A radiosonde observation method is presented, consisting of simultaneous radiosonde observations at closely spaced multiple sites using balloons with varied buoyancies. This method was employed during a strong wind event (Suzuka-oroshi) on the lee side of the Suzuka mountain range, Japan, to derive the detailed structure of the wind as it crossed the mountains. Batches of six radiosondes were launched simultaneously from a line of four sites, using balloons with three different degrees of buoyancy. The four sites were 13 km apart along a 35-km-long transect roughly aligned with the prevailing wind. The observations documented two flow regimes: a downslope flow perpendicular to the mountain range, similar to a windstorm, and an unexpectedly strong low-level jet flowing parallel to the mountain range. The method was more successful at delineating the first regime than the second. The first regime was well simulated by a numerical experiment, but the second regime was not. The vertical wind associated with the downslope windstorm was inferred from the changing slopes of potential temperature isentropes. Comparison of the balloon ascent rates with these isentropes meanders and the simulated vertical wind showed that fluctuations in balloon ascent rate provide reliable information on the vertical direction of the wind. An analysis of the second regime using a long-term meteorological dataset shows that the onset of the low-level jet is related to the synoptic-scale shift in vorticity from positive to negative in the observation area. This vorticity shift appears to be a useful indicator for the low-level jet regime.
Radiosondes are one of the most accurate tools for observing the vertical structure of the atmosphere. For example, a single radiosonde sounding, recording a single atmospheric profile along its trajectory, is able to provide the propagation direction of a gravity wave (e.g., Vincent et al. 1997). However, a single radiosonde sounding cannot provide the data needed to map two-dimensional airflows over a mountain range, such as the structure of a downslope wind. To construct such a map, the ideal observation program would be to launch many radiosondes simultaneously from a line of stations parallel to the prevailing wind that extends across a mountain range, including its windward and leeward sides. This scheme could provide a detailed profile of a narrow, mountain-induced strong wind in a vertical plane. In an attempt to realize this scheme, we launched six radiosondes simultaneously from a line of four closely spaced (~13 km apart) sites that was aligned with the wind direction, using a range of buoyancies at some sites to better sample the vertical space. The purpose of this paper is to document this observational method and its results, and to apply the results in an investigation of a local wind, the Suzuka-oroshi, by observational, numerical, and statistical approaches.
Previous observational studies of strong winds over a mountain range, such as downslope winds, have employed multiple direct observations by radiosondes, dropsondes, or airplanes in traverses across a mountain range. A famous study of this type, which captured the flow over a mountain (Lilly and Zipser 1972), combined an airplane traverse with radiosonde launches from a single site on the lee side of the Front Range of the Rocky Mountains. The resulting vertical–horizontal schematic map of a downslope windstorm is still recognized as exemplary. Later projects involving observations of orographic flow produced such maps on the basis of observation campaigns that included an airplane. For example, Jiang and Doyle (2004) used global positioning system (GPS) dropsonde and flight-level data to document the hydraulic jump associated with wave breaking in the eastern Alps during the Mesoscale Alpine Program. Doyle et al. (2005) similarly observed large-amplitude wave breaking in Greenland during the Fronts and Atlantic Storm Track Experiment, creating a three-dimensional model of the flow structure. The Terrain-Induced Rotor Experiment combined in situ and airborne measurements to capture the vertical atmospheric structure in the Owens Valley of California, on the lee side of the Sierra Nevada (Grubišić et al. 2008; Wang et al. 2009). However, the costs of aircraft, in particular, make such projects difficult to carry out. Although the combination of GPS dropsondes with airplanes is effective for delineating vertical–horizontal flow, the observations cost much more than those made by GPS radiosondes. We, therefore, designed a low-cost observation scheme based solely on GPS radiosondes.
Researchers have devised methods of manipulating balloon trajectories to investigate air flows across mountain ranges. Booker and Cooper (1965) pioneered the use of constant-volume superpressured balloons to investigate trajectories related to mountain waves. Vergeiner and Lilly (1970) used this method to capture part of the structure of a lee wave in the Front Range of the Colorado Rockies. They concluded, however, that constant-volume balloons were inferior to instrumented aircraft. Glenn Shutts introduced multiple sounding techniques that relied on control of balloon trajectories. During a field experiment in the Welsh Mountains in October 1989, three radiosondes were launched at 10-min intervals from a single station (Shutts et al. 1994). The buoyancies of the three radiosondes were varied by inflating them with different amounts of helium. Shutts (1992) examined fluctuations in the ascent rates of the three radiosondes and compared them with ascent rates calculated with a nonhydrostatic mesoscale model to demonstrate the feasibility of using them to map lee-wave structures. Shutts and Broad (1993) employed five radiosondes with different buoyancies to observe waves over the Lake District in northern England and constructed a map of the vertical velocity field from radiosonde and airplane observations. All of these studies used multiple sounding balloons launched from a single site. Avoiding the use of airplanes would require many radiosondes to be launched simultaneously from a line of stations across a mountain range parallel to the prevailing wind. To attempt such a program, we designed a method that can be described briefly as simultaneous multiple-site radiosonde observations with varied buoyancies. To provide wider horizontal coverage in the troposphere than done by Shutts et al. (1994), the method combines multiple soundings at a single station and multipoint sounding stations along a transect across a mountain range. Modern radiosondes with GPS receivers provide more accurate location data than those used by previous studies.
