1. Motivation
Orographic precipitation supplies mountain glaciers and rivers and provides water for irrigation, hydropower, and human consumption. Air mass drying by orographic precipitation reduces the humidity and precipitation in downwind regions. Orographic precipitation is thought to occur in all latitudes and climate zones on earth, but the physical mechanisms may vary. Many examples of orographic precipitation have been studied over the last 50 yr [see reviews by Banta (1990), Smith (2006), and Rotunno and Houze (2007), among others], but almost all these studies have been in midlatitudes. A common element of previous studies is that orographic precipitation events were forced, either by a weather disturbance that is already precipitating, (e.g., frontal cyclone, squall line, easterly wave, or hurricane) or by the diurnal cycle of solar heating. We have two objectives: (i) to examine the physics of orographic precipitation in the tropics and (ii) to identify a location with pure orographic precipitation with a dominance of mechanical lifting and where neither weather disturbances nor diurnal forcing are needed.
One possible location for pure tropical orographic precipitation is the big island of Hawaii at 20°N (e.g., Woodcock 1960; Esteban and Chen 2008). During the long warm season, precipitation is found almost every day on the windward (east) coast. There is, however, a distinct diurnal cycle to the precipitation there and arguments persist about the relative importance of forced lifting by terrain and thermally induced circulation as a cause of precipitation. The mountain peaks are so high, and the trade wind inversion so strong, that most air goes around the island rather than over it. On the sheltered west coast, a diurnal land–sea breeze occurs in the warm season. In wintertime, cyclonic disturbances influence precipitation.
A second well-studied place is Taiwan at 23°N latitude (e.g., Yeh and Chen 1998). Although it does not lie in the trade wind belt, it is subject to nearly steady periods of monsoon wind. On the whole, however, the location and size of Taiwan give it a complex precipitation pattern, including frequent cyclonic disturbances and a diurnal cycle.
Our search for a pure example of tropical orographic precipitation led us to the island of Dominica at 15°N in the West Indies. It has a strong orographic enhancement and the processes involved repeat themselves day after day. Another motivation for a Dominica study is the dominance of orographically triggered convection there. For the most part, previous studies of orographic precipitation have ignored the details of triggered convection because neither the observations nor the models could resolve it. Some relevant exceptions in midlatitude observations are Browning et al. (1974) for South Wales, Smith et al. (2003) for the Alps, and Colle et al. (2008) in western North America. No such observations are known for the tropics. Cloud-resolving models have recently begun to capture triggered convection over terrain (e.g., Kirshbaum and Durran 2005; Fuhrer and Schar 2005; Kirshbaum and Smith 2008), but none of these numerical studies was focused on the tropics.
2. Dominica as a natural laboratory
a. Geography of the island
The volcanic island of Dominica (15°25′N, 61°21′W) is the most mountainous island in the Lesser Antilles chain and receives the most precipitation (Reed 1926; Lang 1967). It lies midway between French islands of Guadeloupe and Martinique in the easterly trade winds. The island terrain forms a simple north–south ridge (Fig. 1) but with higher peaks in the north and south (e.g., Morne Diablotins, 1447 m and Morne Trois Pitons, 1424 m). As shown in Fig. 2, these mountains are higher than the lifting condensation level (LCL) in the region (∼600 m) but lower than the trade wind inversion (∼1800 m). The dimensions of the island are about 17 km in the east–west direction by 45 km north–south. Vegetation on the island grades from tropical rain forest on the east coast to dry grasslands on the west coast. The island is known for its waterfalls, rushing rivers, and flash floods. About half of the electrical energy for 72 000 inhabitants comes from hydropower.
Reed (1926) was the first to note the remarkable rainfall gradients in Dominica. Reviewing the climate of all the Caribbean islands, he wrote: “The most striking example of a great difference in precipitation within a few miles is found in Dominica. Roseau, on the western coast, at an elevation of 25 feet, has a mean annual precipitation of 78 inches, while Shawford, about 3 miles to the northeast, at an elevation of 560 feet, has a mean annual amount of 185 inches.”
In 1935, Harrison (1935) published a geographic description of Dominica. She included a table showing significant terrain enhancement of rainfall. A more complete analysis of the precipitation distribution was provided by Lang (1967) for the period 1920 to 1965. He used monthly data from 80 rain collectors at plantations on the island to draw contours of annual rainfall. His data show a strong correspondence between rainfall and elevation.
The lack of upstream terrain contributes to the laboratory value of Dominica. The easterly trades blow rather steadily against the island in all seasons (Fig. 3). Several dozen back trajectories were run from Dominica using the National Oceanic and Atmospheric Administration (NOAA) Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model, based on the gridded National Centers for Environmental Prediction (NCEP) reanalysis. These calculations show that air parcels take about 4 days to cross the 4000-km width of tropical Atlantic from the Canary or Azore Islands to Dominica (i.e., about the same path as Columbus’ second voyage, but 10 times faster). Most of the trajectories show slow subsidence, with parcels typically descending 1 or 2 km during the Atlantic crossing.
b. Instrumentation
The Yale Dominica Precipitation Project began in March 2007, with the objective of understanding the mechanism of the orographic enhancement in tropical Dominica. Since then, eight HOBO tipping-bucket rain gauges have been installed across the high mountains in the southern part of the island (Table 1; Fig. 1). These gauges record the time of each tip corresponding to 0.2 mm of precipitation. When data are downloaded every 5 months, the logger clocks have been found accurate to better than 1 min, allowing detailed analysis of the rain rate throughout the test period. The location of Freshwater Lake station is taken as a reference point in later diagrams. In Table 1, the total 12-month precipitation is given for four stations, and a 6-month total is given for all eight stations during the dry season.
