The Contribution of Eastern North Pacific Tropical Cyclones to the Rainfall Climatology of the Southwest United States

a. Storm tracks, genesis, and motion

The NHC/TPC tracks of the 35 TCs that brought rainfall to the southwest United States are shown in Fig. 3. The tracks in black (the storms for which precipitation plots are shown in Figs. 8 and 10) are labeled by the first letter of the storm name and year (e.g., K76 is Hurricane Kathleen 1976) at the beginning and end of the track. There is a cluster of genesis points between 10° and 15°N and centered on ∼97°W, with a second larger cluster around 17°N and just west of 105°W. The storms that form farther to the south initially move just north of west before turning toward the northwest around the latitude of the northern genesis region. As required by the criterion outlined in the last section, all of the NAM region rainmakers cross 25°N as TCs (with a generally northward motion), but only slightly more than half cross 30°N, and a mere 26% have the end of their tropical tracks over the continental United States. Twenty-two TCs recurved (i.e., acquired an eastward component to their motion) and 26 storms made landfall in Baja California, western Mexico, or both, whereas only one TC in this study (TC Hyacinth 1972) made landfall solely along the U.S. coastline. Dividing the eastern Pacific HURDAT tracks into 1° × 1° grid boxes indicates that the return rate for a TC of any intensity to the U.S.–Mexico border is once every 5 yr.

To put the rare NAM region rainmaking storms in context with TC activity in the eastern Pacific as a whole, Fig. 4 shows the tracks of all TCs that formed east of 140°W and did not cross the date line from 1 August to 15 October 1958–2003. There were a total of 340 TCs during the period, meaning that just 10% of all eastern Pacific TCs and their remnants brought rainfall to the United States (the black tracks in Fig. 4). This number is not surprising considering the mean genesis position (14.4°N, 109.2°W) and motion (slightly north of due west) of all eastern Pacific storms. TCs that eventually affect the southwest United States form ∼6° farther east than the mean (significant above the 99.9% level) but only 1° farther north (significant at the 98% level). This result indicates that storms that form closer to the Mexican coast are much more likely to acquire northward and eastward components to their motion. As we shall see below, this result makes sense in the context of the large longitudinal extent of the eastern Pacific where TCs can form, but the relatively narrow latitudinal range where genesis is favored.

The main genesis region, storm motion, and sharp decrease in TCs north of 25°N can be explained by examining the climatological sea surface temperatures (SSTs), vertical wind shear, and midlevel steering flow for the month of September (the month with the highest incidence of TC rainfall; Table 1) shown in Fig. 5. Looking first at the SST pattern, the warmest waters are found immediately along the western coast of Mexico, extending north to a maximum of over 30°C in the shallow waters of the Gulf of California, and to the west with temperatures greater than 28°C in a swath between 10° and 15°N from the Mexican coastline past 130°W. SSTs decrease south of ∼5°N and also along a sharp SST gradient north of 20°N and west of ∼112°W. Taking into account the nominal 26.5°C for tropical cyclogenesis set forth by Gray (1968), the potential genesis area in the eastern Pacific extends from the Mexican coast westward to 130°W and beyond but is limited in the north–south direction by the cool waters of the two Pacific Ocean eastern boundary currents, California and Peru. Thus, in a climatological sense, TCs that form and track closer to the Mexican coast have a better chance of surviving the trip northward, which fits the NAM region rainmaking TC tracks that are predominantly east of 115°W (Fig. 3).

Figure 5 shows that the small latitudinal area with favorable SSTs for TC genesis is further limited at its southern edge by values of 850–200-hPa vertical wind shear exceeding 12–15 m s⁻¹, above which Zehr (1992) found that TCs failed to develop. This is consistent with Fig. 4, which shows that only a handful of TCs formed south of 10°N following the approximate latitude of the 12 m s⁻¹ shear line. As for the TCs that do develop and track toward the north and west, if the storms remain close enough to the Mexican coastline to avoid the cool SSTs from the California current, the vertical wind shear starts to increase rapidly north of ∼27°N. Shear as weak as 5 m s⁻¹ is enough to produce a strong wavenumber-1 asymmetry in TC convection (Frank and Ritchie 1999; Corbosiero and Molinari 2002); a 10 m s⁻¹ shear weakened the simulated storms of Frank and Ritchie (2001) within 18–24 h of the imposition of the shear. Thus, as potential rainmaking TCs approach and cross 25°N, a combination of climatologically cool SSTs and strong vertical wind shear work in concert to rapidly weaken the storms and limit their lifetimes as tropical systems.

