The topography in and around the Intermountain West strongly affects the genesis, migration, and lysis of extratropical cyclones. Here intermountain (i.e., Nevada or Great Basin) cyclone (IC) activity and evolution are examined using the ECMWF Re-Analysis Interim (ERA-Interim) the North American Regional Reanalysis (NARR), and the NCEP–NCAR reanalysis from 1989 to 2008, the period during which all three are available. The ICs are defined and tracked objectively as 850-hPa geopotential height depressions of ≥40 m that persist for ≥12 h.
The monthly distribution of IC center and genesis frequency in all three reanalyses is bimodal with spring (absolute) and fall (secondary) maxima. Although the results are sensitive to differences in resolution, topographic representation, and reanalysis methodology, both the ERA-Interim and NARR produce frequent IC centers and genesis in the Great Basin cyclone region, which extends from the southern “high” Sierra to northwest Utah, and the Canyonlands cyclone region, which lies over the upper Colorado River basin of southeast Utah. The NCEP–NCAR reanalysis fails to resolve these two distinct cyclone regions and produces less frequent IC centers and genesis than the ERA-Interim and NARR.
An ERA-Interim-based composite of strong ICs generated in cross-Sierra (210°–300°) 500-hPa flow shows that cyclogenesis is preceded by the development of the Great Basin confluence zone (GBCZ), a regional airstream boundary that extends downstream from the Sierra Nevada across the Intermountain West. Cyclogenesis occurs along the GBCZ as large-scale ascent develops over the Intermountain West in advance of an approaching upper-level trough. Flow splitting around the high Sierra and the presence of low-level baroclinicity along the GBCZ suggest that IC evolution may be better conceptualized from a potential vorticity perspective than from traditional quasigeostrophic models of lee cyclogenesis. Although these results provide new insights into IC activity and evolution, analysis uncertainty and the cyclone identification criteria are important sources of ambiguity that cannot be fully eliminated.
The Intermountain West, which lies between the Sierra Nevada and Cascade Mountains to the west and the Rocky Mountains to the east (Fig. 1), has long been recognized as a region of frequent cyclone activity (e.g., Petterssen 1956; Klein 1957; Whittaker and Horn 1981; Lee 1995). Hazardous weather produced by intermountain (i.e., Nevada or Great Basin) cyclones (ICs) includes high winds, blowing dust, wildfire runs, and dramatic cold-frontal temperature falls (e.g., Shafer and Steenburgh 2008; Steenburgh et al. 2009; West and Steenburgh 2010). For example, from 15 to 16 April 2002, a powerful IC produced the second lowest reduced sea level pressure (982 hPa) ever recorded in Utah (West and Steenburgh 2010). Winds accompanying the storm reached 35 m s−1 in Salt Lake City, Utah, and 42 m s−1 in the nearby Wasatch Mountains. During the height of the storm, 95 000 residences were without power, tractor trailers were toppled along Interstate 80, blowing dust forced the closure of Interstate 15, and a wind-driven wildfire forced the evacuation of Parowan, Utah.
However, a lack of consensus exists in previous studies with regard to the frequency, seasonality, and geographic distribution of IC centers, genesis, and lysis (cf. Petterssen 1950, 1956; Klein 1957; Reitan 1974; Zishka and Smith 1980; Whittaker and Horn 1981). This lack of consensus likely arises from differences in cyclone identification criteria, analysis methodology, study period, and dataset resolution. A heterogeneous mix of months or seasons presented and a lack of focus on the Intermountain West, serve as additional sources of uncertainty that cloud the interpretation of these studies.
Petterssen (1950, 1956) first identified cyclone-center and genesis frequency maxima over the Intermountain West during the winter and summer (months not specified) using cyclone-center data obtained from daily (∼1300 UTC) Historical Weather Maps analyses covering a 40-yr period (1899–1939) and aggregated into 5° latitude–longitude grid cells. Petterssen (1956) attributed the IC center and genesis frequency maxima to the influence of the Sierra Nevada during winter and the presence of thermal lows during summer. In both seasons, the IC center and genesis frequencies exceeded those found to the lee of the Colorado Rockies. Petterssen (1956) also observed that there was little downstream extension of the IC center maximum, suggesting that many ICs are short lived and decay before leaving the Intermountain West. Subsequent analysis of monthly cyclone activity by Klein (1957), which concentrated on a 20-yr “best coverage” subset of the Historical Weather Maps (1909–14 and 1924–37) and subjectively removed thermal lows and quasi-stationary cyclones (methodology not detailed), revealed a broad maximum in the frequency of IC centers from February to April, whereas genesis peaked sharply in June.
In contrast, Reitan (1974), who examined North American cyclone activity from 1951 to 1970, failed to identify maxima in cyclone-center or genesis frequency over the Intermountain West during any of the five months investigated (January, April, June, July, and October). This lack of IC activity may reflect the large grid cells in which cyclone statistics are aggregated (∼740 km on a side, roughly 7° × 9° latitude–longitude at 40°N), or the characteristics of the underlying National Weather Service (NWS) cyclone-track dataset. This cyclone-track dataset derives from 6-h sea level pressure analyses with isobars at intervals of 3, 4, or 5 hPa [depending on the analysis period (Zishka and Smith 1980; Peyrefitte and Astling 1981; Zishka and Smith 1981)] and requires that cyclones persist for at least 24 h with the center moving “through a prescribed area” (Reitan 1974, p. 861). However, using the same NWS cyclone-track dataset, but with a longer study period (1950–77) and smaller grid cells (2° latitude–longitude), Zishka and Smith (1980) did find a weak cyclogenesis maximum over the Intermountain West during January. They also found a dearth of IC activity in July, but did not present statistics for other months, including those identified by Klein (1957) as having the greatest IC center (February–April) and genesis (June) frequencies.