The values in Table 1 confirm the relationship between rainfall and elevation found by Lang (1967). For stations along the coast, however, Lang’s values are 30% greater than our 2007/08 values. In addition to showing the elevational control on precipitation, Table 1 compares three pairs of gauges at similar elevation. On the east coast, Rosalie and LaPlaine (5 km apart) receive similar amounts of rain. On the highest terrain, Freshwater Lake and Boeri Lake (1.5 km apart) are in very good agreement, including the day by day and hour by hour variations. On the west coast, Botanical Garden and Canefield Airport (5 km apart) are quite different. Canefield lies in the rain shadow of the high southern hills, whereas Botanical Garden further south does not.
A key aspect of Dominica that makes it an effective natural laboratory is its proximity to the two conventional Météo-France 2.8-GHz S-band weather radars on Guadeloupe and Martinique. These two radars are located approximately 60 km north and south of the rain gauge line across Dominica. The Guadeloupe radar, with a beam angle of 1.2°, is especially helpful because it has very little beam blockage to the south (with the exception of azimuth 160° from true north). The 5-min, 200-km Guadeloupe radar scan is used in this paper with the Z–R relationship Z = 85R1.2 from Regina (2007), where Z is the radar reflectivity and R is the rain rate in mm h−1. This Z–R relationship was recently optimized for Hurricane Dean using with rain gauge data from Martinique, and it agrees well with the rain gauges in the present study. Other S-band radars, such as the Weather Surveillance Radar–1988 Doppler (WSR-88D), use a similar exponent (i.e., 1.2) in the tropics. For the standard 0.5° PPI scan, the beam is about 1.5 km above sea level at the distance of Dominica. A moderate ground clutter–contaminated echo can be detected from the northern peaks (Morne Diablotins, 1447 m) and occasionally from the southern peaks (Morne Trois Pitons, 1424 m). Echo over both peaks was removed in our analysis. The Guadeloupe radar is not very sensitive to low precipitation rates (the reflectivity threshold is 28 dBZ, or 5 mm h−1 with our Z–R relation), but this was not a serious drawback in the present application.
To observe the nature of convection and air mass transformation over the island, small temperature and humidity loggers by Lascar and Kestrel and time lapse cameras were deployed across the island for short periods. Other useful datasets come from the two airport weather stations, twice-daily rawinsondes from Guadeloupe and Barbados, and daily MODIS images.
c. BOMEX results
Another advantage of Dominica as a natural laboratory is that the region of the ocean just upstream of the island was studied intensively in the May–June 1969 Barbados Oceanographic and Meteorological Experiment (BOMEX; Bean et al. 1972; Delnore 1972; Holland and Rasmussen 1973). The square BOMEX array extended from 12° to 16°N and from 54° to 59°W, centered 450 km east of Dominica. More recently, numerous attempts to simulate the heat and moisture fluxes and the trade wind clouds in the BOMEX region using convection-resolving models have been reported by Siebesma et al. (2003), de Roode and Bretherton (2003), and Kuang and Bretherton (2006), among others. A useful result from BOMEX is the estimate of heat and water vapor fluxes from the ocean. Several flux estimation methods were used, including atmospheric budget (Holland and Rasmussen 1973), ocean budget (Delnore 1972) and eddy correlation from aircraft (Bean et al. 1972). All agreed that the water vapor flux was about 6 mm day−1 and the average sensible and latent heat fluxes were about 10 and 150 W m−2, respectively. The Bowen ratio is thus quite small (i.e., β = 10/150 ≈ 0.07). The impact of the Bowen ratio on convection is discussed by Lewellen et al. (1996). Typical soundings from the BOMEX campaign are described by Siebesma et al. (2003) and in section 10.
The results from the Atlantic Trade Cumulus Experiment (ATEX) further east (February 1969 at 35°W) are also useful for defining our upstream conditions (Stevens et al. 2001). The recent Rain in Cumulus over the Ocean (RICO) project (November 2005 to January 2006; Rauber et al. 2007) provides further insight into the undisturbed ocean convection in the Dominica region. Interannual variation in Caribbean climate is discussed by Taylor et al. (2002).
3. Seasons in Dominica
The seasonality in the region provides a “control variable” with which to examine how orographic precipitation processes vary with upstream parameters. In Figs. 4a–d, we show how a number of parameters vary with season, using 3 yr of data from the Guadeloupe and Barbados soundings. The integrated water vapor PWAT shifts from about 50 mm in the summer and fall to 30 mm in the winter and early spring. With similar timing, CAPE varies from 1500 to 100 J kg−1. Likewise, the temperature at cloud base near 925 hPa (i.e., about 7°C cooler than the surface ocean temperature) varies from 22° to 19°C. By contrast, the wind speed (average of 925- and 850-hPa values) has an almost semiannual cycle between 6 and 10 m s−1.
