Nocturnal low-level jet and ‘atmospheric streams’ over the rain shadow region of Indian Western Ghats



Spatial and temporal characteristics of a nocturnal low-level jet (LLJ) on the east side of the Western Ghat mountain range over India's west coast and processes leading to the formation of the jet are discussed. The boundary-layer jet has a regional scale extent, as revealed by high-resolution Advanced Research Weather Research and Forecasting (ARW) model simulations, and contributes to the formation of ‘atmospheric streams’ of water vapor over the selected land regions. Simulations indicate that the formation of LLJ is mainly attributed to the baroclinicity of the valley atmosphere due to the gently rolling terrain, which is assisted by the persistence of an unstable residual layer above the developing stable boundary layer in the valley and cooling over the slopes. Prior to the formation of LLJ, the boundary layer is dominated by deep roll circulations. The LLJ followed a gust front zone associated with a mountain wave. The low-level flow below the jet is decoupled from the upper-level flow as a result of strong vorticity below the jet and suppression of turbulence at the jet core. A conceptual model for the boundary layer interactions, dynamics of the mountain wave, LLJ, etc. are proposed for Western Ghat region. Copyright © 2011 Royal Meteorological Society

1. Introduction

Low-level jets (LLJs) play a crucial role in the transport and mixing of water vapor and pollutants. The LLJs have different spatial and temporal scales and are caused by a variety of mechanisms (Blackadar, 1957; Holton, 1967; Garrat, 1992). Some of the widely reported LLJ formation mechanisms include inertial oscillation (Blackadar, 1957; Bonner, 1968), change in surface and terrain characteristics and diurnal cycle as in the case of a coastal jet (Smedman et al., 1993, 1995), slope and valley wind (King and Turner, 1997), large-scale baroclinity by sloping terrain (Li et al., 1983; Gerber et al., 1989), severe weather conditions (Arrit etal., 1997), frontogenesis during frontal passages (Whiteman et al., 1997; Banta et al., 2002; Lundquist, 2003), geostrophic adjustment as in the case of a barrier jet (Parish, 1983; Li and Chen, 1998), etc. or a combination of these mechanisms. Nocturnal decoupling of the flow from the surface and an imbalance between the pressure gradient and Coriolis forces introduces an inertial oscillation of winds and is considered to be a major contributing factor for LLJ formation.

Large-scale features of the synoptic jet over the Indian west coast are studied extensively in the context of the ‘Somali jet’, which is a climatological feature during the pre-monsoon and monsoon seasons (Joseph and Raman, 1966; Findlater, 1969). This jet is one of the major characteristics of the Indian monsoon, with a ‘cross-equatorial current from the southern Indian ocean to the central Arabian sea’ (Krishnamurthy et al., 1976), providing a moisture supply over the land regions, fuelling convection and rainfall (Sikka and Gadgil, 1980). Findlater's jet characterizes considerable spatial and temporal variations and oscillations (Goswami et al., 1998) on an intraseasonal scale coupled to the active and break periods of monsoon. Findlater's jet, observed below 700 hPa, is associated with strong horizontal and vertical wind shear, but deviates from the definition of a boundary-layer LLJ as it is typically found above 1 km, in association with synoptic scale flow. Boundary-layer jets over the Indian region are less explored. Boundary-layer LLJs are low-level wind maxima observed below 1 km typically as a nocturnal phenomenon where shear-induced turbulence predominates and the jet nose often coincides with nocturnal inversion layers. LLJs transport moisture (McCorcle et al., 1988; Mitchell et al., 1995; Parker et al., 2005) and pollutants (Beyrich et al., 1994; Corsmeier et al., 1997; Banta et al., 1998; Seaman and Michelson, 2000; Bao et al., 2008) from far-away places and often contribute to surface-level concentrations by vertical mixing. Vertical mixing is accomplished by an upside-down boundary-layer structure attributed to the shear generation of turbulence at the jet height (Mahrt and Vickeres, 1999; Banta et al., 2002; Mahrt and Vickers, 2002; Prabha et al., 2007, 2008; Karipot et al., 2008).

The presence and characteristics of nocturnal LLJ on the east side of the Indian Western Ghats (WG) is not documented, although flow dynamics and moisture transport in this region are expected to play a significant role in the pre-monsoon thunderstorms and also contribute to the diurnal cycle of convection and precipitation. The need to understand LLJ in this region is manifold because of its importance in several applications such as weather prediction, weather modification, moisture and pollution transport assessment, wind power assessment and aviation safety. This region is also influenced by the transport of dust from the Arabian mainland and maritime aerosols. The knowledge of moisture distribution along with that of the aerosols will help determine the areas of potential convective activity in this rain shadow region. This information is also instrumental for an ongoing Cloud Aerosol Interaction and Precipitation Enhancement EXperiment (CAIPEEX) in the region.

The current study is intended to explore the nocturnal LLJ formation mechanism and evolution during a dry period before the pre-monsoon thunderstorm formation. The main emphasis is on the boundary layer characteristics leading to the formation of the jet and associated features. The investigation is based on numerical simulation and observational data analysis. An attempt is also made to introduce a conceptual model for the formation and evolution of the LLJ.