Whittaker and Horn (1981) reconcile some of the confusion created by Reitan (1974) and Zishka and Smith (1980) by presenting monthly cyclone statistics compiled for the Intermountain West (their Great Basin) from the NWS cyclone-track dataset (1958–77). Their analysis, which shows a pronounced IC genesis maximum in April and a secondary maximum in October, suggest that the January and July periods examined by Zishka and Smith (1980) fail to capture the primary IC genesis seasons, whereas the extremely coarse grid used by Reitan (1974) inadequately resolves IC events.
Lee (1995) produced an IC climatology using 3-h surface analyses produced by the National Meteorological Center [(NMC), now the National Centers for Environmental Prediction (NCEP)], with the most relevant figures reproduced here because they have not been published previously in the peer-reviewed literature. Lee (1995) identified cyclones as closed low centers (based on sea level pressure contours at 4-hPa intervals) that are migratory and endure for at least 5 consecutive analysis times (12 h). The 12-h criterion is shorter than used in many studies (see Raible et al. 2008), but reflects the brief life cycle of many ICs. The resulting analysis of cyclogenesis frequency for 1984–86 features a maximum that extends meridionally through eastern Nevada (Fig. 2), but is substantially weaker than that found in the lee of the Colorado Rockies, in contrast to Petterssen (1950, 1956), Klein (1957), and Whittaker and Horn (1981). This contrasting result may reflect the short analysis period or differing methodologies for removing thermal or other quasi-stationary lows. The monthly frequency distribution of Nevada cyclogenesis [i.e., cyclones that form specifically over Nevada] for a longer 11-yr period (1976–86) features an absolute maximum in March and secondary maximum in November (Fig. 3). The absolute minimum is observed in July, with a secondary minimum in December. This bimodal distribution resembles that observed by Whittaker and Horn (1981), but the spring maximum is a month earlier and the fall maximum later and more pronounced. Lee (1995) hypothesized that this cyclogenesis bimodality is caused by the equatorward migration of the polar jet stream across the Intermountain West with the approach of winter and poleward migration with the approach of summer. A strong correlation between the frequency of IC genesis and upper-level trough passages provides support for this hypothesis (Fig. 3).
More recent cyclone climatologies combine objective methods [either cyclone dectection and tracking or the dynamic storm-track approach (e.g., Blackmon 1976; Wallace et al. 1988)] with global or regional reanalyses [see Raible et al. (2008) for a review], but none have examined the Intermountain West in detail. Often these studies, which frequently concentrate on sea level or 1000 hPa, do not provide analyses over mountainous regions like the Intermountain West because of issues related to pressure reduction (e.g., Wernli and Schwierz 2006).
In this paper we attempt to resolve some of the contradictions and interpretation challenges described above by presenting an IC climatology based on three modern reanalyses, which enables a partial assessment of analysis uncertainty arising from inadequate terrain representation, observing system limitations, and reanalysis methodology. The climatology provides new insights into the subsynoptic characteristics of IC activity, but analysis uncertainty and cyclone identification criteria remain sources of ambiguity that warrant further investigation.
2. Data and methods
a. Reanalysis data
The three reanalyses used are the European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis Interim (ERA-Interim) global reanalysis, the National Centers for Environmental Prediction (NCEP) North American Regional Reanalysis (NARR), and the NCEP–National Center for Atmospheric Research (NCEP–NCAR) reanalysis. Providing analyses for the “data rich” period beginning in 1989, the ERA-Interim produces analyses using a version of the ECMWF Integrated Forecast System (IFS) configured with 60 vertical levels (extending to 0.1 hPa), T255 triangular truncation for dynamical fields, and a reduced Gaussian grid with ∼80-km spacing for surface and other gridpoint fields (Uppala et al. 2008; Berrisford et al. 2009). The ∼80-km Gaussian grid spacing provides a reasonable estimate of the effective grid spacing of the ERA-Interim in physical space (Kanamitsu 1989). Nearly all the ERA-Interim data used here covers 1989–2008 and was downloaded from the ECMWF data server at 1.5° grid spacing and 6-h intervals. The lone exception is model topography, which was obtained from NCAR on a transformed grid with comparable resolution to that of the original ERA-Interim (0.7°).
The NARR provides analyses using a version of the NCEP Eta Model configured with 45 levels and 32-km grid spacing (Mesinger et al. 2006). The data was downloaded from the National Oceanic and Atmospheric Administration (NOAA) National Operational Model Archive and Distribution System (NOMADS) at 32-km grid spacing, but, with the exception of topography, which was plotted at near-native grid spacing, was bilinearly interpolated to the ERA-Interim 1.5° grid prior to analysis for comparison purposes. This likely reduces the total cyclone count compared to the native-resolution dataset, although it also limits the influence of small-scale, quasi-persistent, thermally forced lows that are commonly found in the NARR but are not associated with large-scale cyclogenesis. Although the NARR provides analyses every 3 h beginning in 1979, for comparison purposes we use the same analysis increment and period as the ERA-Interim.
The NCEP–NCAR reanalysis generates analyses using a version of the NCEP global spectral model (Kanamitsu 1989; Kanamitsu et al. 1991) configured with 28 vertical levels (extending to 3 hPa) and T62 triangular truncation (Kalnay et al. 1996; Kistler et al. 2001). It has an effective grid spacing of ∼210 km, although the data used here, including model topography, was downloaded from the NOAA Earth System Research Laboratory and processed at 2.5° grid spacing and 6-h intervals. The period of record begins in 1948, but we concentrate on 1989–2008 for comparison purposes.
