The Dominica Experiment (DOMEX) took place in the eastern Caribbean from 4 April to 10 May 2011 with 21 research flights of the Wyoming King Air and several other observing systems. The goal was an improved understanding of the physics of convective orographic precipitation in the tropics. Two types of convection were found. During a period of weak trade winds, diurnal thermal convection was seen over Dominica. This convection caused little precipitation but carried aloft air with island-derived aerosol and depleted CO2. During periods of strong trades, mechanically forced convection over the windward slopes brought heavy rain to the high terrain. This convection was “seeded” by trade-wind cumuli or neutrally buoyant cool wet patches of air. In this mechanically forced convection, air parcels did not touch the island surface to gain buoyancy so no island-derived tracers were lofted. With fewer aerosols, the mean cloud droplet diameter increased from 15 to 25 μm. Plunging airflow and a wake were found in the lee of Dominica. The DOMEX dataset will advance our understanding and test our theories of cumulus triggering and aerosol influence on precipitation.

The surprising sensitivity of the wind field, aerosol concentration, cloud physics, and precipitation over Dominica to trade-wind speed arises from a change from thermally to mechanically triggered convection and from nonplunging to plunging airflow.

Orographic precipitation influences water resources, flooding and landslides, regional climates, and global water budgets. Furthermore, the study of mountain-induced lifting and precipitation serves as a prototype for the study of other types of airmass lifting in the atmosphere (e.g., frontal, cold pool, dryline). Until now, orographic precipitation has been widely studied only in midlatitudes with field projects such as the Sierra Cooperative Pilot Project (SCPP), the Alpine Experiment (ALPEX), Southern ALPEX (Wratt et al. 1996), the Improvement of Microphysical Parameterization through Observational Verification Experiment (IMPROVE; Stoelinga et al. 2003), the Mesoscale Alpine Programme (MAP; Bougeault et al. 2001; Rotunno and Houze 2007), Cumulus Photogrammetric, In situ and Doppler Observations (CuPIDO; Geerts et al. 2008), and others. Summaries can be found in Smith (1979 in Smith (2006), Roe (2005), and Houze (2012). Mountain effects in the subtropics have been studied in Taiwan and Hawaii (e.g., Kuo and Chen 1990; Yang and Chen 2008). These locations still have some midlatitude climatic influences and under typical conditions airflow goes around rather than over these tall terrains. Progress has also been made on thermally driven island convection in the tropics (e.g., Carey and Rutledge 2000; Wilson et al. 2001; Sobel et al. 2011; Robinson et al. 2011). In contrast, the Dominica Experiment (DOMEX) was designed to study orographic precipitation in the tropics with cumulus triggering by forced ascent in unblocked flow.

Dominica (15°N, 61°W) lies in the eastern Caribbean in the Lesser Antilles chain of volcanic islands (Fig. 1). Its terrain is dominated by the 1.4-km peaks of Mt. Diablotin in the north and Trois Piton in the south. Between the peaks is a saddle with an elevation of 600 m. Orographic precipitation is a key issue for Dominica. It generates about half of its electricity from hydropower and there are plans to export bulk fresh water to drier Caribbean islands. Frequent heavy rain events cause flooding and landslides on the island. From many interesting sites in the tropics, Dominica was selected as our natural laboratory for the following reasons: heavy convective orographic precipitation (Fig. 2), steady trade-wind flow, unobstructed upstream environment, simple twin-peak terrain, two off-island French radars, nearby balloon soundings, and road access to high mountain sites. Previous studies of the tropical Atlantic in the Barbados Oceanographic and Meteorological Experiment (BOMEX; Holland and Rasmusson 1973), Rain in Cumulus over the Ocean (RICO; Rauber et al. 2007), and the Barbados Aerosol Cloud Experiment (BACEX; B. Albrecht 2012, personal communication) helped to define the upstream physical environment for Dominica.

Fig. 1.

Landsat-7 image of Dominica in Feb 2000 (day of year = 50). Airflow is from right to left. Cumulus rolls were amplified over the island.

Fig. 1.

Landsat-7 image of Dominica in Feb 2000 (day of year = 50). Airflow is from right to left. Cumulus rolls were amplified over the island.

Fig. 2.

Guadeloupe radar (TFFR) -derived map of annual rainfall (mm day−1) for Dominica for 2007 (from Smith et al. 2009a). Note the small upstream enhancement and dry rain shadow.

Fig. 2.

Guadeloupe radar (TFFR) -derived map of annual rainfall (mm day−1) for Dominica for 2007 (from Smith et al. 2009a). Note the small upstream enhancement and dry rain shadow.

The Dominica Experiment took place in two phases. From 2007 to 2011, data from installed rain gauges, the Guadeloupe radar, and soundings [Pointe-à- Pitre International Airport (TFFR)] were studied to understand the climatology of the island (Smith et al. 2009a). The orographic enhancement associated with the passage of Hurricane Dean was also studied with these instruments (Smith et al. 2009b). Modeling and theoretical studies were carried out as well (Kirshbaum and Smith 2009, hereafter KS09).

