1. Introduction
Mineral dust plays an important but poorly quantified role in the climate system (Ramaswamy et al. 2001). It is one of the most abundant aerosol species in the atmosphere and has an important but complex role in determining the shortwave and longwave radiation budgets (Penner et al. 2001). Dust also has an indirect role through its influence on cloud properties (Rosenfeld 2006). Many studies have indicated that most of the world’s mineral dust is emitted from a number of preferential source regions, notably within the Sahara (e.g., Washington et al. 2003). There is considerable evidence from satellite data that the Bodélé depression in northern Chad (centered near 17°N, 18°E; Fig. 1a) is the world’s most intense dust source (Goudie and Middleton 1992; Herman et al. 1997; Herrmann et al. 1999; Brooks and Legrand 2000; Prospero et al. 2002; Zhang and Christopher 2003; Washington et al. 2003). Spatially continuous dust plumes originating in the Bodélé extend for hundreds of kilometers (Figs. 1b,c) and occur with a remarkable frequency of ∼100 yr−1 (Washington and Todd 2005; Koren et al. 2006; Washington et al. 2006a; Todd et al. 2007). Todd et al. (2007) estimate that during dust plume events about 1.2 Tg of dust is emitted per day, leading to a contribution of between 6%–18% to the total global dust emission. This dust is transported across West Africa (Pinker et al. 2001; Washington et al. 2006a) and the tropical Atlantic. Transport and deposition as far as the Amazon basin has also been documented (Swap et al. 1992) and Koren et al. (2006) suggest that about 50% of all mineral dust transported to the Amazon originates from the Bodélé. This is corroborated by global dust model simulations that indicate that dust from the Bodélé is responsible for about 40% of total aerosol optical thickness over the tropical Atlantic and Amazon basin during winter (Tegen et al. 2006).
Washington et al. (2006b) have argued that two factors account for the primacy of the Bodélé depression as a dust source. First, it contains large amounts of erodible sediment comprising diatomite deposited from paleo Mega-Lake Chad during past wet periods, the last of which ended circa 5000 yr BP (Gasse 2002; Drake and Bristow 2006). The main area of diatomite sediment covers an area of approximately 10 800 km2 (the bright area within the Bodélé depression in Fig. 1a) and has a surface elevation <200 m. The present-day hyperarid climate ensures an absence of vegetation and high erodibility. Second, the Bodélé low-level jet (LLJ) is focused onto these erodible sediments and is responsible for dust emission and its subsequent long-range transport. This LLJ is embedded in the mean low-level northeasterly Harmattan winds (Washington and Todd 2005; Washington et al. 2006a, b). To the northeast of the Bodélé the Harmattan is strongly constrained by an approximately 300-km-wide gap or channel between the Tibesti Mountains to the north (with peaks up to 3415 m) and the Ennedi Mountains to the east (∼1500 m) (Fig. 1a). The surface elevation descends from ∼500 m in the gap southwestward to <200 m in the Bodélé (Fig. 1a). Koren and Kaufman (2004) and Washington et al. (2006b) have argued that this channel accelerates the near-surface flow. The LLJ is known to exhibit strong diurnal (Washington et al. 2006a) and intraseasonal (Washington and Todd 2005) variability, which controls dust emission and long-range transport.
An ability to simulate dust emission and the associated climate impact is crucial to climate modeling efforts, particularly in climate change integrations. A fundamental requirement is the accurate representation of emissions from preferential source regions, which, in turn, requires realistic representation of the wind field. For the Bodélé depression, the challenge is to simulate the mesoscale Bodélé LLJ feature and its relationship with surface winds. Pronounced diurnal variability in surface winds means that dust emission occurs in pulses (Koren et al. 2006; Todd et al. 2007), which the models must resolve. In this paper we compare regional climate model (RCM) simulations with meteorological observations from the Bodélé, which were obtained during the Bodélé Dust Experiment (BoDEx 2005; Washington et al. 2006a; Todd et al. 2007), the first meteorological field campaign to this remote region. Experiments are run with both the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) and regional climate model, version 3 (RegCM3). The first major aim is to evaluate the ability of regional models to simulate this key dust-generating atmospheric feature. Winds in most dust models are derived from global analyses or reanalyses typically with horizontal resolution of 2.5° or 1° at best (e.g., Ginoux et al. 2001; Tegen et al. 2002; Zender et al. 2003) and are known to seriously underestimate winds over the Bodélé (Koren and Kaufman 2004; Washington et al. 2006a; Todd et al. 2007). We answer a number of key questions: (i) How well do the RCM simulations of winds over the Bodélé match the observations? (ii) How do RCM model resolution and parameterization schemes influence the agreement with the observed? (iii) What RCM resolution is necessary to realistically simulate the behavior of the Bodélé LLJ? The second aim is to diagnose the causes of the LLJ and its diurnal cycle. We consider the roles of two primary elements, namely, orography and boundary layer mixing, and compare situations with and without strong synoptic forcing of the LLJ.