The TCs that develop in the eastern Pacific form in a region of climatologically weak midlevel easterly flow less than 7.5 m s⁻¹ (black barbs in Fig. 5). Identified by Chan and Gray (1982) as the level that best correlates with the motion of TCs, the 500–700-hPa layer averaged winds are almost due east at, and south of, 15°N, only acquiring a slight northward component at 20°N. Farther north, the September climatological flow weakens further (<5 m s⁻¹), and turns toward the northwest and eventually north-northeast by 30°N. The motion of the 35 TCs that affected the southwest United States is generally consistent with the climatological steering flow, but with a northward component to their tracks at lower latitudes than normal and with a steeper northward component, the reasons for which will be explored in the next section.

The TCs that affect the southwest United States are unique not only in their genesis locations and storm motion but also in the time of the TC season when they are most likely to impact the region. Figure 6 shows (top) the number of days on which rainfall associated with a TC fell in the southwest United States and (bottom) the number of days at least one TC was active in the eastern Pacific. Consistent with previous studies, the most likely time for TC rainfall in the southwest United States is during the month of September. This peak in rainfall occurs during the late season secondary peak in the number of east Pacific TC days in late August and September. About equal numbers of TC rainfall days are experienced in the NAM region during the last two weeks in August and the first two weeks in October, but the number of eastern Pacific TC days in the later period is only about two-thirds of that in the earlier period. There was only one TC rainfall day during the first two weeks of August, which coincides with the midseason minimum in TC activity in the basin, consistent with the timing of the midsummer drought (Magaña et al. 1999). The peak in the eastern Pacific TC season is during the last two weeks of July, but no TCs or their remnants during the 46 yr studied impacted the southwest United States during this time. This discrepancy suggests that the large-scale flow pattern necessary to steer the storms northward is much more likely at the end of the TC season, as the atmosphere transitions to a fall pattern in the North Hemisphere.

b. Composite flow

As noted in the last section, the TCs that brought rainfall to the southwest United States turned northward sooner (i.e., at lower latitudes) and with a faster speed than the climatological steering flow off the Mexican coast (Fig. 5). To investigate the reasons for this behavior, Fig. 7 shows the composite 500-hPa heights and standardized anomalies for the 20 hurricanes between 1971 and 1999 (see Table 1) to bring rainfall to the southwest United States. The composites are calculated from the NCEP–NCAR reanalysis, are relative to the day the TC crossed 25°N (day t), and were calculated for days t ± 5, although Fig. 7 only shows days t − 3, t − 1, t, t + 1, and t + 3 for brevity.

On day t − 3 (Fig. 7a), a broad area of below-normal 500-hPa heights between 10° and 20°N and 105° to 120°W was the NCEP–NCAR representation of the TC about to impact the southwest United States. The TC was steered toward the north around the continuous belt of high pressure to the east representing the NAM and Bermuda subtropical anticyclones. In the midlatitudes, a relatively weak trough was digging southward off the California coast, while farther north a wave train was present from just south of the Aleutian Islands across Canada and over to Greenland. Also of note on day t − 3 was the small patch of anomalously high heights centered on 25°N and 145°W, which increased significantly in both scale and magnitude by day t − 1 (Fig. 7b). An examination of the origins of these anomalously high heights in individual cases revealed influences from both convection in the central Pacific and cases of downstream development associated with the extratropical transition (ET) of western Pacific TCs. Although beyond the scope of the current study, work is presently under way to study the downstream impacts of events in the western and central Pacific on the recurvature of eastern Pacific TCs. In the current study, whatever their origin, the anomalously high heights in the composite aided in the downstream development of a deeper trough just off the California coast, a building ridge over the southern U.S. Plains, and the anomalously low heights off the southeast U.S. coast. The addition of the deeper California trough to the stationary monsoon anticyclone steered the eastern Pacific TC toward the north to a location just off Baja California on day t − 1. The anomaly had increased significantly in magnitude and merged with the California trough to produce one large region of below-normal heights in the eastern Pacific, with one lobe at the base of the trough and a second, stronger lobe just southwest of Baja California, representing the TC.

