• Back, L. E., , and C. S. Bretherton, 2005: The relationship between wind speed and precipitation in the Pacific ITCZ. J. Climate, 18, 43174328, doi:10.1175/JCLI3519.1.

    • Search Google Scholar
    • Export Citation
  • Banta, R. M., , and C. L. B. Schaaf, 1987: Thunderstorm genesis zones in the Colorado Rocky Mountains as determined by traceback of geosynchronous satellite images. Mon. Wea. Rev., 115, 463476, doi:10.1175/1520-0493(1987)115<0463:TGZITC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Barcilon, A., , J. C. Jusem, , and P. G. Drazin, 1979: On the two-dimensional hydrostatic flow of a stream of moist air over a mountain ridge. Geophys. Astrophys. Fluid Dyn., 13, 125140, doi:10.1080/03091927908243765.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-L., , and A. J. Nash, 1994: Diurnal variation of surface airflow and rainfall frequencies on the island of Hawaii. Mon. Wea. Rev., 122, 3456, doi:10.1175/1520-0493(1994)122<0034:DVOSAA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-L., , and J. Feng, 1995: The influences of inversion height on precipitation and airflow over the island of Hawaii. Mon. Wea. Rev., 123, 16601676, doi:10.1175/1520-0493(1995)123<1660:TIOIHO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Damiani, R., and et al. , 2008: The cumulus, photogrammetric, in situ, and Doppler observations experiment of 2006. Bull. Amer. Meteor. Soc., 89, 5773, doi:10.1175/BAMS-89-1-57.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Durran, D., , and J. Klemp, 1982b: On the effects of moisture on the Brunt-Väisälä frequency. J. Atmos. Sci., 39, 21522158, doi:10.1175/1520-0469(1982)039<2152:OTEOMO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 913928, doi:10.1175/2007MWR2059.1.

    • Search Google Scholar
    • Export Citation
  • Fuhrer, O., , and C. Schär, 2005: Embedded cellular convection in moist flow past topography. J. Atmos. Sci., 62, 28102828, doi:10.1175/JAS3512.1.

    • Search Google Scholar
    • Export Citation
  • Gaberšek, S., , and D. R. Durran, 2004: Gap flows through idealized topography. Part I: Forcing by large-scale winds in the nonrotating limit. J. Atmos. Sci., 61, 28462862, doi:10.1175/JAS-3340.1.

    • Search Google Scholar
    • Export Citation
  • Houze, R. A., Jr., 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Janjic, Z. I., 1996: The surface layer in the NCEP Eta model. Proc. 11th Conf. on Numerical Weather Prediction, Norfolk, VA, Amer. Meteor. Soc., 354–355.

  • Jensen, J., , and B. S. Lee, 2008: Giant sea-salt aerosols and warm rain formation in marine stratocumulus. J. Atmos. Sci., 65, 36783694, doi:10.1175/2008JAS2617.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, H., , G. Feingold, , and A. Sorooshian, 2010: Effect of aerosol on the susceptibility and efficiency of precipitation in warm trade cumulus clouds. J. Atmos. Sci., 67, 35253540, doi:10.1175/2010JAS3484.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, Q., 2003: Moist dynamics and orographic precipitation. Tellus, 55, 301316, doi:10.1034/j.1600-0870.2003.00025.x.

  • Kirshbaum, D. J., 2011: Cloud-resolving simulations of deep convection over a heated mountain. J. Atmos. Sci., 68, 361378, doi:10.1175/2010JAS3642.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and D. R. Durran, 2005: Observations and modeling of banded orographic convection. J. Atmos. Sci., 62, 14631479, doi:10.1175/JAS3417.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and R. B. Smith, 2009: Orographic precipitation in the tropics: Large-eddy simulations and theory. J. Atmos. Sci., 66, 25592578, doi:10.1175/2009JAS2990.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and C.-C. Wang, 2014: Boundary-layer updrafts driven by airflow over heated terrain. J. Atmos. Sci., 71, 14251442, doi:10.1175/JAS-D-13-0287.1.

    • Search Google Scholar
    • Export Citation
  • Klemp, J. B., , J. Dudhia, , and A. D. Hassiotis, 2008: An upper gravity-wave absorbing layer for NWP applications. Mon. Wea. Rev., 136, 39874004, doi:10.1175/2008MWR2596.1.

    • Search Google Scholar
    • Export Citation
  • Knievel, J. C., , G. H. Bryan, , and J. P. Hacker, 2007: Explicit numerical diffusion in the WRF model. Mon. Wea. Rev., 135, 38083824, doi:10.1175/2007MWR2100.1.

    • Search Google Scholar
    • Export Citation
  • Mårtensson, E. M., , E. D. Nilsson, , G. de Leeuw, , L. H. Cohen, , and H.-C. Hansson, 2003: Laboratory simulations and parameterization of the primary marine aerosol production. J. Geophys. Res., 108, 4297, doi:10.1029/2002JD002263.

    • Search Google Scholar
    • Export Citation
  • Miglietta, M. M., , and R. Rotunno, 2009: Numerical simulations of conditionally unstable flows over a mountain ridge. J. Atmos. Sci., 66, 18651885, doi:10.1175/2009JAS2902.1.

    • Search Google Scholar
    • Export Citation
  • Minder, J. R., , R. B. Smith, , and A. D. Nugent, 2013: The dynamics of ascent-forced orographic convection in the tropics: Results from Dominica. J. Atmos. Sci., 70, 40674088, doi:10.1175/JAS-D-13-016.1.

    • Search Google Scholar
    • Export Citation
  • Monahan, E. C., , and I. O’Muircheartaigh, 1980: Optimal power-law description of oceanic whitecap coverage dependence on wind speed. J. Phys. Oceanogr., 10, 20942099, doi:10.1175/1520-0485(1980)010<2094:OPLDOO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Nuijens, L., , B. Stevens, , and A. P. Siebesma, 2009: The environment of precipitating shallow cumulus convection. J. Atmos. Sci., 66, 19621979, doi:10.1175/2008JAS2841.1.

    • Search Google Scholar
    • Export Citation
  • Pierrehumbert, R. T., , and B. Wyman, 1985: Upstream effects of mesoscale mountains. J. Atmos. Sci., 42, 9771003, doi:10.1175/1520-0469(1985)042<0977:UEOMM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rasmussen, R. M., , P. Smolarkiewicz, , and J. Warner, 1989: On the dynamics of Hawaiian cloud bands: Comparison of model results with observations and island climatology. J. Atmos. Sci., 46, 15891608, doi:10.1175/1520-0469(1989)046<1589:OTDOHC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rauber, R. M., and et al. , 2007: Rain in shallow cumulus over the ocean: The RICO Campaign. Bull. Amer. Meteor. Soc., 88, 19121928, doi:10.1175/BAMS-88-12-1912.

    • Search Google Scholar
    • Export Citation
  • Reisner, J. M., , and P. K. Smolarkiewicz, 1994: Thermally forced low Froude number flow past three-dimensional obstacles. J. Atmos. Sci., 51, 117133, doi:10.1175/1520-0469(1994)051<0117:TFLFNF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 602618, doi:10.1175/2010JAS3558.1.

    • Search Google Scholar
    • Export Citation
  • Russotto, R. D., , T. Storelvmo, , and R. B. Smith, 2013: Modeling aerosol activation in a tropical, orographic, island setting: Sensitivity tests and comparison with observations. Atmos. Res., 134, 1223, doi:10.1016/j.atmosres.2013.07.017.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., , and M. L. Weisman, 2009: The impact of positive-definite moisture transport on NWP precipitation forecasts. Mon. Wea. Rev., 137, 488494, doi:10.1175/2008MWR2583.1.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and et al. , 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]

  • Smith, R. B., 1989: Mountain-induced stagnation points in hydrostatic flow. Tellus,41A, 270–274, doi:10.1111/j.1600-0870.1989.tb00381.x.

  • Smith, R. B., 2011: Regime diagrams for K-theory dispersion. Bound.-Layer Meteor., 139, 501519, doi:10.1007/s10546-011-9594-4.

  • 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, 353378, doi:10.1002/qj.49710845605.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , and V. Grubišić, 1993: Aerial observations of Hawaii wake. J. Atmos. Sci., 50, 37283750, doi:10.1175/1520-0469(1993)050<3728:AOOHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , P. Schafer, , and D. J. Kirshbaum, 2009a: Orographic precipitation in the tropics: Experiments in Dominica. J. Atmos. Sci., 66, 16981716, doi:10.1175/2008JAS2920.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , P. Schafer, , D. J. Kirshbaum, , and E. Regina, 2009b: Orographic enhancement of precipitation inside Hurricane Dean. J. Hydrometeor., 10, 820831, doi:10.1175/2008JHM1057.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., and et al. , 2012: Orographic precipitation in the tropics: The Dominica experiment. Bull. Amer. Meteor. Soc., 93, 15671579, doi:10.1175/BAMS-D-11-00194.1.

    • Search Google Scholar
    • Export Citation
  • Smolarkiewicz, P. K., , R. M. Rasmussen, , and T. L. Clark, 1988: On the dynamics of Hawaiian cloud bands: Island forcing. J. Atmos. Sci., 45, 18721905, doi:10.1175/1520-0469(1988)045<1872:OTDOHC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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.

  • Tabary, P., 2007: The new French operational radar rainfall product. Part I: Methodology. Wea. Forecasting, 22, 393408, doi:10.1175/WAF1004.1.

    • Search Google Scholar
    • Export Citation
  • Tian, W.-S., , and D. J. Parker, 2002: Two-dimensional simulation of orographic effects on mesoscale boundary-layer convection. Quart. J. Roy. Meteor. Soc., 128, 19291952, doi:10.1256/003590002320603476.