We conducted a successful pilot experiment using this method on 21 March 2010 in Japan, observing the local strong mountain wind called the Suzuka-oroshi (Yoshino 1975). The area of our observational study, the Suzuka Mountains, is famous in Japan for the occurrence of strong local winds in winter. These are locally called Suzuka-oroshi, oroshi being a term for a strong, cold downslope wind. An oroshi is, therefore, similar to the Bora wind of eastern Europe (e.g., Smith 1987; Grisogono and Belušić 2009). Owada and Harada (1978) suggested from a study of the orientation of wind-sculpted trees that the Suzuka-oroshi is a westerly lee wave over the Suzuka Mountains. Owada (1994) described the general features of the Suzuka-oroshi and reported that six different synoptic-scale weather patterns are associated with it. However, the mechanism, process, and vertical structure of the Suzuka-oroshi remain poorly known. For example, it is not clear whether the Suzuka-oroshi is similar to other typical downslope wind storms.
The first objective of this paper is to explain our radiosonde observation strategy. The second objective is to show its results in a detailed two-dimensional profile of a strong wind crossing a mountain range and to compare the results to those of an analogous numerical experiment. We show that simultaneous multiple-site radiosonde soundings can directly capture a downslope windstorm. We also show that the ascent rate of a radiosonde balloon can provide reliable estimates of the vertical wind direction through a comparison with estimates based on the changing slopes of observed potential temperature isentropes and estimates based on numerical experiments. In addition, our observations captured an interesting strong wind regime associated with a low-level jet, which was not anticipated. This paper describes the two-dimensional structure of this regime. The third objective is to discuss the synoptic conditions associated with the onset of the low-level jet regime. We describe our statistical investigation of climatological conditions associated with the low-level jet and compare these conditions with the synoptic conditions during our observations and with the results of a numerical simulation.
Section 2 of this paper describes the field experiment, our method, and the resulting data. Section 3 shows examples of the flow structures captured by our method and presents a comparison with numerical experiments to clarify the observational results. Section 4 is a discussion of the results as informed by an additional statistical analysis. Section 5 presents our conclusions.
2. Observation method and targeted phenomena
a. Simultaneous multiple-site radiosonde observations with varied buoyancies
Our experimental method (Fig. 1) expands upon the multiple sounding method employed by Shutts et al. (1994), in which balloons with different buoyancies were launched at 10-min intervals from a single site. At seven separate times during the same day, we launched six GPS radiosondes simultaneously at four sites: the Iga Research Institute of Mie University (IGA), the Aoyama-kogen Wind Farm (AOYAMA), the Experimental Farm of Mie University (FARM), and Mie University (UNIV) (Fig. 2). Ideally, the location and spacing of launching sites are determined by the horizontal scale or the wavelength of targeted phenomena. However, in the Suzuka Mountains the choice of launching sites is restricted by various requirements such as sufficient space for preparing and launching balloons, electrical power sources for radiosonde receivers, and authorization to conduct observations. It was impossible to set up a strictly linear array across the mountain range during our experiment. However, the four selected launching sites formed a roughly linear west–east transect about 35 km long with Mt. Kasatori in the center. During westerly or northwesterly synoptic winds, sites UNIV and FARM are on the leeward side of the mountains, IGA is on the windward side, and AOYAMA is at the crest of Mt. Kasatori. The distance between sites along the prevailing wind direction was extremely short, about 13 km. The transect was roughly parallel to the wintertime prevailing winds in this region. The topography around sites FARM and UNIV, on the leeward side, is nearly flat and thus local effects on the wind in the lower boundary layer were negligible.
In addition to the simultaneous launches at four sites, the amount of helium in the balloons was varied to control the balloon trajectories and thus provide better coverage in the vertical plane of the transect. During our observations, we deployed three sizes of balloons: large, normal, and small. Their rates of ascent were fast, normal, and slow, respectively. We ensured that the buoyancy of each type of balloon was consistent at all launch sites by using a weight measurement scale. First we injected helium into the balloons, with a radiosonde and parachute attached, inside a shelter until they reached neutral buoyancy. The balloons were then inflated further to a buoyancy of 1.0–1.2 kilogram-force (kgf; 1 kgf = 9.8 N) for the fast balloons, 0.6–0.7 kgf for the normal balloons, and 0.5 kgf for the slow balloons. These settings were chosen empirically. Because it is difficult to measure accurately the weight lifted by a balloon, buoyancies could not be kept strictly constant for each size of balloon, but we were able to roughly control their rates of ascent. When launched simultaneously, the slower balloons were more strongly influenced by strong winds in the lower troposphere and moved farther downwind. The average ascent rates in the troposphere (<10 km) were 7.58 m s−1 (fast), 5.59 m s−1 (normal), and 4.78 m s−1 (slow) during our experiment.
Observations were made on 21 March 2010. The launch times and types of balloons are summarized in Table 1. Human errors led to some failures in data transmission after the balloon releases.
b. Construction of transect profiles
Because the balloon trajectories during each set of observations varied temporally and spatially depending on the wind direction, we projected the dataset from each radiosonde release onto a vertical plane parallel to the prevailing wind (Fig. 3). We determined the orientation of this plane subjectively by examining all three-dimensional trajectories and identifying the plane that was in the middle of the trajectories. Figure 4 shows an example of this projection based on the trajectories shown in Fig. 3a. Each balloon’s sounding line was determined from its position information (longitudes, latitudes, and elevations). We projected each balloon location (longitude and latitude) onto the chosen plane (Fig. 4a), then arranged the projected sounding lines into a 2D longitude–altitude section (Fig. 4b). Next, we constructed transect profiles interpolated by the triangulation method of Watson (1982) using Generic Mapping Tools (Wessel et al. 2013). The degree of error in a vector variable (wind) depends on the orientation of the chosen plane, which affects the transverse and parallel components of the wind with respect to the plane. The scalar variables of temperature, relative humidity, and wind direction are unaffected by the choice of profile plane. We verified that the results were not sensitive to small changes in the orientation of the plane. Lacking other instrumental data to compare with the sounding data, we had to assume that atmospheric conditions did not change during the period of roughly 1 h during which the balloons were ascending. The resulting profiles provided snapshots of the two-dimensional vertical and horizontal structure of meteorological variables defined by the several balloon trajectories in each radiosonde release (shaded area in Fig. 1).