More broadly, the work has implications for modeling LLJ features elsewhere, an important consideration given the role of LLJs in modulating deep convection, notably over the Great Plains of the United States (see Stensrud 1996, and references therein), over West Africa in the monsoon season (Parker et al. 2005), and to the east of the Andes Mountains in South America (Paegle 1998; Marengo et al. 2002), as well as their role in determining air quality.
2. Data and methods
a. Regional climate model simulations
Most of the analysis presented here is based on multiple simulations using the MM5 model. MM5 is a nonhydrostatic, primitive equation model using terrain-following coordinates (Dudhia 1993; Grell et al. 1994). Simulations were run over the Bodélé domain for the BoDEx field campaign period, 25 February 2005–13 March 2005. National Centers for Environmental Prediction (NCEP) reanalysis fields (6 h, 2.5° horizontal resolution) provided lateral boundary conditions. In total, more than 30 MM5 experiments were conducted to test the sensitivity of the simulated LLJ feature to the following:
(i) Horizontal resolution. To determine the influence of horizontal resolution the model was configured with three nested domain grids in a two-way nesting. The mother domain has 81-km resolution and covers the region 4.4°–34.2°N, 4.0°–36.0°E, while the nested domains have a resolution of 27 km over 9°–24°N, 9°–24°E and 9 km over 15°–21°N, 15°–21°E, roughly centered on the Bodélé.
(ii) Vertical resolution. The sensitivity of low-level and near-surface winds to model vertical resolution was tested by varying the number of levels in the lowest 1000 m. Model experiments were conducted with 3, 7, 9, 11, 13, and 15 levels below 1000 m. Above this height all model experiments have 21 levels. These experiments were only conducted for the “optimum” model configuration described in section 4.
(iii) Parameterizations. Multiple model runs were conducted with combinations of (a) six planetary boundary layer (PBL) schemes—namely, Blackadar (BLR), Burk–Thompson (BT), eta, MRF, Pleim–Xiu (PX), and Gayno–Seaman (GS); (b) three land surface schemes (none, five layer, and Noah); and (c) three radiation schemes [Simple, Community Climate Model (CCM), and Rapid Radiative Transfer Model (RRTM)]. Clouds and precipitation are not a major issue in this study, so we do not expect sensitivity to cumulus or cloud parameterization.
(iv) Surface roughness. PBL winds are influenced by the frictional drag imposed by the underlying surface, which is a function of surface roughness (Z0). By default MM5 attributes a Z0 value of 0.1 m for desert surfaces, which is really only appropriate for partially vegetated desert surfaces. Barren desert surfaces in the Sahara are likely to have Z0 values closer to those cited by Oke (1987) of 0.0003 m. In our study Z0 for the desert land class, which covers the domain north of ∼15°N, was varied over three orders of magnitude from 0.1 to 0.0001 m. This latter value may not be unrealistic for the hard diatomite “pavement” surfaces of the parts of the Bodélé depression itself, although perhaps not for other regions north of ∼15°N.
(v) Orography. One experiment (referred to as NO-OROG) was conducted in which the surface topography was reduced to a uniform 200-m elevation (roughly the elevation of the Bodélé depression). Otherwise, the configuration is identical to the “optimum” experiment described in section 4a.