From these studies, several general attributes of the precipitation climate of Dominica were established. The orographic enhancement factor, defined as the ratio of precipitation on the high terrain to upstream, varies from 2 to 12 depending on conditions. The smaller factor applies to days with substantial rain over the ocean upwind. While enhancement begins a few kilometers upwind of the eastern coast, it is focused primarily on the highest terrain. The high peaks receive nearly 6 m of rain each year compared with 0.5 m upwind (Fig. 2). The rain typically comes in 10-min showers from liquid-only “warm” clouds extending to an altitude of 3–4 km. A very sharp rainfall gradient is present on the lee slopes. The western coast and the adjacent ocean in the lee of the terrain experience less than 20 cm of annual rainfall. This strong “rain shadow” is not caused by water vapor depletion as the annually averaged drying ratio (i.e., the ratio of total rainfall to incoming water vapor flux) is less than 0.5%. Instead, the rain shadow is probably caused by lee side descent, dry air entrainment, or increased static stability. While many larger tropical islands experience a diurnal cycle in precipitation driven by solar heating, this type of modulation is very small on Dominica. On normal to strong tradewind days, it seems that mechanically forced ascent rather than solar heating triggers the convection and precipitation. To address these and other issues, a second phase of DOMEX was designed: DOMEX-2011.

THE DOMEX-2011 FIELD CAMPAIGN.

The goals of DOMEX-2011 were 1) to understand the physics of mountain-triggered convection and precipitation in the tropics, using Dominica as a natural laboratory; 2) to develop datasets that can be used to test and improve numerical models of convection and precipitation in the tropics; and 3) to better understand and predict the weather and climate of the Lesser Antilles including Guadeloupe, Dominica, and Martinique. The DOMEX-2011 field phase (4 April–10 May 2011) in the drier season allowed a focus on orographic precipitation without the complications associated with organized tropical disturbances. The key observing system was the Wyoming King Air research aircraft with the capability to measure flight-level winds, thermodynamics, carbon dioxide, aerosol, and cloud particles. These quantities were recorded at a rate of 25 Hz which, with an air speed of 90 m s−1, gave a sampling distance of about 4 m. In addition, the King Air carried the 95-GHz Wyoming Cloud Radar (WCR) and the 355-nm Wyoming Cloud Lidar (WCL; Wang et al. 2009). The National Science Foundation allocated 80 flight hours to DOMEX-2011 allowing 21 flights (Table 1). The King Air flew out of the Aime Cesaire airport in Lamentin on nearby Martinique.

Table 1.

Typical flow parameters and characteristics for low and high wind cases. The bold numbers indicate distinctly larger values than the numbers in their opposite wind speed category.

Typical flow parameters and characteristics for low and high wind cases. The bold numbers indicate distinctly larger values than the numbers in their opposite wind speed category.
Typical flow parameters and characteristics for low and high wind cases. The bold numbers indicate distinctly larger values than the numbers in their opposite wind speed category.

The King Air flight pattern consisted of an upstream sounding and six horizontal legs (Fig. 3). An upwind sounding was taken at the start of each flight from an altitude of 4000 m down to 150 m. Legs 1, 2, and 5 were flown over the ocean at two altitudes: 300 and 1200 m. Leg 1 observed undisturbed upstream conditions while leg 2 observed any changes in the airflow reaching the Dominican coast. Leg 2 had a “dogleg” shape to keep it close to the coast.

Fig. 3.

Dominica terrain, deployed instruments, and King Air flight tracks 1–6. The altitudes of each leg are given. Typical order of legs was 1L, 2L, 1H, 2H, 3, 4, 3, 4, 5, 6. Labels H and L refer to higher and lower legs.

Fig. 3.

Dominica terrain, deployed instruments, and King Air flight tracks 1–6. The altitudes of each leg are given. Typical order of legs was 1L, 2L, 1H, 2H, 3, 4, 3, 4, 5, 6. Labels H and L refer to higher and lower legs.

The surface instrumentation on Dominica was enhanced for DOMEX-2011. The number of rain gauges across the southern peaks was increased to 10. A satellite-linked weather station was installed on an east-facing ridge near Freshwater Lake (Fig. 3). A small section of hilltop forest was cleared to provide acceptable wind observations with a 6-m tower. Standard meteorological variables were recorded with 2-min resolution. Three web cameras were installed on the eastern and western coasts for continuous time lapse photography with 30-s resolution. The camera at Rosalie on the eastern coast faced eastward. The two cameras in Roseau on the western coast faced eastward and northward. A Joss–Waldvogel disdrometer was installed at Rosalie. With the assistance of the National Center for Atmospheric Research (NCAR), Geostationary Operational Environmental Satellite (GOES) sector images were captured and archived. The Météo-France Guadeloupe balloon sounding frequency was increased from once to twice per day. Data from four Météo-France surface stations were archived with 6-min time resolution. Scans from Guadeloupe and Martinique 2.8-GHz weather radars were recorded every 5 min. Hourly surface meteorological data are available from the two airports on Dominica: Melville Hall and Canefield.

ROLE OF TRADE-WIND SPEED.

The 21 cases flown with the King Air varied considerably in their ambient trade-wind speed, wind direction, humidity, upstream precipitation, and other parameters (Fig. 4). Any attempt to fully categorize these cases would probably require a multidimensional parameter space. To a first approximation, however, it appears that the wind speed is the dominant upstream control parameter. To illustrate the role of wind speed, we plot four over-island parameters versus upstream wind speed in Fig. 5. The standard deviation of vertical wind speed over the eastern slope (Fig. 5a; leg 3) shows a nonmonotonic behavior. We interpret this pattern as evidence of two modes of convection. For wind less than 5 m s−1, the convection is thermally driven. For winds between 5 and 8 m s−1, the convection is weak. We propose that the wind suppresses the thermal convection while the wind is still too weak to mechanically force energetic convection. When the wind exceeds 8 m s−1, the mechanically forced convection is strong. The straight line in Fig. 5a shows a possible relationship between wind speed and convective amplitude by mechanical forcing as proposed by KS09.