    • Search Google Scholar
    • Export Citation
  • Warner, J., 1968: A reduction in rainfall associated with smoke from sugar-cane fires—An inadvertent weather modification? J. Appl. Meteor., 7, 247251, doi:10.1175/1520-0450(1968)007<0247:ARIRAW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Woodcock, A. H., 1960: The origin of trade-wind orographic shower rains. Tellus, 12, 315326, doi:10.1111/j.2153-3490.1960.tb01316.x.

  • Wulfmeyer, V., and et al. , 2008: The convective and orographically induced precipitation study. Bull. Amer. Meteor. Soc., 89, 14771486, doi:10.1175/2008BAMS2367.1.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Map of Dominica showing terrain elevation, flight tracks with names and elevations, the FWL weather station, and rain gauge locations. Flight legs with two elevations shown are referred to as low (L: 0.3 km) and high (H: 1.2 km). The flight tracks discussed in the text (legs 1, 3, and 4) are in bold. The distance and direction to the Guadeloupe sounding site, Martinique radar, and the Trinite-Caravelle weather station (WS) are shown. [Modified from Smith et al. (2012).]

  • View in gallery

    Time series throughout the DOMEX field project of the variable (top) temperature, (middle) precipitation, and (bottom) wind speed. Temperature and wind speed measurements are from the FWL weather station and from the aircraft along upstream leg 1L. Upstream leg 1H is also included for temperature. Precipitation measurements are from the FWL weather station and as estimated over the mountain region from the Martinique radar (from Smith et al. 2012.)

  • View in gallery

    The idealized conditionally unstable sounding used to initialize WRF shown on a standard skew T–logp diagram with reference lines for the saturation mixing ratio and dry and saturated adiabats Atmospheric temperature (solid) and dewpoint temperature (dashed) are plotted.

  • View in gallery

    A few examples of the wind speed profiles used in WRF simulations. Profiles are named by their surface wind speed Us which stays constant from the surface to z = 600 m (i.e., mixed layer). Wind speeds decrease with varying slopes above 600 m and the profiles intersect at 6 km and −10 m s−1 to create reverse shear aloft. The mixed-layer height (600 m) and mountain height (1 km) are shown. Also included are the mean and standard deviation of surface wind speed from the 2010 and 2011 Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) data upstream of Dominica.

  • View in gallery

    Model average from the 3D simulations showing horizontal wind speed (m s−1) in color, dry potential temperature (°C) contours (thin), and a thick black contour of average cloud location (0.2 g kg−1). Fields are averaged over hours 2–4 and along the center 40 km of the mountain axis in the cross-stream direction. (a) A low-wind case with Us = 2 m s−1 and (b) a high-wind case with Us = 10 m s−1.

  • View in gallery

    A sample image of the 3D domain with contours of cloud liquid water content (qc = 0.2 g kg−1). The surface color indicates the potential temperature (°C) at the lowest model level. (a) A low-wind case with Us = 2 m s−1 and (b) a high-wind case with Us = 10 m s−1 as viewed from the northwest looking southeast.

  • View in gallery

    Model cross section of the mountain. The boxes outline the averaging areas used for calculating the three convective indices: Cloud fraction (CF), divergence (Div), and mountaintop temperature. The L3 and L4 areas are representative of the general region where the aircraft-obtained observations. Three-dimensional simulations use the same averaging regions for consistency but are extended along the mountain axis for the center 40 km of the 50-km-long finite ridge.

  • View in gallery

    Colored shading is the U (m s−1) component of the wind speed from model snapshots. A blue contour of 0 m s−1 is also included. These figures allow a visual of the sense of divergence/convergence above the mountain shown by the black arrows at 2-km height. (a) 2 h into a 2D low-wind case with Us = 2 m s−1; (b) as in (a), but for a 3D simulation; (c) 2 h into a high-wind case with Us = 10 m s−1; and (d) as in (c), but for the 3D simulation.

  • View in gallery

    (a) Divergence/convergence as calculated from aircraft data on legs 3 and 4 using (3) plotted against wind speed measured upstream along leg 1L (from Smith et al. 2012.) (b) Divergence as calculated from 2D and 3D model output using legs 3 and 4 averaging areas shown in Fig. 7 and using (4). Both figures show equations and R values for the best-fit lines to the data.

  • View in gallery

    Snapshot of cloud liquid water in the model. (a) A 2D low-wind case with Us = 2 m s−1, 5 h into the simulation; (b) as in (a), but for a 3D simulation; (c) a 2D high-wind case Us = 10 m s−1, 2 h into the simulation; and (d) as in (c), but for a 3D simulation.

  • View in gallery

    Plotted against various measures of wind speed are (a) cloud fraction using (5) and (b) latent heat flux using (6) from the aircraft along legs 3 and 4 (see Fig. 1). (c) Cloud fraction from 3D and 2D model data averaged over the L3 and L4 averaging-area boxes (Fig. 7); note the change in scale. (d) Mean solar radiation vs mean wind speed in the morning both from the FWL weather station. All panels show equations and R values for the best-fit lines to the data.

  • View in gallery

    Estimates of air parcel warming (Δθ) from the upstream ocean surface to the FWL weather station plotted against wind speed. Potential temperatures and wind speeds were averaged from 1200 to 1300 local time and the average upstream θ from leg 1L was used as a baseline. Data points are labeled as clear or cloudy based on above- or below-average solar insolation. The dashed lines show the model fit from (8) with Q = 50, 100, and 200 W m−2. The solid blue line shows Δθ from 2D simulations averaged over the Mtn Temp box shown in Fig. 7. Three-dimensional simulation values are shown with blue filled circles.

  • View in gallery

    Correlation coefficient using (9) of the daily total precipitation amount over Dominica from the FWL rain gauge with the standard levels of the Guadeloupe sounding against (a) wind speed and (b) relative humidity. Three nonoverlapping 370-day time series segments from 2007 to 2011 are correlated for each variable (thin black lines) showing that correlation strength is not dependent on the time period. The thick black line shows the correlation coefficient for all available data.

  • View in gallery

    Diurnal cycle of precipitation from rain gauges over Dominica for (top to bottom): mountaintop, windward side, and leeside. Two rain-rate curves for each category are shown with symbols referring to their locations. Mountaintop sites are BL: Boeri Lake and FWL: Freshwater Lake. The windward sites are GF: Grand Fond and RO: Rosalie. The leeward coast sites are BG: Botanical Garden and CF: Canefield airport. Diurnal cycles are calculated from 3.3–4.2 years’ worth of hourly data from 2007 to 2011.

  • View in gallery

    Two-dimensional model snapshot of the normalized tracer distribution originating from the mountaintop. (a) A low-wind case with Us = 2 m s−1, 5 h into the simulation; and (b) a high-wind case with Us = 10 m s−1, 2 h into the simulation.

  • View in gallery

    Wind speeds during the DOMEX field project period from the weather stations at FWL and Trinite-Caravelle on Martinique, the Guadeloupe rawinsonde sounding, ERA-Interim data, and aircraft RF leg 1L. See Table A1 for additional information.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 215 215 14
PDF Downloads 159 159 10

Wind Speed Control of Tropical Orographic Convection

View More View Less
  • 1 Yale University, New Haven, Connecticut
  • | 2 University at Albany, State University of New York, Albany, New York
© Get Permissions
Full access

Abstract

This study compares observations from the Dominica Experiment (DOMEX) field campaign with 3D and 2D Weather Research and Forecasting Model (WRF) simulations to understand how ambient upstream wind speed controls the transition from thermally to mechanically forced moist orographic convection. The environment is a conditionally unstable, tropical atmosphere with shallow trade wind cumulus clouds. Three flow indices are defined to quantify the convective transition: horizontal divergence aloft, cloud location, and island surface temperature. As wind speed increases, horizontal airflow divergence from plume detrainment above the mountain changes to convergence associated with plunging flow, convective clouds relocate from the leeward to the windward side of the mountain as mechanically triggered convection takes over, and the daytime mountaintop temperature decreases because of increased ventilation and cloud shading. Possible mechanisms by which wind speed controls island precipitation are also discussed. The result is a clearer understanding of orographic convection in the tropics.

Corresponding author address: Alison D. Nugent, Yale University, Department of Geology and Geophysics, 210 Whitney Ave., New Haven, CT 06511. E-mail: alison.nugent@yale.edu

Abstract

This study compares observations from the Dominica Experiment (DOMEX) field campaign with 3D and 2D Weather Research and Forecasting Model (WRF) simulations to understand how ambient upstream wind speed controls the transition from thermally to mechanically forced moist orographic convection. The environment is a conditionally unstable, tropical atmosphere with shallow trade wind cumulus clouds. Three flow indices are defined to quantify the convective transition: horizontal divergence aloft, cloud location, and island surface temperature. As wind speed increases, horizontal airflow divergence from plume detrainment above the mountain changes to convergence associated with plunging flow, convective clouds relocate from the leeward to the windward side of the mountain as mechanically triggered convection takes over, and the daytime mountaintop temperature decreases because of increased ventilation and cloud shading. Possible mechanisms by which wind speed controls island precipitation are also discussed. The result is a clearer understanding of orographic convection in the tropics.

Corresponding author address: Alison D. Nugent, Yale University, Department of Geology and Geophysics, 210 Whitney Ave., New Haven, CT 06511. E-mail: alison.nugent@yale.edu

1. Introduction and background

Two mechanisms initiate atmospheric convection over mountains: surface heating and forced ascent (Houze 1993). The former is driven by diurnal solar radiation on mountain slopes. Buoyant air rises up the slopes, converging to form localized updrafts and clouds. In contrast, strong wind approaching a mountain is forced to rise by the terrain slope. The air reaches the lifting condensation level and clouds form. In a conditionally unstable atmosphere, cloudy air will accelerate vertically and convection will begin. While both mechanisms have been observed and modeled, the transition between them deserves closer analysis and quantification.