2. Data and method

The Advanced Research Weather forecasting model (WRFV3.1.1) was configured with three two-way nested domains (30, 7.5 and 1.875 km resolution). Two nested domains are presented in Figure 1, showing the surface elevation that is resolved in the model at respective resolutions. The lowest model level is kept at approximately 7 m above the surface and the top layer was kept at 19 km. The boundary layer was parameterized with the Yonsei University (YSU) non-local PBL scheme. The lower part of the model atmosphere (<1.6 km) was represented by 15 vertical layers with varying vertical resolution. The PBL scheme was coupled with the Monin Obukhov (MO) similarity scheme for the surface layer and Noah land surface scheme (Chen and Dudhia, 2001) using the updated MODerate Resolution Imaging Spectroradiometer (MODIS) land use dataset. The innermost domain has a cloud resolving model and the outer domains are parameterized with the Kain and Fritsch cloud scheme. The WRF double moment (WDM-6) cloud microphysics scheme was used with modified boundary conditions. The new parameterization of Klemp etal. (2008) is used for damping vertically propagating waves in the upper layers. The model is initialized with final reanalysis products (FNL) from the National Center for Atmospheric Research (NCAR) and outer boundary updates are done every 6 hours. Hourly model output is used in the analysis. Simulations are carried out for the period from 0000 UTC (5.30 IST; IST = 5.30 + UTC) 12 May until 1200 UTC (17.30 IST) 14 May and hourly model outputs are analyzed to investigate the diurnal signature of the LLJ.

Figure 1.

Topographic height in model domains 2 and 3. (PUN, Pune on the east side, and DAP, Dapoli on the west side of WG mountain range, are shown. Four valleys in the domain are marked as A, B, C and D.

The observational data used in this study include hourly surface observations from India Meteorological Department (IMD), radiosonde observations carried out at Pune (73.86°E, 18.5°N, 570 m above mean sea level (msl), surrounded by hilly terrain with elevations reaching up to 710 m). The radiosonde observations were carried out as part of the pilot observations for Cloud Aerosol Interaction and Precipitation Enhancement EXperiment (CAIPEEX). Sodar observations of three-dimensional wind components at the same location are also used in the study. The sodar (Sameer inc., India) was operated at a frequency of 1.8 kHz. Observations of all three wind components were averaged for 5 min intervals from 19 s sampling with an accuracy of 0.1 ms−1. Observations were taken at 30 m above the surface to 900 m, at 30 m vertical resolution and were used in the study.

A wavelet analysis using the Morlet wavelet was applied on the sodar wind observations to investigate periodicities in the wind components and associated variances. Continuous sodar data for only 2 days were available during the study period, which was insufficient for looking at oscillations with longer periodicity (more than a day). An extended study period of 5 days' data (30 May to 3 June) with similar dry conditions was used to examine periodic fluctuations (greater than a day) in vertical velocities. MODIS observations of precipitable water at 5 km spatial resolution were used in the study. The MODIS level II precipitable water data at 11.30 IST (Indian Standard Time) were used to compare with the vertically integrated precipitable water (IPW) from model simulation in the inner domain.

In addition to the observational data, the Modern Era Retrospective-analysis for Research and Applications (MERRA; reanalysis data (with a resolution of half a degree along latitude and two-third degrees along longitude) was used to compare high-resolution Advanced Research Weather Research and Forecasting (ARW) results against a coarse model and also for energy balance comparisons as observational energy balance data are not available. Hourly MERRA data for 2 m level air temperature and mixing ratio, 10 m level winds, radiation and energy balance components, and vertical profiles of wind and temperature at 11 IST from grid points adjacent to radiosonde observation locations are used.

The LLJ is identified in this study with a maximum wind at a certain height with decreasing wind speeds of at least 2 m s−1 both below and above that height (Andreas et al., 2000). Only jets identified below 1.5 km are considered to be LLJ in this study. Other higher elevated jets present in the study domain are considered as an extension of the Findlater jet and are not given emphasis in this paper.

3. General characteristics of the study period

The study period chosen for detailed analysis was 12–14 May 2009, which was a dry period over the study region with northwesterly/westerly mid-level flow and dry easterly/northeasterly flow in the upper layers (above 5 km), characterizing pre-monsoon dry condition before the thunderstorm events. The northwesterlies were associated with the high-pressure system situated over the Middle Eastern region. The subtropical jet is a characteristic feature of northern latitudes over the Himalayan region. Wind fields over the study region veered with height, indicating warm air advection from the northern/northeastern drylands.

The study domain shows considerable variation in topographic height (Figure 1(a)) from a few meters to a kilometer above msl. There are four main valleys represented in domain 3 (indicated as A, B, C, and D in the figure), which are oriented westnorthwest to southeast on the eastern slopes of (WG). The rainfall distribution in this region is quite heterogeneous, with heavy precipitation concentrated on the west side of the mountain and a reduction to 70–80% (of the windward side rainfall) within a distance of 100 km on the east side of the Ghats (Gunnell, 1997). Seabreeze events typically noted over the western side of the mountain ranges do not usually penetrate inland to distances beyond 200 km. A line of clouds is typically noted near the mountain summit along with an area parallel to the coastline without clouds on the upwind area, indicative of the presence of mountain wave and associated subsidence.