c. Sounding data
We used radiosondes (RS-06G, Meisei Electric Co., Ltd., Japan) that record temperature, relative humidity, wind, and air pressure. The GPS receiver mounted on each radiosonde measured highly accurate longitudes, latitudes, and elevations. According to the manufacturer’s specifications, the accuracies of GPS height and pressure are ±5 m and ±1 hPa, respectively. Data were recorded at intervals of 1 s. To make the vertical resolution of the radiosonde data uniform, we averaged the data vertically over intervals of 20 m. In addition, we smoothed the data by the weighted running mean over 100-m intervals to remove high-frequency perturbations.
The structure of atmospheric vertical motion provides important information such as the structure of gravity waves and the location of hydraulic jumps. Although radiosonde trajectories do not indicate atmospheric vertical motions directly, we calculated vertical motions by the following procedure. First, we calculated the fluctuations in the balloon ascent rate, from which the atmospheric vertical motion can be deduced with considerable accuracy (e.g., Corby 1957; Reid 1972; Reeder et al. 1999). We defined the simplified ascent rate (m s−1) as zi+1 − zi, where zi+1 and zi are the (i + 1)th and ith height measurements (m), respectively. Because the height was also recorded at 1-s intervals, the ascent rate data yield the vertical displacement per second of the balloon. Then we calculated the ascent rate averaged over 20-m intervals and smoothed it by the weighted running mean over 100-m intervals. Second, we calculated the anomaly of the ascent rate relative to the average ascent rate in the troposphere (altitude <10 km) for each launch. Figure 5 shows examples of the smoothed and average ascent rates at two observation times for IGA and FARM. Note that we assumed here that the anomalies in the ascent rates represent the relative upward or downward motion compared with the mean ascent rate, rather than actual vertical motions. A method previously used to evaluate vertical velocity from sounding data is that of Wang et al. (2009), who estimated vertical velocities by subtracting the balloon rise–fall rate in still air from the actual rise–fall rate. That method is better for estimating actual vertical velocity, because the rise–fall rate in still air includes the effects of changing air density with height. However, the rise–fall rate in still air is subject to uncertainty (e.g., in the drag coefficient and the balloon volume), and the method used by Wang et al. (2009) may not have been appropriate for our campaign because we had no parameters for calculating the rise–fall rate in still air. In this study we sought to deduce vertical motions by two independent ways: by the anomaly in the balloon ascent rate and by isentropic analysis. The structure of potential temperature in a vertical plane can be used to diagnose the trajectory of a particular air mass because the potential temperature within the air mass is conserved if diabatic heating is sufficiently small and the flow is steady. Because the air mass travels parallel to an isentropic surface if the flow is both steady and adiabatic, the ascent or descent of an isentropic surface provides the vertical displacement of an air parcel.
a. Synoptic-scale atmospheric conditions and surface winds
Figure 6 displays the pattern of sea level pressure on 21 March 2010, derived from the Japanese 25-year Reanalysis Project (JRA-25)/JMA Climate Data Assimilation System (JCDAS) (Onogi et al. 2007). On that day, an intense cyclone was on the east side of Japan in a configuration similar to the common winter sea level pressure pattern around Japan known as “west high and east low,” associated with winter monsoon winds. The probability of a Suzuka-oroshi event was high because the synoptic conditions resembled those described by Owada (1994).
Besides the radiosonde sounding, we used observations of surface winds at Tsu and Mt. Kasatori to evaluate the occurrence of the Suzuka-oroshi (Fig. 2). The city of Tsu, where there is a meteorological station of the JMA, is commonly affected by the Suzuka-oroshi (Owada and Harada 1978). At the peak of Mt. Kasatori, an instrumented tower records wind speed and direction at 25, 45, 60, and 75 m above the ground. The wind speed at the Tsu station and Mt. Kasatori strengthened suddenly between 0400 and 0500 Japan standard time (JST, UTC + 9 h) (Fig. 7). The wind at Tsu was generally westerly from 0400 to 1200 JST and northwesterly from 1200 to 2400 JST. The maximum wind speed at Tsu, 13.0 m s−1, was observed at 1650 JST during the latter period. Conversely, at Mt. Kasatori the maximum wind speed, with the largest vertical wind shear, was observed between 0400 and 0500 JST, after which the wind gradually abated until 1500 JST, then intensified again.
b. Flow regimes
We constructed longitudinal profiles using the data from each of our seven radiosonde releases during the Suzuka-oroshi event of 21 March 2010. The profiles showed two distinct flow regimes during the course of the day, the first resembling that of a downslope windstorm and the second resembling that of a low-level jet. Figure 8 displays vertical profiles of wind speed and wind direction for each site in the two different flow regimes. At elevations below 2000 m in the first regime (Fig. 8a), wind speeds at the mountain top (AOYAMA) and in its lee (FARM, UNIV) were higher than on the upwind side of the mountain (IGA). This difference was greatest (17 m s−1) at 1100-m elevation, where wind gusts at AOYAMA reached 36 m s−1. The average wind direction was westerly (Fig. 8b). In the second regime, wind speeds were nearly the same at all four stations below 3000 m, except at UNIV below 500 m, where the maximum speed reached about 19 m s−1 and the wind direction was northwesterly (Figs. 8c,d).