All objective analyses provide an estimate of the atmospheric state that reflects the quality and representativeness of the observations, the capabilities of the data assimilation scheme, and the skill and bias of the underlying numerical model. Specific concerns over the Intermountain West include 1) very limited use of satellite radiance data due to uncertainties in land surface emissivity, especially over high and irregular orography (Derber and Wu 1998; McNally et al. 2000; McNally et al. 2006); 2) the low density and limited use–representativeness of surface observations at the scales resolved by the three reanalyses (Horel et al. 2002; Myrick et al. 2005; Myrick and Horel 2008; Tyndall et al. 2010); and 3) model terrain representation, which influences the background analysis to which corrections are applied.
Of the three analyses, the ERA-Interim is the most recently developed (2006) and advanced (Uppala et al. 2005, 2008; Berrisford et al. 2009). It utilizes a 12-h four-dimensional variational data assimilation (4DVAR) system that assimilates a comprehensive suite of data including cloud-track winds, satellite radiances (limited over the Intermountain West, see above), radio-occultation measurements (since 1996), and operational radiosonde (wind, temperature, and humidity), and surface observations. In the case of the latter, the atmospheric analysis assimilates only surface pressure, with temperature and humidity used only for the land surface analysis and surface wind not assimilated. The ERA-Interim provides a representation of the regional orography, but fails to fully resolve the height of the Sierra Nevada, particularly the southern “high” Sierra, or the many isolated ranges of Nevada (cf. Figs. 1b and 4a). Instead, a broad plateau extends from a flattened Sierra Nevada northeastward across central Nevada. The topography of Utah is also crudely resolved, however, there is some representation of the lowland Great Salt Lake Basin, the high terrain in the central part of the state, and the lowland Canyonlands region in the southeast.
The NARR utilizes the NCEP Eta Model three-dimensional variational data assimilation (3DVAR) system with a 3-h forecast serving as the first guess for each assimilation cycle (Mesinger et al. 2006). Assimilated data includes cloud-track winds, satellite radiances (limited over the Intermountain West), and operational radiosonde (temperature, wind, and moisture) and surface (pressure, wind, and moisture) observations. Compared to the ERA-Interim, the NARR better (but not fully) resolves the Sierra Nevada, central Utah mountains, and lowland regions of the Great Salt Lake Basin and Canyonlands region (cf. Figs. 4a,b). Much of the narrow basin-and-range topography of Nevada, however, remains poorly represented as a broad plateau.
The NCEP–NCAR reanalysis is the oldest of the three reanalyses and uses spectral statistical interpolation, an approximate form of 3DVAR (Kalnay et al. 1996; Kistler et al. 2001). Assimilated data includes satellite-derived vertical temperature profile retrievals (only above 100 hPa over the Intermountain West and other land areas), cloud-track winds, and operational radiosonde (temperature, wind, and specific humidity) and surface observations (pressure only). Compared to the ERA-Interim and NARR, the NCEP–NCAR reanalysis has a coarser grid spacing and more limited terrain representation (Fig. 4c).
b. Cyclone identification
Previous studies employ a range of approaches to identify and track cyclones. Here we use a three-step approach based in part on Wernli and Schwierz (2006) to identify cyclones at 850 hPa, which is near the mean elevation of the valleys and basins of the Intermountain West and reduces (but does not eliminate) issues related to pressure reduction. First, we identify height minima as grid points where the geopotential height is lower than the eight surrounding grid points. Next, we determine if the height minimum is fully enclosed by a geopotential height contour 40 m greater in magnitude than the height minimum. For a pressure of 850 hPa and temperature of 10°C, this +40-m threshold is roughly analogous to a 4-hPa contour interval (based on the hydrostatic assumption), comparable to that commonly used in sea level cyclone identification. If enclosed by this +40-m contour, the height minimum is considered a low center; otherwise, it is discarded. If multiple low centers are enclosed by the same +40-m contour, the height minimum with the lowest value is selected. In the case of sea level cyclones, for which we present limited statistics to illustrate the influence of thermal lows and sea level reduction, low-center identification follows the same procedure except a +5-hPa closed contour is required. We use 5 hPa instead of 4 hPa since the former is comparable to a 40-m height gradient at 1000 mb assuming a temperature of 10°C.
Low centers are classified as cyclones if they persist for 12 h or more (i.e., 3 or more analysis times). This is a shorter life cycle criterion than used in many cyclone studies (e.g., Raible et al. 2008), which concentrate on cyclones with a life cycle of at least 72 h, but is consistent with Lee (1995) and reflects the brief life cycle of many ICs. Low centers at successive analysis times are considered to be the same cyclone if they are located within 8.5° (∼950 km) of arc length of each other. Given the 6-h analysis increment, this represents an unrealistic upper bound for continuous low-center speed (∼150 km h−1), but was developed subjectively after examining many cases to account for the discontinuous movement of cyclones (i.e., dissipation of one low center dissipates while a new one forms elsewhere) across the region’s mesoscale topography. If, however, the low centers are more than 8.5° apart, they are considered different (discrete) systems, even if they are induced by the same upper-level trough.
Cyclone-center density is the number of cyclone low centers observed in each grid box, expressed as low centers (100 km)−2. It includes all low centers observed during each identified cyclone’s life cycle, but does not include low centers that fail to meet the cyclone criteria described above. Similarly, mean cyclone-center frequency reflects the average number of cyclone low centers each month. Cyclogenesis (cyclolysis) density and frequency are based on the first (last) low center observed for each cyclone. Cyclone amplitude is the difference between the geopotential height of the low center and the outermost closed contour (at 10-m intervals). Maximum amplitude is the largest amplitude observed during a cyclone’s life cycle.
To provide a large-scale perspective, cyclone statistics are presented initially for a region encompassing the eastern Pacific Ocean, the western United States, and parts of western Canada and Mexico. We then examine statistics for the Intermountain West, defined to be between 35° and 42.5°N and 119° and 108°W. We compare results in this region to a comparably sized “ocean” domain directly west (upstream) over the Pacific Ocean (Figs. 1a,b).