This study examines convection over the Commonwealth of Dominica—a small mountainous island in the eastern Caribbean Sea (15°N, 61°W). Convection over Dominica is sometimes driven by surface heating and sometimes by mechanical uplift or forced ascent (Smith et al. 2012). Two recent studies have focused on mechanically driven convection over Dominica through observations and modeling (Kirshbaum and Smith 2009; Minder et al. 2013). Here, based on observations discussed in section 2 and comparisons with a numerical model described in section 3, we explore the characteristics of both types of convection and the role of wind speed.

a. Thermal convection

Studies involving thermal forcing of orographic convection include continental and maritime mountains and both deep and shallow convection (e.g., Banta and Schaaf 1987; Damiani et al. 2008; Wulfmeyer et al. 2008; Robinson et al. 2011; Kirshbaum 2011). For example, the Cumulus, Photogrammetric, In Situ, and Doppler Observations Experiment (CuPIDO) in Arizona (Damiani et al. 2008) described deep convection over arid continental terrain, while Robinson et al. (2011) and Sobel et al. (2011) considered tropical island convection with a variety of scales. Sobel et al. (2011) used Tropical Rainfall Measuring Mission (TRMM) data to conclude that islands with a large land area (>315 km2) are more likely than small islands (<315 km2) to have a pronounced thermal effect on rain rate and rain frequency. Precipitation over small tropical islands may require mechanical forcing from upslope flow because of their small surface area. We expand this view for a midsized island (754 km2) by illustrating how both thermal forcing and mechanical forcing play a role.

b. Mechanical convection

For mechanical forcing to occur, air must pass over rather than detour around a mountain. For this, the nondimensional mountain height (Nh/U)—where N (s−1) is the Brunt–Väisälä frequency, h (m) is the mountain height, and U (m s−1) is the wind speed—sometimes referred to as the inverse Froude number, needs to be small; some combination of a moderate mountain height, weak stratification, and strong wind speed are needed. This measure is especially useful for dry flow with a simple vertical structure. Air lifted over a mountain will form either smooth stratiform or turbulent convective clouds. The latter is more common in the tropics as conditional instability is common in the lower free troposphere.

Numerical studies have examined how various parameters affect the strength and organization of mechanically forced convection over an isolated mountain (e.g., Tian and Parker 2002; Fuhrer and Schär 2005; Kirshbaum and Durran 2005; Miglietta and Rotunno 2009). These parameters have included the height and shape of the mountain, temperature and stability of the upstream sounding, wind speed and wind shear, as well as simulation parameters like domain length and resolution.

c. Previous wind speed comparisons

Surface heating and mechanical lifting can combine to induce convection over an isolated mountain (e.g., Tian and Parker 2002; Kirshbaum and Wang 2014). The relative importance of each mechanism depends on the wind speed. A nondimensional number introduced by Smith and Lin (1982) can help determine the relative importance of the surface heating (Q) and forced orographic ascent (mtn) in producing a vertical air displacement η:
e1
where is the surface sensible heating (W m−2), g is the gravitational constant (9.81 m s−2), b is the heating width (m), ρ is the air density (kg m−3), Cp is the specific heat capacity of dry air at constant pressure (1004 J kg−1 K−1), is the average column temperature (K) and is the average wind speed (m s−1). Values of for the current study are discussed in section 4c.

The role of wind speed on orographic convection has been investigated numerically by Tian and Parker (2002). They compared low wind speeds (2 m s−1) and high wind speeds (10 m s−1) in 2D idealized simulations with various hill widths and heights (up to 500 m) and surface heating magnitudes (up to 480 W m−2). They did not consider the effects of latent heat on their solutions and focused instead on the turbulence closure schemes and mesoscale boundary layer convective features. They found thermal effects of the hill were dominant under low-wind conditions and with stronger background flow, mechanical forcing by the hill was more important for boundary layer convective features.

Miglietta and Rotunno (2009) conducted a number of idealized simulations with conditionally unstable flow to generate orographically triggered deep moist convection. No surface sensible heating was applied. They tested intermediate (10 m s−1), high (15 and 20 m s−1), and low wind speeds (2.5 and 5 m s−1) in a 3D model with infinite mountain ridges of various heights and widths. At low wind speeds, cold-air outflow at the surface propagating far upstream was important for cell redevelopment but prevented any steady convection from forming over the mountain. At intermediate and high wind speeds, steady rain was produced over the windward mountain slope. A cold pool was generated at intermediate wind speeds but only on the downstream side of the ridge. The mechanical lifting role of the orography in forcing precipitation became more important with stronger wind speeds.

Hawaii is the most studied tropical island (20°N, 158°W) for orographic processes and numerous field studies, including the Hawaiian Rainband Project (HARP; 1990) field campaign, have observed how airflow and precipitation evolve over and around the island (e.g., Woodcock 1960; Smith and Grubišić 1993; Chen and Nash 1994; Esteban and Chen 2008). It has peaks that rise over 4 km in height, well above the trade wind inversion (~2 km). Because of its high upstream static stability and altitude, its nondimensional mountain height (Nh/U) is generally around five, which results in a stagnation point upstream of and diverging airflow around the island (Smolarkiewicz et al. 1988; Rasmussen et al. 1989; Reisner and Smolarkiewicz 1994). Hawaii is therefore not an ideal study location for a simple case of pure mechanically forced convection.

The precipitation distribution on Hawaii depends on diverging airflow and diurnal heating and cooling. For example, Esteban and Chen (2008) found that precipitation is strongly influenced by nighttime katabatic flows that collide with the trade winds upstream of Hawaii and initiate convection. The relative strength of the trade winds in comparison to the strength of the katabatic flow determines the precipitation location. On the windward slopes, stronger trade winds generally correspond to higher daily rainfall amounts. The trade wind inversion height may also play a role in precipitation; higher inversion heights were found to have higher precipitation amounts because of a decrease in airflow divergence around Hawaii and thus a larger volume of ascending air on the windward slopes (Chen and Feng 1995).

The literature reviewed above suggests that island convection is sensitive to island height, area, conditional instability, surface temperature, cold pool dynamics, and wind speed.

2. DOMEX

The Dominica Experiment (DOMEX) field project provides a new observational dataset for a wind speed sensitivity analysis. The project centered on Dominica in the tropical trade wind belt (Fig. 1). Its mountainous peaks extend to 1400 m—typically above the lifting condensation level (LCL) but below the trade wind inversion. Orographic convective clouds above Dominica generally remain shallow (<5 km).

Fig. 1.
Fig. 1.

Map of Dominica showing terrain elevation, flight tracks with names and elevations, the FWL weather station, and rain gauge locations. Flight legs with two elevations shown are referred to as low (L: 0.3 km) and high (H: 1.2 km). The flight tracks discussed in the text (legs 1, 3, and 4) are in bold. The distance and direction to the Guadeloupe sounding site, Martinique radar, and the Trinite-Caravelle weather station (WS) are shown. [Modified from Smith et al. (2012).]

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

The DOMEX field campaign during April and May of 2011 consisted of 21 research flights (RFs) on fixed flight tracks over Dominica (Fig. 1) with the University of Wyoming King Air aircraft. The flight pattern contained an upstream sounding (from 4 km to 150 m) and six horizontal legs. Legs 1, 2, and 5 were flown over the ocean at 300 m and 1.2 km. Legs 3 and 4 were flown over the island at 1.7 km. Leg 1 observed undisturbed upstream conditions while legs 3 and 4 measured over island convection. The aircraft recorded in situ quantities including the three components of velocity, temperature, moisture content, droplet size, and aerosol concentration at a frequency of 1 Hz (Δx ~ 90 m) and 25 Hz (Δx ~ 4 m).

To gain a more quantitative understanding of the effects of wind speed on convection, we define three indices of the convective type: (i) the horizontal airflow divergence above the mountain convection, (ii) the location of convection over the windward or leeward slopes, and (iii) the mountaintop temperature. The use of these indices helps to quantify the nature of the flow field and to aid comparison between DOMEX data and idealized 3D and 2D Weather Research and Forecasting (WRF) simulations described in section 3.

The trade winds varied in magnitude from day to day during the field project. DOMEX cases included a range of trade wind speeds from 1 to 12 m s−1, measured at 300 m height upstream over the ocean. The aircraft’s upstream leg 1L (Fig. 1) provided a reliable measure of wind speed during each RF, and was comparable to observations from other wind measuring platforms (see the appendix for details). The easterly trade winds near the surface typically decrease in strength with height and reach a critical level around 2–3 km. The trade wind inversion in temperature is usually found near 2 km height (Smith et al. 2009a). Coastal surface observations show that a katabatic downslope wind sometimes occurs in the night and early morning on Dominica’s east coast but always disappears by 0800 or 0900 local time. Offshore flow was never seen by the aircraft on leg 2L which carefully follows the windward coastline. These observations seem to rule out any cold pool or katabatic flow dynamics as an initiation mechanism for convection as suggested by Miglietta and Rotunno (2009) and the HARP field studies.

Figure 2 presents wind speed, precipitation, and temperature observed during DOMEX from 5 April to 10 May 2011. Over island precipitation from the Martinique radar (estimated via the methodology of Tabary 2007) and the Freshwater Lake (FWL) weather station located at 850-m height is included. Wind speed and temperature from FWL, and upstream aircraft leg 1L (300 m) are shown along with temperature from leg 1H (1.2 km). A diurnal temperature cycle over the island is evident, as are a number of high and low wind periods, and wet and dry periods. Here, high wind refers to upstream wind speeds in the boundary layer below 600 m that are above the climatological average of ~6 m s−1 and low wind refers to wind speeds below the climatological average.

