A large-amplitude lee-wave rotor event observationally documented during Sierra Rotors Project Intensive Observing Period (IOP) 8 on 24–26 March 2004 in the lee of the southern Sierra Nevada is examined. Mountain waves and rotors occurred over Owens Valley in a pre-cold-frontal environment. In this study, the evolution and structure of the observed and numerically simulated mountain waves and rotors during the event on 25 March, in which the horizontal circulation associated with the rotor was observed as an opposing, easterly flow by the mesonetwork of surface stations in Owens Valley, are analyzed.
The high-resolution numerical simulations of this case, performed with the Coupled Ocean–Atmosphere Mesoscale Prediction System (COAMPS) run with multiple nested-grid domains, the finest grid having 333-m horizontal spacing, reproduced many of the observed features of this event. These include small-amplitude waves above the Sierra ridge decoupled from thermally forced flow within the valley, and a large-amplitude mountain wave, turbulent rotor, and strong westerlies on the Sierra Nevada lee slopes during the period of the observed surface easterly flow. The sequence of the observed and simulated events shows a pronounced diurnal variation with the maximum wave and rotor activity occurring in the early evening hours during both days of IOP 8.
The lee-wave response, and thus indirectly the appearance of lee-wave rotor during the core IOP 8 period, is found to be strongly controlled by temporal changes in the upstream ambient wind and stability profiles. The downstream mountain range exerts strong control over the lee-wave horizontal wavelength during the strongest part of this event, thus exhibiting the control over the cross-valley position of the rotor and the degree of strong downslope wind penetration into the valley.
Atmospheric rotors are low-level horizontal vortices that form downstream of and parallel to the crest of a mountain range in close association with large-amplitude mountain waves. High levels of turbulence characterize rotors, which are known to pose severe aeronautical hazards and have been cited as contributing to accidents involving commercial, military, and civilian aircraft (Carney et al. 1997). On the dry lee sides of mountain ranges, rotor circulations can be important for the lofting and transport of aerosols and chemical and biological contaminants (Raloff 2001). This is particularly true over the dry Owens Lake bed located in the lee of the southern Sierra Nevada (Reid et al. 1993). Despite their considerable impact on human activity, and in contrast to the attendant mountain waves, rotors remain relatively poorly understood atmospheric phenomenon. One reason that so little is known about the dynamics of rotors and their internal structure is that they are too small in spatial scale to be routinely sampled by conventional observing networks. Complex terrain settings in which they occur, high levels of turbulence, and intermittent internal structure make rotors difficult to sample using in situ aircraft measurements.
A brief historical account of the early twentieth-century studies of rotors is given in Hertenstein and Kuettner (2005). The term rotor was first applied by Küttner (1938, 1939) following his observational studies using sailplanes over the Riesengebirge in the Sudeten Mountains on the border between Poland and the Czech Republic. In the following decades, theories of atmospheric rotors were advanced, and Kuettner (1959) lists six prominent ones. With the advancement of atmospheric numerical models and the increase in computing power, recent dynamical studies of rotors increasingly rely on numerical modeling (e.g., Doyle and Durran 2002; Vosper 2004; Hertenstein and Kuettner 2005).
The observational documentation of atmospheric rotors was considerably advanced during the 1950s Sierra Wave Project in Owens Valley, California, in the lee of the southern Sierra Nevada (Holmboe and Klieforth 1957). While primarily designed as a study of the mountain lee-wave phenomenon, data on atmospheric rotors were collected as well because the primary instrumentation in both phases of this project were instrumented sailplanes (Grubišić and Lewis 2004). A number of important observations relating to the mountain wave dynamics were made during this project. The strongest wave events were correlated with strong winds at the mountain crest level, a pronounced inversion layer between 3600 and 5800 m (approximately at or 2 km above mountain crest level), and large vertical shear in the lower troposphere. Atmospheric rotors and associated roll (rotor) clouds were classified into two types. A typical roll cloud, associated with a lee-wave rotor, was found to be located under the crest of a lee wave, paralleling the topography and its curvature. The roll cloud of a severe rotor formed an almost vertical wall a considerable distance downstream from the lee slope. This more severe case contained no trailing edge to the roll cloud and the airflow in the lee was found to be similar to a hydraulic jump (Kuettner 1959).
Another important outcome of the Sierra Wave Project was identification of the synoptic setting in which strong wave events form in the lee of the Sierra Nevada. The strongest wave events were found to be associated with 1) an upper-level pressure trough along the Pacific coast with strong westerly flow across the Sierra, and 2) a cold or occluded front approaching California from the northwest, placing Owens Valley in the prefrontal environment. In addition, the jet stream typically was found to cross Oregon or northern California during strong wave events, though some events did occur with a westerly jet stream crossing the Sierra. It was also observed that cold air damming occurred to some extent along the upwind side of the Sierra Nevada (Holmboe and Klieforth 1957).
The Sierra Rotors Project (SRP) represents phase I of a new coordinated multiyear effort to study atmospheric rotors and related phenomena in complex terrain. Phase II of this effort is the recently completed Terrain-Induced Rotor Experiment (T-REX; Grubišić and Doyle 2006). The SRP Special Observing Period (SOP) took place during March and April 2004 in Owens Valley, California. Only ground-based instruments were deployed in SRP, and the instrumentation suite consisted of mesoscale surface observations, GPS radiosonde systems, and wind profiling radars. Sixteen Intensive Observing Periods (IOPs) were conducted during the 2-month field campaign. An intense mountain wave and rotor event occurred during IOP 8 on 24–26 March. In this study, analysis of all available observations and results of numerical model simulations of IOP 8 mountain wave and rotor event are presented to determine the extent to which a high-resolution mesoscale numerical model simulation can reproduce the observed features of a strong rotor event and to identify environmental factors that foster strong rotor development over Owens Valley.