A major concern for the IC climatology is the influence of thermally induced areas of low pressure (i.e., thermal lows), which can form from intense surface heating. Over the desert Southwest, Rowson and Colucci (1992) found that thermal low occurrence is rare from December to April and most common in July and August. At 850 hPa, most thermal lows produced by the three analyses do not maintain an amplitude of more than 40 m for 12 h and are largely, but not completely eliminated by our cyclone amplitude and life cycle requirements.
Any cyclone identification algorithm necessarily involves the use of thresholds, whereas a continuous spectrum is observed in nature. The threshold to which our method is most sensitive is the +40-m amplitude requirement, which was selected for its similarity to the 4-hPa contour interval commonly used in synoptic surface analysis, and because it removed many thermal lows while preserving those that are dynamically forced. Lowering this threshold leads to gains in cyclone frequency in all months and at all locations, but especially during the summer when thermal lows are common. A higher threshold lowers frequencies in all months, but also magnifies the bimodal annual distribution of cyclone events by leaving only the strongest storms, which occur preferentially in spring and fall, as will be shown later.
a. Western United States and the surrounding region
Over the western United States and surrounding region, the 850-hPa cyclone-center density is largest over the northeastern Pacific Ocean (off the coast of British Columbia and southeast Alaska), southern Canada downstream of the Canadian Rockies, the Intermountain West, and downstream of the Colorado Rockies (Figs. 5a–c). High cyclone-center density in the former two regions reflects the Pacific and “Alberta Clipper” storm tracks identified in previous studies of sea level (or 1000 hPa) cyclone activity (e.g., Petterssen 1956; Zishka and Smith 1980; Sickmöller et al. 2000; Hoskins and Hodges 2002; Wernli and Schwierz 2006; Thomas and Martin 2007; Raible et al. 2008). Although previous studies have identified frequent cyclone activity over the Intermountain West and downstream of the Colorado Rockies (see the introduction), the ERA-Interim and NARR cyclone-center density analysis exhibits considerable mesoscale structure (Figs. 5a,b) that has not been documented previously and will be examined in greater depth in section 3b. In contrast, the low-resolution NCEP–NCAR reanalysis produces only a broad cyclone-center density maximum over the Intermountain West and a corridor of high cyclone-center density that begins downstream of the Colorado Rockies and extends eastward over the central United States (Fig. 5c).
The ERA-Interim sea level cyclone-center density is similar to that at 850 hPa over the eastern Pacific, although it does not feature the neck of high cyclone-center density found at 850 hPa just off the California coast (cf. Figs. 5a,d). Over the lower Colorado River Valley and in the lee of the Colorado Rockies, the sea level cyclone-center density is much larger than found at 850 hPa. This likely reflects the greater amplitude of warm-core thermal lows near sea level than at 850 hPa over the lower Colorado River Valley, but is an apparent artifact of pressure reduction in the lee of the Colorado Rockies. Over the Intermountain West, the sea level cyclone-center density is lower than at 850 hPa in the direct lee of the Sierra Nevada, but larger over other areas, including near the western edges of the Great Salt Lake Basin and Snake River Plain. If a 4-hPa closed contour is used instead of 5 hPa, the sea level cyclone-center density is higher throughout the Intermountain West, including in the direct lee of the Sierra Nevada (not shown).
The mean monthly 850-hPa cyclone-center frequency distribution for the western United States and the surrounding region is bimodal with a broad absolute maximum in early spring (April) and a weaker secondary maximum in fall (November; Fig. 6a). The NCEP–NCAR reanalysis has fewer events than the ERA-Interim or NARR, with weaker bimodality and a less abrupt decrease in events from spring to summer. In all months there are substantially more ERA-Interim cyclones at sea level than at 850 hPa.
Cyclogenesis density is largest over the northeast Pacific, downstream of the Canadian Rockies, over the Intermountain West, and downstream of the Colorado Rockies (Fig. 7). Curiously, the cyclogenesis density maxima in the NARR are somewhat weaker than produced by the ERA-Interim downstream of the Canadian Rockies and Colorado Rockies (cf. Figs. 7a,b). Over the Intermountain West and downstream of the Colorado Rockies, the ERA-Interim and NARR cyclogenesis density analyses exhibit considerable mesoscale structure as noted previously for cyclone-center density and discussed later in section 2b. Consistent with cyclone-center density, the ERA-Interim sea level cyclogenesis density is much greater than found at 850 hPa over much of western North America (cf. Figs. 7a,d).
The mean monthly cyclogenesis frequency distribution for the western United States and the surrounding region is similar in shape to that of cyclone-center frequency (i.e., bimodal), but less amplified (cf. Figs. 6a,b). The ERA-Interim and NARR produce broad maxima in the spring and fall. Minima occur in December and August. The NCEP–NCAR reanalysis produces the fewest events. The ERA-Interim generates far more sea level than 850 hPa cyclogenesis events in all months. Given the apparent artificial influence of thermal lows and pressure reduction on cyclone-center and genesis occurrence, the remainder of this paper concentrates on 850-hPa cyclone statistics.