Fig. 4.

Temperature, precipitation, and trade-wind speed during DOMEX-2011. Solid curves are from the FWL mountain weather station. The symbols are from upstream aircraft data. Dashed green line is the average precipitation rate over Dominica derived from the Guadeloupe radar. Note the period 14–20 Apr with weak wind and precipitation and increased diurnal temperature range.

Fig. 4.

Temperature, precipitation, and trade-wind speed during DOMEX-2011. Solid curves are from the FWL mountain weather station. The symbols are from upstream aircraft data. Dashed green line is the average precipitation rate over Dominica derived from the Guadeloupe radar. Note the period 14–20 Apr with weak wind and precipitation and increased diurnal temperature range.

Fig. 5.

Role of ambient trade-wind speed in (a) convection intensity on leg 3 (linear fit for U > 5 only), (b) aerosol number concentration on leg 3, (c) cloud particle number concentration on leg 3, and (d) upper wind divergence between legs 3 and 4 at z = 1700 m.

Fig. 5.

Role of ambient trade-wind speed in (a) convection intensity on leg 3 (linear fit for U > 5 only), (b) aerosol number concentration on leg 3, (c) cloud particle number concentration on leg 3, and (d) upper wind divergence between legs 3 and 4 at z = 1700 m.

The aerosol and cloud number concentrations trend downward with wind speed in Figs. 5b and 5c. As shown below, this trend is probably due to island-derived aerosol carried upward in the thermal convection. Mechanically driven convection in the strong wind cases does not elevate this aerosol. In Fig. 5d, the difference in wind between legs 3 and 4 at z = 1700 m is plotted versus wind speed. In low wind cases, the difference is positive, indicating diverging flow from the cloud top detrainment. In high wind cases, the difference is negative, indicating that the easterly flow found on leg 3 plunges beneath leg 4.

To further illustrate the role of wind speed, we have identified four high wind cases (U > 8 m s−1) and two low wind cases (U < 5 m s−1) that have common characteristics. Their general flow properties are summarized in Table 1.

The upstream wind direction, dry and moist stabilities, and reverse shear are similar for the low and high wind cases. The dry Brunt–Vaisala frequency above cloud base is about N = 0.012 s−1. The equivalent potential temperature decreases by about 20° from z = 500 to 3500 m, indicating conditional instability. The upstream conditions differ most in the low-level wind speed and the height of the wind reversal.

The nondimensional mountain height ĥ = Nh/U is 0.9 in the high wind case and 3.2 in the low wind case. In the latter case one would expect the airflow to deflect around the mountain (Smith 1979), but this is not observed. Perhaps the thermal heating of the slopes or the latent heat aloft draws air up the eastern slopes and keeps the air moving towards the eastern coast of the island (Reisner and Smolarkiewicz 1994). The adiabatic layer beneath cloud base may also help the air to lift and avoid deceleration.

The convection over the island seems to be of two types, as shown in Table 1. The cloud fraction is greater at flight level over the eastern slopes (leg 3) for the high wind case but greater over the western slopes (leg 4) for the low wind case. This shift in convection location is not clear in the standard deviation of w′, as clear air turbulence associated with plunging flow on leg 4 can be as strong as the convection. The shift in convection location is more clearly seen in vertical velocity skewness S = ∫(w′)3dx/[∫(w′)2dx]3/2 and the latent heat flux (Table 1). Positive skewness indicates that the updrafts are narrow and strong with weak broad downdrafts, which is a characteristic of moist convection. The legs with larger skewness also have larger latent heat flux, confirming the shift in convection with wind speed (Table 1).

It is perhaps counterintuitive that the weaker wind case would push the convection farther west than the high wind case, but this discovery reinforces the idea that two different convection mechanisms are at work. The strong wind case generates convection over the windward slope (leg 3) with the mechanical ascent mechanism and suppresses lee convection with descent. In the weak wind case, warm thermal boundary layer flow is pushed westward to the vicinity of leg 4 by the prevailing winds. Judging from the latent heat fluxes, the weak wind convection is at least as strong as the strong wind convection.

Another distinct difference between the high and low wind cases is seen in the cloud particle properties (Table 1). In the weak wind cases, due to lofted island aerosol, the cloud number concentration is large and the mean droplet size is small. The smaller droplet size could explain the lighter precipitation seen in the weak wind cases. It may be more difficult for collision–coalescence to grow raindrops (e.g., Warner 1968; Gunn and Phillips 1957; Blanchard and Spencer 1957). This difference in droplet size is similar to those found in recent numerical simulations of clouds in clean and polluted air (Morrison and Grabowski 2007; Muhlbauer and Lohmann 2008). In this environment, however, the effects of hygroscopic nuclei (e.g., Jensen and Lee 2008) and turbulence (e.g., Wang and Grabowski 2009) must also be considered. Differences in cloud depth and dry air entrainment should also be considered.