To minimize the risk of basing conclusions on a single model, an experiment was conducted using the RegCM3model (Giorgi et al. 1993a, b) over the same period and domain configured with a horizontal gridcell resolution of 20 km and 23 vertical levels. The physics options include the Holstag PBL scheme and the Grell convection scheme.
b. Comparative in situ observational data
Model-simulated fields were compared with observational data collected during BoDEx 2005 (Giles 2005; Washington et al. 2006a; Todd et al. 2007), a multidisciplinary field experiment conducted from 27 February to 13 March 2005. This was the first project to obtain in situ observations of meteorology in the Bodélé depression. The field observations were made at the far eastern margin of the massive diatomite surface of paleo Lake Mega-Chad near 16°53′N, 18°33′E, at an altitude of 179 m, immediately east of the sand sea of the Erg d’Djourab (Fig. 1a). This remote location, indicated as “Chicha” on Institut Géographique National (IGN) 1:1000000 maps, marks the southeastern extremity of the consistently sharp edge of dust plumes from the Bodélé evident on Moderate Resolution Imaging Spectroradiometer (MODIS) true color imagery (Figs. 1b,c).
Surface meteorology was recorded using a Davis Vantage Pro automatic weather station providing 2-min averages of temperature, dewpoint temperature, wind speed and direction, pressure, solar, and UV radiation at a height of 2 m. Time–height profiles of wind speed and direction were derived from pilot balloon (PIBAL) ascents. PIBALs have formed an important component of numerous field campaigns (e.g., Marengo et al. 2002; Egger et al. 2005). In BoDEx 2005 PIBAL ascents were made each day from Chicha at 0000, 0600, 0700, 0830, 1000, 1230, 1500, 1700. and 2100 UTC except during extreme dust events, notably 10–12 March. Washington et al. (2006a) provide a full description of the PIBAL method used and the characteristics of the data. These PIBAL observations of wind speed were averaged into bins with 3-hourly time interval and 20-hPa vertical pressure intervals. Unfortunately, vertical temperature and humidity profiles are not available due to the failure of a kite-based radiosonde on its first ascent.
It should be noted in comparison of model-simulated fields with in situ observations that the data represent averages of varying areas. The surface data are essentially point data, the model data represent an area average of 81 km2 for the smallest (9-km resolution) domain and 729 km2 for domain 2 (27-km resolution), while the PIBAL data represent a vertical transect through varying horizontal distance. Thus the observational data are likely to include far greater influence of local effects.
To supplement the in situ field data, we use wind speed estimates derived from the displacement of dust plumes from the Bodélé depression evident in successive MODIS satellite images (Figs. 1b,c). On most days two MODIS images are available from the Terra and Aqua satellites, typically some 3 h apart. Using the feature-tracking technique of Koren and Kaufman (2004), the displacement of the dust plume front (typically located some 500 km downwind of Chicha) can be mapped enabling the derivation of wind speed at the height of the plume. This source of information on boundary layer winds is important on dust event days during which time no PIBAL observations were possible. Finally, wind fields from ECMWF operational analyses on a 1° grid with 21 vertical sigma levels were utilized along with information on the large-scale circulation from NCEP reanalyses (2.5° grid; Kalnay et al. 1996).
c. Historical large-scale atmospheric fields
Information on historical large-scale atmospheric conditions was obtained from reanalysis data, specifically the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40; available on a 1° grid; Uppala et al. 2006) and the NCEP reanalysis. To minimize the effect of discontinuities in the data associated with abrupt changes in the data sources used in the model assimilation we employ only the period 1979–2001 to derive the mean climatological state. Both the ERA-40 and NCEP data have three vertical levels below 800 hPa (1000, 925, and 850 hPa).
