a. Numerical model

A two-dimensional (2D) coupled land–atmosphere (cloud resolving) numerical model with open lateral boundary conditions is used for all of the simulations performed in this work. Prior studies have shown that the time-mean kinematic and thermodynamic properties of convective lines are well simulated in 2D models, and they have the advantage of computational speed and efficiency when many simulations are required, such as in a sensitivity study (Nicholls et al. 1988; Ferrier et al. 1996; Tao et al. 1996; Xu and Randall 1996; Grabowski et al. 1998; Lucas et al. 2000; Mohr et al. 2003). The Goddard Cumulus Ensemble (GCE) is a nonhydrostatic, anelastic, numerical cloud-resolving model that includes variables for horizontal and vertical velocities, potential temperature, water vapor mixing ratio, and turbulent kinetic energy (Tao and Simpson 1993). The subgrid-scale turbulence used in the GCE model is based on work by Klemp and Wilhelmson (1978). In their approach, one prognostic equation is solved for subgrid turbulent kinetic energy (TKE), which is then used to specify the eddy coefficients. The effect of condensation on the generation of subgrid-scale kinetic energy is also incorporated in the model (Soong and Ogura 1980). The cloud microphysics include a parameterized two-category liquid water scheme (cloud water and rain), and a parameterized three-category ice-phase scheme (cloud ice, snow, and hail/graupel; Lin et al. 1983; Rutledge and Hobbs 1984). Solar and infrared radiation parameterizations are also included in the model (Tao et al. 1996).

In this study, GCE is coupled to the Parameterization for Land–Atmosphere–Cloud Exchange (PLACE) (Wetzel and Boone 1995). The PLACE model is a detailed, interactive process model of the heterogeneous land surface (soil and vegetation) and adjacent near-surface atmosphere. PLACE emphasizes the vertical transport of moisture and energy through a five-layer soil moisture and a seven-layer soil temperature column to the overlying surface. The additional two soil temperature layers are used to aid in resolving large temperature gradients near the surface. Momentum, sensible, and latent heat fluxes are calculated using similarity relationships (Zilitinkevich 1975; Businger et al. 1971). In PLACE, each model grid cell contains a specified fraction of fully vegetated surface, with the rest of the surface treated as nonvegetated bare soil. Separate evapotranspiration equations are used to compute supply- and demand-limited rates for vegetation and bare-soil areas. The demand-limited rate is calculated using the mixing-ratio gradient between the ground and the surface layer and the magnitude of the canopy resistance (Noilhan and Planton 1989). Supply-limited rates are dependent upon the gradient between the water potentials of the soil and plant (Wetzel and Boone 1995). The resulting latent heat flux in a grid cell is then computed as a weighted sum of the fluxes from the vegetated and bare-soil areas. The exchanges of radiation and heat, momentum, and moisture fluxes couple the GCE and PLACE models (Tao et al. 2003). This same two-dimensional version of GCE–PLACE has previously been used to investigate both landscape-generated deep convection (Lynn et al. 1998) and the sensitivity of West African convective line water budgets to land cover (Mohr et al. 2003).

A horizontal domain of approximately 1000 km divided into 2048 grid points with 1990 inner points at 0.5-km spacing and a stretched (1:1.06) grid on either side is used in all simulations. This domain consists of only tiger bush vegetation and does not represent the entire area outlined in Fig. 1, as this area contains several other land-cover types. The model consists of 33 grid points in the vertical starting at 0.0 km and ending at 21.5 km. This grid was also stretched with highest resolution in the lowest levels with the spacing between levels ranged from 80 m at the surface to 1200 m at the top of the troposphere. GCE operated on a 5-s time step and called PLACE for surface fluxes at 3-min intervals. All simulations started at local midnight and ran for 24 h, allowing for the examination of a full diurnal cycle. No lifting mechanism, such as a warm bubble or cold pool, was applied so that the initial atmospheric structure could evolve over the course of the day.

b. Initial conditions and numerical experiments

The vegetation, soil type, soil moisture, and soil temperature data used to initialize the model were derived from the hydrologic HAPEX-Sahel field experiment. HAPEX-Sahel was conducted during 1991–92 within a 1° by 1° square in southwestern Niger (2°–3°E and 13°–14°N). The goal of HAPEX-Sahel was to improve the parameterizations of land surface–atmosphere interactions at the global circulation model scale, because of the hypothesized link between land degradation and climatic changes in the Sahel. The experiment combined remote sensing and ground-based measurements and contained an intensive observation period that captured the transition of the wet season to the dry season in August through October of 1992. All of the land surface data originated from within the southern supersite within the HAPEX-Sahel study area during the intensive observation period. This site was predominantly composed of the tiger bush land cover but also contained areas of millet, grassy savanna, and fallow bush. A full description of the site and the experimental methods therein can be found in Wallace et al. (1994).