The role of wind speed can be further illustrated by comparing two flights with contrasting wind speed. Research flight 7 on 18 April had weak trade winds of about 2 m s−1 with westerlies above 1500 m (Fig. 6a). Research flight 13 on 27 April had stronger trade winds of nearly 12 m s−1 with westerlies above 3500 m (Fig. 6b). Photographs of the two cloud systems are shown in Fig. 7 and the precipitation patterns are in Fig. 8. Aloft at 1,700 m, the two cases had different wind fields. In the weak wind case, legs 3 and 4 show a divergence associated with detrainment from the cumulus line (Fig. 9a). In the fast wind case, legs 3 and 4 show a convergence (Fig. 9b). The fast air found along leg 3 drops below leg 4 and stagnant and slow turbulent air is found on leg 4 (Smith 1987).

Fig. 6.

Upstream aircraft soundings for the low wind (RF07) and high wind (RF13) cases: (a) cross-mountain wind speed and (b) potential and equivalent potential temperature. Note the reverse shear and the conditional instability.

Fig. 6.

Upstream aircraft soundings for the low wind (RF07) and high wind (RF13) cases: (a) cross-mountain wind speed and (b) potential and equivalent potential temperature. Note the reverse shear and the conditional instability.

Fig. 7.

Photographs of Dominica convection taken from the King Air for a weak and strong wind case, (a) RF07 and (b) RF13, respectively.

Fig. 7.

Photographs of Dominica convection taken from the King Air for a weak and strong wind case, (a) RF07 and (b) RF13, respectively.

Fig. 8.

Precipitation over Dominica for the 4-h flight period from the Guadeloupe radar: (a) low wind case (RF07) and (b) high wind case (RF13). Note change in color scale.

Fig. 8.

Precipitation over Dominica for the 4-h flight period from the Guadeloupe radar: (a) low wind case (RF07) and (b) high wind case (RF13). Note change in color scale.

Fig. 9.

Horizontal winds at z = 1700 m for (a) a low wind case (RF07) and (b) a high wind case (RF13). With low ambient winds, winds diverge from convective detrainment whereas with high ambient winds, winds appear to converge flow from lee-side plunging.

Fig. 9.

Horizontal winds at z = 1700 m for (a) a low wind case (RF07) and (b) a high wind case (RF13). With low ambient winds, winds diverge from convective detrainment whereas with high ambient winds, winds appear to converge flow from lee-side plunging.

The Wyoming Cloud Radar shows the vertical distribution of large cloud droplets and small raindrops. Over the eastern windward slope (leg 3), the radar shows more cells and stronger reflectivity on the strong wind day (Fig. 10). Over the western slopes (leg 4), the radar shows weak updrafts on the weak wind day but suppressed convection and a 600-m-deep layer of spillover rain on the strong wind day (Fig. 11).

Fig. 10.

Reflectivity from the 95-GHz Wyoming Cloud Radar on upwind leg 3, for (a) a low wind case (RF07) and (b) a high wind case (RF13). The high wind case has much stronger reflectivity but some beam attenuation too.

Fig. 10.

Reflectivity from the 95-GHz Wyoming Cloud Radar on upwind leg 3, for (a) a low wind case (RF07) and (b) a high wind case (RF13). The high wind case has much stronger reflectivity but some beam attenuation too.

Fig. 11.

Reflectivity from the 95-GHz Wyoming Cloud Radar on downwind leg 4, for (a) a low wind case (RF07) and (b) a high wind case (RF13). The high wind case shows spillover and “book end” convection.

Fig. 11.

Reflectivity from the 95-GHz Wyoming Cloud Radar on downwind leg 4, for (a) a low wind case (RF07) and (b) a high wind case (RF13). The high wind case shows spillover and “book end” convection.

The mean droplet size in high wind case (RF13) is compared with the slow wind case (RF07) in Fig. 12. This diagram includes all the cloud penetrations over the island for each flight. The probability peak occurs at diameters of 15 and 25 mm for the slow and fast wind cases, respectively. We argue below that the droplet size is influenced by island-derived aerosols.

Fig. 12.

The probability density for mean droplet size for the low wind (RF07) and high wind (RF13) cases. The breadth of each peak is mostly due to entrainment. The shift arises from island source aerosols on the low wind day.

Fig. 12.

The probability density for mean droplet size for the low wind (RF07) and high wind (RF13) cases. The breadth of each peak is mostly due to entrainment. The shift arises from island source aerosols on the low wind day.

ISLAND-DERIVED TRACERS.

The aerosol distribution along legs 3 and 4 is shown in Fig. 13 for the weak and high wind examples (i.e., RF07 and RF13). In this figure, the aerosol number concentration and CO2 concentration are plotted versus time while the aircraft flies two complete “racetrack” circuits around legs 3, 4, 3, and 4 at z = 1700 m (see Fig. 3). Brief cloud penetrations are excluded from these plots so the data represent air outside of clouds. At the northern and southern ends of the racetrack, the aircraft is outside of the influence of the island, so the observations represent the undisturbed environment. In the middle of legs 3 and 4, the aircraft flies through the detraining air from clouds over the island. For RF07, the aerosol number concentration over the island rises well above the ambient level whereas for RF13 the aerosol concentration is constant along each leg.

Fig. 13.

CO2 and aerosol for the racetrack leg sequence 3, 4, 3, 4 at z = 1700 m, outside of cloud, for (a) a low wind case (RF07) and (b) a high wind case (RF13). Leg numbers indicate the middle of each leg when the aircraft is over the island. Island terrain is shown at the bottom. On the low wind day, the convection seems to be detraining aerosol enriched and CO2-depleted air.