Fig. 2.
Fig. 2.

Time series throughout the DOMEX field project of the variable (top) temperature, (middle) precipitation, and (bottom) wind speed. Temperature and wind speed measurements are from the FWL weather station and from the aircraft along upstream leg 1L. Upstream leg 1H is also included for temperature. Precipitation measurements are from the FWL weather station and as estimated over the mountain region from the Martinique radar (from Smith et al. 2012.)

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

The nondimensional mountain height is relevant to the question of whether airflow goes over a mountain and whether plunging flow occurs on the lee side. Estimates for Dominica under high wind speed conditions yield values of Nh/U = 0.71–1.91 (Minder et al. 2013) which are marginal for both phenomena (Smith 1989; see also section 4c). Other factors such as sensible heat flux, latent heating due to condensation of clouds, and reverse shear aloft reduce the utility of the nondimensional mountain height. Minder et al. (2013, their Fig. 5) includes vector wind diagrams from DOMEX observations from aircraft leg 2L which show only minor flow deflection around the island. They attribute this lack of blocking to latent heat release.

3. Numerical simulations

Numerical simulations were performed using WRF (Skamarock et al. 2008) version 3.2 to compare with DOMEX observations and to explore convection with a full range of wind speeds. WRF is a fully compressible nonhydrostatic model on a mass-based terrain-following grid. The entire set of 3D and 2D simulation setups was designed to be as idealized as possible while still capturing trends with wind speed. The 3D simulations with 200-m grid spacing focus on one low- and one high-wind case and are useful for accurately capturing convective-scale features. A larger set of 2D simulations with 1-km grid spacing acts as a complimentary wind speed sensitivity study that spans the wind speed spectrum. The main differences between the 3D and 2D simulations are highlighted in Table 1 and discussed below. As will be shown, even though 2D simulations are highly idealized, they are still able to capture the basic dynamics of the flow regimes present.

Table 1.

The main differences between 3D and 2D simulations. See the text for further information.

Table 1.

In all simulations, a mountain was centered in a 200-km domain with a 1-km maximum height (h0) and a 10-km half-width (a). The 3D simulations have a 200-km square domain and a mountain with the same horizontal cross section as the 2D case but a finite length of 50 km. The terrain is given by (2) [used by Gaberšek and Durran (2004), among others]:
e2
where a constant 200 W m−2 sensible heat flux and surface drag were added over the terrain, parameterized by the Eta surface scheme (Janjic 1996) based on Monin–Obukhov theory using a roughness length of 80 cm. We include no sensible heat flux over the ocean, and neither radiation nor PBL parameterizations are used.

An idealized sounding (Fig. 3) was used to initialize the model (from Minder et al. 2013). Below 4 km, the temperature and moisture fields in the idealized sounding match a typical DOMEX aircraft sounding while the region above 4 km has uniform stratification. Thermodynamic variables in the idealized sounding were based on a high-wind research flight (RF 13). Soundings from low-wind research flights had nearly identical temperature profiles to high-wind soundings but tended to be drier in the cloud layer. The sounding is conditionally unstable, the LCL is located at 933 hPa (762 m) and it contains 4 cm of precipitable water, 553 J kg−1 of surface based convective available potential energy (CAPE), and a level of free convection at 980 m. CAPE in this sounding is higher than most days during the field project, but CAPE varies widely from day to day. CAPE is not very useful here since we are interested in shallow convection capped by the trade wind inversion. In the process of simplifying the sounding a negligible amount of convective inhibition was added below the LCL (≤5 J kg−1). The wind speed is constant in the lowest 600 m and is varied from 0 to 15 m s−1. Above this height the wind speed decreases linearly to −10 m s−1 at 6 km to where it remains constant (Fig. 4). Westerlies aloft (i.e., antitrades) were common during the DOMEX period and climatologically (Smith et al. 2009a). On the high-wind days, the critical level is higher and the wind shear is stronger than on low-wind days. The critical level is the level of no wind and may be important to the dynamics of plunging flow.

Fig. 3.
Fig. 3.

The idealized conditionally unstable sounding used to initialize WRF shown on a standard skew T–logp diagram with reference lines for the saturation mixing ratio and dry and saturated adiabats Atmospheric temperature (solid) and dewpoint temperature (dashed) are plotted.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Fig. 4.
Fig. 4.

A few examples of the wind speed profiles used in WRF simulations. Profiles are named by their surface wind speed Us which stays constant from the surface to z = 600 m (i.e., mixed layer). Wind speeds decrease with varying slopes above 600 m and the profiles intersect at 6 km and −10 m s−1 to create reverse shear aloft. The mixed-layer height (600 m) and mountain height (1 km) are shown. Also included are the mean and standard deviation of surface wind speed from the 2010 and 2011 Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) data upstream of Dominica.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

A 1.5-order turbulent kinetic energy (TKE) closure was used along with a monotonic sixth-order filter to damp grid-scale variability (Knievel et al. 2007). Periodic lateral boundary conditions are used, though open boundary conditions provided nearly identical results. The Coriolis force is included on perturbations to the flow using an f-plane approximation, with f = 3.9 × 10−5 s−1. Time stepping was done with a third-order Runge–Kutta method, while advection used fifth-order horizontal and third-order vertical methods. A positive-definite limiter was applied during the advection of microphysical variables (Skamarock and Weisman 2009). A sponge layer above 6 km damps vertical velocities to avoid unrealistic reflection off the model top at 12 km (Klemp et al. 2008). This is justified as the atmospheric processes above 6 km are of minor importance to the present problem. The simulations were run for a minimum of 3 h although low wind speed simulations were run for a longer time. Solutions only reach a quasi-steady state so simulation times of 2–5 h are typically used for analysis during which the change in conditions is acceptably small.

Other main differences between 3D and 2D simulations include horizontal and vertical grid spacing and aspects of the model setup like seeding and lateral damping, which are forced by the dimensionality. While both 3D and 2D simulations use variable vertical resolution, the number of vertical levels and height of the lowest model level differ. To break symmetry in the 3D simulations, the model is seeded over the entire domain at startup by random temperature perturbations of the range ±1 K between z = 1 and 4 km height (e.g., Kirshbaum and Smith 2009). At the upstream and downstream edges of the 2D domain, a 20-gridpoint-wide region of damping linearly relaxes the winds, temperature, and moisture in the model back toward the initial sounding. Steady state is more closely reached in the 3D simulations. As is common with 2D models, there is a slight drift in solutions as a weak blocking wave moves upstream (e.g., columnar disturbance) (Pierrehumbert and Wyman 1985).

4. Characteristics of the model flow fields

Characteristics of the flow field differ strongly between the high and low wind speed simulations. These differences are summarized below.

a. Low-wind thermal convection

Forced by surface heating, air rises up the windward and leeward sides of the mountain and converges near the mountaintop (Fig. 5a). Shallow convective clouds (3–4-km tops) are found over the mountain, with a maximum shifted slightly toward the leeward side (Figs. 5a and 6a). Divergence is found aloft associated with cloud-top outflow.

Fig. 5.
Fig. 5.

Model average from the 3D simulations showing horizontal wind speed (m s−1) in color, dry potential temperature (°C) contours (thin), and a thick black contour of average cloud location (0.2 g kg−1). Fields are averaged over hours 2–4 and along the center 40 km of the mountain axis in the cross-stream direction. (a) A low-wind case with Us = 2 m s−1 and (b) a high-wind case with Us = 10 m s−1.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Fig. 6.
Fig. 6.

A sample image of the 3D domain with contours of cloud liquid water content (qc = 0.2 g kg−1). The surface color indicates the potential temperature (°C) at the lowest model level. (a) A low-wind case with Us = 2 m s−1 and (b) a high-wind case with Us = 10 m s−1 as viewed from the northwest looking southeast.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Three- and two-dimensional simulations have similar low wind flow features but as expected, 3D simulations with smaller grid spacing have more detail and finer-scale structures. Convection especially is more realistic in 3D low-wind simulations as convective cells are generated by heating over the windward and leeward sides of the island instead of being concentrated in a single leeside convective plume (see sections 5a and 5b and figures therein). In addition, the cloud base is slightly higher in 3D simulations.

b. High-wind mechanical convection

Convection is initiated just upstream of the island where air begins to be lifted (Fig. 5b). Stratified fluid dynamics (i.e., mountain waves and hydraulics) play a large role; air accelerates over the mountain peak and down the leeward slopes to where it undergoes a turbulent “hydraulic jump” in the lee. The hydraulic jump strength and location are sensitive to the strength of friction over the island and to the critical-level height (not shown). Shallow convective clouds are found almost exclusively over the windward side and peak of the mountain (Figs. 5b and 6b). Convergence is found aloft associated with the plunging flow (see section 5a).

Three- and two-dimensional simulations of high wind flow are also similar. One exception is the plunging flow in the lee of the mountain, which is more accurately captured in the 3D than the 2D simulations. In 3D, the jet stays close to the ground level as it does in DOMEX observations instead of separating from the surface (see section 5a and figures therein).