The tiger bush land cover is deeply rooted, highly reflective, sparsely vegetated, and able to withstand the highly oxidizing and leaching Sahel climate. The land cover in the study simulations was specified so that on average ∼67% was bare crusted soil and the rest vegetation (based on aerial photographs from the experiment site). The vegetated strips within the tiger bush land cover are about 10–30 m wide by 100–300 m long and consist primarily of 2–4–m-tall shrubs (Guiera Senegalensis and Combretum micranthum) (Wallace and Holwill 1997). There exists very little herbaceous growth with the exception of the upslope edge of the shrubs. A large portion of the bare-soil transects in the tiger bush land cover are strongly crusted and have very low infiltration rates and very low saturated hydraulic conductivity (Bromley et al. 1997). Below the surface crusting, the soil texture is principally sandy loam (Wallace and Holwill 1997). Hence, the soil type used in this study is a sandy loam soil.

Initial soil moisture conditions define the simulations used in this study. A wet regime refers to a simulation that contains a wet soil moisture profile and a dry regime refers to a simulation that contains a dry soil moisture profile. Table 1 lists the initial soil moisture profiles and parameters, and Table 2 lists the vegetation parameters used in both numerical case studies. Soil moisture, soil temperature, and vegetation cover were randomly varied across the model domain to account for the natural variability of the tiger bush land cover. In addition, the saturated hydraulic conductivity of the first soil moisture layer was decreased in order to account for the strongly crusted uppermost soil layer within PLACE. The wet soil moisture profile is typical of a soil column that has been exposed to several days of consecutive precipitation, whereas the dry soil moisture profile is more representative of soils that have been drying for approximately one week. This was based on observations of soil transects within the southern supersite. Initial volumetric soil moisture in the wet regime ranges from 0.26 cm³ cm⁻³ (65% saturation) in the lowest layers to as little as 0.14 cm³ cm⁻³ (35% saturation) in the near-surface layers. Initial volumetric soil moisture in the dry regime ranges from 0.15 cm³ cm⁻³ (38% saturation) in the lowest layers to as little as 0.04 cm³ cm⁻³ (10% saturation) in the near-surface layers.

The sounding data used to initialize the numerical simulations came from field observations during HAPEX-Sahel and operational rawinsonde launches from Niamey, Niger. The field observations were used in conjunction with the operational soundings to make complete and detailed initial soundings for the simulations. The initial soundings were then smoothed to remove any sharp changes in wind speed, wind direction, and humidity to prevent numerical instability within the GCE model.

Before attempting a 24-h simulation, the atmosphere and the land surface must be in quasi equilibrium in order to avoid generating large nonphysical surface fluxes. First-guess soil moisture and temperature profiles were created, and PLACE was run offline forced by actual surface observations in the southern supersite of pressure, temperature, wind speed, specific humidity, and radiation from midnight to 0300 LST. The resulting soil moisture and temperature profiles at the end of the first 3-h offline simulation became the initial soil moisture and temperature profiles for a repeat of the 3-h simulation. The 3-h simulation was repeated (about 20 iterations) until the initial and final soil moisture profiles were nearly identical (soil moisture within ±0.001 cm³ cm⁻³ and soil temperature within ±0.1 K). The resulting initial soil moisture values are listed in Table 1. This methodology was applied to both soil moisture profiles within each numerical case study so that each regime (wet and dry) had its own initial soil temperature profile. By changing only the initial soil moisture profile, we can isolate the physical processes important in boundary layer growth and convective initiation within the tiger bush environment.