On day t (Fig. 7c), when the TC crossed 25°N in both the HURDAT and NCEP–NCAR reanalyses, the anomaly associated with the TC was now located at the base of the trough and had its largest magnitude of more than three standard deviations below the mean. At this, and subsequent times, it was no longer possible to separate the trough and the TC in the coarse reanalyses, but as we will see in the case studies below, they often remain separate entities for many days after their interaction begins. This interaction likely strengthened the downstream ridge through diabatic processes and the trough over the southeast United States through downstream development. This trough–ridge–trough pattern is similar to the 500-hPa composites for recurving eastern Pacific TCs shown by Farfán (2004) and Ritchie and Szenasi (2006) (see Fig. 2 in each of these works) but is slightly deamplified because of the larger number of cases here versus Farfán (2004) (20 versus 6) and the inclusion of all storms, not just those with a strong trough interactions as in Ritchie and Szenasi (2006).

On day t + 1 (Fig. 7d), the deepening east coast trough impinged on the subtropical ridge to its south and pinched the ridge into two separate anticyclones, as the west coast trough–TC anomaly started to lift and move ashore as the anomalously high heights over the southern Plains retreated to the south. The ridge was further weakened on day t + 3 (Fig. 7e) as the trough–TC remnant crossed the intermountain west onto the High Plains as another strong ridge built in the eastern Pacific. The climatologically low heights over the southeast United States had weakened and retrogressed to the west on day t + 3, but they eventually moved slowly up the eastern seaboard in advance of the progressive Pacific trough/TC remnant (on days t + 4 and t + 5; not shown).

As will be explored below, the exact scale of and phasing between the trough and TC were different for each case and led to very different interactions, precipitation amounts, and patterns, but the features common to all events were the trough off the west coast of the United States steering the TC toward the north and the downstream ridge built by the trough–TC interaction over the continental United States.

c. Precipitation patterns and climatology

The 35 TCs and their remnants that brought rainfall to the southwest United States possess a wide variety of rainfall distributions from southwest–northeast-oriented streaks extending well into the central United States to isolated pockets along the U.S.–Mexico border. The precipitation pattern depends on a number of factors including topography, the scale and nature of the west coast trough (progressive or cutoff), and the proximity of the trough to the TC. The dynamic evolution of three representative case studies will be explored in the next section, but examples of three common precipitation patterns and the average contribution of TCs to the warm-season rainfall totals over the southwest United States are presented here. (To view plots of storm total rainfall and percent contribution to the warm-season precipitation for all 35 TCs studied, see http://www.atmos.ucla.edu/~kristen/monsoon/monsoon.html.)

Figure 8 shows the storm total precipitation (mm) and the contribution of the days with TC rainfall to the warm-season precipitation (%) for Hurricane Doreen (1977), Hurricane Lester (1992), and Tropical Storm Hyacinth (1968). The warm-season contribution was simply calculated by dividing the total precipitation on days when a TC was present by the warm-season total at each grid point. As clearly evidenced by the large midlatitude trough off the California coast in the composite shown in Fig. 7, not all of the precipitation that fell on the days with a TC in the vicinity was directly associated with the TC. It is impossible, however, to separate the TC-induced rainfall from the trough-induced rainfall; thus, all rainfall that fell when a TC tracked into the NAM region will be referred to as TC rainfall.