The paper is organized as follows. In section 2 we describe the instrumentation and observations from the Sierra Rotors Project and the design of our numerical simulations. The synoptic setting, upstream atmospheric structure, and surface and wind profiler observations during a three-day period of disturbed weather surrounding IOP 8 are presented in section 3. Observations and numerical simulation of the temporal evolution and spatial structure of the flow on the lee side of the Sierra Nevada during the core period of this event are presented in section 4. The role of ambient upstream conditions and downstream topography in determining the mountain wave response and appearance of rotors in this case are discussed in section 5. Section 6 concludes the paper.
2. Field measurements and numerical simulations
a. Sierra Rotors Project: Field site and instrumentation
Owens Valley is a nearly north–south-oriented rift valley located to the lee of the Sierra Nevada in eastern California (Fig. 1a). The Sierra Nevada forms the valley’s western boundary. Its eastern boundary is formed by two mountain ranges: the White Mountains to the north and the Inyo Mountains to the south connected at Westgard Pass east of Big Pine (Fig. 1b). The width of the valley is approximately 30 km from ridgetop to ridgetop and 15 km at the valley floor. The southern half of the valley, between the towns of Big Pine and Lone Pine, features some of the steepest relief in the contiguous United States (Fig. 1c). The ridge line of the Sierra Nevada reaches an average elevation of 3500 m, and the tallest peaks exceed 4000 m, including the tallest mountain in the contiguous United States, 4418-m Mount Whitney. There are a few, lower-elevation passes along the Sierra Nevada, including Kearsarge Pass, located west of Independence, and Sawmill Pass, located further to the northwest (Fig. 1c). In contrast to the tallest peaks, the valley floor lies at an average elevation of approximately 1150 m. This 3000-m elevation change occurs in less than 10 km of horizontal distance, with the resulting 30% eastern Sierra Nevada slopes. A similar slope exists along the Inyo Mountains on the eastern side of the valley. The valley bottom is made up of a gradually sloping (5%) alluvial fan on the western portion of the valley floor and a more flat portion to the east (Fig. 1c). Owens (dry) Lake bed, the largest point source of particulate matter in the United States, lies just to the south of this area. As steep lee side slopes are excellent mountain wave generators and aerosol lofting and transport is an important problem related to mountain waves, rotors and downslope windstorms, Owens Valley was an ideal setting for the project. The ground-based measurement systems deployed in SRP were located both on the lee side of the Sierra Nevada in Owens Valley as well as on the upwind side in San Joaquin Valley.
At the core of the SRP instrumentation in Owens Valley was a long-term network of automatic weather stations (AWS) installed and operated by the Desert Research Institute (DRI) for the study of atmospheric rotors. The DRI network consists of 16 AWS with telemetry arranged in three approximately parallel rows that extend across the valley south of Independence (Fig. 1c). The average separation between individual stations along these three lines is approximately 3 km. Each station has a standard 10-m meteorological tower, and sensors for wind, temperature, relative humidity, and pressure. All quantities are measured to World Meteorological Organization (WMO) standards. Sensors are sampled every three seconds, and these readings are temporally averaged over 30-s nonoverlapping intervals. Thirty-second data are transmitted via radio communication to the base station in Independence, and from there, via Internet to a central repository at DRI. (The observations are available in near–real time at http://www.wrcc.dri.edu/trex/.) The DRI network has been in continuous operation since late February 2004.
During the two months of SRP, two National Center for Atmospheric Research/Earth Observing Laboratory (NCAR/EOL) Integrating Sounding Systems (ISS) were deployed in Owens Valley. The two NCAR ISS were the Multiple Antenna Profiler ISS (MAPR; Cohn et al. 2001) and the Mobile Integrated Sounding System (MISS; Cohn et al. 2004). Both systems contain a 915-MHz boundary layer radar wind profiler, a Radio Acoustic Soundings System (RASS), radiosonde launch capability, and an automated surface station that measures temperature, relative humidity, pressure, wind speed and direction, rainfall accumulation, and radiative fluxes. MAPR was positioned in the center of the valley near the southern line of AWS (Fig. 1c). Since MAPR contains multiple vertically pointing antennae, it can obtain a horizontal wind vector and vertical measurement without the need to scan in multiple beam directions, resulting in a higher temporal resolution (approximately 5 min in the horizontal and 30 s in the vertical). MISS is a mobile system that could be transported to the area of strongest wave activity. During IOP 8, MISS was located at the airport north of Independence (Fig. 1c).
Given the importance of documenting the upstream thermodynamic and kinematic flow structure for interpreting the mountain wave response, substantial effort was made to obtain adequate upwind upper-air sounding data. During SRP, two supplemental upstream sounding sites were located in San Joaquin Valley. The NCAR’s Mobile GPS Advanced Upper Air System (MGAUS) was based in Fresno with an ability to move to an optimal upstream location depending on the wind direction. The fixed system was operated at the Naval Air Station (NAS) Lemoore, located southwest of Independence (Fig. 1b). During IOP 8, MGAUS was located in Fresno.
Wave activity was observed during 9 of the 16 IOPs conducted during the 2-month field compaign, and potential rotor activity was observed during three of these events.
b. Numerical simulations
The numerical model used in this study is the Naval Research Laboratory’s (NRL) Coupled Ocean–Atmosphere Mesoscale Prediction System (COAMPS; Hodur 1997), a limited area, nonhydrostatic, fully compressible prognostic model with multiple nesting capability and an atmospheric data assimilation system based on multivariate optimum interpolation analysis (Barker 1992). The physical parameterizations used in this study include subgrid-scale turbulent mixing based on a prognostic equation for the turbulent kinetic energy (Mellor and Yamada 1982), cloud microphysics (Rutledge and Hobbs 1983), surface fluxes (Louis 1979), and the short- and longwave radiation (Harshvardhan et al. 1987).