Two numerical case studies are presented in this paper, a convective case with precipitation, 25 August 1992, and a fair weather case, 1 September 1992. The objectives of the fair weather case study are to analyze how much wet soil might increase the potential for convective initiation and how much dry soil might reinforce fair weather conditions. The fair weather and convective cases will be compared to diagnose the influences of the different soil moisture regimes within these environments.

a. Initial atmospheric conditions

The 0000 UTC 25 August 1992 sounding is depicted in Fig. 2. This sounding has several key attributes. The environment is unstable with an initial convective available potential energy (CAPE) value of 1860 J kg⁻¹ and convective inhibition (CIN) of 80 J kg⁻¹. Both CAPE and CIN were calculated by lifting the average of the lowest 50 hPa.

The core of the African easterly jet is located around 630 hPa, and the core of the tropical easterly jet is located near 180 hPa (Fig. 2b). The location and strength (15 m s⁻¹) of the African easterly jet in this environment creates significant low-level wind shear. There is approximately −5.7 × 10⁻³ s⁻¹ of shear from height of the maximum low-level wind at 0.47 km to 3.65 km (location of the African easterly jet). The meridional component of the wind is southerly from the surface to about 900 hPa, indicating strong moist monsoonal flow at this time.

b. Surface fluxes

PLACE calculates the surface energy budget in the vegetation layer at the air–ground interface according to Eq. (1):

View Expanded

In Eq. (1), the left-hand side represents the effect of energy storage by the biomass. The water equivalent heat capacity of the biomass is b_M (J K⁻¹ m⁻²) and T_s represents the temperature (kelvin) of the material that lies between the atmosphere and the soil. The right-hand side of the equation contains the surface flux terms. These are net radiation (R_net), ground heat flux (G), latent heat flux (E), and sensible heat flux (H). Figure 3 is the time series of domain-averaged surface fluxes throughout the simulation. The difference between the net radiation and the ground heat flux (R_net − G) is also plotted in Fig. 3 to show the amount of available energy for altering the diurnal cycle of boundary layer moist entropy. The wet-regime simulation received slightly more net radiation due to the lower albedo and cooler temperatures (up to 4 K) of the wetter soils through associated decreases in upwelling shortwave and longwave radiation, respectively. The wet regime also had more ground heat flux than the dry regime (nearly double) because of the higher heat conductivity of the moister soils, resulting in a smaller available energy term.

In both the wet and dry regimes, the sensible heat flux was larger than the latent heat flux. The sensible heat flux in the dry regime was nearly 50% higher than the sensible heat flux in the wet regime. Conversely, the wet regime had 60% more latent heat flux than the dry regime. In both regimes, the latent heat flux peaks in the morning were associated with maxima in evapotranspiration. In the afternoon, evaporating rainfall at the surface was the principal cause, allowing latent heat flux to continue after sunset.

c. Boundary layer development

The vertical distribution of MSE, or moist entropy, is easily related to surface fluxes and is useful in evaluating the potential for deep moist convection within each regime. Strong vertical gradients in MSE encourage the development of moist convection to reestablish a neutral distribution. The equation for MSE is as follows:

View Expanded

In the above equation, g is the acceleration due to gravity, z height, C_p specific heat capacity at constant pressure, T temperature, L latent heat of vaporization, and w water vapor mixing ratio.

Equivalent potential temperature (θ_e) is a proxy for identifying the distribution of MSE within each regime. Figure 4 is the distribution of θ_e of the lowest 3 km in the model at hour 13, just prior to the triggering of deep convection. The wet regime has a 1-km layer of air with θ_e greater than 342 K, and almost the entire surface layer exceeds 344 K. In both of these layers, the dry regime θ_e is at least 2 K cooler than the wet regime. In the surface layer the dry regime θ_e does not exceed 344 K. The boundary layer depth is defined as the depth of the well-mixed layer. In Fig. 5, the boundary layer in the wet regime is shallower (1.39 km versus 1.78 km) and exhibits a stronger gradient of θ_e (MSE) between 1 and 3 km in the early afternoon. The evapotranspiration in the wet regime produced a much larger latent heating term (Lw), despite a smaller sensible heating term (C_pT), than the dry regime. The boundary layer of the dry regime grew faster than the boundary layer of the wet regime and thus was more prone to the entrainment of drier air from the free atmosphere (Fig. 4). The increased entrainment of drier air in the dry regime boundary layer contributed to its weaker vertical gradient of MSE throughout the simulation.