3.1. ARW, observations and MERRA comparisons

A model verification is done with the help of India Meteorological Department (IMD) surface observations at Pune and simulated values from MERRA reanalysis products (Figure 2) for 2 days. Although the diurnal temperature variation is reproduced well in the ARW simulation, the model results show lower maximum temperature compared to observations. MERRA temperature shows a higher maximum temperature. Hourly nocturnal temperature from ARW has a slight warm bias (≤2°C), in contrast to MERRA, which shows a cold bias during the night. A warm bias at night points to less radiative cooling and more down-mixing of warm air from above. A comparison of ARW water vapor mixing ratio showed a dry bias (2–4 g kg−1) against most of the hourly observations (Figure 2(b)). MERRA showed reasonable agreement with water vapor mixing ratio observations during the daytime. Night-time differences exceed 2 g kg−1. Wind speed comparisons show considerable variations in the observations compared with the simulation (Figure 2(c)). The increase in daytime wind speed on the second day would mean that there is more turbulence. However, temperature also increased in association with this. The nocturnal surface level wind was always above 2 m s−1, both in the model and observations. The wind direction showed a clear diurnal variation, changing from northwesterly during the day to westerly during the night. MERRA does not reproduce this feature on the first day of the simulation; however, it gave a good comparison on second day. The backing of wind with time is noted here.

Figure 2.

Comparison of simulated and observed 2 m level temperature (a), 2 m level water vapor mixing ratio (b), 10 m level wind speed (c) and wind direction (d) at station Pune. Respective data from MERRA reanalysis are also shown.

The differences in the temperatures and mixing ratio between the models could be examined with the help of a comparison of energy balance comparison from ARW and MERRA (Figure 3). The incoming southwest radiation flux from the ARW model is slightly higher than that of MERRA (Figure 3(a)); however, there is no significant difference between the sensible heat flux of both models (Figure 3(b)). Sensible heat flux stayed positive throughout the night due to weak convective conditions in the model. It may also be noted that the land use is ‘urban and built-up land’ and soil type is ‘clay loam’. This behaviour is possible because of these conditions. However, it needs further investigations with direct observations of energy balance and soil characteristics. The ground heat flux from the ARW model is large compared to the MERRA, during both day and night. The ground heat flux differencesduring the midday period (less heat is stored in MERRA) are consistent with higher temperatures in MERRA. It should also be noted that the ARW model has predicted a lower temperature on the first day, which is mainly due to a higher ground heat flux on this day compared to the second day. The latent heat flux played a minor role in the energy balance (MERRA represented low latent heat flux <60 Wm−2 compared to the dry conditions in WRF), the primary influence on the temperatures being from partitioning of energy into ground heat flux.

Figure 3.

Comparison of WRF simulated and MERRA incoming short-wave (SW) and radiation at the surface (a), sensible (H) heat flux (b), ground heat (G) flux (c), and long-wave (LW) radiation (d) at the surface for Pune.

The model results at 1130 IST on 13 May 2009 are compared with the radiosonde profile observed over Pune (Figure 4) and MERRA reanalysis profiles. The simulated profiles and observations are generally comparable; except that a jet observed above (at 3 km) the boundary layer was at a higher elevation in the observations. This jet is an extension of the Findlater jet. Between the layers 3.5 and 7 km, wind direction changes to northeasterly, indicating advection from land regions. The radiosonde wind ascent throughout the 17 km layer lasted approximately 1 hour (starting at 1130 LST) and showed several oscillations during the upward flight, indicative of the gravity waves. The radiosonde flight was conducted during the development of the convective boundary layer. A comparison of water vapor mixing ratio profiles indicated that the dry surface bias in the water vapor content (Figure 4(b)) pertained not only to the surface, but extended up to 6 km (Figure 4(b)). In addition, the water vapor mixing ratio (Figure 4(b)) in the updrafts/weak downdrafts (Figure 4(c)) was higher than that over downdraft regions, indicative of water vapor transport through updrafts. Strong downdrafts (Figure 4(c)) are characterized by dry air (Figure 4(b)). It should be noted that both profiles (in the updraft and downdraft) deviated considerably from the radiosonde observations.

Figure 4.