Two important derived parameters (not plotted) are the horizontal water vapor flux 
Precipitation also has a seasonal cycle (see Figs. 4e and 5). Over the ocean to the east of Dominica, where average conditions are relatively uniform, satellite-derived National Aeronautic and Space Administration (NASA) Global Precipitation Climatology Project (GPCP) data provide estimates of the seasonal cycle of precipitation (Adler et al. 2000). The precipitation for 3 yr for two grid cells centered at 58.75°W and at 13.75°N and 16.25°N, respectively, is shown in Fig. 4e. The average annual rainfall is about 700 mm. Although these satellite estimates have their own errors, they provide another basis for judging the orographic enhancement over Dominica. This rainfall, about 2 mm day−1, is less than the BOMEX estimate of evaporation (6 mm day−1) cited earlier.
The radar data from Guadeloupe (Fig. 5a) show rainfall in four sampling areas for the period March 2007 to March 2008. The undisturbed ocean patch (75 km northeast of Dominica) shows about 8 mm day−1 in the summer and fall, with only 1 mm day−1 in the winter and spring. It somewhat exceeds the GPCP satellite rainfall data in Figs. 4e and 5.
When averaged over the entire island of Dominica (Fig. 5a), the precipitation is much larger, with 20 and 4 mm in the rainy and dry seasons, respectively. If we sharpen the focus to a few pixels over the southern hills (Mountain in Fig. 5a), we get a still larger precipitation: 30 mm day−1 in the wet season and 8 mm day−1 in the dry season. The driest region is found over the ocean downstream of the island (Fig. 5a). This oceanic “rain shadow” receives 5 mm day−1 in the wet season and 0.5 mm day−1 in the dry season. As shown in the dashed curves (Fig. 5a), the wet season 5 mm day−1 in the rain shadow is mostly due to a few large events, such as Hurricane Dean.
A similar seasonal cycle of precipitation is seen in the data from four rain gauges (Fig. 5b). Generally speaking, the radar and rain gauge values agree fairly well. For example, at Freshwater Lake, the August area-averaged radar value is 29 mm day−1, while the rain gauge gives 37 mm day−1. The radar value may be lower because it includes some adjacent pixels over lower terrain with less rain. In general, from Fig. 5 one gets the impression that it rains heavily in all seasons on the high terrain in Dominica. The coastal sites and nearby oceans have a stronger seasonal cycle.
4. Orographic enhancement
The spatial pattern of orographic enhancement shown in Fig. 6 is derived from a full year of Guadeloupe’s 5-min radar scans during easterly flow. Several days with noneasterly flow were removed from the analysis, as were Hurricane Dean on 17 August 2007 (Smith et al. 2009) and a large easterly wave event on 10 September 2007. A slight upstream enhancement is seen, extending about 10 km upwind of the east coast. Over the island, the average rain rate roughly follows the terrain, with isolated maxima over the high terrain in the north and south. Precipitation drops off steeply on the western leeward slopes and a dry rain shadow is seen over the ocean 25 km west of the island. The only significant precipitation in the rain shadow occurs during those few disturbed days each year when the wind is not easterly.
Another view of the enhancement is shown in the east–west transect across the southern mountains (Fig. 7). In this diagram, days have been divided into three types according to the amount of precipitation over the sea upwind (type 1: <2 mm day−1; type 2: 2–10 mm day−1; type 3: 10–25 mm day−1). In general, the radar and rain gauge data agree well. Under dry type-1 conditions (most days), the average rain rate increases from about 1 mm day−1 upwind to 8 mm day−1 over the highest terrain. The enhancement factor (i.e., the ratio of mountain to upstream precipitation) is 8. On wetter days, the enhancement as an absolute amount is greater, but the enhancement factor is smaller, about 2 to 3. Again, the lee region is much dryer than the undisturbed upwind region.
Monthly enhancement factors are given in Table 2. For rain gauge data we use LaPlaine as the upstream reference whereas for radar data we use the upstream ocean region. The former enhancement factor ranges from 2 to 5; the latter, with a dryer reference point, ranges mostly from 4 to 10. The rain gauge data show a larger enhancement in the dryer months of January through April.
5. Diurnal modulation
In many parts of the world, over both land and sea, the diurnal cycle of solar heating influences the occurrence of convective precipitation (e.g., Bell and Reid 1993). Diurnal control in Dominica would not be surprising because the island’s surface has a large diurnal range in temperature, reaching 7°C on the lower slopes (Smith et al. 2009). To determine the diurnal cycle in precipitation near and over Dominica, we use the radar and rain gauge data (Fig. 8). To reduce noise, the rain amount is summed into twelve 2-h segments. Several large events such as Hurricane Dean on 17 August were removed. Because only 1 yr of data was analyzed, there are statistical limitations on how well we can constrain the amplitude of the diurnal cycle. In Fig. 8, neither the radar nor rain gauge data shows a significant diurnal cycle, although we cannot rule out a possible 20% morning maximum in the mountain precipitation. The implication of the absent diurnal modulation in precipitation is that the triggering of convective precipitation over Dominica is purely due to mechanical lifting of the airstream. The convection caused by daytime solar heating plays little role.
6. Rain rate statistics
Visitors to Dominica quickly discover that rain there comes in short bursts, as it does in many other tropical locations. Extensive time-lapse photography of clouds over the windward slopes confirms the rapid development of convective cells over the island, with little occurrence of stratiform clouds. The dominance of convective precipitation can be shown with a rain rate histogram (Fig. 9). The ordinate of this diagram is the percent that each rain rate contributes to the total annual precipitation. By this measure, the rain gauge (Fig. 9a) and radar (Fig. 9b) data agree. Both show a peak in the rain rate histogram near 40 mm h−1. This large rain rate suggests a brevity to the rainfall. At Freshwater Lake, for example, the typical duration of heavy precipitation is the ratio of average rain rate to the typical rain rate; that is, T = (0.6 mm h−1)/(40 mm h−1) = 0.015 h h−1, or 22 min day−1.