3. The mean state of the Bodélé LLJ and the large-scale circulation
In winter North Africa is dominated by the northeasterly Harmattan winds, which occur in the lowest 100 hPa of the atmosphere. These winds are related to the north–south pressure gradient between the high-pressure ridge centered over Libya (the Libyan high) (Fig. 2a) and the monsoon trough/intertropical front to the south. In the ERA-40 reanalysis data, the mean winter (October–March) Bodélé LLJ is a pronounced feature embedded within the Harmattan. At 925 hPa, the height of the peak velocity, the LLJ appears as a NE–SW-oriented band of enhanced northeasterly wind overlying the Bodélé region with a maximum of ∼12 m s−1 near 17°N, 19°E (Fig. 2b). Wind speeds at 925 hPa in excess of 11 m s−1 extend over a large region from northern Chad to Lake Chad, a distance of some 1200 km. A secondary wind maximum is apparent to the west of Lake Chad and to the south of the Aïr Mountains centered near 15°N, 10°E. The high wind speeds to the northwest of the Tibesti (centered near 22°N, 14°E) indicate a split in the Harmattan wind associated with the presence of the mountains. In the early winter months the mean wind speed at Chicha in the jet core at 925 hPa is ∼12 m s−1 but only ∼2 m s−1 at 800 hPa (Fig. 2c). There is a marked annual cycle in the strength of the Bodélé LLJ with a maximum (minimum) in the winter (summer) months (Fig. 2c). This is associated with the annual cycle of the Hadley circulation with a marked meridional migration of the ITCZ and modulation of the strength of the subtropical anticyclones over North Africa. During the summer months of the West African monsoon the intertropical front, which marks the discontinuity between the dry northeasterlies and the moist southwesterlies, moves north to lie near 17°N over Chad, bringing light and variable winds to the Bodélé.
Differences between the ERA-40 winds and the lower-resolution NCEP winds most likely result from the representation of the Tibesti and Ennedi massifs. In the NCEP data the LLJ is displaced some 300 km to the south, extending in a broad band from eastern Chad west to western Niger (Fig. 2d). Wind speeds peak at 9 m s−1 to the north of Lake Chad and are only ∼8 m s−1 over the Bodélé depression. There is little evidence of a north–south split in the LLJ associated with the Tibesti. Koren and Kaufman (2004) have shown that the NCEP reanalyses data underestimate 925-hPa wind speed during dust plume events in this region by ∼60%, a finding corroborated by Washington et al. (2006a) and Todd et al. (2007) during the BoDEx 2005 campaign. We hypothesize that the underestimation of the LLJ wind speeds in the NCEP relative to the ERA-40 reanalysis is a result of the coarse resolution of the NCEP model. From this it is plausible that a model resolution of 2.5° is not sufficient to provide a realistic representation of the Bodélé LLJ location and intensity. We are able to explicitly evaluate the role of resolution in the suite of regional modeling experiments.
4. Evaluation of regional model simulations of the LLJ and surface winds during BoDEx 2005
The fidelity of each model simulation of winds over the BoDEx period is evaluated by comparison with the available in situ and satellite-derived wind observations. The key criteria for evaluation are the mean vertical position, strength, diurnal cycle, and day-to-day variability of the LLJ. Particular attention is given to the LLJ characteristics simulated under strong synoptic forcing, characteristic of major dust plume events. In addition, the ability of the models to resolve near-surface winds is evaluated, as this is the key control on dust emission. The MM5 simulation that provided the most accurate simulation of the mean LLJ and its variability over the BoDEx period was configured as follows: 36 vertical levels with 15 below 1000 m, eta PBL scheme, no land surface scheme, RRTM radiation, mixed phase moisture scheme, no convection scheme, and a Z0 value for desert surfaces set to 0.001 m. We refer to this experiment at the “optimum” configuration or OPT. This section begins with an overview of the LLJ characteristics as simulated by OPT with 27-km horizontal resolution (section 4a) followed by a summary of the sensitivity of the model LLJ to the various model configurations and a short comparison to the RegCM3 simulation (section 4b).
a. Comparison of winds simulated by the “optimum” model configuration with observations
1) The mean LLJ during BoDEx 2005
The vertical profile of mean wind speeds at Chicha as simulated in experiment OPT (over the period coincident with PIBAL data 28 February–9 March 2005) shows clear evidence of a low-level jet feature (Fig. 3a) that is very similar to that in PIBAL observations. The mean jet-core lies some 30 hPa (∼400 m) above the surface, and the altitude where wind speeds are ∼50% of the core maximum is ∼100 hPa above the surface, which agree exceptionally closely with PIBAL observations (Fig. 3b). There is a pronounced diurnal cycle in LLJ speeds whose phase is well represented in OPT with a nighttime (daytime) maximum (minimum) resulting in strong (weak) vertical shear between the surface and the jet. The ratio of wind speeds in the jet core at 0600 UTC to that at 1500 UTC (the “diurnal ratio”) is ∼4 in the PIBAL data and ∼3 in OPT, and both indicate a rapid collapse in LLJ strength between 0600 and 1200 UTC. There is some evidence in the PIBAL observations at the very lowest layers (below 980 hPa) that the LLJ is progressively mixed down to the surface after sunrise between 0600 and 1500 UTC, which is also apparent in the model simulation. In the PIBAL observations there is a region of higher wind speeds at ∼840 hPa that is not apparent in the model. Moreover, the MM5 OPT experiment underestimates the LLJ mean maximum (at 0600 UTC) by about 15%, with a maximum mean wind speed of 14.5 m s−1 compared to 17 m s−1 in PIBAL observations. Possible reasons for this underestimation are discussed in section 4b. The ECMWF operational analyses (Fig. 3c) and NCEP reanalysis (not shown) data broadly capture these diurnal characteristics, except that in both cases (i) the peak nighttime LLJ velocity is underestimated by ∼50% and (ii) the magnitude of the diurnal cycle is substantially damped compared with the observed data (diurnal ratio ∼1.5).