In all three reanalyses, 850-hPa cyclolysis density is greatest in the favored regions for cyclone centers and genesis noted above (Fig. 8). Lysis is, however, somewhat more common than genesis over the northeast Pacific (especially near the coast), whereas genesis is more common than lysis immediately downstream of the Canadian Rockies (cf. Figs. 7 and 8). The high cyclolysis density over the Intermountain West reflects the locally high cyclone activity, but also the fact that many ICs decay within the region or cannot be traced as coherent +40-m height minima into the lee of the Rockies, as discussed further in the next section. The monthly frequency of lysis resembles that of genesis (cf. Figs. 6b,c).
b. IC centers, genesis, and lysis
The analysis described above suggests that cyclone statistics over the Intermountain West exhibit considerable mesoscale structure that has not been documented in previous studies. Indeed, the 850-hPa cyclone-center density analysis over the Intermountain West is strongly influenced by the regional orography, but is also sensitive to the resolution, terrain representation, and assimilation methodology used in each reanalysis. In the ERA-Interim, cyclone centers are most commonly found in a region that extends from the lee of the high Sierra northeastward into northwest Utah and then northward into eastern Idaho (Fig. 9a). We refer to the region of high cyclone-center density between the high Sierra and northwest Utah, which has a clear analogy in the NARR (Fig. 9b), as the Great Basin cyclone region. Within the Great Basin cyclone region the ERA-Interim produces cyclone-center density maxima over southern Nevada (somewhat downstream from the immediate lee of the high Sierra) and along the western edge of the Great Salt Lake Basin (Fig. 9a). Farther east, another region of locally high cyclone-center density lies over the upper Colorado River basin of southeast Utah, a lowland region between the high plateaus and mountains of central Utah and the Colorado Rockies that we refer to as the Canyonlands cyclone region (see also Fig. 1). High cyclone-center densities extend from this region into north-central Wyoming. Although each distinct cyclone region may be interpreted as a preferred mesoscale cyclone track, this is not necessarily the case. The two regions lie in or downstream of areas of preferred orographic vortex stretching downstream of high topography and it is possible for cyclone centers to move discontinuously from one region to the other (e.g., a low center dissipates in the Great Basin cyclone region as a new low center forms in the Canyonlands cyclone region).
The NARR also produces high cyclone-center densities in the Great Basin and Canyonlands cyclone regions, but with some differences compared to the ERA-Interim (cf. Figs. 9a,b). In the Great Basin cyclone region, the cyclone-center density maximum over southern Nevada is shifted toward the high Sierra, there is no secondary maximum along the western edge of Great Salt Lake Basin, and the extension of high cyclone-center densities into eastern Idaho is weaker. In the Canyonlands cyclone region, the NARR cyclone-center densities are much larger than produced by the ERA interim. In contrast, in eastern Colorado, the NARR cyclone-center density maximum is weaker than produced by the ERA-Interim. It is unknown if these disparities reflect differences in the resolution and topographic representation, or differences in the observations and assimilation techniques used by the ERA-Interim and NARR. In comparison, the coarse resolution NCEP–NCAR reanalysis fails to capture the mesoscale details found in the ERA-Interim and NARR (Fig. 9c). Instead, a broad region of high cyclone-center density covers the Intermountain West, with a broad corridor of high cyclone-center density extending downstream from eastern Colorado (see also Fig. 5).
The mean monthly 850-hPa IC center frequency distribution (i.e., for cyclone centers in the intermountain region identified in Fig. 1) is bimodal in all three reanalyses, with an absolute maximum in May and secondary maximum in September or October (Fig. 10a). Minima occur in July or August and November or December. The bimodality in the ERA-Interim and NARR is much stronger than produced by the NCEP–NCAR reanalysis and more amplified than found for overall for the western United States and the surrounding regions (cf. Figs. 6a and 10a). Lee (1995) attributes this bimodality to the semiannual migration of the polar jet stream across the Intermountain West as it moves southward with the approach of winter and then northward with the approach of summer. Greater activity in spring compared to fall reflects decreased static stability and enhanced coupling with the upper levels. The bimodality, however, stands in stark contrast to the mean monthly 850-hPa cyclone frequency over the ocean domain (identified in Fig. 1), which we use as a proxy for the upstream cyclone activity, and features a single maximum in February, a minimum in June, and only a weak secondary maximum in cyclone activity in the spring (Fig. 10d, see Fig. 1 for domain location). Additional research is needed to reconcile these contrasts.
The ERA-Interim produces the highest IC genesis density in the Great Basin cyclone region, downstream of Sierra Cascade ranges in northwest Nevada, and within the Canyonlands cyclone region (Fig. 11a). The strongest maximum is found over southern Nevada, within the Great Basin cyclone region and somewhat downstream of the poorly resolved high Sierra. A secondary maximum is found near the western edge of the Great Salt Lake Basin. Although the NARR also produces the highest IC genesis density in the Great Basin and Canyonlands cyclone regions, there are differences compared to the ERA-Interim (cf. Figs. 11a,b). First, the NARR does not generate an area of high genesis density downstream of the Sierra Cascade ranges in northwest Nevada. Second, within the Great Basin cyclone region, the primary cyclogenesis maximum is shifted into the direct lee of the high Sierra, and there is no secondary maximum near the western edge of the Great Salt Lake Basin. Finally, the cyclogenesis density is greater in the Canyonlands cyclone region. While these changes may reflect differences in resolution and terrain representation, lower cyclogenesis density downstream of high topography and within the better resolved lowland regions of northwest Nevada and the Great Salt Lake Basin is inconsistent with this observation. We hypothesize that these differences also reflect differences in the observations and techniques used during assimilation. Nevertheless, the ERA-Interim and NARR agree on the existence of two mesoscale regions of high IC genesis density, whereas the NCEP–NCAR reanalysis generates high IC genesis density only in the Great Basin cyclone region (Fig. 11c).
The mean monthly ERA-Interim–NARR IC genesis frequency distribution is bimodal, similar to that of IC center frequency (Fig. 10c). The genesis frequency maxima occur in May (primary) and September or October (secondary). The NCEP–NCAR reanalysis distribution is also bimodal, but with less amplitude. From 1989 to 2008 the ERA-Interim produces 283 cases of IC genesis (14.2 yr−1), whereas the NARR produces 334 (16.7 yr−1). For comparison, Lee (1995) identified 250 cyclogenesis events in sea level pressure analyses from 1976 to 1986 (22.7 yr−1). As with cyclone frequency, the strong bimodality in IC genesis stands in stark contrast to the cyclogenesis frequency upstream over the eastern Pacific (Fig. 10d). Based on these results, and the quasi-modal distribution of cyclogenesis observed downstream of the Canadian Rockies by Whittaker and Horn (1981, see their Fig. 6), we hypothesize that the cumulative bimodal cyclone-center and genesis frequency distributions for the western United States and the surrounding region (Fig. 6a) reflects the superposition of strongly bimodal cyclone-center and genesis frequency distributions over the Intermountain West and Colorado with weakly modal cyclone-center and genesis frequency distributions in other cyclonically active regions.