For reference purposes, we include in Fig. 9 a rain rate histogram from North Haven, Connecticut (41°N), from the same instrument over a 9-month period. The peak rate of 3 mm h−1 is an order of magnitude weaker than the Dominica peak rain rate.
The statistical relationships between pairs of stations are also informative. Gauge data from each site were averaged over different intervals (e.g., 1 h, 12 h, etc.) and then compared with other sites using scatterplots and correlations. The rainfall at Rosalie and LaPlaine on the east coast (5 km apart) correlate with R-squared values of 0.28, 0.37, 0.41, 0.61, and 0.61 for averaging intervals of 1,12, 18, 24, and 48 h, respectively. Thus, these two stations seldom experience the same convective cell, but they do experience the same cluster of cells or weather systems. By contrast, Freshwater and Boeri Lakes, 1.5 km apart on high ground, have R-squared values of 0.75 and 0.9 for 1 and 12 h; thus, they usually experience the same cells. If the typical cell diameter is about 3 km (i.e., between 1.5 and 5 km) and the cell speed is 7 m s−1, the typical shower duration would be 430 s. Using the daily duration of 22 min, the typical daily number of events at a particular point is N = 1320 s/430 s ≈ 3. Other statistical methods might yield different estimates.
7. Properties of the incoming flow
As we develop hypotheses for the precipitation physics in Dominica, it is necessary to have some information about the fluctuations in the incoming flow. In addition to the BOMEX literature, the simplest and least expensive method for characterizing the inflow is to monitor the temperature, humidity, and wind at the east coast beach. A portable Kestrel 4500 weather logger (manufactured by Nielsen–Kellerman) was placed at LaPlaine beach for this purpose. The sensor height was 2 m and it was placed 3 m horizontally from the water line. At this location some errors could arise from nearby breaking ocean waves and spray, but good dewpoint agreement was found with higher sensors and with sensors further back from the beach.
The Kestrel sensors can record several hours of data with a sampling interval of 10 s and they have a NIST traceable calibration. Prior to the experiment, using square wave inputs with moderate ventilation, we determined that the Kestrel wind sensor responds within 3 s to sudden changes whereas the temperature and humidity sensors require about 15 s for an 80% approach to the correct value. The mean measured temperature and humidity values were verified with portable Lascar sensors.
A short time series from LaPlaine beach for midday on 29 June 2008 (Fig. 10; Table 3) shows little variation in temperature T but does indicate dewpoint fluctuations Td of 1°C. Wind speed varied between 2.5 and 4.5 m s−1. The small relative magnitude of the temperature fluctuations compared to the dewpoint fluctuations (see Woodcock 1960) is consistent with the small Bowen ratio found in BOMEX. The large eddies are transporting more water vapor than heat. Similar Kestrel deployments in other locations have always given a ratio of temperature to dewpoint variance in qualitative proportionality to the Bowen ratio. The inverse correlation between dewpoint and wind speed indicates, as expected, that convection over the sea is transporting water vapor upward and westward momentum downward.
The observed fluctuation in dewpoint is relevant to convective triggering because the dewpoint depression (DPD = T − Td) is linked to the lifting condensation level [i.e., ZLCL(m) ≈ 120 × DPD]. With a constant temperature of 28°C, parcels with dewpoint of 24° or 25°C will have LCLs of 480 or 360 m respectively. Sudden larger shifts in dewpoint are sometimes observed at LaPlaine, associated with the arrival of small offshore air masses.
The qualitative nature of the incoming cloud field was determined using many hours of time-lapse photography from a viewing site at Riviere Cyrique (Fig. 1). Looking east, groups of cumulus castellanus (Fig. 11) drift against the island with no evident deflection. Only a small fraction of these clouds precipitate over the sea. About half form small anvils at the trade wind inversion. Although some of the anvils have virga, virga are not seen below cloud base. A few clouds penetrate the inversion to higher altitudes. Aloft, cirrus layers drift in different directions on different days. Looking north over the windward slopes of the island, the cloud bases thicken and rapid tower growth occurs.
8. Air mass drying
Another measure of air mass drying is the change in specific humidity at the surface across the island. The specific humidity across the southern part of the island was determined during a 1-week deployment of six Lascar temperature/humidity sensors and a number of shorter deployments of Kestrel sensors at locations across the island. Whereas other sites show significant diurnal modulation of dewpoint, the LaPlaine beach site on the east coast and the high-elevation saddle at Freshwater Lake do not. Typical properties at the beach and mountain sites for the period 29 June–4 July 2008 are given in Table 3.
The dramatic 17% decrease in specific humidity between the east coast beach and mountain top has two possible explanations. A sharp isolated hill with flow splitting would penetrate into the atmosphere, sampling dryer air aloft. Freshwater Lake, however, is a saddle in the terrain at 800 m, surrounded by peaks reaching 1200 to 1400 m. Air is more likely to stream toward this site than to split around it. A second and more likely explanation is that the surface air has lost water vapor by vertical mixing. The air aloft is generally quite dry in the “trade wind tropics” due to subsidence. Mixing will transport water vapor upward and bring dry air to the surface. The mixing arises from the same vigorous convection that brings precipitation to the peaks.