The baseline IOP 8 simulation was run using a 200-s time step, and five nested domains with horizontal grid increments of 27, 9, 3, 1, and 0.33 km placed within the outer domain with grid spacing of 81 km (Fig. 2). The number of horizontal grid points in each domain is 61 × 61, 82 × 82, 109 × 100, 124 × 142, 127 × 160, and 190 × 190. In the vertical, 60 sigma levels were used with variable spacing from 10 m at the surface to 3.75 km at the model top, which was set at 34.8 km with a sponge layer applied to the upper 13.3 km in which all prognostic fields were relaxed to their smoothed values. Lateral boundary conditions were specified using the Naval Operational Global Atmospheric Prediction System (NOGAPS) forecast fields, which were applied every 6 h throughout all simulations. The simulation consisted of a 12-h spinup run initialized at 1200 UTC 23 March, and three subsequent 24-h runs initialized at 0000 UTC 24, 25, and 26 March. The initial fields for the 24-h runs were created by blending the previous COAMPS forecast with the NOGAPS analysis and available operational upper-air soundings and surface data.
In addition to the baseline run, we have conducted additional experiments in order to investigate sensitivity of our numerical results to the effects of radiational heating as well as downstream topography. In the sensitivity run for radiation, the radiation scheme was applied every 15 min, compared to every 60 min in the baseline run. In the sensitivity run for downstream topography, the White and Mountains were replaced in all model domains with flat terrain, matching the height of the lowest part of Owens Valley.
3. SRP IOP 8
SRP IOP 8 extends over 2 days of a 3-day period of disturbed weather in the southern Sierra Nevada. The extended IOP was conducted from 1800 UTC 24 March to 1800 UTC 26 March 2004.1 The core observing period of IOP 8, centered around 0000 UTC 26 March, was 24 h long (1200 UTC 25 March–1200 UTC 26 March). During the core period, the GPS radiosondes were launched every 3 h, both upstream of the Sierra Nevada and in Owens Valley. In the following, we describe the synoptic setting, upstream environment, and surface and wind profiler observations during this IOP.
a. Synoptic overview
The synoptic situation at 0000 UTC 26 March, the central time of the core IOP 8 period, is shown in Fig. 3. An occluded surface low is located off the coast of British Columbia (Fig. 3a). The cyclone’s trailing cold front has started to move inland at this time and is now passing through Northern California and the San Francisco Bay. This front would pass over Owens Valley between 0900 and 1200 UTC 26 March.
At 700 hPa, which is approximately at the height of the Sierra Nevada crest (∼3000 m), the upper-level low is producing strong west-southwest (WSW) flow and moisture advection across the Sierra Nevada at this time. The two operational radiosonde locations closest to the Sierra mountains are located in Reno and Desert Rock, Nevada. At 0000 UTC 26 March, the 700- and 500-hPa wind speeds at Reno were both 27.5 m s−1, indicating there was little wind shear over the northern Sierra Nevada at this time. The dewpoint depression in this sounding had also decreased substantially from 12 h earlier. At the same time, the Desert Rock sounding indicates an increase of wind with height from 10 m s−1 at 700 hPa to 17.5 m s−1 at 500 hPa across the southern Sierra. Twelve hours later, following the cold-frontal passage, this shear would also decrease over the southern Sierra Nevada. The 250-hPa chart (Fig. 3d) shows a southwest (SW)–northeast (NE)-oriented jet stream, with the core located over the northern Sierra. This synoptic environment is consistent with conditions favorable for mountain wave development identified by Holmboe and Klieforth (1957).
b. Upstream atmospheric structure
The time–height section of the upstream conditions during the core IOP 8 period, constructed using three-hourly MGAUS soundings, is shown in Fig. 4. Prior to 0300 UTC 26 March, the upstream profiles of horizontal wind and stability are characterized by strong vertical wind shear and a distinct two-layer stability structure below tropopause at approximately 11 km MSL. The two-layer stability structure consists of a lower layer of stronger stability separated from a weaker stability layer aloft by a sharp inversion with the base at approximately 5 km MSL. Increased amount of moisture associated with the cold front appears at first as a tongue of increased moisture between 4 and 5 km MSL in 1800 and 2100 UTC 25 March soundings. In subsequent soundings the depth of this moist layer increases significantly, leading to a large amount of precipitation and freezing rain conditions in the western Sierra foothills, cause of the missing data in 0300 and 0600 UTC 26 March soundings. The cold-frontal passage on the upwind side of the Sierra occurred between 0500 and 0600 UTC 26 March, followed by a sharp drop in moisture above 3 km MSL, tropopause descent to below 10 km MSL, and homogenization of the vertical stability profile and reduction of vertical wind shear.
Details of the 1800 UTC 25 March and 1200 UTC 26 March soundings are shown in Figs. 5 and 6, which also show corresponding soundings from the model simulation. In the prefrontal sounding (1800 UTC 25 March; Fig. 5), the mountain normal wind speed component2 increases sharply from 2.5 to 20 m s−1 between 2000 and 5000 m MSL. The difference in the potential temperature lapse rate above and below 5000 m MSL is clearly evident as is a 7-K potential temperature jump in between these two layers (Fig. 5c). The sharp inversion that produces this strong jump in potential temperature is less distinct in earlier and later prefrontal soundings. The two-layer stability structure with an inversion can also be seen in the Brunt–Väisälä frequency profile with average values of N 21 = 2 × 10−4 s−2 in the lower layer and N 22 = 1 × 10−4 s−2 in the upper layer (Fig. 5d). In the postfrontal sounding (1200 UTC 26 March; Fig. 6), the observed mountain normal wind speed does not exceed 20 m s−1 within the troposphere (Fig. 6a), and the lapse rate remains relatively constant until the tropopause is reached (∼8–9 km), resulting in a relatively uniform Brunt–Väisälä frequency profile (Figs. 6c,d).