This same high-wind case has been previously modeled by Minder et al. (2013) with an identical thermodynamic sounding. While their setup included slightly more complexity than ours with three nested grids going down to 100-m grid spacing, real Dominican topography, and the Milbrandt and Yau microphysics scheme, our results are consistent. They focus on airflow deformation, plunging flow, the downwind wake and airmass transformation in high wind simulations. Here we stress the differences in the island-scale flow and drivers of convection with changes in wind speed.

c. Nondimensional parameters

The model flow fields discussed above are roughly consistent with the nondimensional numbers governing the airflow over Dominica (shown in Table 2). Consider first the nondimensional mountain height (Nh/U). While it is large in low wind conditions, there is little flow deflection around the mountain ridge in the 3D simulations owing in part to the surface heating and the conditionally unstable sounding. The nondimensional mountain height is calculated with a dry N instead of a moist Nm as some prior studies have suggested (e.g., Barcilon et al. 1979; Durran and Klemp 1982b; Jiang 2003). In practice, the use of Nm requires a stable saturated flow with a uniform lapse rate, neither of which applies to the present case study. In DOMEX, the variations of Nm in the vertical are too large to give an unambiguous value of Nm or Nmh/U useful for a measure of flow dynamics. The values for nondimensional mountain width (Na/U) suggest nearly hydrostatic flow except in convection and turbulence.

Table 2.

Nondimensional parameters estimated for Dominica observations and simulations. Measures of nonlinearity, hydrostaticity, advective to convective time scales, and the role of heating to orographic lifting [see (1)] are included. Low wind corresponds to Us = 2 m s−1 and high wind corresponds to Us = 10 m s−1. The temperature structure of the background flow does not change significantly from day to day, so for order of magnitude estimates, we can assume a typical value of N = 0.01 s−1. See section 4c for details.

Table 2.

Another important but nontraditional parameter included in Table 2 is the ratio of advection time a/U to the approximate time scale for the development of convection, or Tadv/Tconv = Nma/U. This parameter is a simplification of the control parameter used by Kirshbaum and Smith [2009, their (12)] neglecting entrainment and dry descent. Our estimate of Nm uses a θe gradient of −7.7 K km−1 measured upstream of Dominica and neglects virtual temperature effects for a simple approximation. The values in Table 2 show that even in the high wind days there is sufficient time for convection to grow over the island. The parameter from (1) is estimated using typical values (, b = 1000, ρ = 1.2, T = 300, and N = 0.01). This parameter clearly captures the wind speed transition: heating dominates at low wind speeds when > 1 while mechanical lifting dominates at high wind speeds when < 1 (see also Reisner and Smolarkiewicz 1994).

5. Indices of convective transition

The three indices described in section 2—horizontal divergence aloft, cloud fraction, and mountaintop temperature—are used to quantify the distinct regimes of convection and test the numerical model. They are determined from aircraft observations and model simulations. The model index values are computed using averaging areas representative of the region measured by the aircraft (Fig. 7). In the case of 3D simulations, the same averaging area is used for consistency but the area is also averaged in the along-ridge direction over the center 40 km of the 50-km ridge.

Fig. 7.
Fig. 7.

Model cross section of the mountain. The boxes outline the averaging areas used for calculating the three convective indices: Cloud fraction (CF), divergence (Div), and mountaintop temperature. The L3 and L4 areas are representative of the general region where the aircraft-obtained observations. Three-dimensional simulations use the same averaging regions for consistency but are extended along the mountain axis for the center 40 km of the 50-km-long finite ridge.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

a. Divergence aloft

The first index is the horizontal flow divergence above the mountain. Figure 8 helps to illustrate this index by showing xz cross sections of wind speed (U) from 2D (Figs. 8a,c) and 3D (Figs. 8b,d) simulations. In the low-wind case (Figs. 8a,b: Us = 2 m s−1), upslope flows are driven by heating on the mountain slopes and converge at the mountaintop. To compensate for convergence at low levels, divergence is found aloft. The height of this cloud top outflow is likely determined by the trade wind inversion height. In contrast, the high-wind case (Figs. 8c,d: Us = 10 m s−1) develops a plunging layer where air is lifted over the mountain and accelerates downslope in the lee. Above the mountain is a region of convergence where the fast-moving plunging air meets stagnant air aloft in the lee.

Fig. 8.
Fig. 8.

Colored shading is the U (m s−1) component of the wind speed from model snapshots. A blue contour of 0 m s−1 is also included. These figures allow a visual of the sense of divergence/convergence above the mountain shown by the black arrows at 2-km height. (a) 2 h into a 2D low-wind case with Us = 2 m s−1; (b) as in (a), but for a 3D simulation; (c) 2 h into a high-wind case with Us = 10 m s−1; and (d) as in (c), but for the 3D simulation.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Using aircraft data from legs 3 and 4, a divergence/convergence index was computed where the overbar denotes an average along the over-mountain part of the indicated flight leg (L):
e3

The crosswind component was found by rotating the coordinate system of the U and V components of wind speed measured by the aircraft gust probe to align with the island axis. Both flight legs are located over the island at 1.7-km height. Leg 3 is slightly upstream of the peaks on Dominica while leg 4 is slightly downstream (see Fig. 1). When the upstream trade wind speed along leg 1L is below ~(4–6) m s−1, airflow above the island shows a sense of divergence, changing gradually to convergence with increasing upstream trade wind speed (Fig. 9a). This change of the divergence index quantifies the change from cloud-top outflow at low wind speeds to plunging flow at high wind speeds.

Fig. 9.
Fig. 9.

(a) Divergence/convergence as calculated from aircraft data on legs 3 and 4 using (3) plotted against wind speed measured upstream along leg 1L (from Smith et al. 2012.) (b) Divergence as calculated from 2D and 3D model output using legs 3 and 4 averaging areas shown in Fig. 7 and using (4). Both figures show equations and R values for the best-fit lines to the data.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Divergence in the model was calculated in a similar way to (3) but used the Div(L3) and Div(L4) regions in Fig. 7 averaged over 2 h of model simulation time:
e4

A pattern similar to the observations was found; with increasing upstream wind speed, divergence changes gradually to convergence in the wind speed range from 4 to 6 m s−1. The 3D and 2D results are similar showing that the flow is well approximated even in a 2D domain. The simulated flow transitions from divergence to convergence at slightly lower wind speeds than observations. This is likely due to overly efficient heat transport in the numerical model when no boundary layer parameterization is used. This topic is further explored in section 5c.

b. Cloud fraction and latent heat flux

Cloud location is the second convection index. At low wind speeds, convection is dominated by surface heating over the mountain as the winds are too slow to cause ascent-forced convection and too slow to advect heat away from the mountain surface. The opposite is true at high wind speeds where surface heating has a much reduced effect and convection is initiated by orographic uplift on the windward side of the mountain (Tian and Parker 2002). These differences in convective initiation lead to differing cloud locations (Figs. 5 and 10). Curiously, the convective location shifts upwind with increasing wind speed. While the higher resolution 3D simulations have more detail, both 3D and 2D simulations show similar features. The upstream shift of clouds with strong wind speed is further evidence (see section 2) that cold pool dynamics are not active.

Fig. 10.
Fig. 10.

Snapshot of cloud liquid water in the model. (a) A 2D low-wind case with Us = 2 m s−1, 5 h into the simulation; (b) as in (a), but for a 3D simulation; (c) a 2D high-wind case Us = 10 m s−1, 2 h into the simulation; and (d) as in (c), but for a 3D simulation.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Cloud fraction (CF) along a flight leg is used as an index of convective location. The Forward Scattering Spectrometer Probe (FSSP-100) on board the aircraft measures the size distribution of cloud droplets in the range of 2–47 μm. Using a minimum cloud droplet concentration of 15 cm−3 to indicate the presence of a cloud, CF is computed for a given leg by the following equation, where X indicates the number of 1-s observations along the flight leg:
e5
CF was calculated for the over-island portion of both legs 3 and 4 (Fig. 11a). On upwind leg 3, CF increases with increasing wind speed. On lee side leg 4, CF decreases with increasing wind speed. Correlation coefficient R values are low but the two best-fit lines are significantly different from each other and from zero at a confidence interval of 99% based on a two-sample t test. While the sampling statistics of this quantity from our aircraft observations is poor, it has consistency with other included measurements and allows insight into this important aspect of the Dominican climatology.
Fig. 11.
Fig. 11.

Plotted against various measures of wind speed are (a) cloud fraction using (5) and (b) latent heat flux using (6) from the aircraft along legs 3 and 4 (see Fig. 1). (c) Cloud fraction from 3D and 2D model data averaged over the L3 and L4 averaging-area boxes (Fig. 7); note the change in scale. (d) Mean solar radiation vs mean wind speed in the morning both from the FWL weather station. All panels show equations and R values for the best-fit lines to the data.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

In addition to the CF, latent heat flux (LHF in watts per square meter) can also be computed along a flight leg with 1-s flight level data. LHF is a measure of the strength of moist convection calculated by (6) where the primes indicate that the leg mean has been removed:
e6

Here, w is vertical velocity (m s−1), q is specific humidity (kg kg−1), ρ is air density (kg m−3), Lυ is the latent heat of vaporization (J kg−1), Δx is the distance between measurements, and l is the flight leg length (both in meters). Figure 11b shows LHF as calculated on legs 3 and 4 for comparison with Fig. 11a. A similar pattern is seen in the LHF; it increases with increasing wind speed on leg 3 and decreases on leg 4.

The CF was measured in the model simulations in a comparable way using (5), but XCloud and XTotal refer to the number of cloudy and total grid points in the averaging areas labeled CF(L3) and CF(L4) in Fig. 7. Model results (Fig. 11c) have the same general pattern found from observations. The controlled nature of the experiment and the larger sample volume reduce the variability of this computed index. Three-dimensional simulations are able to capture the magnitude of the cloud fraction from observations in both the L3 and L4 regions. Clouds simulated in 2D are wider because of the larger grid spacing and have less variable spacing owing to the restricted dimensionality. In both simulation sets, CF increases with increasing wind speed in the upstream area and decreases on the lee side.