This drop in surface specific humidity (17%) is more than an order of magnitude greater than the drying ratio derived from the average precipitation rate (2), suggesting that far more water is transported upward than is precipitated (see also Smith et al. 2003; Kirshbaum and Smith 2008). Because neither the precipitation nor the air mass drying varies diurnally, the vertical mixing must be driven by mechanical lifting and instability rather than solar heating. Further drying may occur along the lee slopes, but we have not measured this change.
9. Case studies
To improve our understanding of precipitation physics in Dominica, we used the rain gauge data from Freshwater Lake in 2007 to select, from hundreds of precipitation events, nine events that coincided with Moderate Resolution Imaging Spectroradiometer (MODIS) overpasses. Thermal band data with 1-km resolution and reflective band data with 500-m resolution were examined. In addition, the 5-min radar scans from Guadeloupe were studied for each case. These cases are summarized in Table 4. The upstream ocean rainfall (i.e., upstream index) and the cloud top temperatures vary widely, and neither correlates with Freshwater Lake rainfall. Cases 2 and 4 with the largest CAPE have the coldest cloud top temperatures. Otherwise, the cloud tops were little colder than the inversion.
The example of 29 September 2007 (case 4 in Table 4) was reasonably representative of days with little ocean precipitation and repeated precipitation events over the island. The accumulated gauge precipitation (Fig. 12) shows the brevity and frequency of rain pulses characteristic of Dominica and the large orographic enhancement. The 24-h accumulated radar precipitation “streakplot” (Fig. 13) shows the generation and movement of convective systems near the island. A wind-oriented Hovmöller plot through Freshwater Lake (Fig. 14) shows at least six independent cells forming and dying over the island. An instantaneous MODIS snapshot is shown in Fig. 15. By using mid-IR bands we can differentiate ice and liquid clouds. The rain from the large east–west-oriented cell south of Dominica in the image (Fig. 15) can be identified in Fig. 13. Overall, the frequent weak convective events over the island are more relevant to the full precipitation climatology than these occasional larger ocean events (Fig. 6).
10. Conceptual model of convective triggering
To understand the mechanism of convective development over Dominica, we consider two theories, both related to the conditional instability in the ambient flow. Before discussing these slightly different theories, we describe the properties of the ambient flow and the nature of the lifting by the terrain.
According to Holland and Rasmussen (1973) and Siebesma et al. (2003), the lower part of the typical BOMEX sounding consists of three layers (Fig. 16). From the surface to about 500 m, the lapse rate is dry adiabatic, with temperature dropping from 28°C at the sea surface to 23°C. The relative humidity increases from about 80% at the sea surface to 90% at the top, while the specific humidity decreases upward from about 17 g kg−1 at the sea surface to 16 g kg−1. In the cloud layer, from 500 to 1500 m, the average dewpoint depression is about 2° and the relative humidity is about 90% (i.e., unsaturated). The lapse rate is about −6.6°C km−1 and the moist adiabat has a slope of −4.6°C km−1, making the layer conditionally unstable. The specific humidity decreases upward from 16 to 11 g kg−1. The temperature at the top of the cloud layer is about 17°C. The third layer is the trade wind inversion, from 1500 to 2000 m, with potential temperature increasing rapidly and specific humidity decreasing upward. This three-layer structure is the balanced result of competing influences of large-scale subsidence and advection and small-scale convection (Siebesma et al. 2003).
As the surface air in the easterly trade wind encounters the ∼0.15 slope of the island’s higher terrain, it rises with a typical vertical speed of 1 m s−1, cools adiabatically, and quickly reaches the LCL. Aloft, near the inversion, we assume that that the air does not lift appreciably. In subcritical hydraulic flow, a stable layer will rise or descend only slightly as the air approaches a ridge top (see the appendix). In between, the lower half of the cloud layer from 500 to 1000 m lifts with a slower vertical velocity; cooling slightly and reaching saturation (Fig. 16).
There are two weaknesses to this theory. First, with a no explicit proposal for the source of the small fluctuations, no estimates of convection strength can be made. Thus, it is an incomplete theory. The observed ambient temperature fluctuations in Fig. 10 are very small and not accurately known. Second, as the convection begins to grow, downdrafts may develop fairly early in the development. If downdrafts occur, they will quickly desaturate the air so that our assumption of a fully saturated layer is violated. The moist stability frequency (3) would not give an accurate growth rate estimate in a cloud field with saturated updrafts and dry downdrafts. If the spatial variation in humidity is sufficiently large, there never will be a moment when the lifted layer is fully saturated.
The second theory we consider is the bulk lifting of the conditionally unstable layer with embedded saturated and unsaturated parcels (Woodcock 1960). Two lines of evidence support the existence of these humidity fluctuations. As seen in Fig. 11, there are trade wind cumuli, with small cloud fraction, constantly drifting toward the island. These clouds have greater humidity than the surrounding air. Furthermore, according to the surface data in Fig. 10, there are dewpoint fluctuations of about 1°C in the ambient flow, even at the surface. These fluctuations are caused by the weak trade cumulus convection over the sea upwind. As this heterogeneous layer is lifted by the terrain, the cloudy and dry air will cool adiabatically at different rates, quickly generating temperature differences approaching 1°C. A temperature anomaly of 1° produces an acceleration of g′ = g(ΔT/T) ≈ 0.03 m s−2. After only 100 s, this acceleration will produce a vertical velocity of 3 m s−1. Similarly, if heterogeneous air in the subcloud layer is lifted by the terrain, the moist parcels will reach their LCLs first. A dewpoint difference of 1°C gives an LCL difference of 120 m. As moist and dry parcels lift together by 120 m moving along different adiabats, they quickly generate a temperature difference approaching 1°C. Unlike the first theory, this theory is not invalidated by the coexistence of saturated and unsaturated air.