1) Downslope windstorm
The profile at 0530 JST depicts a structure similar to that of a downslope windstorm (Fig. 9). During this radiosonde sounding, the surface wind was strong at both stations, and the instrumented tower at Mt. Kasatori recorded a large vertical wind shear between 25 and 75 m (Fig. 7). The prevailing wind was southwesterly, so the vertical structure was profiled from southwest to northeast (Fig. 3a). Figure 9a shows the potential temperature and the anomalies in ascent rate along this profile plane. Some of the balloon trajectories crossed, but on the whole our launching strategy succeeded in covering an expanded area of the atmospheric field in detail.
Figure 9a shows that the isentropic surfaces inferred from our data formed a wave pattern in the troposphere. It is noteworthy that the ascent rate anomaly along the sounding line, interpreted as the relative vertical velocity, agrees with the flow structure expected from the curvature of isentropes. For example, a strong negative ascent rate anomaly was inferred above the top of Mt. Kasatori that extended up to the 294-K isentropic surface. The fact that isentropes from 292 to 300 K were consistently concave downward over the top of the mountains is consistent with the negative ascent rate anomaly. This correspondence of ascent rate anomalies and isentropic slopes was also apparent in other areas. The strong positive ascent rate anomalies along the FARM-Fast balloon trajectory from 292 to 300 K are consistent with the upward slope of the isentropes on the leeward side of the mountains, which signify the presence of updrafts there. At higher altitudes, for example along the 312-K contours, this same consistency was apparent. The fact that the slopes of the isentropes, both at lower and upper altitudes were in accord with the positive and negative signs of the ascent rate anomaly, signifies the presence of updrafts and downdrafts associated with a gravity wave. This agreement indicates that our method for estimating atmospheric vertical velocity from balloon ascent rate is reliable. Quantitative estimates of radiosonde-derived vertical velocity are described in section 3c and compared with a numerical simulation. It should be noted that the wave potential temperature contours tended to have large amplitudes at levels where vertical stratification was strong.
Figure 9b shows details of the 0530 JST profile in the lower troposphere, including the derived wind speed anomaly. Positive wind speed anomalies reached two peaks under 2000 m, one where the isentropes descended in the lee of the mountain and the other where the isentropes regained their elevation, around the FARM station. This strong wind corresponded with the downward wind leeward of the mountain top, where the isentrope descended sharply and the ascent of the balloon was strongly inhibited.
Figure 9c shows the wind direction along the profile and the transverse component of the wind, defined as the degree to which it blew to the right of the plane of the profile as seen facing leeward. The troposphere had a two-layered wind flow in which the wind was west-southwesterly in the upper layer and westerly in the lower layer (thus having a large transverse component). The resulting shear at the boundary between the two layers, at about 2000 m, caused the boundary to form wave patterns.
Figure 9d displays profiles of relative humidity and equivalent potential temperature. High humidities showed that a cloud layer was present around 2000–3500 m, where the potential temperature was not conserved along the path of an air parcel. The equivalent potential temperature, which is conserved under moist adiabatic processes, showed better agreement with fluctuations of the ascent rate anomaly than the potential temperature. Especially in the unsaturated layer below the cloud base (<2500 m), both the equivalent potential temperature and the potential temperature had almost the same isentropic structure. Both the isentropes descended in the lee of the mountain and regained altitude farther downwind; thus, the isentropes of the potential temperature (Fig. 9b) and equivalent temperature (Fig. 9d) were interchangeable below 2500 m. It is reasonable to assume that the air parcel traveled along the isentrope, at least below the cloud base.
The lower layer corresponded to the layer with positive wind speed anomaly (Fig. 9b) where the westerly wind approached the mountain perpendicularly and generated strong winds on the downstream side. The rise–fall pattern of the isentrope associated with a strong wind in the lower layer resembles the pattern of a hydraulic jump. The large transverse component of the lower wind (Fig. 9c) raises the possibility that the structure is distorted by its projection onto the profile plane. However, the westerly wind was closely aligned with the observation sites (Fig. 3a), and the results shown in Fig. 9 were not sensitive to differences in the chosen azimuth of the profile plane.
Figure 9 implies that our observations qualitatively captured the downslope windstorm. Atmospheric conditions on the windward side of a range are known to control the flow regimes across a mountain barrier (e.g., Vosper 2004; Reinecke and Durran 2008). The vertical profiles of zonal wind, wind direction, relative humidity, and potential temperature observed at the westernmost (upwind) end of our transect are shown in Fig. 10. The profiles at 0530 JST had two distinguishing features compared to those at other times. First, a strong zonal wind layer, corresponding to westerly winds, was observed at 1500–2500-m height at 0530 JST (Figs. 10a and 10b). These winds were perpendicular to the mountains, were associated with very dry air, and were located below the cloud base (Fig. 10c). Whereas the other potential temperature profiles showed well-mixed air overlain by an intense inversion layer (Fig. 10d), the 0530 JST profile displayed warmer conditions and stable stratification above 1500 m. The level of the inversion layer corresponded to the level of incoming dry air.
Using these upwind soundings, we evaluated the conditions associated with the occurrence of downslope winds in terms of the Froude number (Fr), the ratio of the flow velocity divided by the phase speed of a gravity wave. In terms of shallow-water wave theory, the occurrence of a downslope windstorm is associated with a transition from subcritical (Fr < 1) to supercritical (Fr > 1) flow (Durran and Klemp 1987; Durran 1990). We assumed a simple two-layer flow, with the layers divided by the change in gradient of the isentrope. For the calculation of Fr,
where the parameter U is the averaged zonal wind component in H, the thickness of the fluid from the surface to the altitude associated with the inversion layer; is the mean potential temperature in the fluid and is the difference in potential temperature between the two layers; and is the acceleration of gravity. The term is the reduced gravity, from which the phase speed is calculated. Table 2 shows the values of Fr and each relevant parameter during the observation period. The value of Fr at 0530 JST was nearly 1.0, meaning that the state of flow over the mountain was nearly critical and suggesting that atmospheric conditions were likely to induce a downslope wind. At all other times, however, Fr was less than 0.5 because the phase speed was greater than at 0530 JST; thus, the upstream flow was in a subcritical state.