The observed and simulated soundings shown are in a fairly close agreement, in particular in the prefrontal environment. The wind profiles in the prefrontal environment are particularly in close agreement below 5000 m MSL. Larger discrepancies exist between this altitude and the tropopause, where the model simulation is underpredicting the strength of the barrier normal component of the wind below 10 km MSL, and thus underpredicting the positive shear of the observed wind profiles. The increase of wind, as well as wind shear, with time at those altitudes is captured by the simulation, although with a delay of approximately three hours. The prefrontal simulated stability profiles have the same two-layer stability structure discussed above but lack the inversion around 5000 m, most likely due to the lack of needed vertical resolution to represent this finescale feature. In the postfrontal environment, the discrepancy between the modeled and observed winds and potential temperature profiles is larger, in particular in the 0600 and 0900 UTC soundings. The simulated winds (as well as wind shear) and stability are stronger, and temperature and humidity higher, indicating again approximately 3-h delay in the timing of cold-frontal passage in the model simulation. Some of those discrepancies are seen in the 1200 UTC soundings as well, although by this time the model simulation has once again approached the observed atmospheric structure.
c. Surface and wind profiler observations
Surface observations from Owens Valley for the period 1200 UTC 24 March to 1200 UTC 27 March are illustrated in Figs. 7 and 8. The two figures show station time series of wind, temperature, relative humidity, and pressure for stations 1 and 4 of the DRI network. Station 1 is located on the alluvial fan on the western side of the valley, whereas station 4 lies on the flat part of the valley floor (Fig. 1c). For comparison, in the same figures, we also show corresponding time series for a 3-day period of undisturbed weather in early March 2004.
The undisturbed time series for station 1 shows sharp transitions of wind direction from easterly southeasterly flow during the day to westerly wind at night (Fig. 7d). While some elements of that diurnal pattern can be identified in the time series of wind during the disturbed period (Fig. 7b), that one appears dominated by prevalence of much stronger and gustier westerlies. The diurnal temperature amplitude in the disturbed period is significantly reduced compared to the quiescent period, particularly after frontal passage early on 26 March (cf. Figs. 7a,c). For station 4, the thermally forced circulation pattern is characterized by a transition from a southerly, upvalley flow during the day to a generally northerly, downvalley flow at night (Fig. 8d). As with station 1, some of the characteristic thermal circulation wind directions for this station can also be identified in the 24–27 March period, but strong westerlies on 24 March and northwesterlies throughout 25 and 26 March prevail (Fig. 8b). During the first day of IOP 8, the temperature at this station shows little difference with respect to the quiescent period, indicating prevalence of clear skies above this station at that time (Fig. 8a versus Fig. 8c). During the second and third day of the disturbed period, the temperature amplitude was progressively reduced, showing, respectively, the effect of clouds and the presence of colder postfrontal air mass at this station.
Wind profiler observations for the same time period are shown in Fig. 9, which shows a time–height series of vertical velocity and spectral width measured by MAPR. While vertical velocities are normally less than about 1 m s−1 in the boundary layer, except perhaps for short periods within thermal plumes, the measurements during 24–27 March show many multiple-hour periods during which the wind profiler consistently recorded strong positive or negative vertical motions (Fig. 9a). The sign and strength of these measurements greatly depends on the phase of the wave above the profiler location. The first of these periods starts around 2000 UTC on 24 March, the time of the onset of strong westerly winds seen at the surface stations, with an updraft exceeding 2 m s−1 for more than 3 min at the lowest wind profiler gates (between 0.5 and 1 km AGL). This is followed by an observation of persistent large vertical velocity aloft beginning around 2200 UTC on 24 March, with persistent positive vertical motions continuing through about 0200 UTC 25 March and descending in altitude. Two periods of very strong vertical motions were measured during the core IOP 8 period on 25 March. The strong vertical motions persisted well into 27 March. The vertical velocity maxima measured by the wind profiler during this 3-day period were in the range 6–8 m s−1, and minima in excess of −8 m s−1. The width of the Doppler spectrum, which represents velocity variation within the radar pulse volume during the integration period (typically about 30 s), but is also affected by the horizontal wind speed, seemed to increase during periods of strong vertical motion reaching values in excess of 2 m s−1 (Fig. 9b). Large spectral width is characteristic of turbulence which would be generated by rotors or wind shear.
The diurnal variation in surface wind regimes, which were observed during the core IOP 8 period on 25 March, are also apparent during the previous day in the time series for both stations 1 and 4 (Figs. 7 and 8). This includes a sharp transition to the reversed easterly flow at station 4 immediately following the afternoon peak in the strength of the westerly winds. The pattern repetition is evident in the wind profiler data as well (Fig. 9), where the descent of the strong vertical motions from about 3 to 4 AGL toward lower elevations within the valley, and in particular increase of spectral width below 1 km AGL, is seen both around 0000 UTC 25 March and 0000 UTC 26 March. These observations and their relation to the evolution of mountain waves and rotors above Owens Valley are discussed in more detail in the next sections.
Figure 10 shows the comparison between the observed and simulated wind speed and direction for the network stations 1 and 4 during the core IOP 8 period. For both stations, the simulated wind follows the observed variation of wind regimes, beginning with light and variable winds in the early morning, transitioning to thermally forced circulations during the midday, and abruptly shifting to strong westerlies or northwesterlies in the afternoon. However, they also show that the model is several hours late in producing the morning transition, and about 1 h early and 1 h late, respectively, in the transition to strong westerlies (at station 1) and to northwesterlies (at station 4) in the afternoon. In addition, the strength of the midday thermal circulation at station 4 is underpredicted, the strength of westerlies at station 1 is overpredicted in the late afternoon and evening hours, and the model simulation result does not contain a several-hours-long period of observed easterly winds at station 4, which, as will be shown in the next section, represents the ground evidence of a rotor.
In the same diagram shown are also results of the two sensitivity experiments. Overall, the differences between these runs and the baseline run are relatively small. The more frequent application of radiation leads to an hour earlier establishment of thermal flow pattern at station 4, but delays its appearance at station 1. The lack of Inyo Mountains appears to have similar effects on the establishment of thermal circulation as the radiation, but it leads to stronger overprediction of downslope flow at station 1 by additional 25%. Further discussion of the effects of downstream topography can be found in section 5b.