One additional measure of cloud location comes from the FWL weather station on an east-facing windward ridge on Dominica. Solar radiation reaching this station in the morning as the sun rises can be considered a proxy of cloudiness over the windward slope. In Fig. 11d, mean shortwave insolation is plotted against mean wind speed, as measured by the FWL weather station. In the morning (0700–0900 local time) solar radiation is highest at low wind speeds and decreases with increasing wind speed. This is consistent with expectations since cloud fraction is highest on the windward side at high wind speeds, blocking solar radiation from reaching the FWL weather station. No significant correlation between afternoon insolation and wind speed is observed, probably because the east-facing location of the weather station has a poor view of the western sky. The solar radiation, cloud fraction, and latent heat flux data all support the idea that the clouds shift upwind with increasing wind speed.

c. Mountaintop temperature

The change in mountaintop temperature is the final index showing the role of wind speed. Surface temperature is the driver for thermal convection. Close inspection of Fig. 2 shows a clear difference in the temperature range at FWL between high-wind and low-wind periods. During low-wind periods, the temperature at FWL shows a large diurnal range of 7°C (i.e., from 17° to 24°C). During high-wind periods, the daily temperature change is less than 2°C. We hypothesize that the temperature signal is larger on low-wind days owing to less ventilation, giving a greater potential for thermally driven convection. The effect of cloud shading also plays a pronounced role in the temperature signal.

The rise in air temperature inland of the windward coast is proportional to the sensible heat flux from the island surface (Q) and controlled by advection of heat downstream and the turbulent vertical transport of heat aloft. Because of the rough forested surface combined with moderate or strong wind, forced convection dominates in the lower portion of the boundary layer such that heat is mostly a passive tracer there. Estimates of the Monin–Obukhov length, indicating the depth of the forced convection region, generally exceed 50 m. On the very-low-wind days, this assumption is less justifiable.

According to Smith (2011), a theoretical model of forced convection must take into account the vertical gradient of the turbulent diffusivity. An extension of his forced convection formulation gives a useful expression for the potential temperature rise Δθ at a distance x from the windward coast:
e7
In this formulation, Q is assumed to be constant. The length scale Lμ = K0U/μ2, and the vertical profile of turbulent diffusivity is assumed to be K(z) = K0 + μz. The diffusivity gradient is μ = dK/dz = ku*, K0 = ku*z0, the friction velocity is , the von Kármán constant is k = 0.4, and A ≈ 0.57 is a dimensionless constant. For the present case, we choose a forest roughness length z0 = 0.8 m with a drag coefficient of CD = 0.03 so . The length scale can be estimated m. For a point x = 8000 m downstream, the length ratio in (7) becomes x/Lμ = 696, which gives
e8
For a sensible heat flux of Q = 200 W m−2 and wind speed of U = 10 m s−1, (8) yields Δθ = 1.6°C.

The prediction of (8) is compared with data from the FWL mountain station in Fig. 12. The quantity Δθ = θFWLθsurf is computed at FWL using the upstream surface (θsurf) as a baseline. Assuming that the lower boundary layer is well mixed, the average aircraft-derived θ on leg 1L at 300-m height gives a good estimate of θsurf. The magnitude of Δθ shows the warming caused by the island.

Fig. 12.
Fig. 12.

Estimates of air parcel warming (Δθ) from the upstream ocean surface to the FWL weather station plotted against wind speed. Potential temperatures and wind speeds were averaged from 1200 to 1300 local time and the average upstream θ from leg 1L was used as a baseline. Data points are labeled as clear or cloudy based on above- or below-average solar insolation. The dashed lines show the model fit from (8) with Q = 50, 100, and 200 W m−2. The solid blue line shows Δθ from 2D simulations averaged over the Mtn Temp box shown in Fig. 7. Three-dimensional simulation values are shown with blue filled circles.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

At the FWL weather station, Δθ decreases with increasing wind speed (Fig. 12). Data points from cloudy and clear days are separated for additional information; a cloudy day is defined as one with below average solar radiation measured at the FWL weather station while a clear day has above average solar radiation. Clear days have higher Δθ values than cloudy days. The theoretical model (8) fits the observational data qualitatively and shows the strong sensitivity to wind speed and cloudiness. Both effects can easily change Δθ by ~5°C.

WRF predictions of FWL temperature are also shown in Fig. 12 with a specified sensible heat flux of 200 W m−2. This flux value was not measured; it was chosen based on reasonable estimates of radiation partitioning in a tropical forested location. The 2D model simulations (solid black) show the same sign of the relationship between Δθ with wind speed but are qualitatively different. The quantity Δθ from the simulations is computed in the lowest model grid box (below 28 m: 3D and 40 m: 2D) just upwind of the mountain peak (Fig. 7, labeled Mtn Temp) and the upstream surface θ is used. Increasing wind speed leads to a decreasing Δθ in the model. However, the temperature differences from 2D simulations have a much weaker wind speed dependence. The 3D simulations are slightly better at capturing the rise in temperature at low wind speeds. Overall, neither the 3D nor 2D model setup is appropriate for surface temperature prediction. With coarse resolution and no boundary layer parameterization, the heat is transported over the first few grid spaces crudely with subgrid parameterization. Farther aloft, we believe that convection and heat transport processes are properly resolved. The theoretical model (8) is more convenient and more accurate than WRF for Δθ. We did not use a conventional PBL parameterization because of uncertainty in how it would handle the sudden growth of the internal boundary layer starting at the windward coast.

6. Influence of wind speed on precipitation

Wind speed has been shown to control convection over the island of Dominica in sections 4 and 5, but its influence on precipitation is unresolved. Our current limited understanding of this influence is summarized below.

Trends during the 6-week DOMEX field campaign (Fig. 2) show higher amounts of precipitation on Dominica under high wind conditions. Long-term statistics developed from 3 years of daily data from 10 rain gauges (shown in Fig. 1) and Guadeloupe rawinsonde data show this same trend. The correlation coefficient (CC) is defined in the usual way (9) as the ratio of the covariance of the daily rain gauge (RG) precipitation time series with some aspect of the Guadeloupe sounding (Guad) data time series, Cov(RG, Guad), and the two standard deviations, σ:
e9

This CC was computed between precipitation and 30 parameters from the Guadeloupe sounding. These parameters ranged from measures of precipitable water to measures of stability over multiple levels in the atmosphere. It was found that the 925-hPa wind speed (U925, CC = 0.325) and the 850-hPa relative humidity (RH850, CC = 0.371) were most closely correlated with daily precipitation amount at the mountain peak on Dominica. The modest magnitude of these correlations may be associated with complicating factors such as tropical Atlantic weather disturbances passing over Dominica (Smith et al. 2009b). Days with no rain are included in the calculation of the CC. An attempt to compute the correlations using only rainy days showed a slight decrease in the CC.

The correlation between precipitation and low level wind speed reinforces the importance of wind speed in controlling the convection type over Dominica. It is also significant that the maximum correlation of precipitation and wind speed occurs at 925 hPa (Fig. 13a). The lowest level (1000 hPa) of the sounding is affected by surface properties near the sounding launch site on Guadeloupe but, higher aloft at 925 hPa, a valid measure of trade wind speed is possible. Three curves are included in Fig. 13a for three different 370-day time periods. The correlation is equally strong in all three periods showing the robustness of this result. A fourth line showing the correlation for the entire time period is also included. Relationships between precipitation and trade wind speed have also been found over the open ocean (e.g., Back and Bretherton 2005; Nuijens et al. 2009). It is likely that high-precipitation disturbances like easterly waves are accompanied by higher wind speeds, but the correlations in Fig. 13a are robust even when high-precipitation days are removed.

Fig. 13.
Fig. 13.

Correlation coefficient using (9) of the daily total precipitation amount over Dominica from the FWL rain gauge with the standard levels of the Guadeloupe sounding against (a) wind speed and (b) relative humidity. Three nonoverlapping 370-day time series segments from 2007 to 2011 are correlated for each variable (thin black lines) showing that correlation strength is not dependent on the time period. The thick black line shows the correlation coefficient for all available data.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

In addition to the wind speed, a similarly high correlation exists between relative humidity at 850 hPa and precipitation at Dominica’s mountaintop (Fig. 13b). This relationship adds another dimension to the understanding of precipitation controls; the cloud-layer [i.e., ~(960–820) hPa] moisture is also important in determining precipitation amount. A drier cloud layer could suppress rain development through strong dry-air entrainment.

The correlation between island precipitation and lower-troposphere U and RH (Fig. 13) raises the question of whether one of these is an indirect correlation. To test for this, the correlation between U and RH for the same three periods was computed (not shown). At 925 hPa where the U effect on precipitation is the strongest, the U to RH correlation is very weak. At 850 hPa, where the RH effect on precipitation is strongest, the U to RH correlation is moderate and positive. Higher wind speeds over the sea may moisten the upper part of the trade wind cloud layer, slightly enhancing cloud and precipitation development (Nuijens et al. 2009). This dryness effect was not included in the model setup (section 3). The simulations for high- and low-wind cases used identical thermodynamic soundings in order to isolate the role of wind speed.

The diurnal cycle of precipitation over the open ocean in the tropics has an early morning maximum (Rauber et al. 2007) while large tropical islands have a strong midafternoon maximum from solar heating (Sobel et al. 2011). The diurnal precipitation maximum on Dominica is weak but more closely related to the diurnal maximum of precipitation over the open ocean (Fig. 14). Six rain gauges are shown from three separate regions: the mountaintop, the windward side, and the lee side. The strong orographic effect is easily seen; the mountaintop receives a majority of the rainfall while the windward side receives less, and the lee side receives the least.

Fig. 14.
Fig. 14.