2) Low-level jet
At 1430 JST, the Tsu station observed strong surface winds whereas the wind at Mt. Kasatori was near its weakest point during the windstorm (Fig. 7). The profile at that time (Fig. 11a; orientation shown in Fig. 3b) consisted of two layers separated by a stratified layer at about 3000–4000 m featuring a steep gradient in potential temperature and wavy isentropic lines. In the lower layer, below 3000 m, there was little atmospheric stability. The upper layer had a stratified wavy structure and was located above the cloud top. The equivalent potential temperature in the upper layer had almost the same structure as the potential temperature, although with weaker amplitude (Fig. 11d). The balloons showed little fluctuation in their ascent rates and had fairly linear trajectories, unlike the large fluctuations seen in the 0530 JST profile. A positive wind anomaly at low elevations on the leeward side (Fig. 11b) corresponded to the strong low-level winds observed at UNIV (Fig. 8c). In addition, a cold air mass was observed only over the leeward side where the low-level air was well mixed (Fig. 11b). The wind direction data were also peculiar (Fig. 11c): whereas westerly winds dominated the flow over the mountains, a strong northwesterly wind was recorded at low levels on the downwind side that was almost perpendicular to the profile plane.
Figure 12 shows time series of wind speed, wind direction, potential temperature, and meridional wind observed at UNIV on the leeward side and IGA on the windward side. The vertical structures of wind direction at both sites rotated clockwise with time similar to the surface winds shown in Fig. 7a, indicating that synoptic-scale conditions associated with a dominant wind gradually changed with time. A northwesterly wind was observed in the lower boundary layer at both sites, but it appeared earlier on the leeward side (UNIV) than on the windward side (IGA) (Figs. 12a and 12c). The meridional wind component was northwesterly at UNIV forming a jetlike structure associated with a large wind shear below 3000 m, where the lower and upper layers had strongly differing wind directions (Fig. 12b). Because the topography is nearly flat around UNIV, this northwesterly wind was not of local origin. The potential temperature at both sites gradually decreased during the day, but the cold air mass associated with the northwesterly was also observed at UNIV earlier than at IGA (Figs. 12b and 12d). It should be noted that the potential temperature in the lower layer was lower at 1430 JST, in daytime, than it was around dawn at 0530 JST. Therefore the near–mixed layer was not the result of the diurnal temperature cycle. These results imply that the northwesterly wind observed at UNIV was a low-level jet that did not originate on the windward side (IGA) of the mountain range. Note that the strong wind accompanying the low-level jet occurred when the Froude number was small, which suggests that the Froude number alone is not a strong indicator of whether a strong wind event will occur on the lee side of the mountain.
c. Numerical simulation of two flow regimes
1) Model configuration
We conducted a numerical experiment by using the Advanced Research version of the Weather Research and Forecasting Model, version 3.5.1 (Skamarock et al. 2008), to simulate the two observed flow regimes (Figs. 9 and 11) for comparison with the profile plains constructed from radiosonde observations, to confirm the basic feasibility of the radiosonde technique and to assess the reliability of the observations. Four nested domains (Fig. 13) were used with horizontal grid sizes of 27, 9, 3, and 1 km. The innermost of these domains covers the windward and leeward side of the Suzuka Mountains. We used a two-way nesting simulation, in which the results of a child domain were fed back to a coarser-resolution domain. The time step of each domain was 162, 54, 18, and 6 s, respectively. The model top was at 50 hPa (approximately 20 km), and the model contained 50 vertical layers with thicknesses that increased with altitude from approximately 39 m at the bottom to 490 m in the middle troposphere. The initial and lateral boundary conditions for the simulation were taken from the National Centers for Environmental Prediction’s Global Forecast System Final Operational Global Analysis (FNL). For topography, we used the U.S. Geological Survey’s Global 30 arc s elevation dataset (GTOPO30), in which the horizontal scale is approximately 1 km. The physical schemes chosen for the simulation variables are summarized in Table 3. The simulation started at 2100 JST 20 March 2010, which was 8 h before the first radiosonde observations during the downslope windstorm. The period of the simulation was 30 h, covering the entire day of radiosonde observations. Here we focus on the innermost domain at the same times that the downslope windstorm (0530 JST) and low-level jet (1430 JST) were observed.
2) Comparison with observations
To compare the observations (Figs. 9 and 11) with the simulation, we adopted the same projection plane as was used for the observations (Fig. 3). Figure 14 presents the simulation results that correspond to the observations shown in Figs. 9b,c and 11b,c. The simulation data in Fig. 14 were picked along the sounding lines. The vertical wind anomaly shown in Fig. 14 was calculated as the difference from the average simulated vertical wind profile from the surface to 10-km height, analogous to the calculation of the balloon ascent rate anomaly used for the radiosonde observations.
In the case of the downslope windstorm (Figs. 14a,b), the numerical model simulated the observations well. The isentrope descended sharply in the lee of the mountain and then regained altitude where the strong wind was accompanied by meandering of the isentrope. The vertical velocity anomaly was in accord with the shape of the horizontal isentropes and also showed good agreement with the observed ascent rate anomaly (cf. Figs. 14a and 9b). The westerly wind in the lower troposphere was also well simulated in the numerical model (cf. Figs. 14b and 9c).
