mb heights (dark gray contours), 500 mb heights (light gray contours), and surface low and high pressures (L and H symbols) for atmospheric river conditions over Bloomington, IN in (a) DJF, (b) MAM, (c) JJA, and (d) SON.The dashed black curve in (d) shows the location of the transect in Figure 3.

Figure 3 .
Figure 3. Composite transects of potential temperature (shading), transect-normal winds (blue curves), specific humidity (dotted green curves), moisture flux at the 60 g m kg 1 s 1 level (red contour), and the 2 PVU potential vorticity contour (black curve) for atmospheric river conditions over Bloomington, IN in (a) DJF, (b) MAM, (c) JJA, and (d) SON.Thick, transparent blue curves in all four panels show frontal zones.The trace of the transect is shown in Figure 2d.

Figure 4 .
Figure 4. Composite of upward vertical velocity at 700 mb within regions of IVT higher than 250 kg m 1 s 1 , CAPE (colored contours), and the location of lows, highs, and the transect trace as in Figure 2.