4. Downstream evolution during core IOP 8
In this section we focus on the downstream evolution during the core IOP 8 period. The discussion is divided into four key periods that illustrate well the evolution and structure of this strong wave and rotor event. The numerical simulation results are used in conjunction with the surface and wind profiler observations to reveal the full three-dimensional structure of the flow and the observed phenomena.
a. Small-amplitude trapped lee waves decoupled from valley circulation
The first period we highlight falls during the morning and midday hours of 25 March, from approximately 1430 through 2200 UTC. Figure 11a shows the objective analysis of horizontal wind, temperature, and pressure for the middle of this period at 2000 UTC 25 March. Pressure has been reduced to a horizontal plane that is tangential to the lowest point on the valley floor using the hydrostatic pressure relation and the cross-valley temperature measurement as a proxy for the vertical temperature profile. This objective analysis reveals a thermally forced valley circulation with southerly upvalley winds in the flat part of the valley, and southeasterly upslope winds on the sloping, western half of the valley (Fig. 10).
While light winds of the thermal circulation pattern were observed at the surface, there was clear observational evidence of wave activity aloft (Fig. 9). During the morning hours of 25 March (from 1300 to 2300 UTC), there is a strong layer of returns between 3 and 4 km AGL with alternating positive and negative vertical velocities in excess of 2 m s−1. Broad spectral widths of more than 2 m s−1 are also present in this layer, indicating strong shear or turbulence, both of which enhance atmospheric reflectivity, making this layer visible to MAPR. At the same time very low values of spectral width are observed below 2 km AGL, indicating near quiescent conditions. At about 3 km MSL, 1 km below the layer of strong wind profiler returns, the radiosonde released at 2100 UTC 25 March from Owens Valley captured a sharp increase of moisture (Fig. 12a). Not only did the relative humidity increased by over 60% in this spike, but the balloon descended by 85 m in the same layer, suggesting that the radiosonde passed through a developing low-level cloud. The 1-km visible satellite imagery near this time shows a single crest wave cloud located over Owens Valley (not shown).
Simulated winds along the lowest COAMPS (10 m) sigma surface at 2100 UTC 25 March (Fig. 13a) show the thermally forced circulation pattern, somewhat weaker than observed. The thermal circulation at the ground does develop in the model simulation in Owens Valley at 2000 UTC, but this flow becomes better developed by 2100 UTC. Figure 14a shows the vertical cross section of model-predicted potential temperature, horizontal wind, and the east–west wind component of the wind along the northern line of stations at 2100 UTC 25 March. A series of trapped, short wavelength lee waves near the crest of the Sierra Nevada is clearly seen in this vertical cross section. These weak waves do not penetrate into Owens Valley, and wave activity is significantly weakened downwind of the Inyo Mountains. The updraft–downdraft pattern associated with these short waves over Owens Valley can be also seen in the model simulated vertical velocity field at 4 km MSL (Fig. 13a).
b. Large-amplitude mountain wave and strong downslope winds
The second highlighted period falls in the afternoon hours of 25 March, from about 2200 UTC 25 March to 0300 UTC 26 March. At 2200 UTC 25 March, the strong westerly winds aloft begin to penetrate into the valley. By 0100 UTC 26 March, both at stations 1 and 4 of the ground network there is an abrupt shift to westerly winds and a gradual increase in wind speed to over 10 m s−1 (Fig. 10). The objective analysis of the surface network observations at this time shows strong southwesterly winds on the western end of the network, and northwesterly winds on the eastern side of the valley leading to an anticyclonic curvature of the flow field that is particularly evident at the north end of the network (Fig. 11b).
The wind profiler observations during this period show strong downdrafts, well in excess of 3 m s−1 between 3 and 4 km AGL and the broad spectral widths observed at these altitudes extending down to the valley floor (Fig. 9). While the spectral widths below 2 km AGL intensify between 0100 and 0300 UTC 26 March, the vertical velocities aloft decrease during this time only to intensify again toward the end of this period. The radiosonde released from Owens Valley at 0000 UTC 26 March experienced a descent of about 1700 m at speeds of up to 6 m s−1 after entering the wave cloud region (Fig. 12b). Given the ascent rate of 4 m s−1, the downdraft may have been as strong as 10 m s−1.
Wave clouds associated with strong vertical motions detected during this period and a dust storm over dry Owens Lake bed associated with strong westerly winds can be seen in a photo from Owens Valley from 2330 UTC 25 March (Fig. 15b). In addition to two stacked lenticular clouds, the photo shows a well-developed roll cloud underneath the lower lenticular cloud. By 0100 UTC 26 March, the visible satellite image shows that the wave clouds extend the whole length of Owens Valley on the lee side of the Sierra (Fig. 15a).
In the model simulation, the southerly flow remains in place through much of the valley at 2300 UTC. The strong westerly flow in the lee of the Sierra during this period appears first in the northern part of the valley, extending gradually to the south. In addition to a wide band of downslope flow on the Sierra lee slopes, westerly winds in the valley appear as gap jets formed by Sawmill and Kearsarge passes, extending from the alluvial slope further east into the flat part of Owens Valley. At this time, the gap jets are still being deflected northward by the southerly flow in the valley (not shown). Figure 13b shows the simulated horizontal winds at 0100 UTC 26 March. The simulated strong winds (∼10–15 m s−1) on the lee slopes of the Sierra Nevada extend down to the leading edge of the area of 8 m s−1 updraft at the foot of the eastern Sierra slopes, at the western end of the area covered by the surface network. These simulated winds are in good agreement with the observed as is the anticyclonic curvature of the flow at the northern end of the surface network area. The model-simulated wind field reveals that this curvature is part of the southern of the two counterrotating vertical vortices associated with the Kearsarge gap jet. The vertical cross section from the model simulation at 0100 UTC 26 March shows a large-amplitude wave whose first crest is located over the center of the valley (Fig. 14b). The horizontal wavelength of this wave is about twice that of the small-amplitude waves in the preceding period and is approximately equal to the ridge-to-ridge width of the valley (∼30 km). Both the large vertical deflection of isentropes and a thin zone of slower zonal winds over the Sierra lee slopes reaching up to 10 km MSL indicate that intense vertical motions are taking place in the immediate lee of the Sierra Nevada. Much smaller vertical motions that were detected by MAPR at this time appear consistent with its location underneath the wave crest.