Comparison of vertical profiles of MERRA and WRF model horizontal wind speed profile (a) and water vapor mixing ratio (b) derived from radiosonde observations (RS) at Pune on 13 May 2009 at 1100 LST. Model-derived vertical velocity (c) and corresponding mixing ratios in the updraft (WRF Updraft) and downdraft (WRF downdraft) are also shown. Wind direction from radiosonde is given in (a) top axis. This figure is available in colour online at

3.2. Comparison with sodar observations

A comparison of sodar observations at Pune and model results is shown in Figure 5. Weak boundary layer winds are noticed both in the sodar observations and in the model (Figure 5) during daytime. Sodar observations indicated a nocturnal low-level jet with a mean height of 200 m (Figure 5(a)). Jet height and speed were variable and showed a close resemblance to the model simulated jet (Figure 5(b)). Sodar wind profiles at 15-minute intervals are presented, which showed a clear indication of the non-stationary behaviour of the jet, with considerable changes in the height and speed of the jet. Simulated results are presented for hourly intervals and appear smoother than the observations; however, general characteristics of the jet such as height and speed, time of onset and jet decay are simulated reasonably well. It is apparent that the model has simulated some of the general characteristic of the observed LLJ. These characteristics over a regional scale/horizontal extent of the jet will have a significant effect on the moisture distribution over the region. The temperature and mixing ratio variations in the simulation showed development of a strongly stable layer below the jet and a strong vertical gradient of water vapor (Figure 5(b)). Daytime boundary layer development showed indications of alternating updrafts and downdrafts that became stronger and had a longer period until the jet initiated at 18.30 IST on both days examined here (Figure 5(c)). The long-period updrafts/downdrafts are associated with the onset of the gust front. Indications of this gust front are noted at 1500 IST on 12 May and 1600 IST on 13 May, with a strong updraft on both the days. The updraft/downdraft pairs become weaker and disappear as the LLJ sets in, indicating stable boundary layer development. Overall, WRF model comparison with observations indicated better agreement with MERRA and suitability for further detailed analysis.

Figure 5.

Time–height cross-section of wind speed (m s−1) from sodar located at Pune (a) and wind speed from simulations (b) in color shades, potential temperature (continuous lines) and mixing ratio (dash-dotted lines). Time–height cross-section of vertical velocity (m s−1) from sodar located at Pune (c) and from simulations (d). Arrows indicate onset of the gust front.

4. Temporal evolution of the boundary layer leading to jet formation

The discussion in this section is based on intricate details of the flow images of vertical velocities derived from simulations at 700 m and 1500 m above msl (Figure 6) and vertical transects along the Pune latitude (Figure 7). The characteristics of the convective boundary layer, its evolution and decay, leading to the formation of the nocturnal jet, are discussed. The convective boundary layer at 1330 IST is dominated by boundary layer roll circulations off the coastal areas over the Arabian Sea (Figure 6(a) and (b) shows a horizontal distribution of vertical velocity at 700 m and 1500 m and Figure 7(a) shows a vertical transect at 1330 IST). A front is noticed along the coastal areas 73–74°E with high vertical velocities, apparently due to the sea breeze penetration inland. During the development of the daytime boundary layer (Figure 5(a)), pairs of strong updrafts and downdrafts develop over the eastern slopes and inland areas and transform to roll-type circulations, which are elongated and aligned parallel to the wind (Figure 6(a) and (b)). Convection over inland regions is dominated by closely spaced updrafts and downdrafts which are deeper compared to that over the slopes (Figure 7(a)); at 1330 IST, the valley boundary layer has stronger upward vertical velocities (>1 m s−1) and greater vertical mixing (reaching 3.5 km in height) over a larger volume, unlike over the slopes and hill tops, with updraft/downdraft pairs reaching 2–2.5 km. A sea breeze front (note the north–south orientation of this front parallel to the coastline) is noticed along 74°E, but has not advanced inland. The convective rolls that occupied the valley boundary layer inhibited the progression of this sea breeze front. Rolls are responsible for the vertical mixing of the water vapor and pollutants and ventilating the valley.

Figure 6.

Spatial distribution of simulated vertical velocity at three times (1330, 1830, 2030 IST; 800, 1300, 1500 UTC) at 700 m and 1500 m above msl on 13 May. (Dark gray indicates vertical velocities >−1 m s−1 and white indicates show maximum updrafts > 1 ms−1). Convergence lines with high vertical velocity are noted with dashed lines (bottom right).

Figure 7.

Vertical distribution of simulated vertical velocity at four times—(a) 1330 IST, (b) 1830 IST, (c) 2030 IST, and (d) 2230 IST on 13 May—above msl. (Black/dark gray indicates downdrafts >−1 m s−1 and white indicates show maximum updrafts > 1 m s−1). This figure is available in colour online at

At 1830 IST, the organized convection began disintegrating into more disorganized patterns which are more prevalent over the valley atmosphere (Figure 6(c) and (d)). This was also observed with the help of photographs of clouds taken over Pune. A comparison of those cloud patterns and temperature patterns is provided at∼majfiles/QJRMS_thara/cloud-patterns-valley.pdf. Meanwhile, the vertical cross-section shows that there are deeper and wider convective cells (Figure 7(b)) compared to that during 1350 IST. The rolls subsequently disappear as a stable layer is formed close to the surface, as revealed in Figure 6(e) at 2030 IST. The roll circulations are replaced by wide updrafts along the valley bottom and downdrafts over the slopes and hill tops. A residual layer is present above the stable layer (Figure 7(c) and (d)), with characteristics similar to the disintegrated convective boundary layer, with dominating updrafts; however, rolls are not noticed in that layer.