Fig. 13.

CO2 and aerosol for the racetrack leg sequence 3, 4, 3, 4 at z = 1700 m, outside of cloud, for (a) a low wind case (RF07) and (b) a high wind case (RF13). Leg numbers indicate the middle of each leg when the aircraft is over the island. Island terrain is shown at the bottom. On the low wind day, the convection seems to be detraining aerosol enriched and CO2-depleted air.

Our interpretation is as follows. In the case of thermal convection in weak ambient wind (e.g., RF07), the air detraining from the plume must have come in contact with the island surface, or at least a shallow island boundary layer, in order to gain sensible heat and buoyancy. Using a diffusivity K ≈ 100 m2 s−1, we estimate that the depth d of this internal boundary layer grows downwind of the coast and reaches d = (Kx/U)1/2 < 150 m. Aerosols released from the island surface thereby provide a tracer of thermally driven convection. Conversely, in the case of mechanically driven convection during high winds (e.g., RF13), the plume buoyancy arises “internally” from latent heat release in moist parcels several hundred meters above the island surface. Air in these rising plumes shows no indication of contact with the island surface.

The carbon dioxide concentration provides an independent indication of air with recent contact with the island surface (Fig. 13). In the detraining air over the island on legs 3 and 4 on RF07, the CO2 concentration drops below ambient by about 2 ppmv. We interpret this drop as evidence of air that has recently lost CO2 by forest uptake on Dominica. A nearly identical pattern of island tracers is seen on the other weak wind case RF08. The other strong wind cases (i.e., RF12, RF16, and RF17) show no island-derived tracers at 1,700 m.

UPSTREAM SEEDS OF MECHANICALLY FORCED CONVECTION.

Our theory of mechanically driven convection relies on the idea that the ambient upstream air contains clouds and other fluctuations in humidity (Woodcock 1960; KS09). When the airstream is quickly lifted by the terrain, cloudy and dry air parcels ascend along different adiabats. After a few hundred meters of lift, significant differential buoyancy is generated and strong ascending plumes are created. With the aircraft data on legs 1L and 1H in DOMEX-2011, we have the possibility of observing the humidity seeds upstream of the island and possibly deducing their origin.

The upstream horizontal aircraft legs in DOMEX-2011 are similar to many over-ocean research flights in previous projects (e.g., Bean et al. 1972; Donelan and Miyake 1973). Our preliminary analyses of these legs confirm three interesting features seen in these earlier studies. First, while the latent heat fluxes are all positive, most of our sensible heat fluxes at 300 m are negative (e.g., Nicholls and Lemone 1980; Siebesma et al. 2003). If the convection is locally driven, this result implies that the buoyancy force driving the subcloud convection comes from the difference in molecular weight between water vapor and air (i.e., the “virtual effect”). This is “compositional” convection rather than the usual “thermal” convection. Second, the horizontal scales of vertical velocity fluctuations differ greatly from the dominant scales of temperature and water vapor fluctuations (Nicholls and Lemone 1980). Using lag autocorrelation, we found w scales of 500–1,000 m but T scales and q scales of 3–8 km (i.e., almost a factor of 10 difference). Third, in the upper part of the subcloud layer, the temperature and specific humidity fluctuations are strongly anticorrelated (e.g., Shinoda et al. 2009). One explanation for this observation involves turbulent entrainment of potentially warm dry air at the top of the mixed layer, but Siebesma et al. (2003) did not see this in their marine boundary layer simulations. Alternatively, evaporation of falling rain might create cool moist patches (Paluch and Lenschow 1991).

In this third respect, however, our results differ from previous results. We note not only that T and q are anticorrelated, but their contributions to virtual temperature and air density nearly cancel. This is shown in Fig. 14, where the scatter of points lies along a reference line for constant virtual temperature (TV = T + 0.61q). Also, there is little vertical velocity on these scales. We wonder whether these neutral buoyancy patches are the remnants of old detrainment or cold pool events that underwent “buoyancy sorting.” Dense parcels fell and less dense parcels rose until lateral buoyancy variations vanished. Once this adjustment has occurred, the patches drift along as passive tracers. Upon orographic lifting, they become dynamically active as the wet patches develop positive buoyancy. Their initial scale over the sea (i.e., 3–8 km) may determine the scale of convection over Dominica.

Fig. 14.

Scatter diagram for temperature and specific humidity along leg 1L at z = 300 m for RF07. The reference line represents constant virtual temperature and therefore constant buoyancy. The cooler, wetter parcels in the lower right of the cluster may be the seeds of the island convection. The steep line with slope L/Cp indicates how the points would scatter if the cause of the variation was evaporating rainfall.

Fig. 14.

Scatter diagram for temperature and specific humidity along leg 1L at z = 300 m for RF07. The reference line represents constant virtual temperature and therefore constant buoyancy. The cooler, wetter parcels in the lower right of the cluster may be the seeds of the island convection. The steep line with slope L/Cp indicates how the points would scatter if the cause of the variation was evaporating rainfall.

THE PLUNGE AND THE WAKE.