c. Rotor footprint
The third highlighted period is from 0300 to 0500 UTC 26 March, during which the rotor activity reached the maximum strength. The observed wind at station 1 remained westerly during this time, though it had decreased in speed to below 5 m s−1 (Fig. 7). However, at station 4 the wind shifted to an easterly direction at 0300 UTC, suggesting a presence of the rotor circulation at the valley floor. Easterly winds of about 3 m s−1 persisted for about 2 h at this location. The objective analysis of the mesonet observations shows that easterly winds were observed at several stations in the central part of the valley at this time (Fig. 11c). This change of wind over the ground network appears dynamically consistent with the observed change in the pressure pattern, in which low pressure developed at the western end of the valley, underneath the trough of the deepening wave, while higher pressure developed further east, underneath the wave crest. The surface pressure and wind field at 0300 UTC 26 March suggest that the postwave boundary layer separation, induced by a counter pressure gradient set up by the lee wave, has possibly occurred (Baines 1995; Doyle and Durran 2002). The flow over the surface network area, however, is highly unsteady during this time and exhibits a high degree of spatial variation, including flow reversals and convergences.
Although the wind profiler signal strength weakens aloft during this period and the measurements are less continuous, around 0300 UTC 26 March, the time of the flow reversal at the ground, broad spectral widths measured by MAPR have appeared below 1 km AGL. The transverse, cross-valley flow3 derived from wind profiler data from MISS, located at the northern end of the surface network (cf. Fig. 1c), reaches a minimum of −6 m s−1 at about 1 km AGL shortly after 0400 UTC (Brown et al. 2005). At this time, the negative cross-valley flow, which is present above MISS between 0200 and 0600 UTC, extends to the ground where it was evidenced in the MISS surface wind measurement as a short-lived 3 m s−1 easterly flow during the period of generally westerly winds at this site (not shown).
The simulated flow in the vertical cross section over the northern line of ground stations at 0400 UTC 26 March shows that the mountain wave has grown in amplitude, and attendant strong westerlies are being advected further into the valley along the lee Sierra Nevada slopes (Fig. 14c). Elevated area of negative zonal wind speed is present underneath the mountain wave crest during most of the time during this period. Large values of turbulent kinetic energy are also predicted at the leading and upper edge of the wave. Although there appears to be no sizable area of the reversed flow at the ground underneath the mountain wave crest at this time, surface flow deceleration and reversals are present in this cross section at other times during this period. Figures 16a,b show the incipient stages of the boundary layer separation at this location, in which the appearance of the surface flow reversal is preceded by a strong surface flow deceleration. While over the surface network area, the Kearsarge gap jet, the strongest of Owens Valley gap jets, appears to be disrupting the wave structure and rotor circulation (Fig. 13c), farther to the north the model simulation produces a more sustained reversed, easterly flow underneath the wave crest and convergences at the valley floor (Figs. 16c,d). To the north of the surface network area, the separation also occurs earlier (around 0100 UTC), which is consistent with the southward propagation of the frontal zone along the Sierra crest (cf. Figs. 13b,c). The existence in the model results of the surface and elevated flow reversals underneath the lee-wave crest during the time when the reversed, easterly winds were observed by the surface network and broad spectral widths and cross-valley flow were measured by the wind profiler below and at about 1 km AGL supports the interpretation that these features represent the signature of a turbulent rotor in the model simulation.
d. Vertical wave energy propagation and wave breaking
The final period of strong wave activity recorded during the core IOP 8 occurred from approximately 1000 to 1400 UTC 26 March. The time–height series from MAPR indicates that the strongest vertical velocities of the core IOP 8 period occurred around 1200 UTC 26 March, after frontal passage (Fig. 9). The layer of large vertical velocities between 2 and 4 km AGL is again coincident with an area of broad spectral widths, which in this case most likely represents strong wave-induced turbulence. Strong positive (+10 m s−1) and negative ascent rates (over −3 m s−1) were also experienced between 2 and 4 km AGL by the radiosonde released from the valley floor at 1200 UTC 26 March. Again, assuming an ascent rate of 4 m s−1 this would indicate that the sonde passed through waves with positive and negative vertical motions of about ± 6–7 m s−1 (not shown).
The model simulation results during this time are highly unsteady, showing a transition between wave regimes. At 1200 UTC (cf. Fig. 19e) 26 March, the wave over Owens Valley weakens as the westerly flow enters the valley, forcing a stronger wave response over and to the lee of the Inyo range. This strong westerly jet appears to detach from the ground at the foot of the Sierra lee slope and extend across the valley. An incipient rotor with surface flow deceleration underneath the lee-wave crest is seen over the valley center. At the same time more wave energy is seen to propagate above 7000 m MSL. By 1400 UTC, nearly vertical isentropes directly above the lee slopes from 4 to 9 km MSL indicate that wave breaking is occurring at these altitudes. The intense turbulence and the strongest updrafts and downdrafts measured by MAPR in IOP 8 appear to be related to the still strong, but rapidly changing structure of the wave and rotor over the valley (Fig. 14d). Shortly after this time, upper-level winds had shifted to northwesterly, and terrain channeling led to strong northerly winds in excess of 10 m s−1 at the valley floor. The terrain channeled northerlies, somewhat weaker than observed, were also captured by the model simulation at this time (Fig. 13d).
5. Environmental factors affecting rotor development
a. Ambient upstream conditions
The importance of an upstream temperature inversion at or above the mountain crest level for the lee-wave structure was advanced in recent studies by Vosper (2004), Mobbs et al. (2005), and Hertenstein and Kuettner (2005). In particular, based on idealized 2D numerical simulations, Hertenstein and Kuettner (2005) suggest that positive wind shear across the mountain-top inversion leads to type I or lee-wave rotor events, while the absence of shear leads to type II or a hydraulic-jump rotor. In this section, we examine the role of this inversion for the development of lee waves and rotors during SRP IOP 8.