Differences in boundary layer growth and development led to differences between the two regimes in the temperature and moisture stratification of the free atmosphere. In Fig. 6, the CAPE drops early in both regimes due to the formation of a small radiation inversion at the surface. The CAPE then builds up in the morning until the onset of convection, which then erodes the CAPE values throughout the afternoon. By noon, the wet regime had 2220 J kg⁻¹ of CAPE (Fig. 6a), while the dry regime (Fig. 6b) had 1530 J kg⁻¹ (32% less). The slower-growing boundary layer of the wet regime was less prone to the entrainment of drier air aloft. In the warmer, moister, shallower, wet regime boundary layer (Fig. 5a), significant CAPE was produced and remained higher than in the dry regime (Fig. 5b) from midmorning onward. In Figs. 5 and 6, the wet regime appears more capable of sustaining deep convection with heavy precipitation than the dry regime.

d. Convection, precipitation, and final water budgets

Enhanced cumulus began forming around hour 13 in the dry regime and hour 14 in the wet regime. Figure 7 is the surface precipitation rate recorded at every 5 min across the entire model domain. Convection and precipitation in the dry regime started slightly earlier (hour 13.5 versus hour 14). Faster boundary layer growth in the dry regime lifted parcels to their level of free convection sooner. However, all precipitation ended around hour 18. Most of the convective lines in both regimes lasted no longer than 1 h. By using the term “convective lines,” we are referring to lines of cumulonimbus clouds in space and time as depicted in Fig. 7 and not lines in space only. Fewer convective lines developed in the wet regime, but several lasted over 2 h. Both regimes exhibited a second onset of convection between hours 16 and 17. Once initiated, the wet regime supported convection until the end of the simulation whereas all but one of the convective lines in the dry regime lasted less than 2 h.

The duration, spatial distribution, and intensity of convection varied between the two regimes. Figure 8 is a time series of domain maximum updrafts for both regimes. The peak vertical velocity in the wet regime was 1 m s⁻¹ lower than the dry regime, but it maintained strong updrafts (above 20 m s⁻¹) until the end of the simulation. Updrafts above 10 m s⁻¹ ended between hour 16 and 17 in the dry regime. The differences in updrafts revealed that convection was generally more intense in the wet regime, producing differences in the final precipitation rate distribution (Fig. 9). The wet regime has a higher number of counts in all of the categories, especially the >50 mm h⁻¹ bin, and had 77 counts over 100 mm h⁻¹ compared to 9 counts in the dry regime. Table 3 is the frequency distribution of surface precipitation rates from Fig. 9. The wet regime exhibits a higher percentage of precipitation rates in all of the binned categories except for the 0–5-mm bin where the dry regime has slightly over 3% more in this category. Noteworthy in this table is the higher percentage of precipitation rates falling within the largest precipitation rates bin. The percentage of precipitation rates within the wet regime is more than double those produced in the dry regime, highlighting the more intense precipitation produced in the wet regime.

The distribution of precipitation rates in the wet regime is the result of more intense, longer-lasting convection, arising from the manner in which the soil moisture influenced boundary layer growth and development. High sensible heat flux and significant entrainment into the dry regime boundary layer resulted in convection that was typically weak and short lived.

The condensed water budget integrated over time and over the full model domain shows how much more efficient the wet regime was in producing precipitation:

View Expanded

Here, C is the total condensation plus deposition, E the total evaporation plus sublimation, R the total surface rainfall, and S_c the mass of suspended condensate. The precipitation efficiency (PE) is defined as the ratio of total rainfall to total condensation and deposition:

View Expanded

Table 4 displays the final water budgets for wet and dry regimes. The wet regime had more condensation than the dry regime and approximately the same amount of evaporation. Less condensation in the dry regime resulted in 55% less precipitation. The higher and drier boundary layer in the dry regime had a lower PE than the wet regime (15% versus 27%). More nonprecipitating clouds occurred in the wet regime, giving the higher value for suspended condensate.

a. Initial atmospheric conditions

The objective of this case study is to observe how the different soil moisture regimes alter fair weather conditions and which regime would favor more intense convection should preexisting disturbances such as a gust front from a transient squall line perturb the environment. In addition, we reference results from the previous case study to assess the impact of the land surface in convective versus fair weather environments. The 0000 UTC sounding used to initialize the wet and dry regime simulations is depicted in Fig. 10. This sounding is typical of a fair weather sounding within the tiger bush environment. The lower-tropospheric subsidence inversion and associated dry air intrusion at 850 hPa generate a significant amount of CIN, 250 J kg⁻¹, in this environment. The CAPE of this sounding is only 470 J kg⁻¹, and coupled with the large amount of CIN, creates a stable environment.