Over the lee slope of the island, clear evidence of descent and acceleration was found, even though the aircraft could not fly safely at low altitude. Clues to the descent were 1) sloping cloud tops, 2) slow reverse flow along leg 4 indicating that the main airstream has plunged below, 3) clear air turbulence on leg 4 from the strong reverse wind shear below, 4) cloud clearing near the coast, 5) dual Doppler measurements of accelerated flow below leg 6 over the lee slope, 6) a rise in q on legs 4 and 6 associated with descent, 7) spillover of rain from upwind, and 8) convergence of the horizontal wind vectors between legs 3 and 4. By these criteria, plunging air was found on at least four of the DOMEX flights (RF12, RF13, RF16, and RF17). The Roseau cameras show a foehn-wall cloud over the ridgeline on those days, and a shallow high-speed layer of air moving downslope. On occasion, the plunging easterly flow persists all the way to Canefield airport on the western coast of Dominica (e.g., the period 1400–2200 UTC 27 April 2011 including flight RF13).

The wake legs 5L and 5H revealed wake structures on most strong flow flights. There were usually two or three wakes, each from a different mountain peak on Dominica. In addition, the air outside the wake was accelerated by several meters per second relative to the upstream wind speed. Preliminary analysis seems to confirm the hypothesis of Smith et al. (1997) that wakes have constant pressure but a Bernoulli function deficit. After correcting the pressure for diurnal and semidiurnal tidal oscillations and GPS-derived aircraft altitude changes, the pressure appears generally depressed to the west of Dominica but constant across the wakes themselves. This observation provides a partial explanation of both the general regional acceleration and the wake itself. The wake also has a deficit in carbon dioxide of 2 ppmv, indicating that wake air has touched the forests of Dominica.

CONCLUSIONS.

The strategy of using similar flight patterns around Dominica in different ambient flows helped us identify the role of trade-wind speed. Two types of convection occur over Dominica, depending on the trade-wind speed. Under weak trades (U < 5 m s−1), diurnal thermal convection occurs with the strongest updrafts located over the ridgeline and lee slopes (Fig. 15a). Under strong trades (U > 7 m s−1), mechanically driven convection occurs with the strongest updrafts over the windward slopes (Fig. 15b). Under intermediate flow speeds (5 < U < 7 m s−1) the thermal convection is suppressed by the trade winds but the mechanically forced convection is still weak.

Fig. 15.

Schematic of the two types of convection found over Dominica: (a) diurnal thermal convection found with weak trade winds, and (b) mechanically driven convection found with strong trade winds. Heavy up arrows are the strongest updrafts. The heavy down arrow and curved dotted streamline show the plunging flow over the lee slopes. Thin arrows show the origin of buoyancy. Filled dots are aircraft legs into the page. Symbols S, T, and W describe measured properties of the flow: S = humid seeds for convection; T = island derived tracer; W = wake. Inverted Vs indicate clear-air turbulence.

Fig. 15.

Schematic of the two types of convection found over Dominica: (a) diurnal thermal convection found with weak trade winds, and (b) mechanically driven convection found with strong trade winds. Heavy up arrows are the strongest updrafts. The heavy down arrow and curved dotted streamline show the plunging flow over the lee slopes. Thin arrows show the origin of buoyancy. Filled dots are aircraft legs into the page. Symbols S, T, and W describe measured properties of the flow: S = humid seeds for convection; T = island derived tracer; W = wake. Inverted Vs indicate clear-air turbulence.

In the strong wind cases, the mechanically forced convection builds over the windward slope and brings heavy rain to the highest terrain. These clouds are quickly dissipated over the lee slope by plunging airflow. The lee slope environment is characterized by fast descending flow, clear air turbulence, and raindrops spilling over from the windward side convection. Near the ridge ends, convection is still generated by the terrain but the plunging flow is absent, causing “book-end” convection. The mechanically forced convection is triggered either by upstream cumulus clouds or upstream subcloud moisture anomalies (KS09). The subcloud moisture anomalies are neutral patches with compensating temperature and humidity so that the virtual temperature is constant. We speculate that these patches result from buoyancy sorting in the upwind oceanic trade-wind boundary layer.

Surprisingly, the nature of the convective triggering is reflected in the transport of island-source tracers. On strong wind days, plume buoyancy is probably generated a few hundred meters above the Earth's surface by the orographic lifting of ambient moisture anomalies. As these buoyant parcels never enter the island internal boundary layer, they do not gather island source tracers. In contrast, on weak wind days, plume parcels gain their buoyancy from the sun-heated island. Air detraining from these plumes carries high aerosol and reduced carbon dioxide concentrations derived from the island surface.

Even more surprising is the impact of the aerosol tracer on cloud microphysics and precipitation. On weak wind days, with high island-derived aerosol concentration, the cloud droplet number density is much larger and the mean cloud droplet diameter much smaller than on the high-wind low-aerosol days (i.e., 15 vs 25 μm). As a result perhaps, little precipitation falls from the convective clouds on the low wind days. This result is consistent with the lack of a climatological diurnal cycle in precipitation. While low wind days allow a diurnal cycle of convection to occur over Dominica, they contribute minimally to the annual rainfall amount. Other factors that might suppress precipitation on weak wind days are drier conditions aloft, reduced sea salt nuclei, or the limited frequency of weak wind days.

Many questions remain unanswered about Dominica's airflow, clouds, and precipitation. What is the role of upstream precipitation? Does the convection seeding come from the cloud or subcloud layer? What determines the scale of the convection? Does the plunging require a critical level in the upwind flow? How does the plunge dissipate the convection? What is the source of the island derived aerosol? How does the aerosol suppress precipitation? How is the wake generated? Our answers to these questions may improve short-term forecasting in the eastern Caribbean.