The three storms chosen for Fig. 8 were selected because of their very different precipitation patterns and endpoints of their tropical tracks. Hurricane Doreen (1977) can be seen emerging from the tangle of tracks off Baja California around 28°N; it propagated toward the northwest, eventually weakening below tropical depression status at 0000 UTC 18 August (the end of the solid black track in Fig. 8a) just off the coast of California and below the ERA-40 relative vorticity threshold of 4 × 10⁻⁵ s⁻¹ over interior southern California (the end of the dashed track) 72 h later. The precipitation pattern in Fig. 8a shows no distinct swath of heaviest rainfall and most of the local maxima are tied to the topographical features noted in the discussion of Fig. 2a. When the rainfall associated with Doreen is compared with the warm-season rainfall total, however, a different picture emerges. Because portions of southern California and Nevada typically receive less than 20 mm of rainfall during the summer (Fig. 2a), an event like Doreen accounts for greater than 50%, and in some locations 100%, of warm-season rainfall. TCs that produce their heaviest rainfall west of ∼115°W and/or have the greatest contribution to the warm-season precipitation over California and Nevada will be referred to as California–Nevada track TCs. The synoptic evolution of another of the nine storms in this category (see Table 1), Hurricane Kathleen (1976), will be investigated in section 4a.

The second and largest (with 15 members; see Table 1) category of storms is called the southwest–northeast swath because of their elongated strips of heavy precipitation and long-lived vorticity maxima that strongly interact with mobile midlatitude troughs. The evolution of these storms looks most like the composite shown in Fig. 7, with the TC recurving and propagating toward the northeast in front of and along the forward edge of the trough, similar to the “full trough interaction” cases of Ritchie and Szenasi (2006). Figure 8c shows the quintessential example in this category, Hurricane Lester (1992), which produced a large swath of greater than 30 mm of rain over a 5-day period to the left of its track, covering four states from Arizona to Nebraska. Consistent with its left of track precipitation maximum (Atallah et al. 2007), Lester has been documented to be the first case of an eastern Pacific storm to undergo ET (Dickinson et al. 2004). Because Lester will be the focus of a separate, forthcoming manuscript, the evolution of another southwest–northeast swath case, Hurricane Hilary (1993), will be explored in section 4b.

In comparison with both Doreen and Lester, the precipitation associated with Tropical Storm Hyacinth (1968) is much more isolated and confined to the end of its track along the southern Arizona–New Mexico border (Fig. 8e). A second, larger swath of precipitation associated with a midlatitude trough can be seen across the Pacific Northwest and northern Rockies. In this case, the phasing between the trough and Hyacinth was such that the trough moved onshore too far north and too early to interact with Hyacinth more than steering the TC northward (analogous to the “missed interaction” category of Ritchie and Szenasi 2006). The cases that possess a missed or weak trough interaction and/or that weaken over the NAM region will be referred to as isolated or weakening in situ. Hurricane Marie (1984) will serve as an example of one of these 11 storms (see Table 1) in section 4c.

Although there are large variations in the distribution of precipitation from TCs in the southwest United States, it is useful to average the percent contribution from TC rainfall to the warm-season precipitation in the 27 yr when there was at least one TC rainfall event (see Table 1) to get a broad-brush picture of their impact on the NAM region (Fig. 9). The average contribution of TCs to the warm-season precipitation increases from east to west, accounting for less than 5% of the rainfall in eastern New Mexico and increasing to more than 20% in extreme western Arizona, southern California, and northern Baja California (similar to Fig. 2 of Larson et al. 2005). This east to west gradient is the mirror image of the gradient in average warm-season precipitation in the region (Fig. 2a), which ranges from over 200 mm in the east to less than 25 mm along the Pacific coast. Even though the plot is averaged over all 35 TCs, two of the precipitation categories defined above are evident in the percent contribution map: one extending northeast from southwest Arizona, through the northwest corner of New Mexico, and into southwestern Colorado, and the other extending north-northwest from southern to central California and into southern Nevada.

a. California–Nevada track

Hurricane Kathleen (1976) is the eastern Pacific TC to travel the farthest north over the United States as it was tracked by NHC/TPC well into southern Nevada (Fig. 3). The storm produced in excess of 40 mm of rain over all of southern California during the 3-day period 10–12 September, which accounted for upward of 70% of the warm-season rainfall over a large portion of the region (Fig. 10b). The heaviest rainfall occurred along and to the left of the track; a distinct vorticity signature of the storm (>4 × 10⁻⁵ s⁻¹) could be tracked into southwestern Montana (Figs. 10a and 11g).