Petterssen (1956) suggested that most cyclones that develop in the Intermountain West also decay there. In the ERA-Interim (NARR), 36% (42%) of the cyclones that form in the Intermountain West decay within the region. On the mesoscale, IC lysis density maxima in the Great Basin and Canyonlands cyclone regions reflect the fact that many cyclones remain resident in these regions for their complete life cycle (Figs. 12a,b). On the other hand, weak IC lysis density maxima downstream of the Colorado Rockies and over New Mexico reflects the movement of some ICs out of the region, although it is possible that this movement is not continuous (e.g., a cyclone may move discontinuously from the Canyonlands cyclone region to eastern Colorado, but be counted as a single cyclone since that is within the confines of our 8.5° search radius). The IC lysis density analysis from the NCEP–NCAR reanalysis lacks such mesoscale detail, but is also largest over the Intermountain West (Fig. 12c). The monthly IC lysis frequency, by definition, resembles that of IC genesis and is not shown.
Consistent with Lee (1995), higher-amplitude cyclones (e.g., maximum amplitude >100 m) are most common over the Intermountain West from April to June in both the ERA-Interim and NARR (Figs. 13a,b). Nevertheless, the ERA-Interim and to a lesser degree the NARR, produce some deep storms in February and March.
c. Composite evolution of high-amplitude intermountain cyclone events
A subset of strong ICs (i.e., within the 90th percentile of maximum amplitude for a total of 31 cyclones) was selected to classify the large-scale flow (relative to the orientation of the Sierra Nevada) during cyclogenesis and to generate a composite that summarizes key aspects of IC evolution. At genesis time, 23 of the events (74%) feature mean cross-barrier south-southwesterly to west-northwesterly (210°–300°, hereafter SW) 500-hPa flow at the three ERA-Interim grid points immediately upstream of the Sierra Nevada (Fig. 14, see Fig. 4a for gridpoint locations and Table 1 for event dates and flow classification). The remaining eight events (26%) feature west-northwesterly to northerly flow (301°–5°), most with either along-barrier or reversed cross-barrier (i.e., from the climatological lee) flow. Lee (1995) performed a similar but subjective event categorization using the 250-hPa flow (with 280° used instead of 300° as the discriminator between SW and NW events) and found that 75% and 24% of the events fell into his SW and NW categories, respectively.
We use the 23 SW flow events to produce an IC composite of events that are likely strongly influenced by the Sierra Nevada. For the times presented, the composite dynamic tropopause potential temperature, 700-hPa temperature, and 850-hPa geopotential height across most of the analysis domain deviates significantly (>95% confidence based on the two-sided Student’s t test; Wilks 2006, p. 139) from a monthly weighted ERA-Interim climatology (1989–2008) that accounts for the proportion of events from each month.
At 24 h prior to cyclogenesis (t = −24 h), a cyclonic potential vorticity (PV) anomaly (as indicated by low potential temperature on the dynamic tropopause; Martin 2006, section 9.2) and upper-level short wave trough axis lie just upstream of the Pacific Northwest, with weak 500-hPa ascent found downstream over much of the western United States (Fig. 15a). West to southwesterly 700-hPa flow impinges on the Sierra Nevada (the Sierra crest lies just above this level), with a broad baroclinic zone draped across the northern portion of the Intermountain West (Fig. 15b). At 850 hPa, flow splitting occurs upstream of the high Sierra over central California (Fig. 15c). Downstream of the high Sierra, the Great Basin confluence zone (GBCZ), an airstream boundary that frequently contributes to cold-frontal development over the Intermountain West and separates southerly to southwesterly flow over Utah and southern Nevada from westerly flow over northern Nevada (Shafer and Steenburgh 2008; West and Steenburgh 2010), extends across the Intermountain West. Although one might expect a traditional barrier-parallel lee trough as observed during Rocky Mountain lee cyclogenesis (e.g., Steenburgh and Mass 1994), the most pronounced surface trough lies normal to the Sierra Nevada along the GBCZ and near the leading edge of the 700-hPa baroclinic zone.
Cyclogenesis occurs at t = 0 h as the cyclonic PV anomaly amplifies and digs over California and large-scale ascent strengthens over the Intermountain West (Fig. 16a). Despite persistent cross-barrier flow at crest level (Fig. 16b), the composite low center does not form in the direct lee of the Sierra Nevada, but instead ∼300 km downstream over central Nevada along the GBCZ and leading edge of the 700-hPa baroclinic zone, which is amplifying into a frontal wave (Fig. 16c). This evolution is remarkably similar to that observed by West and Steenburgh (2010) during the 2002 Tax Day Storm and suggests that the orographic modification of IC genesis is more complex than may be inferred from studies of Rocky Mountain lee cyclogenesis (e.g., Palmén and Newton 1969; Bannon 1992; Steenburgh and Mass 1994; Davis 1997; Schultz and Doswell 2000; Thomas and Martin 2007). In particular, preexisting low-level baroclinicity over the Intermountain West, which may be concentrated in some cases along the GBCZ (West and Steenburgh 2010), appears to act as a surface thermal anomaly in the classical PV view of cyclone development (Hoskins et al. 1985). This shifts the locus for cyclone development from the Sierra Nevada to farther downstream along the developing GBCZ and frontal boundary.