A long-term objective is to compare DOMEX results with orographic precipitation elsewhere in the tropics. Potential analogues to Dominica include the Central American Cordillera, the Western Ghats of India, the coastal range of Myanmar, the Philippines, Sumatra, and Madagascar. There may also be analogues in the midlatitudes where conditionally unstable airstreams are lifted by terrain or by other means.

ACKNOWLEDGMENTS

The DOMEX project was funded through the AGS Division at the (U.S.) National Science Foundation with advice from Brad Smull. The King Air participation was funded through the NSF deployment pool. The Yale staff were supported by a grant to Yale University (0930356). The installation of instruments in Dominica was assisted by the Forestry Division (David Williams, Ashton Lugay), Met Office (Marshall Alexander), Agriculture (Rickey Brumant), Fisheries (Andrew Magloire, Derrick Theophile), and the Environmental Coordinating Unit (Collin Guiste), Grand Fond School (Mrs. Verna Frederick), and Springfield Plantation (Nancy Osler). A camera and disdrometer were installed at Rosalie Bay Resort (Jan Neel, Ravey Sanchez). The disdrometer was loaned to us by Sandra Yuter. The Ports Authority (Benoit Bardouille) gave permission for flight operations over the island. On Martinique, Météo-France supported DOMEX in several ways including surface and radar data, daily weather briefings, extra 0000 UTC soundings, and office space. Thanks to Patrick Van Grunderbeeck, Daniele Carnino, Christian Massip, Niels Oger, Eloi Regina, and others. The Wyoming King Air staff (Larry Oolman, Brett Wadsworth, Ahmad Bandani, Sam Haimov, and others) provided a high level of support to the project. The aircraft and the airborne instruments worked reliably and the staff helped to solve several problems regarding instrument selection, flight track design, ATC clearances, and variable weather conditions. Larry Bonneau at the Yale Center for Earth Observation helped with satellite images, maps, and webcam installation. Greg Stossmeister and Glenn Randers-Pehrson assisted with GOES image archiving and animation. Bill Boos made useful suggestions for the research. Sigrid R.-P. Smith assisted with logistics.