Figure 17a shows the heights, strengths, and associated wind shears of inversions from the available MGAUS and NAS Lemoore soundings during the extended IOP 8 period. The Lemoore soundings have lower vertical resolution than the MGAUS soundings, leading likely to some of finescale inversions and isothermal layers remaining undetected in the time period between 1800 UTC 25 March and 0600 UTC 26 March. Solid boxes in this figure outline time periods when westerly winds observed at surface network stations 1 and 2, located furthest up the Sierra lee slope, exceeded simultaneously 7 m s−1. We take these periods to represent a proxy for the periods of strongest wave activity over the valley, which is a reasonable assumption given that the leading edge of the large-amplitude wave never reached that far west, staying located between stations 3 and 4, as evidenced in correlation patterns of cross-valley winds at these two lee slope stations with those further into the valley (not shown). Dashed boxes in Fig. 17a outline times in which the modeled wind speeds exceeded 7 m s−1 at stations 2 and 3. The latter two stations were selected for the model diagnostics as westerly winds were over predicted at station 1 in the model simulation.
Based on the available soundings, it appears that during the periods of observed wave activity a weak inversion was located near the crest of the Sierra Nevada between 3000 and 4000 m MSL. Additional stronger inversions or isothermal layers were detected further aloft between 4500 and 6000 m MSL. The majority of these inversions are fairly shallow. Excluding the four thickest inversions for which the depth was in the range of 100–250 m, the average inversion depth is approximately 40 m. While the wind shear across these inversions varies between positive and negative values at successive time periods, the wind shear across the strongest of the inversions between 4500 and 6000 MSL was positive (at 1800 UTC 25 March), which perhaps was the key time period that determined the subsequent development of this wave and rotor system into a lee-wave rotor. Displayed in Fig. 17b is also the time series of the barrier normal component of wind at 3000 and 5000 m MSL, which shows a continuous increase of wind speed with time in the prefrontal environment, in which the maximum was reached just before frontal passage prior to 0600 UTC 26 March. The positive wind shear between these two altitudes also increases with time, reaching the maximum during the strongest wave and rotor activity between 0000 and 0600 UTC 26 March. Overall this sounding summary appears consistent with both Vosper (2004) and Hertenstein and Kuettner (2005) findings.
While the vertical resolution of the model grid is not fine enough to resolve any of the observed inversions (cf. Fig. 5), the larger-scale structure of the ambient upstream conditions is properly represented, including a distinct two-layer stability structure clearly evident in all prefrontal soundings. This leads to the ability of the model simulation to correctly capture various wave and rotor regimes and observed transitions. This is evidenced in the simulated periods of very strong westerlies on the Sierra lee slopes being in close agreement with the observed ones.
The transition from trapped lee-wave regime (with the rotor) to the vertical wave energy propagation should also be well captured in the change of the Scorer parameter profile with time. In an idealized atmosphere with constant wind speed and a two-layer stability structure, Scorer (1949) showed that trapped lee waves will occur if the Scorer parameter, given by
decreases sharply with height. In Eq. (1), U(z) is the barrier normal component of the wind. The Scorer parameter profiles for the two soundings from Figs. 5 and 6 are shown in Fig. 18. The prefrontal profile derived from observations (Fig. 18a) reflects the inversion, two-layer stability structure, and strong positive shear, leading to a sharp decrease of Scorer parameter with height between approximately 3000 and 8000 MSL, which clearly favors trapped lee-wave development. Aside from being smoother, the simulated profile has the same shape (Fig. 18a). The Scorer parameter profiles from the postfrontal period are more uniform with height (Fig. 18b), in particular the one from the model sounding. This clearly is consistent with the transition from wave energy trapping at lower levels to vertical wave propagation obtained by the model simulation.
b. Downstream topography
While the Sierra Nevada is the primary generator of mountain waves over Owens Valley, at only 30 km from the Sierra Nevada, the White–Inyo range is expected also to exert strong influence on the wave response over Owens Valley. The two mountain ranges together form an almost perfect double mountain system. To explore the effect of the downstream mountain range on the lee-wave rotor development in SRP IOP 8, we have carried out the experiment without the terrain of the White–Inyo range. In this experiment, both the White–Inyo range and the valley to the east of it were removed, and replaced by flat terrain of the height of Owens Valley. Effectively, this more than doubles the width of the resulting valley and halves the height of the downstream mountain range. The results from this sensitivity experiment are shown in Figs. 19 and 20.
For the short lee-wave pattern that remained decoupled from the valley circulation, the effect of the Inyo range appears minimal (Fig. 14a versus Fig. 19a). That effect is much stronger for the large-amplitude long wave of 0400 UTC 26 March (Fig. 14c versus Fig. 19b), suggesting that this 30-km long wave represents a resonant wave response to the presence of the Inyo range. In the absence of the downstream range, the horizontal wavelength is clearly shorter, and updrafts over Owens Valley somewhat weaker (Fig. 20). Two additional examples of constructive interference in which the change of the horizontal wavelength is minimal, yet the change of flow over Owens Valley is significant, are shown in Figs. 19c–f. In Figs. 19c,d, the rotor underneath the first wave crest in the Sierra lee is better developed in the presence of the Inyo range, even though the first updraft is stronger in the no-Inyo run. In Figs. 19e,f the constructive interference leads to the wave amplitude increase over Owens Valley and formation of an incipient rotor there that is absent in the no-Inyo run.
The summary of the effects of the downstream mountain range on the lee-wave horizontal wavelength, the strength of the maximum updraft over Owens Valley, and the eastward extent of the westerly wind penetration into the valley along the Sierra Nevada lee slope within the innermost domain of the model is shown in Fig. 20. The horizontal wavelength was determined at approximately the ridge height of the Sierra. The altitude range searched for the maximum updraft is between the ground and 8 km MSL. The maximum updraft is the first updraft in the Sierra lee at all times shown in this diagram. The position of the maximum westerly wind penetration along the Sierra lee slopes is defined as the easternmost position of a continuous band of westerly winds in excess of 5 m s−1 along the lowest model surface between 36.6°N and the northern border of the domain. The values shown in Fig. 20c represent the projection of this position along the valley axis onto the latitude of 36.8°N. It is clear that the downstream mountain range exerts a strong control on the wavelength of the resonant lee wave between 0100 and 0500 UTC 26 March. The effect on the maximum updraft is weaker, but clearly the absence of the Inyo range (or widening of the valley) leads to the increase of the eastward extent of strong downslope winds along the Sierra lee slopes. The larger eastward extent of the downslope flow in the no-Inyo run is consistent with the overprediction of westerly wind at station 1 shown earlier in section 3c.