As the upper-level cyclonic PV anomaly and 500-hPa ascent slide southeastward by t = +12 h (Fig. 17a), the composite surface low shifts into central Utah while strong troughing remains quasi-stationary over southern Nevada (Fig. 17c). The complex 500-hPa vertical motion pattern at this and other times reflects both large-scale forcing and orographic effects. For example, ascent over the four corners region is enhanced by strong upslope flow over the Colorado Rockies (as resolved by the ERA-Interim). A sharp 700-hPa wind shift with cold advection in its wake lies over central Nevada, consistent with either in situ cold frontogenesis along the GBCZ, as described by West and Steenburgh (2010), and/or the GBCZ being overtaken by the 700-hPa wind shift that was located over eastern Oregon 12 h earlier (Fig. 16b). Although limitations of compositing preclude a definitive conclusion, the latter appears unlikely since temperatures along the 700-hPa wind shift are 3°–6°C warmer than 12 h earlier (cf. Figs. 16b and 17b) and the 350–600-km difference in trough position is inconsistent with the 110–325 km expected based on advection by the 2.5–7.5 m s−1 winds along and behind the wind shift. Thus, it appears that the composite reflects cold frontogenesis along the GBCZ during IC genesis.
The cyclonic PV anomaly and 500-hPa ascent continue to slide southeastward through t = +24 h (Fig. 18a). An elongated surface trough now extends across southern Nevada and Utah with the surface low center over the Canyonlands cyclone region detached from the surface wind shift (Figs. 18b,c). Inspection of individual events contributing to the composite shows a rich spectrum of cyclone positions and intensities at this time (not shown). Thus, even though the composite deviates significantly from climatology at a significance level greater than 95%, individual cyclone tracks may vary from the composite, with low centers often moving discretely between preferred lowland regions of high cyclone activity.
The composite evolution described above captures many of the salient characteristics of IC evolution as described in the recent case study of West and Steenburgh (2010). In particular, the composite features the development of the GBCZ downstream of the high Sierra, with the GBCZ becoming the locus for surface cyclogenesis as an upper-level cyclonic PV anomaly, concomitant trough, and related large-scale ascent move over the Intermountain West. The composite GBCZ also appears to develop the character of a cold front (i.e., cold advection and a cyclonic wind shift), consistent with the West and Steenburgh (2010) case study and the climatological study of strong cold fronts by Shafer and Steenburgh (2008), which identified confluence downstream of the Sierra Nevada as an important mechanism for intermountain cold-frontal development. Although the surface evolution of individual cases may vary, the composite illustrates that the GBCZ is an important regional airstream boundary that accompanies many ICs and may serve as a locus for surface cyclogenesis and frontogenesis.
The composite evolution also indicates that the Sierra Nevada plays an important role in IC genesis, but in ways that differ from traditional quasigeostrophic models of lee cyclogenesis (e.g., Tibaldi et al. 1990; Bannon 1992; Davis 1997; Davis and Stoelinga 1999) or from observational studies of Rocky Mountain lee cyclogenesis (e.g., Palmén and Newton 1969; Steenburgh and Mass 1994; Schultz and Doswell 2000; Thomas and Martin 2007). Rocky Mountain lee cyclogenesis typically occurs within a barrier parallel lee trough, with the overall cyclone evolution across the western North America cordillera described as “amoeba-like” (Palmén and Newton 1969; Bannon 1992). This evolution is well captured by quasigeostrophic models. In the case of IC evolution, however, the dominant surface feature is not a barrier-parallel lee trough, but the GBCZ, which is oriented normal to the high Sierra and extends downstream across the Intermountain West. Strong flow splitting around the high Sierra during IC genesis is indicative of orographic blocking, which is not accounted for by quasigeostrophic models. For example, quasigeostrophic models are quantitatively invalid if the mountain aspect ratio, H/L is larger than f/N, where H and L are the mountain height and length scales, f is the Coriolis parameter, and N is the Brunt–Väisälä frequency (Davis 1997). In the case of the high Sierra, a narrow, steeply sloped range, H ∼ 4 km, L ∼ 100 km, f ∼ 10−4 s−1, and N ∼ 1.2 × 10−2 s−1, yielding a mountain aspect ratio that is ∼5 times larger than f/N.
Instead, we propose that IC genesis might be better conceptualized through an adaptation of the Mattocks and Bleck (1986) model of Alpine lee cyclogenesis. In their model, the Alps block the movement of an impinging low-level cold front as an upper-level trough and cyclonic PV anomaly move relatively unimpeded into the lee. Disturbance of the vertical alignment between the cold front and upper-level PV anomaly disrupts thermal wind balance, leading to vigorous geostrophic adjustment, vortex stretching, and cyclogenesis as the upper-level PV anomaly moves into the lee of the Alps.
There are three important differences, however, with IC genesis (as inferred from the composite). The first difference is the orientation of Sierra Nevada, which are nearly orthogonal in orientation relative to the Alps (i.e., from south-southeast–north-northwest rather than mostly west–east with a hook on the west end). The second difference is the large-scale pattern, which features a low-level baroclinic zone that does not impinge on the Sierra Nevada, but rather has a leading edge that drapes across the high Sierra prior to cyclogenesis. As a result, as the cyclonic PV anomaly and large-scale ascent intensify and move across the Sierra Nevada, they are interacting with a low-level baroclinic zone that is being concentrated by confluence downstream of the high Sierra. This represents an additional surface (i.e., surrogate) PV anomaly in the classic PV view of cyclogenesis (Hoskins et al. 1985) that is not accounted for by the Mattocks and Bleck (1986) model of Alpine lee cyclogenesis. The third difference is the presence of substantive downstream orography. Future work could use a similar modeling strategy to Mattocks and Bleck (1986), but with a mountain barrier oriented like the Sierra Nevada, to develop a dynamically based conceptual model for IC genesis and perhaps provide a robust comparison with Alpine lee cyclogenesis. This work could also explore how IC life cycles are modified by downstream orography.