The core of the African easterly jet is located around 650 hPa and the core of the tropical easterly jet is located near 170 hPa. There are −6.2 × 10⁻³ s⁻¹ of shear from the height of the maximum low-level wind at 0.5 to 3.5 km (location of the African easterly jet). This value of shear is higher than calculated in the convective case study and is due to the lower altitude of the African easterly jet and stronger southwesterly flow near the surface. The meridional component of the wind is southerly to 880 hPa yielding a slightly deeper monsoon flow deeper than the convective case.

b. Surface fluxes

The time series of domain-averaged surface fluxes for both regimes are in Fig. 11. The wet regime had slightly more net radiation than the dry regime. As in the previous case study, this was from a combination of cooler soil temperatures (up to 5 K) and the lower albedo of the wetter soils. The wet regime exhibited a 50% larger ground heat flux, due to the higher thermal conductivity of the wetter soils resulting in 13% less available energy than in the dry regime. Increased evapotranspiration in the wet regime produced nearly 3 times the latent heat flux of the dry regime. However, unlike the previous case study, the latent heat flux peaks only in the morning hours because there is no evaporating rainfall in the late afternoon. The sensible heat flux in the dry regime peaked about 200 W m⁻² higher than in the wet regime. This case study had similar amounts of net radiation, ground heat flux, and available energy compared to the wet and dry regimes of the previous case study. The latent heat flux in the wet regime was nearly 22% larger than the previous case study and was produced by a larger surface vapor pressure deficit in the fair weather environment, making higher evapotranspiration rates from the tiger bush land cover possible. In the dry regime latent heat flux was 20% less than in the convective case study because of warmer near-surface air temperatures. Higher air temperatures increased the stomatal resistance of the vegetation in PLACE thereby reducing transpiration. The sensible heat flux in the wet regime was similar to the convective case study, but 11% higher in the dry regime due to a warmer (1–2 K) surface temperature.

c. Boundary layer development

The differences in surface fluxes between the two regimes created two different boundary layers. In Fig. 12a, the wet regime develops a stronger vertical gradient in θ_e (MSE) and, since no convection develops, maintains it through the end of the simulation. The vertical gradient of MSE in the wet regime in Fig. 12a is stronger than the wet regime in the convective case (Fig. 6a). The boundary layer in the dry regime (Fig. 12b) entrained copious dry air, producing a weaker vertical gradient of MSE. The dry regime in this case study behaved similarly to the dry regime in the convective case study, as both were prone to the entrainment of drier air aloft. There is a more obvious difference in the vertical gradients of MSE between the wet and dry regimes in this case study compared to the convective case study.

The CAPE in the dry regime increased to 700 J kg⁻¹ by noon, falling to 20 J kg⁻¹ by the end of the simulation (Fig. 13). The wet regime, where entrainment was not as prevalent, increased to 1230 J kg⁻¹ of CAPE by noon, falling to 780 J kg⁻¹ by the end of the simulation. This is consistent with the previous case study in which the wet regime builds up significantly more CAPE than the dry regime. The fair weather wet regime builds up less CAPE, but since there is no convection to eliminate CAPE, there is more remaining at the end of the simulation.

a. Wet versus dry soil moisture regimes

Surface fluxes within GCE-PLACE have been presented for all of the simulations performed in this study. Compared to observations during HAPEX-Sahel, GCE-PLACE underestimates the overall evaporative fraction [E/(H + E)] of the tiger bush land cover. However, the relative increases in evaporative fraction between wet soil moisture days and dry soil moisture days were consistent with observations. The differences in evapotranspiration from the tiger bush land cover between the two regimes proved to be responsible for differences in boundary layer growth and development in both fair weather and convective cases. The wet regimes produced shallower boundary layers with stronger vertical gradients of MSE (θ_e), more MSE per unit area, and more CAPE throughout the simulation, particularly in the 1 September fair weather case study. The increased sensible heat flux in the dry regime boundary layers resulted in faster boundary layer growth and, in the convective case, more entrainment of dry air into rising parcels. In the 25 August convective case, dry regime parcels reached their levels of free convection faster than in the wet regime, initiating deep convection and precipitation earlier in the afternoon.