References

References
Bean
,
B. R.
,
R.
Gilmer
,
R. L.
Grossman
,
R.
McGavin
, and
C.
Travis
,
1972
:
Analysis of airborne measurements of vertical water vapor flux during BOMEX
.
J. Atmos. Sci.
,
29
,
860
869
.
Blanchard
,
D. C.
, and
A. T.
Spencer
,
1957
:
Condensation nuclei in the vicinity of the island of Hawaii
.
Tellus
,
9
,
525
527
.
Bougeault
,
P.
,
and Coauthors
,
2001
:
The MAP Special Observing Period
.
Bull. Amer. Meteor. Soc.
,
82
,
433
462
.
Carey
,
L. D.
, and
S. A.
Rutledge
,
2000
:
The relationship between precipitation and lightning in tropical island convection: A C-band polarimetric radar study
.
Mon. Wea. Rev.
,
128
,
2687
2710
.
Donelan
,
M.
, and
M.
Miyake
,
1973
:
Spectra and fluxes in the boundary layer of the trade-wind zone
.
J. Atmos. Sci.
,
30
,
444
464
.
Geerts
,
B.
,
Q.
Miao
, and
J. C.
Demko
,
2008
:
Pressure perturbations and upslope flow over a heated, isolated mountain
.
Mon. Wea. Rev.
,
136
,
4272
4288
.
Gunn
,
R.
, and
B. B.
Phillips
,
1957
:
An experimental investigation of the effect of air pollution on the initiation of rain
.
J. Meteor.
,
14
,
272
280
.
Holland
,
J. Z.
, and
E. M.
Rasmusson
,
1973
:
Measurements of the atmospheric mass, energy, and momentum budgets over a 500-kilometer square of tropical ocean
.
Mon. Wea. Rev.
,
101
,
44
55
.
Houze
,
R. A.
,
2012
:
Orographic effects on precipitating clouds
.
Rev. Geophys.
,
50
,
RG1001
,
doi:10.1029/2011RG000365
.
Jensen
,
J. B.
, and
S.
Lee
,
2008
:
Giant sea-salt aerosols and warm rain formation in marine stratocumulus
.
J. Atmos. Sci.
,
65
,
3678
3694
.
Kirshbaum
,
D. J.
, and
R. B.
Smith
,
2009
:
Orographic precipitation in the tropics: Large-eddy simulations and theory
.
J. Atmos. Sci.
,
66
,
2559
2578
.
Kuo
,
Y.-H.
, and
G. T.-J.
Chen
,
1990
:
The Taiwan Area Mesoscale Experiment (TAMEX): An overview
.
Bull. Amer. Meteor. Soc.
,
71
,
488
503
.
Morrison
,
H.
, and
W. W.
Grabowski
,
2007
:
Comparison of bulk and bin warm-rain microphysics models using a kinematic framework
.
J. Atmos. Sci.
,
64
,
2839
2861
.
Muhlbauer
,
A.
, and
U.
Lohmann
,
2008
:
Sensitivity studies of the role of aerosols in warm-phase orographic precipitation in different dynamical flow regimes
.
J. Atmos. Sci.
,
65
,
2522
2542
.
Nicholls
,
S.
, and
M. A.
Lemone
,
1980
:
The fair weather boundary layer in GATE: The relationship of subcloud fluxes and structure to the distribution and enhancement of cumulus clouds
.
J. Atmos. Sci.
,
37
,
2051
2067
.
Paluch
,
I. R.
, and
D. H.
Lenschow
,
1991
:
Stratiform cloud formation in the marine boundary layer
.
J. Atmos. Sci.
,
48
,
2141
2158
.
Rauber
,
R. M.
,
and Coauthors
,
2007
:
Rain in shallow cumulus over the ocean: The RICO campaign
.
Bull. Amer. Meteor. Soc.
,
88
,
1912
1928
.
Reisner
,
J. M.
, and
P. K.
Smolarkiewicz
,
1994
:
Thermally forced low Froude number past three-dimensional obstacles
.
J. Atmos. Sci.
,
51
,
117
133
.
Robinson
,
F. J.
,
S. C.
Sherwood
,
D.
Gerstle
,
C.
Liu
, and
D. J.
Kirshbaum
,
2011
:
Exploring the land–ocean contrast in convective vigor using islands
.
J. Atmos. Sci.
,
68
,
602
618
.
Roe
,
G. H.
,
2005
:
Orographic precipitation
.
Annu. Rev. Earth Planet. Sci.
,
33
,
645
671
.
Rotunno
,
R.
, and
R. A.
Houze
,
2007
:
Lessons on orographic precipitation from the Mesoscale Alpine Programme
.
Quart. J. Roy. Meteor. Soc.
,
133
,
811
830
.
Shinoda
,
T.
,
and Coauthors
,
2009
:
Structure of convective circulation in the atmospheric boundary layer over the northwestern Pacific Ocean under a subtropical high
.
J. Meteor. Soc. Japan
,
87
,
979
996
.
Siebesma
,
A. P.
,
and Coauthors
,
2003
:
A large eddy simulation intercomparison study of shallow cumulus convection
.
J. Atmos. Sci.
,
60
,
1201
1219
.
Smith
,
R. B.
,
1979
:
The influence of mountains on the atmosphere
.
Advances in Geophysics
,
Vol. 21
,
Academic Press
,
87
230
.
Smith
,
R. B.
,
1987
:
Aerial observations of the Yugoslavian bora
.
J. Atmos. Sci.
,
44
,
269
297
.
Smith
,
R. B.
,
2006
:
Progress on the theory of orographic precipitation
.
Tectonics, Climate, and Landscape Evolution
,
Willett
S. D.
et al., Eds., Geological Society of America Special Paper 398
,
1
16
.
Smith
,
R. B.
,
A. C.
Gleason
,
P. A.
Gluhosky
, and
V.
Grubisic
,
1997
:
The wake of St. Vincent
.
J. Atmos. Sci.
,
54
,
606
623
.
Smith
,
R. B.
,
P.
Schafer
,
D. J.
Kirshbaum
, and
E.
Regina
,
2009a
:
Orographic precipitation in the tropics: Experiments in Dominica
.
J. Atmos. Sci.
,
66
,
1698
1716
.
Smith
,
R. B.
,
P.
Schafer
,
D. J.
Kirshbaum
, and
E.
Regina
,
2009b
:
Orographic enhancement of precipitation inside Hurricane Dean
.
J. Hydrometeor.
,
10
,
820
831
.
Sobel
,
A. H.
,
C. D.
Burleyson
, and
S. E.
Yuter
,
2011
:
Rain on small tropical islands
.
J. Geophys. Res.
,
116
,
D08102
,
doi:10.1029/2010JD014695
.
Stoelinga
,
M. T.
,
and Coauthors
,
2003
:
Improvement of Microphysical Parameterization through Observational Verification Experiment
.
Bull. Amer. Meteor. Soc.
,
84
,
1807
1826
.
Wang
,
L.-P.
, and
W. W.
Grabowski
,
2009
:
The role of air turbulence in warm rain initiation
.
Atmos. Sci. Lett.
,
10
,
1
8
.
Wang
,
Z.
,
P.
Wechsler
,
W.
Kuester
,
J.
French
,
A.
Rodi
,
B.
Glover
,
M.
Burkhart
, and
D.
Lukens
,
2009
:
Wyoming Cloud Lidar: Instrument description and applications
.
Opt. Express
,
17
,
13 576
13 587
.
Warner
,
J.
,
1968
:
A reduction in rainfall associated with smoke from sugar-cane fires—An inadvertent weather modification?
J. Appl. Meteor.
,
7
,
247
251
.
Wilson
,
J. W.
,
R. E.
Carbone
,
J.
Tuttle
, and
T. D.
Keenan
,
2001
:
Tropical island convection in the absence of significant topography. Part II: Nowcasting storm evolution
.
Mon. Wea. Rev.
,
129
,
1637
1655
.
Woodcock
,
A. H.
,
1960
:
The origin of trade-wind orographic shower rains
.
Tellus
,
12
,
315
326
.
Wratt
,
D. S.
,
and Coauthors
,
1996
:
The New Zealand Southern Alps experiment
.
Bull. Amer. Meteor. Soc.
,
77
,
683
692
.
Yang
,
Y.
, and
Y.-L.
Chen
,
2008
:
Effects of terrain heights and sizes on island-scale circulations and rainfall for the island of Hawaii during HaRP
.
Mon. Wea. Rev.
,
136
,
120
146
.

Footnotes

A supplement to this article is available online (10.1175/BAMS-D-11-00194.2)

Supplemental Material