A frontal zone noticed along the offshore area has progressed further inland by 2030 IST (Figure 6(e)). The WG mountain range is a strong barrier to the landward progression of sea breeze, initiating mountain waves (Figure 7(b)) that propagate upwards (Figure 7(c)), characterizing strong updrafts and downdrafts (>1.5 m s−1). Figure 7(c) shows a classic picture of the mountain wave propagation. Such a mountain wave formation has been coined an ‘evening wave’ by Roper and Scorer (1952). There have been two seminal studies (Sarker, 1965; De, 1971) investigating the mountain waves over WG using theoretical approaches, which emphasize that mountain waves over WG can have a wavelength of 25–75 km; this is also true in our study. However, these mountain waves were not explored in relationship to its diurnal course. It may be noted that the signature of this mountain wave appears earlier; at 1830 IST, before the convection ceases, a wave gradually propagates upward, which appears as a standing wave. However, as the wave moves inland (as the convection stops), the wave also propagates to greater heights (Figure 7(c)). The mountain wave propagation excites horizontally propagating gravity waves upstream, which reach down to the lower boundary layer (Figure 7(c) and (d)). This causes momentum transfer in the residual boundary layer (RL) and oscillations in the stable boundary layer (SBL). The initiated mountain wave is seen as a gust front at the surface and in the sodar measurements. The SBL wind speeds behind this front align with strong updrafts (>2 m s−1), forming an LLJ, which flows downslope, carrying more moist air.

Maximum LLJ speed noted in the simulation was 10–14 m s−1 along the eastern slopes of WG. The gust front moved at a speed of 100 km h−1; it was situated along 74°E at 1830 IST and shifted to the valley bottom (77°E) at 2230 IST. This may be viewed as the propagation of mountain wave to inland locations (see movie file llj.mpg—refer to Appendix for details of available files). Simulations indicate that the LLJ is formed behind the front, follows the gust front inland and is noticed over the entire model domain. Boundary layer clouds (BLCs) noticed offshore bring supersaturated air to the level where the jet is initiated (Figure 8). There is a second line of clouds along the mountain tops, which is closely associated with the forced lifting. However, the low-level flow is channeled along the coastline, making a barrier between the two lines of BLCs. This cloud-free region is also characterized by descending motions (downdrafts associated with the mountain wave). Upper-level flow is north/northeasterly, showing a veering of wind with height (thick arrows).

Figure 8.

The structure of LLJ and an associated gust front ahead of the jet (jet is shown with isosurface of 10 m s−1). Horizontal streamlines through a coastal location (star) at different vertical levels up to 6 km are shown. Arrows indicate direction of streamlines. Clouds along the coastline and offshore are also shown with white and gray isosurfaces. Thin vectors correspond to lower levels and thick vectors correspond to higher elevations. Location of gust front inland is shown with dashed lines. A temporal evolution of these features is provided in the movie file located at∼majfiles/QJRMS_thara/llj.mpg

5. Spatial distribution of LLJ

Analysis of model domain 2 with a horizontal resolution of 7.5 km is used in this analysis. Figure 9 shows a temporal variation of the jet speed (a), jet height (b), 2 m level water vapor mixing ratio (c) and Brunt–Väisälä frequency for stability (d) at different longitudes. There is a clear indication that LLJ develops early along the coastal areas only along the east side of WG and propagates inland at a speed of 10 m s−1 or more (Figure 9(a)). The initiation of LLJ over locations 300 km inland are delayed by 3–4 hours. Once initiated, the LLJ lasted until the following day's insolation heated the eastern slopes of the WG. The height of the jet also varied between 200 and 600 m depending on location (Figure 9(b)). The 2 m level water vapor mixing ratio (Figure 9(c)) increased at the same time as jet initiation and seemed to be related to LLJ evolution. The low-level atmospheric stability showed more stable conditions developing early at night along locations closer to the coastline and weak stability over the valley (Figure 9(d)). For a given time, stability increased significantly closer to the slopes and the valley bottom was more unstable compared to the slopes. This can be visualized from Figure 9(e) showing the vertical cross-section along the Pune latitude. The presence of LLJ is clearly noted to follow the terrain. The temperature distribution showed strong gradients in temperature, leading to the differences in stability conditions discussed earlier.

Figure 9.

Variation of jet speed (a), jet height (b), 2 m level mixing ratio (c) and Brunt–Väisälä frequency (d) a vertical transect across Pune latitude at 18 IST (e) on 13 May 2009 for wind speed (color), water vapor mixing ratio (dashed contour), perturbation potential temperature (contour) in (e); hatched area shows terrain intersection.