The comprehensive climatology and composite analysis presented in this paper resolves some of the contradictions apparent in previous studies with regard to IC activity and provides new insights into IC evolution. The monthly distribution of IC center and genesis frequency is clearly bimodal with spring (absolute) and fall (secondary) maxima. Lee (1995) suggests that this bimodality reflects the equatorward migration of the polar jet stream across the Intermountain West with the approach of winter and poleward migration with the approach of summer, although the lack of similar bimodality immediately upstream over the Pacific Ocean suggests that the causes may be more complex and warrant further investigation. IC center and genesis density derived from the ERA-Interim and NARR further show that the Great Basin and Canyonlands cyclone regions represent two distinct areas of high cyclone activity that likely result from mesoscale flow-mountain interactions over the region. The former lies downstream of the southern “high” Sierra, the region’s most formidable barrier, whereas the latter is a lowland region between the high mountains and plateaus of central Utah and the Colorado Rockies. Many ICs are relatively short lived and decay within the Intermountain West, although some penetrate into the lee of the Rockies or southeastward into New Mexico.
The composite of the strongest IC events (i.e., 90th percentile for maximum amplitude) produced in large-scale southerly to westerly (SW) flow reveals many key aspects of IC evolution. As cross-barrier flow develops in advance of a digging upper-level cyclonic PV anomaly from the northwest, a regional airstream boundary known as the Great Basin confluence zone (GBCZ) forms downstream of, and normal to, the high Sierra. The GBCZ is the dominant surface feature during the cyclogenesis, not a barrier-parallel lee trough as one might expect from quasigeostrophic theory and observational studies of Rocky Mountain lee cyclogenesis (e.g., Palmén and Newton 1969; Bannon 1992; Steenburgh and Mass 1994; Davis 1997; Davis and Stoelinga 1999; Schultz and Doswell 2000; Thomas and Martin 2007). Cyclogenesis and frontogenesis occur along the GBCZ as ascent ahead of the upper-level cyclonic PV anomaly intensifies over the Intermountain West. The low center does not form in the direct lee of the Sierra Nevada, but ∼300 km downstream over central Nevada. This appears to be a consequence of low-level baroclinicity, which is draped across the Intermountain West, may be concentrated by the GBCZ (e.g., West and Steenburgh 2010), and appears to act as a surrogate cyclonic PV anomaly (Hoskins et al. 1985). After a period of quasistationary development, the GBCZ and low center move downstream, with individual cyclone evolution varying by event depending on the details of the large-scale evolution and front–mountain interactions.
The composite evolution also illustrates that the Sierra Nevada play a key role in IC evolution, but in ways that differ from traditional quasigeostrophic models of lee cyclogenesis. In particular, they act as a steep, quasi-isolated obstacle, resulting in windward flow splitting and the development of the GBCZ to the lee. The GBCZ is the dominant surface feature during cyclogenesis but, unlike a classic lee trough, is oriented normal to the Sierra Nevada. Future work should examine the mechanisms responsible for the GBCZ and determine if IC evolution might be better conceptualized from a PV perspective, possibly involving an adaptation of the Mattocks and Bleck (1986) model of Alpine lee cyclogenesis.
Although these findings provide new insights into IC activity and evolution, analysis uncertainty remains an important source of ambiguity. The NCEP–NCAR reanalysis lacks sufficient resolution to resolve many ICs, but differences between the ERA-Interim and NARR indicate that increasing resolution does not eliminate uncertainties with regard to the regional distribution of cyclone activity. For example, both the ERA-Interim and NARR produce maxima in IC center and genesis density over the Canyonlands cyclone regions, but the NARR maximum is much stronger, indicating more frequent cyclone activity. Thus, we have confidence that this is an area of preferred cyclone activity (which is consistent with our synoptic experience), but its relative importance remains uncertain. Given the dearth of traditional observations over the region, nonoperational mesonet data such as that collected by MesoWest (Horel et al. 2002) could help reduce such uncertainties in future reanalyses.
A second source of ambiguity is the cyclone identification criteria. Our methodology concentrates on a pressure level consistent with the elevation of the valleys and basins of the Intermountain West, which helps reduce (but not eliminate) issues related to pressure reduction. By necessity, however, we have used a relatively short life cycle criterion (≥12 h), which will complicate comparison with other studies, many of which have life cycle requirements of as long as 72 h. Another concern is the 8.5° search radius, which is very large but required to deal with discontinuous low-center propagation. This threshold limits the spatial scale of discontinuous propagation and may eliminate some events that undergo dramatic redevelopment downstream (e.g., from Nevada to the lee of the Colorado Rockies), while perhaps including others that are essentially new cyclones. Finally, identifying and eliminating thermal lows remains a challenging problem. These sources of ambiguity, along with the aforementioned analysis uncertainty, cannot be fully eliminated. They could even become more problematic in higher-resolution reanalyses that contain stronger and smaller-scale thermally and dynamically forced orographic low pressure systems.
We thank Court Strong, Larry Dunn, and John Horel for their suggestions and comments and gratefully acknowledge the provision of datasets, software, or computer time and services by the ECMWF, NCEP, NCAR, Unidata Program Center, Center for Ocean–Land–Atmosphere Studies, and University of Utah Center for High Performance Computing. This research was supported by National Science Foundation Grant ATM-0627937 and the National Weather Service C-STAR program. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation or the National Weather Service.
Corresponding author address: Dr. W. James Steenburgh, Department of Atmospheric Sciences, University of Utah, 135 South 1460 East Room 819, Salt Lake City, UT 84112. Email: email@example.com