The track of Kathleen and its extremely heavy precipitation were due to the interaction of the storm with an unusually strong disturbance on the DT that fractured off of, and was left behind by, the trough located over the Dakotas at 0000 UTC 9 September (Fig. 11a). The fracture occurred 2 days earlier as the trough came ashore and the northern section continued to move steadily eastward and the cutoff disturbance sank southward and retrogressed slightly off Baja California (Figs. 11a–c). The flow around this DT disturbance steered Kathleen (labeled K in Fig. 11a) north through 1200 UTC 10 September (Fig. 11d) when it began to interact with the cyclone. The relative vorticity signature of Kathleen increased markedly to over 10 × 10⁻⁵ s⁻¹ at this time and remained stronger than any value during its time over the eastern Pacific. Although some of this increase may be due to the storm being better sampled while over the continental United States, there is undoubtedly a strong and constructive interaction between the cutoff and Kathleen, as evidenced by the left-of-track precipitation maxima (Atallah et al. 2007), the superposition of the features and kinking of the isobars on the DT in the vicinity of Kathleen, the simultaneous strengthening of the systems, and the strong downstream ridging over the southwest United States shown in Figs. 11d,e. Although no satellite imagery was available for this time period, the strong, bent-back ridge and anticyclonic outflow seen in Figs. 11d–f and the heavy precipitation amounts across the area provide strong evidence for deep convection and the diabatic generation of the ridge. The amplifying effect of downstream development in response to the southwest U.S. ridge eruption can be seen clearly starting on 0000 UTC 11 September and continuing through 0000 UTC 12 September (Figs. 11e–g) as the half wavelength between the ridge and downstream trough over the eastern United States contracts, a new ridge erupts over northern New England, and the next downstream trough and ridge become meridionally elongated.

Kathleen was declared extratropical by NHC at 1800 UTC 10 September and was tracked by the center for another 18 h as it cyclonically rotated around the cutoff cyclone from its eastern to northern edges. During the same time period, the cutoff began to come ashore and lift northward in response to the next upstream trough (Figs. 11d–f). As the cutoff moved ashore, it began to weaken and open up as the remnants of Kathleen weakened initially and then reintensified at 0000 UTC 12 September (Fig. 11g) before weakening again as the DT disturbance and remnants crossed the northern Rockies.

b. Southwest–northeast swath

The storm total and contribution to the warm-season precipitation for Hurricane Hilary (1993) are shown in Figs. 10c,d. NHC/TPC stopped tracking Hilary south of the U.S.–Mexico border (see H93 located at ∼31°N, 113°W in Fig. 2), but the TC remnants could be tracked in the ERA-40 relative vorticity field moving from southwest to northeast across central Arizona and the southern Rockies and into Nebraska (Figs. 10c and 12) during 26–30 August. Maximum precipitation values exceeding 80 mm were mostly confined to the right side of the remnant’s track and contributed over 90% of the summer rainfall in southern California and over 30% along its track through Arizona and New Mexico (Fig. 10d).

The path of Hilary and its interaction with a mobile midlatitude trough can be examined through a series of daily DT and GOES-7 WV images (at 1200 UTC) shown in Figs. 12 and 13. At 1200 UTC 23 August, Hilary was a weakening 50-kt tropical storm located southwest of the tip of Baja California (Figs. 12a and 13a). To the north of the circulation were two features of interest: 1) a strip of cyclonic shear vorticity and a relatively lower (higher pressure) DT along ∼30°N that was a remnant potential vorticity streamer left behind by the trough located over the central United States, and 2) a potent midlatitude trough digging southward and coming ashore over Washington and British Columbia, seen nicely in the WV imagery in Figs. 13b,c. Over the next 48 h (Figs. 12b,c and 13b,c), Hilary slowly crept northward along the west coast of Baja California and the Pacific Northwest trough propagated south and eastward, advecting the cyclonic shear line (which had become two distinct centers by 1200 UTC 24 August in Fig. 12b) eastward and ahead of the main trough.