6. Moisture transport by the LLJ

The spatial distribution of Integrated Precipitable Water (IPW) from both the simulation (Figure 10(a)) and MODIS (Figure 10(b)) at 1130 IST showed significant influence from the topographic features. The valley points showed enhanced IPW compared to summit points. There are four main valleys in domain 3 (marked A–D in the figure) and four undulating spatial patterns of IPW are noticed both in the observations and model simulations. The IPW observations and simulations show enhanced values along the coastal areas. Over the Arabian Sea there are some pixels with very low IPW, attributed to cloud. Moisture transport to the east side of the WG is blocked by high mountain ranges. However, the spatial demarcation in the IPW is noted across a very narrow range of distances (approximately 60 km inland). The orographically forced winds, however, play a crucial role in the transport of moisture to inland locations through relatively lower elevated ‘gaps’ in the mountain barrier. There are three gaps as shown in the inner domain image of water vapor mixing ratio at 700 m above msl (Figure 10(b); black color indicates the terrain intersections and dark gray to white indicates high to low water vapor mixing ratio); and moisture is transported to inland areas through gaps in the barrier as ‘streams’. These streams are sharper in the MODIS data than in the model output. The model output is derived from integrated water vapor mixing ratio over the whole depth of 20 km vertical domain. Another important aspect is that offshore convection associated with the boundary layer clouds (noticed as alternating gray and white patches over the Arabian Sea) increase the moisture at this elevation through saturated updrafts. The increase in moisture over inland locations can also be identified by an increase in IPW in the MODIS data and in the model results (e.g. small lower-atmospheric ‘water streams’ analogous to atmospheric rivers (Ralph et al., 2004) of water vapor). Interestingly these patterns are also noticed in the MODIS aerosol optical depth (not shown), indicating the aerosol accumulation in the valley points. Further details of this important aspect are presented for hourly images of water vapor distribution at 1 km above msl made from the ARW model output as a movie file accessible at∼majfiles/QJRMS_thara/streams.mpg. In the presence of LLJ, water streams appear well along the valleys and four main streams in the study area are found.

Figure 10.

Column integrated precipitable water (cm) from ARW simulation (a) and from MODIS Level II data (b) on12 May at 525 UTC (10.45 IST). A movie file is provided at∼majfiles/QJRMS_thara/streams.mpg which shows evolution of water vapor streams.

7. Mechanism of LLJ formation

The WG is considered to produce upstream blocking effects offshore (Grossman and Durran, 1984). In the lee of this mountain range, various complex interactions could take place. A preliminary analysis of various effects such as the influence of temperature gradients between the valley and slopes, influence of blocking effects with the help of Froude number, slope wind influences, and effect of inertial oscillation are explored in this section.

7.1. Thermal wind

One potential mechanism for formation of LLJ could be associated with the horizontal temperature gradient between valley and slopes. The thermal wind equation (Holton, 2004) could be used to explain the formation of the jet corresponding to this scenario. The vertical shear in the geostrophic v component could be expressed with the help of horizontal temperature gradients as follows:

equation image(1)

where f = 2 Ω sin Φ; g is acceleration due to gravity, Ω is angular velocity vector, Φ is latitude, T is temperature, and x is distance between two east–west points (74°E over the slopes, 79°E over the valley). The temperature gradient between the slopes and the valley atmosphere, jet speed and height, and vertical shear in the geostrophic v wind component are presented in Figure 11(a). Jet speed is enhanced when the temperature gradient between the valley and the slopes increases, and a strong geostrophic shear (>0.015 s−1) is noticed in correspondence with LLJ speeds exceeding 8–12 m s−1 over the Pune location. Two important aspects to be noted are that the air column over the slopes cools faster than the valley atmosphere. A convective residual layer remains over the valley atmosphere as the valley atmosphere cools slowly. A combination of these two effects leads to strong east–west temperature gradients. As Figure 11(a) suggests, increase in v wind shear due to the thermal wind could explain the existence of the jet. The increase in v shear adds a northerly component to the mean flow, which leads to northwesterly LLJ (see movie file streams.mpg for wind direction changes during the jet).

Figure 11.

Variation of potential temperature gradient between the slopes (74°N) and valley (79°N), the slopes and valley atmosphere at 850 m (Δθ = θslopeθvalley_850 mb), slope and valley bottom (Δθs = θslopeθvalley_bottom) LLJ speed (UJ; speed at jet maximum) and LLJ height (HJ; height of jet maximum) are presented in (a). Temporal variation of upstream wind speed, Brunt–Väisälä frequency and Froude number are presented in (b). Slope wind and layer averaged velocity vector, heating rate are presented in (c) over Pune location. This figure is available in colour online at

7.2. Blocking effects of mountain

The Froude number (defined as the ratio of upstream wind speed and the product of Brunt–Väisälä frequency and the width of the mountain barrier ≈ 100 km) was less than one, estimated for a 2-day period (Figure 11(b)). This indicates that on the west side of the mountain blocking effects due to high terrain dominate. This also suggests that flow separations and channeling effects are possible over the west coast. A diurnal oscillation in the Froude number (Figure 11(b)) is due to stability changes associated with the diurnal cycle. Low values (<1) of Froude number do not support amplification of the lee waves and their horizontal propagation. The kinetic energy is rather insufficient for excessive lifting and amplification of downstream propagating mountain waves, rather supporting the vertical propagation of mountain waves.