The spatial structure and temporal evolution of an intense lee-wave rotor event has been examined. This event was documented observationally during 25 March 2004 in IOP 8 of the Sierra Rotors Project over Owens Valley in the lee of the southern Sierra Nevada. The SRP observational data include surface observations by a mesonetwork and 915-MHz wind profilers obtained in Owens Valley, and radiosonde data from Owens Valley downstream as well as San Joaquin Valley upstream of the Sierra Nevada. These data were used in conjunction with the results from the high-resolution COAMPS simulations, run with multiple-nested domains and the finest grid increment of 333 m, to study the structure and evolution of lee waves in which IOP 8 rotors formed and to identify environmental factors that foster rotor formation.
The large-amplitude waves and rotors in this event formed in the pre-cold-frontal environment characterized by strong positive vertical shear of the southwesterly, Sierra-normal winds. Another important characteristic of this prefrontal environment is a two-layer stability structure with the lower, more stable layer separated from the less stable air aloft by a number of inversions, the strongest ones located between 4500 and 6000 m MSL. While waves above the Sierra ridge were present most of the time during the 3-day period of synoptically perturbed weather encompassing this IOP, intense rotors formed only during the first 2 days, the 2 days of IOP 8 (1800 UTC 24 March to 1800 UTC 26 March). Presented in this study is the analysis of the lee-wave rotor that occurred during the second day of IOP 8, during its core period (1200 UTC 25 March to 1200 UTC 26 March), which had the highest density of observations.
The wind profiler measurements, whose utility in documenting episodes of mountain waves, temporal variability of such waves, and their vertical structure has been documented previously (Ralph et al. 1997), were key in identifying periods of strong wave activity above Owens Valley, but also of highly turbulent conditions at low altitudes within the valley. Together with wind profiler measurements, radiosonde data were key in identifying low-level horizontal flow reversals (easterly cross-valley flow during periods of prevailing westerlies) and strong wave up- and downdrafts at higher altitudes. The objective analyses of ground network data yield the flow field and wave-induced pressure perturbations over a large area of the valley floor, and consequently the diagnostics of strong downslope winds and surface wind reversals associated with rotors. The analyses of these observations presented here show that the flow over Owens Valley went through transitions between the number of different regimes during the 24 h of the core IOP 8 period. These include transition from the weak waves above the Sierra ridge decoupled from the thermally forced valley flow during morning and midday hours to the large-amplitude wave with a rotor that formed in the afternoon and reached the maximum strength in the early evening hours. The wave-induced counter pressure gradients led to postwave boundary layer separation and the appearance of the reversed easterly flow at low levels, the hallmark of the rotor. The same diurnal variation of wave regimes with the maximum of rotor activity in the early evening was observed during both days of this IOP. The third transition, which was unique to the core IOP 8 period, was associated with the cold-frontal passage leading to the most intense vertical motions and turbulence observed at the Sierra ridge level during this IOP.
The numerical simulation results were important in reconstructing the full three-dimensional structure and temporal evolution of the lee waves and the rotor. The comparison of the observational analyses with the numerical model results shows that, at the horizontal resolution of 333 m, the COAMPS model reproduces credibly the observed changes in the lee-wave regimes and the formation of the turbulent rotor. The timing of some of the observed flow transitions in the model simulation was, however, found to be slower than observed, including the establishment of daytime thermally forced flow within the valley, and the passage of the upper-level trough and cold front over the Sierra Nevada. Assimilation of radiosonde and wind profiler observations would likely be beneficial in improving this aspect of the simulations. The simulation results show that the IOP 8 rotor formed underneath the crest of a resonant, long lee wave over the center of Owens Valley. The numerical simulation framework has also allowed us to investigate the effect of the downstream topography on the characteristics of the lee-wave response in this case, and to show that the Inyo range exerts a strong control over the wavelength of the resonant lee-wave mode, and thus controls the rotor location, and the degree of penetration and strength of the downslope, westerly winds in the valley.
While the SRP dataset has been invaluable in furthering our understanding of atmospheric rotors, it is a fairly limited dataset for the study of nonstationary lee waves and rotors and flows in Owens Valley that display multitude of spatial and temporal scales and rich three-dimensional structure. The comprehensive T-REX dataset, with multitude of ground-based and airborne, in situ, and remotely sensed measurements, including those by three research aircraft, two Doppler lidars, and an airborne Doppler cloud radar, offers an unprecedented opportunity to achieve further advancements of our understanding of atmospheric rotors and conditions under which they form.
Sierra Rotors was a joint NSF-funded project involving DRI, NRL, University of Washington, and NCAR. We thank all SRP participants for their individual efforts. DRI surface network has been built and is maintained in collaboration with the DRI Western Regional Climate Center. The objective analysis of the Sierra Rotors surface network data was done with the assistance of Ming Xiao, whose help is greatly acknowledged. Stephen Cohn and Bill Brown of NCAR are thanked for providing the wind profiler data and their interpretation for this study. Numerical simulations were run at the Mesoscale Dynamics and Modeling Laboratory (MDML) 68-processor Linux cluster, funded by NSF MRI Grant ATM-0116666. This work was supported in part by NSF Grant ATM-0242886.
Corresponding author address: Vanda Grubišić, Division of Atmospheric Sciences, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512. Email: Vanda.Grubisic@dri.edu
This article included in the Terrain-Induced Rotor Experiment (T-Rex) special collection.
Here, UTC = Pacific standard time (PST) + 8 h.
The 340° azimuth defines the orientation of the Sierra ridge.
The positive axis points eastward, toward the Inyo range.