More convective lines developed in the dry regime but they were generally less intense and short lived compared to the convective lines in the wet regime. For the dry regime, the combined effect of numerous convective cells and increased entrainment into the boundary layer air quickly eroded the vertical gradient of MSE (θ_e) and reduced the CAPE, making it impossible to sustain convection throughout the evening hours. In the wet regime, the boundary layer was able to sustain deep convection until the end of the simulation, producing 2.25 times more precipitation.

b. Fair weather cases versus convective cases

The vertical gradient in vapor pressures at the surface and in the surface layer is a key component of the calculation of evapotranspiration in PLACE. The fair weather case sounding (Fig. 10a) was drier than the convective case sounding (Fig. 2a) and capped by a strong subsidence inversion at 850 hPa. Such a situation would have the potential to produce high evapotranspiration rates should enough soil moisture be available. There was slightly more latent heat flux in the wet regime in the fair weather case compared to the wet regime in the convective case, but similar sensible heat fluxes. The biggest difference in the surface fluxes is between the dry regimes. Maximum sensible heat flux is 20% higher, because of warmer soils and in turn warmer skin temperatures, in the dry regime of the fair weather case, producing a higher boundary layer (Figs. 5b and 12b). Despite the height of the fair weather dry regime boundary layer, it was still well below the level of free convection determined by the fair weather lapse rate.

Differences in surface fluxes between the two regimes were more pronounced in the fair weather case. Because of a stronger surface vapor pressure gradient, the difference in latent heat flux between the wet and dry regimes was 11% higher in the fair weather case than in the convective case. The dry regime in the fair weather case produced faster boundary layer growth from the morning into the early afternoon compared to the dry regime in the convective case. While there was more entrainment in the dry regimes than in the wet regimes in both cases, the boundary layer in the dry regime in the fair weather case entrained more warm, dry air than did the dry regime in the convective case. The dry regime in the fair weather case exhibited a weaker vertical gradient of MSE than the dry regime in the convective case. This resulted in bigger differences between the wet and dry soil moisture regimes in the fair weather case with respect to MSE gradients. In the wet regimes of both cases, the surface fluxes and thus the boundary layer growth were comparable. The CIN and capping inversion in the fair weather sounding made it impossible for rising parcels to reach their level of free convection even in the wet regime. Dry entrainment in the fair weather simulations was responsible for making even shallow clouds rare.

The different soil moisture regimes had a noticeable impact in both case studies, although the initial sounding was the most important variable determining a fair day versus a rainy day. Without a moist southwesterly monsoon flow, precipitation is unlikely in the tiger bush environment even when the soil is wet. That is, land surface processes can modify the development of convection but cannot solely determine convective initiation. The wet regime of the fair weather case presents an interesting situation. The high CAPE of the late afternoon (Fig. 13a) could make it possible for a preexisting disturbance, such as a westward-moving squall line, to trigger new convection and precipitation. It is easier to imagine a squall line meeting its demise in the dry regime environment (Fig. 13b) than creating new cells. For areas such as central Niger where fair weather soundings are common, the results of the fair weather case study suggest land–atmosphere cooperation mechanisms that could produce the mesoscale soil moisture gradients that Taylor and Lebel (1998) observed during HAPEX-Sahel.

Our simulated convective lines are distinctly local, representative of small convective systems that develop from daytime heating of an unstable boundary layer during an active monsoon (Mohr et al. 2003). From TRMM data, Mohr (2004) and Mohr and Thorncroft (2006) observe that a majority of West African convective systems are small, weak, develop near locations of significant orography during active monsoons, and have a life cycle of a couple of hours. The results of the convective case in this study appear to describe the land–atmosphere exchange processes responsible for the development of the convective systems described in the observational studies. The differences between the fair weather case and convective case highlight the importance of the land surface in fair weather conditions via significant differences in CAPE generation between the two soil moisture regimes. Should an organized large convective system interact with a fair weather environment similar to the one presented here, it is reasonable to hypothesize that the system would exhibit comparable sensitivity to the soil moisture conditions.