To confirm that the projected section represented the actual vertical–horizontal structure, we rendered the simulated variables in altitude–longitude sections onto the two different planes shown in Fig. 15: one matching the line used for the observations and the other running due east–west. The results are shown in Fig. 16. Compared to the observations (Figs. 8a and 8b), the simulated wind speeds along the sounding lines from stations AOYAMA and FARM were generally overestimated at every level, and along the sounding line from station IGA a constant westerly wind layer below 2000 m was simulated blowing toward the mountain, in contrast to the observations (Figs. 16a and 16c). Analogous to the observations (Fig. 9b), the wind speed anomaly appeared on the downstream side of the simulated profile, and the isentropic surface also corresponded to the regime of downslope windstorm (Figs. 16b and 16d). Overall, the simulated and observed wind anomaly structures closely matched each other, but the simulated windstorm was more intense than the actual windstorm. The similarity of these structures on the two different planes indicates that the observational results were substantially independent of the profile orientation.
The vertical wind speed can be deduced from the slope of the observed isentropic curve. The isentropic vertical wind , defined as the vertical component of the wind blowing along an isentropic surface, is estimated by the following equations:
The slope angle of the isentrope, , is calculated by the difference in height of a potential temperature between two neighboring sounding lines. For example, between IGA Fast and IGA Normal is estimated by Eq. (2), where the height of a specific potential temperature is at IGA Fast (on the windward side) and at IGA Normal (on the leeward side), and is the horizontal distance of an isentrope between soundings at the ith (windward) and (i + 1)th (leeward) stations. In Eq. (3), is the component of the wind blowing along an isentropic surface with slope , and is the parallel wind component along the plane of the profile. Note that along the plane including UNIV was not estimated because there was no sounding leeward from UNIV.
Figure 17 displays the simulated and observed ascent rate anomalies, along with the vertical wind anomaly estimated with Eq. (4), at stations along the sounding lines. The observed ascent rate and the isentropic vertical wind were smoothed by the running mean over 500-m intervals, and the anomalies were calculated with respect to the vertical average below 10 km for a sounding line. At 0530 JST (Fig. 17a), during the downslope windstorm period, the ascent rate anomalies of the balloons were in agreement with those of the isentropic vertical wind and the numerical simulation at stations on the windward side (IGA and AOYAMA), especially for IGA, where an intense downward wind at 3500–4000 m was seen in all estimates. At FARM, the isentropic vertical wind showed smaller fluctuations than the other measures. Figure 17b shows the correlation coefficient between the ascent rate and isentropic vertical wind anomalies for all the stations, calculated from the bottom to each level. The correlation coefficient exceeded 0.4 below 4000 m and decreased with greater height. It is noteworthy that the isentropic vertical wind calculated by the potential temperature was correlated well with the balloon ascent rate even in the saturated cloud top layer at about 3500 m (Fig. 9d). The simulated, observed, and isentropic estimated vertical wind anomalies were overall in good agreement over the stations, the indication being that the balloon ascent rate measured using our methodology was a reliable indicator of vertical wind speed. To evaluate the reliability of this indicator, we estimated correlation coefficients for each station and for the times of all eight radiosonde launches (Table 4). These correlation coefficients were positive at 0530 JST during the downslope windstorm period, averaging 0.47, and the correlations were strong for IGA (0.74) and AOYAMA (0.8). During the other observation periods, however, the overall correlation between the ascent rate anomaly and the simulated vertical wind was weak except at 0830 JST (Table 4).
Our results show that for the downslope windstorm regime, the technique of multiple sounding with varied buoyancies (Shutts et al. 1994) can be used to better cover the space between multiple observation sites (Fig. 9a). The differing trajectories of fast and normal balloons launched at IGA produced more data in the region between IGA and AOYAMA, and the slow balloon launched at UNIV on the leeward end of the profile expanded the area of observations to the east. This strategy, thus, offers advantages when the possible observation sites for balloon launches are restricted, as is often the case in Japan. It offers a way to achieve higher resolution and cover wider ranges. The 13-km horizontal distance between our observation sites was appropriate to capture the structure of the downslope windstorm. The technique is also significant in enabling us to discern vertical and horizontal air motions on the basis of both isentropes and ascent rate anomalies of radiosonde balloons. It is surprising that the vertical wind field deduced from balloon ascent rates shows good agreement with vertical winds inferred independently from the horizontal shape of isentropes (Figs. 9a and 17a). The numerical experiment supported these results.
The low-level jet regime was not fully replicated by our numerical simulation. The simulated wind speed anomaly and the transversal component show that the wind speed was higher at FARM and UNIV than at IGA, matching the observational results (cf. Figs. 14c,d and Figs. 18 and 11b,c). However, the simulated potential temperatures had a pattern representing a stably stratified layer between the windward and leeward side below 3000 m, in contrast to the mixed layer that was observed. The simulated wind direction on the windward side of the mountains was north-northwesterly, whereas the observations at IGA showed a westerly wind. These disagreements represent a failure of our numerical simulation to replicate the atmospheric conditions during the low-level jet. Because our model configuration was inadequate to resolve the low-level jet, we lack enough evidence to clarify the cause of the jet. The vertical winds estimated from the ascent rate and the slope of the isentrope also disagreed with each other during the low-level regime (Fig. 17c); the correlations were negative (Fig. 17d). This disagreement may reflect the fact that the variation in vertical motion was small, unlike during the period of the downslope wind, such that the balloons were only weakly influenced by vertical wind during their ascent. Johansson and Bergström (2005) have shown that balloon ascent rates tend to be higher within the boundary layer than above it, owing to a decrease in the drag on the balloon due to turbulence. For example, the ascent rate anomaly at IGA Normal (Fig. 17c) was overestimated below 4000 m, a level that corresponded to the top of the constant potential temperature layer (Fig. 10d). This layer was well mixed; thus, the ascent rate anomaly could not be correlated with vertical air motions. Because the slope of the isentrope was also weak at this time, the vertical wind derived from the potential temperature was not reliable.