7.3. Slope flow effects

Another possible cause for the LLJ formation is slope flow on the eastern side of WG. McNider (1982) suggested an oscillatory solution for the slope wind, which depends on the rate of temperature, vertical temperature gradient and terrain slope. Slope wind analytical formulation from McNider (1982) is used to calculate the slope wind component in the layer below 600 m. The downslope velocity component (us) for frictionless conditions is given by

equation image(2)

where, ξ2 = (gγ/θ0)sin2 α, Lc is rate of change of temperature, α is slope, γ is lapse rate and θ0 is ambient temperature. The solution does not explain the strong winds that are noted during the LLJ (Figure 11(c)) on day 1; however, it produces nearly half the wind speed of the LLJ. This result suggests that slope winds are important in the nocturnal boundary layer but are unlikely to be the reason for the existence of the large-scale nocturnal jet.

7.4. Inertial oscillation

The role of inertial oscillation is considered a common reason for the formation of LLJs over several places. To investigate the role of inertial oscillation, wind hodographs are considered. Figure 12(a) and (b) shows hodographs of wind (Holton, 1967) at the surface and at 250 m above the surface, at Dapoli (located on the west coast) and at Pune (on the east side), respectively. In the hodograph, u and v components of wind velocities are plotted, along with their time variation at a constant height. The wind oscillations during the course of the day are described with the help of u and v component variations with time at two different levels (10 m and 250 m above the surface). Each point in the figure corresponds to hourly horizontal wind components from the model, giving information on velocity variation with time. The wind oscillations at both elevations are similar. Dapoli has a typical anticyclonic (veering) wind vector with time, and a cyclonic turning of wind is noted over Pune, which is atypical, compared to LLJs studied over other locations (Thorpe and Guymer, 1977; Ulden and Wieringa, 1996; Baas et al., 2009; Bain et al., 2010). This indicates that inertial oscillation may not be the cause of jet formation over Pune. It is also to be noted that both u and v components of winds increased during the oscillation.

Figure 12.

Hodograph of winds over Dapoli (a) and Pune (b) at 10 m and 250 m level. Numbers associated with the symbols represent each hour in Indian Standard Time (IST). Arrows show the direction of oscillation.

Another investigation was carried out with the help of spectral analysis of sodar vertical velocity. The relationship between periodicity of events associated with LLJ and its variance contribution are shown in Figure 13(a). The spectral analysis for the study period of 2 days is shown with different symbols for three elevations (90 m, 270 m, 420 m). There are three peaks in the spectra corresponding to 2 hours, semi-diurnal and diurnal periodicity. The diurnal periodicity is most dominant at lower levels and might be attributed to the mountain wave propagation. The inertial period corresponding to the Pune latitude is 37.7 hours. In order to check the inertial periodicity, continuous 5-day data (30 May to 3 June) with similar ambient conditions to the 2-day case study period was used to derive spectra. As is evident in Figure 12(a) (lines), the energy corresponding to the inertial period was not significant at lower levels. However, this period was found to be significant above the jet core and has more energy.

Figure 13.

Energy spectra of vertical velocity at different heights (a) and normalized vertical velocity variance at peak periodicities (2, 12, 24, and 38 hours) as a function of height (b). Symbols are used for data during 12–13 May. Time–height cross-section of the vertical component of vorticity (s−1) from simulation at sodar location (c).

7.5. Recirculation zones below the jet

A close examination of w spectra shows a suppression of the variance (and thus less turbulence) at jet height (Figure 13(b)). Spectral analysis for horizontal components of sodar winds at different heights (not presented) also indicated similar semi-diurnal and diurnal periodicities and suppression of energy at jet height. The turbulence is maximal in the lower layers and also increases above the jet. It is noticed from the simulations that this suppression of turbulence at jet height is closely associated with production of strong vorticity below the jet (Figure 13(c)). This could be interpreted as due to a hydraulic jump, as described by Schär and Smith (1993) for sub-critical upstream conditions, but needs further clarification from vorticity budget analysis. The presence of less turbulence at jet level is in contradiction to several other LLJ studies where vertical mixing is accomplished by the generation of turbulence at jet height due to increased shear (Mahrt, 1999; Banta et al., 2002; Mahrt and Vickeres, 2002; Prabha et al., 2007, 2008; Karipot et al., 2008). The increased shear in some cases is also found to suppress turbulence by sheltering large eddies (Smedman et al., 2004; Prabha et al., 2008) from penetrating to lower layers.

8. Conceptual model for LLJ over the WG

The nocturnal LLJ that has formed appears to be due to a combination of effects. During the daytime, the moisture/temperature distribution shows an increase/decrease until the top of the mountain barrier. The wind speed in the lee side is less than 3 m s−1, compared to 9–12 m s−1 on the windward side. The boundary layer roll circulations are the primary mode of convection over inland locations (vertical velocity distribution showed alternating updrafts and downdrafts which scale the boundary layer depth of 3–3.5 km at deep valley points). Events after sunset are summarized with the help of the sketch shown in Figure 14. After sunset, the eastern slopes cool faster than the valley bottom and the valley atmosphere. The disappearance of daytime intense convection over land assists the mechanically lifted air over the mountain to progress inland, in the form of a mountain wave. As the valley bottom cools, a residual layer remains over the developing stable layer. These events lead to a strong temperature gradient (of up to 14 K; see Figure 11(a)) between the slopes and the valley atmosphere (point B is warmer than A in Figure 14). The valley potential temperatures are independent of height over a very shallow stable layer. This indicates the presence of a deep, well-mixed, and warmer residual layer. These events create a pressure gradient between the slopes and valley to drive the LLJ carrying moisture to inland locations. A thermal wind v component is imposed on the mean flow to make it northwesterly over the region. The jet blows from west to east with a northwesterly component along tributary valleys. This change in wind direction was also noticed in the IMD surface observations (Figure 2(d)), as discussed earlier. The mountain wave carries a large vertical flux of horizontal momentum associated with the LLJ. In response to the progress of LLJ inland, moisture is also transported in the lower layers, especially through selected narrow regions as ‘streams’ where the jet is strong. This low-level moisture accumulation is noticed behind the gust front associated with the mountain wave (see movie file at∼majfiles/QJRMS_thara/streams.mpg).