In a companion paper (Kirshbaum and Smith 2009), large-eddy simulations of tropical orographic convection are carried out to discover the mechanism of convective growth over Dominica. Those results favor the latter theory in which the ambient humidity fluctuations play a key role, along with conditional instability. This dependence of orographic precipitation on weak upstream heterogeneity and convection casts doubt on the old paradigm that knowledge of the mean upstream sounding and the terrain geometry is sufficient to predict orographic precipitation. The simulation of Dominica rainfall requires the consideration of the shallow trade wind convection upwind and the related humidity fluctuations.
The sudden disappearance of clouds and precipitation on the lee side of the island is probably due to two effects. First, the convection over the windward slope dries the surface air and removes any column instability (section 8). Second, the plunging of the trade wind inversion over the lee slopes warms the air adiabatically. According to dry hydrostatic “hydraulic” theory, a layer of dense fluid flowing with Fr < 1 will descend slightly over the windward slope and plunge violently (like a waterfall) over the lee slope. With Fr ∼ 0.25–0.5 (section 3), this behavior will be robust (see appendix). A similar plunging inversion was found over the island of St. Vincent (Smith et al. 1997), 200 km to the south.
11. Conclusions
A remarkable example of pure orographic precipitation in the tropics was discovered in Dominica in the Lesser Antilles. The combination of steady trade winds, simple mountain geometry, and off-island scanning radar make Dominica an ideal natural laboratory. Its large orographic enhancement factor (from 2 to 8 depending on weather pattern and season) is caused by rapid growth of convection over the windward slopes. A large rain shadow is located downwind over the ocean. Although the highest terrain receives 7 m of rain per year, the rain pulses are brief and the average drying ratio is small (DR < 1%). By contrast, the surface specific humidity decreases by about 17% from the east coast to mountain top because of vigorous vertical mixing. The lack of diurnal modulation of precipitation indicates that the convection is caused by terrain-forced ascent. The mechanism of convection growth probably involves ambient humidity fluctuations associated with weak open-ocean cumulus convection. As air is lifted by the terrain, wet and dry parcels cool with different adiabatic lapse rates, generating temperature anomalies and vertical accelerations. Downwind, a plunging inversion and reduced instability prevent further precipitation on the lee side.
The theory involving ambient humidity fluctuations might be consistent with the statistics described in sections 4 and 6. The fact that some humidity fluctuations are always present helps to explain why the orographic precipitation occurs even with vanishing precipitation upstream. As long as the warm ocean causes weak atmospheric convection, the process can persist. The open-ocean convection can generate humidity fluctuations either by juxtaposing dry air from aloft with moist air from below or by moistening subcloud air with showers. The ubiquity of humidity fluctuations is reflected in the large enhancement factor (i.e., enhancement ratio) in type-1 conditions in Fig. 7. On the other hand, the orographic precipitation amount increases with upstream precipitation (Fig. 7). With more vigorous convection upwind, the ambient humidity fluctuations will be larger, seeding more convection over the island.
The evidence for convective triggering presented here contrasts with Dominica’s rainfall physics during Hurricane Dean in August 2007 (Smith et al. 2009). With Dean’s high winds and supercritical Froude numbers, no convection could be triggered by the terrain. Orographic enhancement then was due to the seeder–feeder mechanism acting within a hurricane rainband.
Future research on Dominica precipitation will use high-resolution numerical simulation to investigate the details of how convection is triggered in rapidly rising conditionally unstable air with pre-existing perturbations (Kirshbaum and Smith 2009). Direct in situ observations of cloud structure over the island may be required to clarify the physics of convective growth. The suggested role of ambient humidity fluctuations could be tested with upstream monitoring. Cloud microphysics has also been neglected so far. Finally, we would like to know if orographic precipitation in other tropical regions has similar physics to that in Dominica.
Acknowledgments
Thanks to the people and government of the Commonwealth of Dominica for hosting this research, especially the Divisions of Agriculture and Forestry. Special thanks to Arlington James of the Forestry Division who helped to find sites for the gauges and shared existing rain collector data. Thanks also to Nathaniel Isaac from the Met Office, Nancy Osler from the Archbold Tropical Research and Education Center, Verna Dejean Frederick from the Grand Fond School, Joyce Pascal and Jem Winston of Rosalie, and Ricky Brumant from the Agricultural Training Center. Sigrid R.-P. Smith assisted with the field work. Larry Bonneau assisted with the image analysis. The MODIS image was provided by NASA. NOAA supports the HYSPLIT model. Sounding data came from the University of Wyoming website. David Bolvin from NASA assisted with the GPCP data. This research was supported by Météo-France and by a grant to Yale University from the National Science Foundation (ATM-112354).