Diurnal cycle of precipitation from rain gauges over Dominica for (top to bottom): mountaintop, windward side, and leeside. Two rain-rate curves for each category are shown with symbols referring to their locations. Mountaintop sites are BL: Boeri Lake and FWL: Freshwater Lake. The windward sites are GF: Grand Fond and RO: Rosalie. The leeward coast sites are BG: Botanical Garden and CF: Canefield airport. Diurnal cycles are calculated from 3.3–4.2 years’ worth of hourly data from 2007 to 2011.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

From the indices of convection (section 5c), it has been shown that Dominica does have solar heating effects, so why does it not show a stronger midday diurnal maximum? The small amplitude of the diurnal cycle suggests that little precipitation is formed from thermally driven convection. As thermal convection is a weak wind occurrence, this observation is consistent with the preference of precipitation for high-wind days. The relative rarity of weak winds is an additional component. The 2010 and 2011 annual average surface wind speed of ~6 m s−1 (Fig. 4) has dynamics more closely related to an intermediate- or high-wind mechanically driven regime.

A third factor that may affect the role of wind speed on precipitation is lofted island-derived aerosols. Smith et al. (2012) observed island-derived aerosols detraining from convective clouds above Dominica on low-wind days. These aerosols appear to dramatically increase the cloud droplet number concentration and reduce the cloud droplet diameter (i.e., from 25 to 15 μm). The relationship of aerosol number concentration to cloud droplet number concentration in DOMEX data has also been explored by Russotto et al. (2013). To test the dynamics of particle lofting, passive tracers were added into the 2D model as a record of air movement and dispersion. The tracers are continuously emitted at a constant rate from the top of the mountain, are advected with the flow, mixed by the subgrid-scale turbulence parameterization, and do not interact with cloud microphysics. The difference in tracer distribution after 2 h in a low- and high-wind simulation is substantial (Fig. 15). In the low-wind case, tracers are carried into the clouds above the island with the thermally driven flow. In the high-wind case, the tracers are carried downstream into the wake and they avoid rising into the clouds.

Fig. 15.
Fig. 15.

Two-dimensional model snapshot of the normalized tracer distribution originating from the mountaintop. (a) A low-wind case with Us = 2 m s−1, 5 h into the simulation; and (b) a high-wind case with Us = 10 m s−1, 2 h into the simulation.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

This wind speed control of aerosol lofting may be significant for cloud microphysics and precipitation. High aerosol concentrations increase the number of available cloud condensation nuclei (CCN), resulting in a larger cloud droplet number concentration. Assuming a fixed liquid water content, the cloud droplet size will be reduced. This effect is known as the cloud lifetime effect and has the ability to decrease precipitation efficiency in warm tropical clouds (Warner 1968; Jiang et al. 2010). This factor may contribute to low precipitation amounts on low-wind days.

Sea salt, or marine aerosols, may have the opposite effect of increasing precipitation on high-wind days. DOMEX observations detect more haze in the ascending air beneath cloud base on high-wind days suggesting the presence of hygroscopic salt particles (C. D. Watson 2013, personal communication). The literature suggests that higher wind speeds cause white-capping and salt particle emission from the ocean surface (Monahan and O’Muircheartaigh 1980; Mårtensson et al. 2003; Jensen and Lee 2008). Marine aerosols can act as giant nuclei, or large CCN, which may enhance precipitation efficiency in certain cases (Jensen and Lee 2008). The presence of salt particles on high-wind days may be consistent with the observed larger cloud droplet sizes and the reduced precipitation on low-wind days.

7. Conclusions

Both thermally and mechanically driven moist convection were observed over Dominica, apparently controlled by ambient wind speed. Three convective indices were defined from observational data and model simulations to test the hypothesis of wind speed control. Both 3D and 2D idealized model setups were able to capture the trends in these indices with wind speed.

When the wind speed is weak, thermally driven convection is present. Solar surface heating drives a thermal circulation that is modified by the presence of the weak wind. Clouds present in this thermal case are initiated by surface convergence at the mountaintop and are pushed slightly downstream by the mean wind. The observed divergence above the island, the leeward-shifted convection, and the high mountaintop surface temperature in the low-wind case are evidence of the thermal initiation of convection.

When the wind speed is strong, windward-side orographic uplift and leeside plunging flow dominate. Clouds are found on the upstream side of the mountain owing to forced ascent. The leeward side sees fewer clouds as they are evaporated through the strong descent in the plunging flow. The plunging flow causes convergence of air above the mountain and strong winds in the lee. The mountaintop surface temperature is reduced because of enhanced ventilation and cloudiness over the windward slopes.

The cause of the rather dramatic shift in convection type with wind speed probably has to do with how the two types of convection mutually exclude each other. At low wind speeds, the island surface heats up more from the sun and strong thermal convection causes divergence aloft and subsidence upwind and downwind of the island, suppressing forced ascent and enhancing solar insolation. At high wind speeds, island ventilation and cloud shading cool the island surface and weaken any thermal convection. The strong upslope flow both overpowers any weak upstream subsidence and creates over-island subsidence that suppresses thermal convection.

The impact of trade wind speed on island precipitation is unclear although the statistical correlation is clearly positive. While the sensitivity of convection type to wind speed is one likely cause, other factors must be considered. First is the tendency for dryer air in the cloud layer under weak wind conditions. Dryer air will suppress convective precipitation by entrainment and in-cloud evaporation of cloud droplets. Related to this is the statistical connection between high winds and weather disturbances in the tropics.

The second factor related to precipitation is the role of aerosols. According to Smith et al. (2012) and Russotto et al. (2013), clouds during high wind conditions have fewer but larger cloud droplets, probably enhancing precipitation. The root cause of the shift in droplet size with wind speed is either enhanced large size sea salt CCN from sea spray under high-wind conditions or enhanced small-size island-derived CCN under low-wind conditions. More work is required to understand these influences on precipitation.

Acknowledgments

The authors acknowledge funding from the National Science Foundation (NSF Grant 0930356); helpful discussions with Campbell D. Watson, Daniel J. Kirshbaum, William R. Boos, Trude Storelvmo, Richard Rotunno, and the University of Wyoming program manager Jeffrey French; constructive criticism from three anonymous reviewers of a previous version; along with the Yale University High Performance Computing Center for its support with modeling and Météo-France for access to radar and weather station data. We would also like to acknowledge high-performance computing support from Yellowstone (ark:/85065/d7wd3xhc) provided by NCAR’s Computational and Information Systems Laboratory, sponsored by the National Science Foundation.

APPENDIX

Redundant Measurements of Ambient Wind Speed

Given the proposed wind speed control of convection, it is important to understand how well ambient wind speed can be measured by various platforms. Multiple measures of wind speed smoothed by a 48-h running mean are shown in Fig. A1. These include data from the Martinique Trinite-Caravelle weather station, the Guadeloupe rawinsonde sounding at 925 hPa, Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) (Dee et al. 2011) from an upstream region at 10-m height, wind data from the FWL weather station, and the average wind speed from the aircraft upstream leg 1L. General trends of high and low wind speed periods are well captured in all data sources. As expected, differences exist; for example, the FWL weather station has a slower wind speed likely owing to friction from the forested island surface showing that it is not an ideal measure of ambient wind speed. More specifically, the FWL weather station has a fractional relationship of approximately 0.63 × Guadeloupe (wind at FWL has only 0.63 of the value of wind from the Guadeloupe sounding) and 0.87 × Trinite-Caravelle assuming a linear scaling between the paired sites. This scaling information and correlations between all sites are shown in Table A1. All correlations are above 0.78. ERA-Interim agrees most closely with the Trinite-Caravelle and FWL weather stations (correlation coefficient, CC > 0.94), but the Guadeloupe sounding also gives an accurate measure. While the balloon is launched over land, by 925 hPa, the wind measure is mostly outside the influence of the island. For the general purpose of using wind speed as a predictor of convective type, we believe that any of the above wind speed measures would be appropriate.

Fig. A1.
Fig. A1.

Wind speeds during the DOMEX field project period from the weather stations at FWL and Trinite-Caravelle on Martinique, the Guadeloupe rawinsonde sounding, ERA-Interim data, and aircraft RF leg 1L. See Table A1 for additional information.

Citation: Journal of the Atmospheric Sciences 71, 7; 10.1175/JAS-D-13-0399.1

Table A1.

Wind speed observations: The bottom left of the table (roman) shows correlation coefficients between wind speed measurements. The top right of the table (italics) shows multiplicative factors between datasets. For example, FWL = 0.63 × Guadeloupe. The data, from 26 Mar to 9 May 2011, has been interpolated to 2-m intervals and smoothed over 48 h. Correlations and comparisons with the research flight data include only data points close to the research flight time.

Table A1.

REFERENCES

  • Back, L. E., , and C. S. Bretherton, 2005: The relationship between wind speed and precipitation in the Pacific ITCZ. J. Climate, 18, 43174328, doi:10.1175/JCLI3519.1.

    • Search Google Scholar
    • Export Citation
  • Banta, R. M., , and C. L. B. Schaaf, 1987: Thunderstorm genesis zones in the Colorado Rocky Mountains as determined by traceback of geosynchronous satellite images. Mon. Wea. Rev., 115, 463476, doi:10.1175/1520-0493(1987)115<0463:TGZITC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Barcilon, A., , J. C. Jusem, , and P. G. Drazin, 1979: On the two-dimensional hydrostatic flow of a stream of moist air over a mountain ridge. Geophys. Astrophys. Fluid Dyn., 13, 125140, doi:10.1080/03091927908243765.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-L., , and A. J. Nash, 1994: Diurnal variation of surface airflow and rainfall frequencies on the island of Hawaii. Mon. Wea. Rev., 122, 3456, doi:10.1175/1520-0493(1994)122<0034:DVOSAA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Chen, Y.-L., , and J. Feng, 1995: The influences of inversion height on precipitation and airflow over the island of Hawaii. Mon. Wea. Rev., 123, 16601676, doi:10.1175/1520-0493(1995)123<1660:TIOIHO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Damiani, R., and et al. , 2008: The cumulus, photogrammetric, in situ, and Doppler observations experiment of 2006. Bull. Amer. Meteor. Soc., 89, 5773, doi:10.1175/BAMS-89-1-57.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Durran, D., , and J. Klemp, 1982b: On the effects of moisture on the Brunt-Väisälä frequency. J. Atmos. Sci., 39, 21522158, doi:10.1175/1520-0469(1982)039<2152:OTEOMO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 913928, doi:10.1175/2007MWR2059.1.