It appears that two methods of estimating the vertical wind, by the balloon ascent rate and by the slope of the isentrope, worked better during conditions such as the downslope windstorm, when upward or downward motion was vigorous. Although the measurement of vertical winds is one of the most difficult problems in atmospheric science, our method, despite its simplicity, offers a way forward, subject to confirmation by future studies.
Our findings revealed two limitations of our method. The first is that to produce the best results, our method depends on the strength of winds. In our launch at 1430 JST, for example, the two balloons launched from FARM followed almost the same trajectory below 3000 m where the winds were weak (Fig. 8c), although they separated in the much stronger winds at higher elevations (Fig. 11a). There is an optimum range of wind speeds for sampling the vertical–horizontal profile, given that excessively strong winds will carry the balloons downwind too fast to sample the full vertical profile over the mountain. The second weakness is that the position of launch sites restricts the direction of longitudinal profiles, limiting our ability to trace the low-level jet outside the profile plane. A multiple-sounding program should be designed carefully with respect to the orientation of target phenomena.
4. Discussion with statistical analyses
Our observational strategy succeeded in making snapshotlike depictions of the detailed structure of a Suzuka-oroshi event. It is, to our knowledge, the first time that the wavy structure of the potential temperature field over a mountain range at a horizontal scale less than 100 km has been mapped by using only radiosonde soundings. Moreover, our observations revealed two flow regimes associated with the Suzuka-oroshi: a downslope windstorm and a low-level jet. The first regime was characterized by a westerly wind, and the upstream flow represented a critical condition as deduced from the Froude number. The second regime, a jetlike structure flowing parallel to the mountain range, was characterized by a cold northwesterly wind that appeared first on the downwind side of our profile. Owada and Harada (1978) suggested that the Suzuka-oroshi is a northwesterly wind originating as a lee wave over the Suzuka Mountains. The low-level jet regime we documented is consistent with their hypothesis. The occurrence of the downslope windstorm regime is reasonable because the Froude number was near the critical level. However, a strong wind event accompanied the low-level jet despite a small value of the Froude number. Thus, the Froude number alone is not a strong indicator that a strong wind event will occur on the lee side of the mountain.
In an attempt to uncover hidden factors in the occurrence of the low-level jet regime, we performed a statistical data analysis comparing the observational and numerical results for periods of strong northwesterly winds. We began with the record of strong winds in the study area in the winter months (December–March) between 1987 and 2011 (Fig. 19a), defining windy days as days when the hourly mean wind speed exceeded 10 m s−1 at Tsu or at one of the nearby JMA weather stations Ueno, Kameyama, and Yokkaichi (UN, KM, and YK, respectively, in Fig. 2). This definition is the same one used by Owada (1994). The results show that Tsu clearly experienced more days of strong wind than other stations in each of these months, and the number of windy days increased from December to March. The proportion of northwesterly winds during these days of strong wind also increased from December to March. In March, northwesterly winds accounted for approximately 60% of the windy days at Tsu.
Considering the results of that exercise, we selected a larger set of medium or higher wind speed events from the winter months of the 1987–2011 record in which the hourly mean wind speed at Tsu exceeded the average of the wind speeds at the three neighboring stations by more than 5 m s−1. We chose this criterion to match conditions during the low-level jet regime, in which the wind at UNIV, the station nearest Tsu (Fig. 2), was much stronger than in the surrounding area. First we subtracted the hourly average wind speed at the three stations (Ueno, Kameyama, and Yokkaichi) from the wind speed at Tsu for each hour and determined the hour of the day with the greatest difference. Next, we selected the days when the greatest difference in wind speed exceeded 5 m s−1 and defined them as windy days. The wind direction of a windy day was then defined as the wind direction at Tsu during the hour of the greatest difference in wind speed with respect to its neighboring stations. On some of these windy days the hourly mean wind speeds at Tsu were less than 10 m s−1, but they were nevertheless much stronger than those at the surrounding stations. Figure 19b shows the temporal distribution of these windy days with their wind direction indicated. The seasonal variations in the number of windy days and in the proportion of northwesterly winds are similar to those shown in Fig. 19a. Figure 19 thus implies that the wind direction was northwesterly during the period of Suzuka-oroshi in March. The 645 windy days with strong northwesterly winds had strong northwesterly winds only on the lee side of the radiosonde sounding line, and their mean wind speed was 10.3 m s−1. This group of “strong northwesterly days” also included the day we conducted the radiosonde observations and observed the low-level jet. Although not every strong northwesterly day may be associated with a low-level jet, the ensemble of such days should display common features associated with the low-level jet.
To identify the synoptic conditions generating the strong northwesterly winds at Tsu, we prepared maps of the average conditions during the 645 strong northwesterly days by using the 6-hourly product of JRA-25/JCDAS, setting the zero hour as the hour in the JRA25 product nearest to the start of the strong northwesterly wind (Fig. 20). The average sea level pressure pattern at hour 0 was a typical winter monsoon wind pattern called “west high and east low” (Fig. 20a). Figure 20b indicates the average evolution of the geopotential height anomaly (departure from the 24-h average) at the nearest grid point to Tsu over the 24-h period centered at hour 0. The figure shows that the vertical structure of the geopotential anomaly changed from negative to positive at hour 0. These features signify that the vertical structure changed from one associated with a low pressure system to one associated with a high pressure system at the time the strong northwesterly was first detected at Tsu.