REFERENCES
Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and E. J. Nelkin, 2000: Tropical rainfall distributions determined using TRMM combined with other satellite and rain gauge information. J. Appl. Meteor., 39 , 2007–2023.
Banta, R. M., 1990: The role of mountains in making clouds. Atmospheric Processes over Complex Terrain, W. Blumen, Ed., Amer. Meteor. Soc., 59–81.
Bean, B., R. Gilmer, R. Grossman, R. McGavin, and C. Travis, 1972: An analysis of airborne measurements of vertical water vapor flux during BOMEX. J. Atmos. Sci., 29 , 860–869.
Bell, T. L., and N. Reid, 1993: Detecting the diurnal cycle of rainfall using satellite observations. J. Appl. Meteor., 32 , 311–322.
Browning, K. A., F. F. Hill, and C. W. Pardoe, 1974: Structure and mechanism of precipitation and the effect of orography in a wintertime warm sector. Quart. J. Roy. Meteor. Soc., 100 , 309–330.
Colle, B. A., Y. Lin, S. Medina, and B. F. Smull, 2008: Orographic modification of convection and flow kinematics by the Oregon Coast Range and Cascades during IMPROVE-2. Mon. Wea. Rev., 136 , 3894–3916.
Delnore, V. E., 1972: Diurnal variation of temperature and energy budget for the oceanic mixed layer during BOMEX. J. Phys. Oceanogr., 2 , 239–247.
de Roode, S. R., and C. S. Bretherton, 2003: Mass-flux budgets of shallow cumulus clouds. J. Atmos. Sci., 60 , 137–151.
Esteban, M. A., and Y. L. Chen, 2008: The impact of trade wind strength on precipitation over the windward side of the island of Hawaii. Mon. Wea. Rev., 136 , 913–928.
Fuhrer, O., and C. Schär, 2005: Embedded cellular convection in moist flow past topography. J. Atmos. Sci., 62 , 2810–2828.
Harrison, L. C., 1935: Dominica: A wet tropical human habitat. Econ. Geogr., 11 , 62–76.
Holland, J. Z., and E. M. Rasmussen, 1973: Measurements of the atmospheric mass, energy, and momentum budgets over a 500-kilometer square of tropical ocean. Mon. Wea. Rev., 101 , 44–55.
Kirshbaum, D. J., and D. R. Durran, 2005: Atmospheric factors governing banded orographic precipitation. J. Atmos. Sci., 62 , 3758–3774.
Kirshbaum, D. J., and R. B. Smith, 2008: Temperature and moist-stability effects on midlatitude orographic precipitation. Quart. J. Roy. Meteor. Soc., 134 , 1183–1199.
Kirshbaum, D. J., and R. B. Smith, 2009: Orographic precipitation in the tropics: Large-eddy simulations and theory. J. Atmos. Sci., in press.
Kuang, Z., and C. S. Bretherton, 2006: A mass-flux scheme view of a high-resolution simulation of a transition from shallow to deep cumulus convection. J. Atmos. Sci., 63 , 1895–1909.
Lang, D. M., 1967: Soil and land use surveys, No. 21: Dominica. Soil and Land Use Surveys, Trinidad: Imperial College of Tropical Agriculture, 1–58.
Lewellen, D., W. Lewellen, and S. Yoh, 1996: Influence of Bowen ratio on boundary-layer cloud structure. J. Atmos. Sci., 53 , 175–187.
Rauber, R. M., and Coauthors, 2007: Rain in shallow cumulus over the ocean: The RICO campaign. Bull. Amer. Meteor. Soc., 88 , 1912–1928.
Reed, W. W., 1926: Climatological data for the West Indian Islands. Mon. Wea. Rev., 54 , 133–160.
Regina, E., 2007: Analyse de l’episode Dean a partir des données du radar du Diamant et des pluviometres. Direction Interrégionale Antilles-Guyane, Météo-France, 21 pp.
Rotunno, R., and R. A. Houze, 2007: Lessons on orographic precipitation from the mesoscale alpine programme. Quart. J. Roy. Meteor. Soc., 133 , 811–830.
Schär, C., and R. B. Smith, 1993: Shallow-water flow past isolated topography. Part I. Vorticity production and wake formation. J. Atmos. Sci., 50 , 1373–1400.
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., 2006: Progress on the theory of orographic precipitation. Special Paper 398: Tectonics, Climate, and Landscape Evolution, S. D. Willett et al., Eds., Geological Society of America, 1–16.
Smith, R. B., and Y-L. Lin, 1982: The addition of heat to a stratified airstream with application to the dynamics of orographic rain. Quart. J. Roy. Meteor. Soc., 108 , 353–378.
Smith, R. B., and J. P. Evans, 2007: Orographic precipitation and water vapor fractionation over the southern Andes. J. Hydrometeor., 8 , 3–19.
Smith, R. B., A. C. Gleason, P. A. Gluhosky, and V. Grubišić, 1997: The wake of St. Vincent. J. Atmos. Sci., 54 , 606–623.
Smith, R. B., Q. F. Jiang, M. G. Fearon, P. Tabary, M. Dorninger, J. D. Doyle, and R. Benoit, 2003: Orographic precipitation and air mass transformation: An Alpine example. Quart. J. Roy. Meteor. Soc., 129 , 433–454.
Smith, R. B., P. Schafer, D. Kirshbaum, and E. Regina, 2009: Orographic enhancement of precipitation inside Hurricane Dean. J. Hydrometeor., 10 , 820–831.