Figure 14.

Conceptual model for the nocturnal LLJ over the WG.

Numerical models and reanalysis products have difficulty in representing LLJs by their speed, height, horizontal and temporal variations and extent (Kosović and Curry, 2000; Cuxart et al., 2006; Steeneveld et al., 2008). These problems are believed to be due to low spatial and temporal resolution of models (Anderson and Arritt, 2001; Prabha et al., 2011) and to inadequate boundary layer physics (Storm etal., 2009) or influence on boundary layer physics through sensitivity to land surface parameterization (Prabha et al., 2011). Our simulations reinstate model inadequacy, while also illustrating the capability of the model to re-create some of the observed features. Recent numerical model verification studies are unraveling the importance of LLJs and their role in the diurnal cycle of convection and precipitation (Hu 2003; Wang et al., 2009; Pospichal et al., 2010). These jets have a significant impact on the water cycle over the region through atmospheric streams, which are not represented in coarse resolution models. Most importantly, the representation of these jets in high-resolution numerical models also requires improved understanding from observations and simulations, which is inevitable for accurate prediction of cloud and precipitation processes.

9. Conclusions

A high-resolution simulation with the ARW model has been used to highlight some of the characteristic features of the boundary layer over the rain shadow region in the Western Ghats of India. It is shown that roll circulations rooted in the surface layer and reaching to approximately 3.5 km height are the primary mode of convection during the day. The study revealed the events that lead to the formation of an LLJ on the eastern side of the mountain range and their regional-scale characteristics. Some intricate details such as the importance of mountain wave propagation and its association with the LLJ, and the relationship with the residual layer over the valley atmosphere, are revealed for the first time.

The following conclusions are drawn:

  • There exists a nocturnal LLJ on the east side of the WG mountain range over the Indian peninsular region, which is different from the well-studied large-scale jet over the Indian region: the Findlater jet.

  • The nocturnal LLJ is baroclinically driven as a result of the temperature gradients between the valley atmosphere and slopes, and influenced by the progression of mountain wave and an associated gust front.

  • Atmospheric ‘streams’ of water vapor noticed with the LLJ are an integral part of the diurnal cycle, which are important over the eastern slopes of WG. These features might not be resolved in low spatial resolution models, as details of the topographically forced nocturnal jet are not appropriately simulated.

  • A mountain wave with a diurnal periodicity is noticed, which can be recognized in the surface observations with strongly coherent updrafts and downdrafts exceeding 1 m s−1.

  • The upward-propagating mountain wave excites gravity waves which are found to reach down to the nocturnal boundary layer.

  • A deep residual layer (remnant from the CBL) is noticed over the stable boundary layer, which is dominated by upward vertical velocities, making the valley atmosphere less stable and contributing to jet development.

Additional verification studies using more observations are needed to investigate these interactions further. However, this study gives some important details of the boundary layer dynamics and their interaction with the topography over the rain shadow region. This forms a crucial step in designing experiments aimed at studying the boundary layer over the area. Non-local characteristics of the rolls might imply that point measurements can be misleading and careful considerations are necessary to make appropriate measurements and interpretations. The rolls in the boundary layer reach up to 3.5 km in the valley during the day and remain as a convective region through the night as a convectively active residual layer above the developing stable boundary layer, holding the pollutants until the jet dynamics and gravity waves fill the stable valley atmosphere beneath it. This is a very important aspect considering the pollution dynamics and might play a key role in the accumulation of dust and pollution during the dry season.

10. Appendix A: Web Links for Supplementary Material


The Indian Institute of Tropical Meteorology (IITM) and the CAIPEEX experiment are fully funded by the Ministry of Earth Sciences (MOES), Government of India, New Delhi. Computational support for this study is also provided through the CAIPEEX program. Sodar is funded by the Department of Science and Technology (DST, India). The authors thank Dr Anandakumar Karipot, University of Pune, for several useful comments and suggestions on the manuscript. The first author acknowledges discussions on the topic with Prof. Robert Houze and Dr Jimy Dudhia. The authors also thank two anonymous reviewers and the associate editor for several key suggestions, which improved the manuscript and presentation considerably. Mrs V. V. Sapre of IITM and Alan Norton of NCAR are thanked for efforts in making supplemental materials available online, with respective web links. This manuscript is dedicated to Prof. Anna Mani who made the first Indian wind atlas.