Figure 20a also shows the average relative vorticity at 925 hPa during the strong northwesterly days at hour 0. The area of negative vorticity, which corresponds to anticyclonic circulation, covers the west side of Japan, and the Tsu area is at its edge. The evolution of the vertical profile of relative vorticity (Fig. 20b) indicates that the vorticity below 900 hPa shifted sign from positive to negative, similar to the change in the geopotential anomaly. These features show that the shift from cyclonic to anticyclonic circulation is a diagnostic indicator for the outbreak of a strong northwesterly day. This change corresponds to the eastward progress of synoptic-scale anticyclonic circulation, as shown in the eastward extension of the area of negative vorticity in Fig. 20a.
We then investigated whether the low-level jet event of 21 March 2010 satisfies this environmental condition. Figures 20c and 20d depict the conditions in the reanalysis dataset on the day of the radiosonde observations. Hour 0 was set at 1500 JST 21 March 2010, the onset of the low-level jet. At that time the hourly mean wind speed at Tsu, 11.5 m s−1, exceeded the mean of the surrounding three stations by more than 7.3 m s−1, the largest wind speed difference of the day (see bottom of Fig. 20d). An area of negative vorticity covered the west side of Japan with its edge reaching the observation area (Fig. 20c), and the pattern of vorticity elsewhere was also quite similar to the statistical pattern in Fig. 20a. The vertical profiles of the geopotential anomaly and relative vorticity near Tsu that day (Fig. 20d) were also quite similar to the statistical profiles (Fig. 20b), with the vertical structure over Tsu changing from a low pressure system to a high pressure system and shifting to negative vorticity below 900 hPa. The agreement between the observations and the statistical pattern implies that the synoptic conditions just described are important factors for the occurrence of strong northwesterly winds with a low-level jet.
The vorticity patterns on 21 March 2010 in our numerical simulation are shown in Figs. 20e and 20f. Note that the simulation did not reproduce the low-level jet (see Fig. 18). The synoptic features in Fig. 20e roughly correspond to those of the reanalysis dataset in Fig. 20c, but the positive relative vorticity in the observation area does not agree with the negative vorticity in the reanalysis and in the statistical results in Fig. 20a. The vertical profiles were well simulated (Fig. 20f), except that the relative vorticity in the lower layer in the simulation was positive from 0900 to 1500 JST, indicating cyclonic circulation over the Suzuka Mountains, in conflict with the reanalysis (Fig. 20d) and the statistics (Fig. 20b). The reason our numerical experiment failed to simulate the strong northwesterly regime with the low-level jet may be due to its failure to reproduce the synoptic-scale vorticity change from cyclonic to anticyclonic.
This comparison suggests that a synoptic-scale shift in vorticity from positive to negative (from cyclonic to anticyclonic circulation) is a key factor in the occurrence of strong northwesterly winds at Tsu. Our radiosonde observations revealed that the low-level jet regime occurred when strong northwesterly wind was observed only at Tsu, and when synoptic-scale conditions indicated a change in sign of the vorticity.
Our observational strategy of simultaneous multiple-site radiosonde observations with varied buoyancies yielded detailed two-dimensional profiles of the strong Suzuka-oroshi wind crossing the Suzuka Mountains. Our observations also documented two flow regimes associated with the Suzuka-oroshi: a downslope windstorm and a low-level jet. The second of these regimes, although not yet characterized in detail, may offer important insights into the Suzuka-oroshi. Our comparison of the observations with a numerical simulation and a statistical data analysis offers evidence that synoptic-scale negative vorticity (anticyclonic circulation over the domain) is a favorable condition for generating strong northwesterly winds with a low-level jet. The Froude number is conventionally used as an indicator for the occurrence of leeside strong winds, but it failed in the case of the low-level jet regime. Why this shift in the observation area from positive to negative vorticity, corresponding to the onset of anticyclonic circulation, is a useful indicator for the low-level jet regime in Suzuka-oroshi events will be the subject of future work, including comparisons of observational data with mesoscale simulations.
The ascent rate of the radiosonde balloon and the slope of isentropes both provide a reliable indication of the vertical wind direction (upward or downward) during downslope windstorms. This method offers better coverage of a longitudinal wind profile by varying the ascent paths of simultaneously launched balloons, although its advantage is greatest at high wind speeds. By relying exclusively on radiosondes, it offers useful results at relatively low cost.
Note that our observational method performed well for the downslope windstorm but not for the low-level jet regime. Nevertheless, the multipoint sounding method holds promise. Conventional upstream sounding succeeded in predicting leeward strong winds, but the upstream soundings alone were not able to detect the low-level jet regime. The advantage of radiosondes over airplane measurements is that radiosonde observations have high vertical resolution from the surface through the troposphere. The method of multipoint sounding, by taking advantage of this high vertical resolution in the boundary layer, was able to detect a low-level jet regime. For a future field campaign, we would add stations on the northwest side of the Suzuka Mountains to capture the structure of the strong northwesterly winds that accompany the low-level jet.
We thank Yuji Yamada and the employees of Aoyama-kogen Wind Farm Co., Ltd., for their cooperation and the use of their observatory at Mt. Kasatori, without which we could not have made these observations. We also thank the staff of the Experimental Farm of Mie University and the Iga Research Institute of Mie University, who made their facilities available as observational points. Dr. Daniel Kirshbaum and three reviewers made numerous suggestions for improving the paper. Dr. T. Kikuchi, Dr. J. Inoue, and Dr. M. Hori lent us their radiosonde system and some meteorological instruments. The assistance of numerous students of the Geosystem Science course at Mie University was crucial for making these demanding observations. All the sounding data are available at Weather and Climate Dynamics Division, Mie University, Japan. The map, meteorological fields in the profiles, and graphs were drawn with Generic Mapping Tools and Grid Analysis and Display System (GrADS). We used ASTER Global Digital Elevation Map (GDEM), version 2, to depict the mountain range. This study was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) through a Grant-in-Aid for Scientific Research (Grant 24654151).