Stevens, B., and Coauthors, 2001: Simulations of trade wind cumuli under a strong inversion. J. Atmos. Sci., 58 , 1870–1891.
Taylor, M. A., D. B. Enfield, and A. A. Chen, 2002: Influence of the tropical Atlantic versus the tropical Pacific on Caribbean rainfall. J. Geophys. Res., 107 , 3127. doi:10.1029/2001JC001097.
Woodcock, A. H., 1960: The origin of trade-wind orographic shower rains. Tellus, 12 , 315–326.
Yeh, H-C., and Y-L. Chen, 1998: Characteristics of rainfall distributions over Taiwan during the Taiwan Area Mesoscale Experiment (TAMEX). J. Appl. Meteor., 37 , 1457–1469.
Appendix
Airflow and Inversion Dynamics
As a background to the precipitation problem, we outline the fluid dynamics of airflow past Dominica using the idealized hydrostatic equations for a single layer of fluid beneath a density discontinuity (e.g., Schär and Smith 1993; Smith et al. 1997). This formulation should give a fair description of disturbed trade wind flow beneath the inversion, if moisture and convection can be neglected.
The terrain of Dominica with rain gauge locations (see Table 1). Elevations derived from NASA’s Shuttle Radar Topography Mission (SRTM). The Freshwater Lake (FW) station is a reference point for other figures. The viewing site is Riviere Cyrique (Fig. 11).
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Dominica’s terrain profile as seen from the east. Distance is measured from the latitude of the FW station (Table 1). The LCL and trade wind inversion are indicated.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Wind rose for the Guadeloupe upper-air station for the average 925- and 850-hPa level winds. Radius is the relative frequency of soundings in a 10° azimuthal increment over a 3-yr period (2005–07).
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Seasonal cycle of five smoothed environmental variables for 3 yr (2005–07). (a)–(d) Data from the Guadeloupe and Barbados upper air stations: (a) integrated water vapor (mm), (b) CAPE (J kg−1), (c) temperature at the 925-hPa level, and (d) wind speed (m s−1), averaged between the 850- and 925-hPa levels. (e) The satellite-derived precipitation from the GPCP archive for two pixels just upstream of Dominica.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Seasonal cycle of precipitation for 1 yr from (a) Guadeloupe’s radar and (b) four rain gauges (Table 1). In (a), data are averaged over a large area upstream and downstream, namely the area of Dominica and a small area near Mt. Trois Pitons; upstream GPCP data are also included. Dotted curves have removed a few large events in the rainy season.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Averaged daily precipitation over Dominica derived from the Guadeloupe radar for the period March 2007–February 2008 during easterly flow. Contour interval is 2 mm day−1. Maximum rate is 14 mm day−1 (i.e., 5100 mm yr−1). Note the slight upwind enhancement and the ocean rain shadow to the west. The dashed line is the transect shown in Fig. 7.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Averaged daily precipitation along an east–west Dominica transect for the period March 2007–February 2008 derived from the Guadeloupe radar and four rain gauges. Days are divided into three groups based on the intensity of precipitation at sea. Type 1 has < 2 mm day−1, type 2 has 2–10 mm day−1, and type 3 has > 10 mm day−1. Only days with easterly flow are included.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Diurnal cycle of Dominica’s precipitation from (a) Guadeloupe radar and (b) rain gauges. Areas and gauges are the same as in Fig. 5. Data from a 1-yr period is averaged in 2-h blocks. Several large events were removed.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Rain rate histogram from (a) rain gauges and (b) Guadeloupe radar. Areas and gauges are the same as in Fig. 5. Data from three large events on 17 August, 10 September, and 26–27 October have been removed. The rain gauge plot includes data from a midlatitude site (North Haven, CT) over a similar period with the same instrument. The radar plot includes open-ocean data from the Martinique radar.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Short time series of temperature, dewpoint, and wind speed at LaPlaine beach on the east coast of Dominica on 29 Jun 2008.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Photograph of trade wind cumuli to the east of Dominica. These clouds are drifting towards the camera location at the Riviere Cyrique viewing site (Fig. 1).
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Accumulated precipitation (mm) from four rain gauges on 29 Sep 2007.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Accumulated precipitation (mm) near Dominica on 29 Sep 2007 from the Guadeloupe radar. Contour interval is 20 mm. Line shows transect for the Hovmöller plot in Fig. 14.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
A wind-oriented Hovmöller plot through the FW reference point for 29 Sep 2007 (mm h−1). Vertical lines indicate the coasts of Dominica. Horizontal line is the time of the MODIS image in Fig. 15.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Terra MODIS image taken at 1425 UTC 29 Sep 2008. This false color composite image is an RGB-167. Pixel size is 500 m. Liquid clouds are light blue, ice is red, and vegetation is green. The Dominican coast and FW station are marked in yellow.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Schematic of the BOMEX sounding and convection triggering. The upstream sounding has three layers: the subcloud dry layer (0–500 m), the cloud layer (500–1500 m), and the inversion (1500–2000 m). The cloud layer is conditionally unstable but subsaturated. Forced lifting brings the lower part of the cloud layer to saturation, triggering convection.
Citation: Journal of the Atmospheric Sciences 66, 6; 10.1175/2008JAS2920.1
Monthly orographic enhancement factors.
East coast/mountain thermodynamic comparison (typical June/July 2008).
Case studies in 2007.