As the northern portion of the trough propagated quickly eastward through the northern Rockies on 26 and 27 August (Figs. 12d,e and 13d,e), the southern end of the trough thinned, became nearly stationary, and interacted with the rapidly weakening Hilary (NHC issued its last advisory for the storm at 0600 UTC 27 August). The GOES-7 WV imagery during this period shows the deep tropical moisture associated with the remnants of Hilary streaming north and then eastward from Baja California across the Rockies and onto the central Plains in Figs. 13d–f, which likely helped build the skinny fold over the ridge (or anticyclonic wave breaking in the terminology of Thorncroft et al. 1993) spanning the upper Plains and Midwest, southern Ontario, and into New England (Fig. 12d). The remnant Hilary vorticity center, with its rich supply of tropical moisture, moved rapidly to the northeast along the front side of the weak and nearly stationary southern California trough, helping to bring over 100 mm of rain to portions of Kansas and Nebraska on 27 and 28 August. These heavy precipitation amounts continued the streak of greatest precipitation to the right of Hilary’s track, consistent with the composites of Atallah et al. (2007) that showed a stronger interaction with a downstream ridge than a weak, upstream trough for recurving TCs exhibiting precipitation maxima to the right of their tracks.

c. Weakening in situ

The amount and distribution of precipitation for Hurricane Marie (1984) are quite different from the two cases described above. Figure 10e reveals only isolated pockets of rainfall exceeding 30 mm in central Arizona and eastern Nevada that contribute only a small amount to the warm-season totals, except in isolated pockets on the California coast and Arizona’s borders with California and Nevada (Fig. 10f). These small precipitation amounts were due to Marie’s northwestward track—the first indication that it did not interact with a mobile midlatitude trough—around the western edge of an upper-level anticyclone over Mexico as seen in the DT winds at 0000 UTC 8 and 9 September in Figs. 14a,b. Ventilated by this anticyclonic flow aloft and fueled by warm waters below, Marie intensified to a strong 80-kt hurricane early on 8 September and attained an impressive satellite presentation at 0000 UTC 9 September, as seen in Fig. 15b.

As Marie moved northwestward on 9 and 10 September, the storm lost its eye and deep convection (Figs. 15c,d) as it slowly weakened to below tropical depression strength by 0000 UTC 11 September. Despite being downgraded by NHC, the Marie remnant maintained a strong relative vorticity signature exceeding 8 × 10⁻⁵ s⁻¹ through 0000 UTC 13 September (Fig. 14f) that can be seen in the low-level (i.e., less bright white) swirl of clouds off southern California in Figs. 15d,e. As the low-level center moved westward (see track on Fig. 10e), the moisture associated with Marie propagated to the east and help fuel afternoon convection over Arizona (Fig. 15d), which brought widespread rains of greater than 10 mm (Fig. 10e).

The strong anticyclonic circulation on the DT, originally centered east of Marie over northern Mexico (Fig. 14a), moved west and then north with the storm (Figs. 14b–e) and helped the Marie remnants fight off the encroachment of, and interaction with, a small, but long-lived cyclonic anomaly on the DT located near 33°N, 124°W on Fig. 14a that broke off the southern end of the trough over the northern plains. The anticyclonic circulation eventually propagated east and away from the TC remnant by 0000 UTC 14 September (Fig. 14g) in advance of a very strong Pacific trough that advected the cyclonic DT anomaly northeastward (it is located over northern California in Fig. 14h). The digging trough also increased the winds over the Marie remnants, causing the vorticity signature to elongate and weaken below 4 × 10⁻⁵ s⁻¹ as it became sheared out on the forward side of the trough.