    • Search Google Scholar
    • Export Citation
  • Fuhrer, O., , and C. Schär, 2005: Embedded cellular convection in moist flow past topography. J. Atmos. Sci., 62, 28102828, doi:10.1175/JAS3512.1.

    • Search Google Scholar
    • Export Citation
  • Gaberšek, S., , and D. R. Durran, 2004: Gap flows through idealized topography. Part I: Forcing by large-scale winds in the nonrotating limit. J. Atmos. Sci., 61, 28462862, doi:10.1175/JAS-3340.1.

    • Search Google Scholar
    • Export Citation
  • Houze, R. A., Jr., 1993: Cloud Dynamics. Academic Press, 573 pp.

  • Janjic, Z. I., 1996: The surface layer in the NCEP Eta model. Proc. 11th Conf. on Numerical Weather Prediction, Norfolk, VA, Amer. Meteor. Soc., 354–355.

  • Jensen, J., , and B. S. Lee, 2008: Giant sea-salt aerosols and warm rain formation in marine stratocumulus. J. Atmos. Sci., 65, 36783694, doi:10.1175/2008JAS2617.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, H., , G. Feingold, , and A. Sorooshian, 2010: Effect of aerosol on the susceptibility and efficiency of precipitation in warm trade cumulus clouds. J. Atmos. Sci., 67, 35253540, doi:10.1175/2010JAS3484.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, Q., 2003: Moist dynamics and orographic precipitation. Tellus, 55, 301316, doi:10.1034/j.1600-0870.2003.00025.x.

  • Kirshbaum, D. J., 2011: Cloud-resolving simulations of deep convection over a heated mountain. J. Atmos. Sci., 68, 361378, doi:10.1175/2010JAS3642.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and D. R. Durran, 2005: Observations and modeling of banded orographic convection. J. Atmos. Sci., 62, 14631479, doi:10.1175/JAS3417.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and R. B. Smith, 2009: Orographic precipitation in the tropics: Large-eddy simulations and theory. J. Atmos. Sci., 66, 25592578, doi:10.1175/2009JAS2990.1.

    • Search Google Scholar
    • Export Citation
  • Kirshbaum, D. J., , and C.-C. Wang, 2014: Boundary-layer updrafts driven by airflow over heated terrain. J. Atmos. Sci., 71, 14251442, doi:10.1175/JAS-D-13-0287.1.

    • Search Google Scholar
    • Export Citation
  • Klemp, J. B., , J. Dudhia, , and A. D. Hassiotis, 2008: An upper gravity-wave absorbing layer for NWP applications. Mon. Wea. Rev., 136, 39874004, doi:10.1175/2008MWR2596.1.

    • Search Google Scholar
    • Export Citation
  • Knievel, J. C., , G. H. Bryan, , and J. P. Hacker, 2007: Explicit numerical diffusion in the WRF model. Mon. Wea. Rev., 135, 38083824, doi:10.1175/2007MWR2100.1.

    • Search Google Scholar
    • Export Citation
  • Mårtensson, E. M., , E. D. Nilsson, , G. de Leeuw, , L. H. Cohen, , and H.-C. Hansson, 2003: Laboratory simulations and parameterization of the primary marine aerosol production. J. Geophys. Res., 108, 4297, doi:10.1029/2002JD002263.

    • Search Google Scholar
    • Export Citation
  • Miglietta, M. M., , and R. Rotunno, 2009: Numerical simulations of conditionally unstable flows over a mountain ridge. J. Atmos. Sci., 66, 18651885, doi:10.1175/2009JAS2902.1.

    • Search Google Scholar
    • Export Citation
  • Minder, J. R., , R. B. Smith, , and A. D. Nugent, 2013: The dynamics of ascent-forced orographic convection in the tropics: Results from Dominica. J. Atmos. Sci., 70, 40674088, doi:10.1175/JAS-D-13-016.1.

    • Search Google Scholar
    • Export Citation
  • Monahan, E. C., , and I. O’Muircheartaigh, 1980: Optimal power-law description of oceanic whitecap coverage dependence on wind speed. J. Phys. Oceanogr., 10, 20942099, doi:10.1175/1520-0485(1980)010<2094:OPLDOO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Nuijens, L., , B. Stevens, , and A. P. Siebesma, 2009: The environment of precipitating shallow cumulus convection. J. Atmos. Sci., 66, 19621979, doi:10.1175/2008JAS2841.1.

    • Search Google Scholar
    • Export Citation
  • Pierrehumbert, R. T., , and B. Wyman, 1985: Upstream effects of mesoscale mountains. J. Atmos. Sci., 42, 9771003, doi:10.1175/1520-0469(1985)042<0977:UEOMM>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rasmussen, R. M., , P. Smolarkiewicz, , and J. Warner, 1989: On the dynamics of Hawaiian cloud bands: Comparison of model results with observations and island climatology. J. Atmos. Sci., 46, 15891608, doi:10.1175/1520-0469(1989)046<1589:OTDOHC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rauber, R. M., and et al. , 2007: Rain in shallow cumulus over the ocean: The RICO Campaign. Bull. Amer. Meteor. Soc., 88, 19121928, doi:10.1175/BAMS-88-12-1912.

    • Search Google Scholar
    • Export Citation
  • Reisner, J. M., , and P. K. Smolarkiewicz, 1994: Thermally forced low Froude number flow past three-dimensional obstacles. J. Atmos. Sci., 51, 117133, doi:10.1175/1520-0469(1994)051<0117:TFLFNF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 602618, doi:10.1175/2010JAS3558.1.

    • Search Google Scholar
    • Export Citation
  • Russotto, R. D., , T. Storelvmo, , and R. B. Smith, 2013: Modeling aerosol activation in a tropical, orographic, island setting: Sensitivity tests and comparison with observations. Atmos. Res., 134, 1223, doi:10.1016/j.atmosres.2013.07.017.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., , and M. L. Weisman, 2009: The impact of positive-definite moisture transport on NWP precipitation forecasts. Mon. Wea. Rev., 137, 488494, doi:10.1175/2008MWR2583.1.

    • Search Google Scholar
    • Export Citation
  • Skamarock, W. C., and et al. , 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3_bw.pdf.]

  • Smith, R. B., 1989: Mountain-induced stagnation points in hydrostatic flow. Tellus,41A, 270–274, doi:10.1111/j.1600-0870.1989.tb00381.x.

  • Smith, R. B., 2011: Regime diagrams for K-theory dispersion. Bound.-Layer Meteor., 139, 501519, doi:10.1007/s10546-011-9594-4.

  • 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, 353378, doi:10.1002/qj.49710845605.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , and V. Grubišić, 1993: Aerial observations of Hawaii wake. J. Atmos. Sci., 50, 37283750, doi:10.1175/1520-0469(1993)050<3728:AOOHW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , P. Schafer, , and D. J. Kirshbaum, 2009a: Orographic precipitation in the tropics: Experiments in Dominica. J. Atmos. Sci., 66, 16981716, doi:10.1175/2008JAS2920.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., , P. Schafer, , D. J. Kirshbaum, , and E. Regina, 2009b: Orographic enhancement of precipitation inside Hurricane Dean. J. Hydrometeor., 10, 820831, doi:10.1175/2008JHM1057.1.

    • Search Google Scholar
    • Export Citation
  • Smith, R. B., and et al. , 2012: Orographic precipitation in the tropics: The Dominica experiment. Bull. Amer. Meteor. Soc., 93, 15671579, doi:10.1175/BAMS-D-11-00194.1.

    • Search Google Scholar
    • Export Citation
  • Smolarkiewicz, P. K., , R. M. Rasmussen, , and T. L. Clark, 1988: On the dynamics of Hawaiian cloud bands: Island forcing. J. Atmos. Sci., 45, 18721905, doi:10.1175/1520-0469(1988)045<1872:OTDOHC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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.

  • Tabary, P., 2007: The new French operational radar rainfall product. Part I: Methodology. Wea. Forecasting, 22, 393408, doi:10.1175/WAF1004.1.

    • Search Google Scholar
    • Export Citation
  • Tian, W.-S., , and D. J. Parker, 2002: Two-dimensional simulation of orographic effects on mesoscale boundary-layer convection. Quart. J. Roy. Meteor. Soc., 128, 19291952, doi:10.1256/003590002320603476.

    • Search Google Scholar
    • Export Citation
  • Warner, J., 1968: A reduction in rainfall associated with smoke from sugar-cane fires—An inadvertent weather modification? J. Appl. Meteor., 7, 247251, doi:10.1175/1520-0450(1968)007<0247:ARIRAW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Woodcock, A. H., 1960: The origin of trade-wind orographic shower rains. Tellus, 12, 315326, doi:10.1111/j.2153-3490.1960.tb01316.x.

  • Wulfmeyer, V., and et al. , 2008: The convective and orographically induced precipitation study. Bull. Amer. Meteor. Soc., 89, 14771486, doi:10.1175/2008BAMS2367.1.

    • Search Google Scholar
    • Export Citation
Save