In experiment OPT (Fig. 4a) the LLJ is focused to the south of the Tibesti Mountains but extends almost as far south as Lake Chad. Mean wind speeds in the jet region are in excess of 11 m s−1, centered on 18°N, 18°E north of Chicha. The mean wind field over the BoDEx period at 925 hPa from ECMWF analyses (not shown) is structurally very similar to the MM5 27-km OPT experiment (Fig. 4a) and to the long-term mean ERA-40 data (Fig. 2b) with a peak wind speed in excess of 11 m s−1, centered on 18°N, 19°E. This wind speed maximum, however, while similar to that of the MM5 OPT experiment, reflects considerable overestimation (underestimation) during the day (night) (Fig. 3c).
2) Day-to-day variability in LLJ strength
Substantial dust emission occurred on 10–12 March (Fig. 1c), with minor emission on 28 February, and 4 and 9 March (Fig. 1b). Following Washington et al. (2006a) we divide the BoDEx period in two subsamples dependent on the strength of the synoptic forcing of the LLJ. Dust-free days (1–3 and 6–8 March) experienced low pressure systems across the Mediterranean (Fig. 5a) and a blocking anticyclone west of the British Isles, a signature typical of negative NAO months. Under these conditions the north–south pressure gradient across northern Chad was anomalously weak. We refer to these conditions as weak synoptic forcing of the LLJ. The large dust event from 10 to 12 March occurred when the blocking anticyclone over the northeastern Atlantic suddenly migrated eastward and extended a pronounced ridge across North Africa, drawing strong north-northeasterlies across the Bodélé region (Fig. 5b). We refer to these events as associated with strong synoptic forcing of the LLJ. Knippertz and Fink (2006) have analyzed the dynamics of a Saharan dust outbreak event in March 2004, in which similarly explosive anticyclogenesis occurred and caused widespread dust mobilization. They argue that evaporation of precipitation along the southern side of the Atlas Mountains and cold advection generated strong cooling and a density-current-like flow over the Sahara. The case of 10–12 March 2005 is largely consistent with this in that the explosive anticyclogenesis that spreads from the northwest is associated with a major cold surge in which 925-hPa temperature anomalies reach −10 K over an extensive region to the south of the Atlas Mountains (Fig. 5c) resulting in high geopotential over northern Africa (Fig. 5c) and the strongly ageostrophic north-northeasterly flow (Fig. 5b). Station reports (not shown) do indeed indicate rainfall over the Atlas Mountains on 8–10 March. The weaker dust event of 28 February coincided with the passage of a weak high pressure system across southern Algeria and southern Libya, while the brief event of 4 March featured the eastward migration of a slightly stronger high across the central and eastern Mediterranean. These changes in the large-scale circulation are reflected in LLJ strength variability from day to day observed in the PIBAL data (Fig. 6a).
(i) Strong synoptic forcing
The strong LLJ events on days with strong synoptic forcing (i.e., 28 February, 4 and 9–12 March) are very well simulated by the MM5 OPT model experiment (Figs. 6a,b). Simulated LLJ wind speeds in excess of 20 m s−1 at Chicha on those days are markedly stronger than on days with weak synoptic forcing. PIBAL data indicate wind speeds >25 m s−1 on 4 and 9 March, while MM5 simulates 25 m s−1 on the 4th but only 17 m s−1 on the 9th, although in the latter case this is due to a small error in the horizontal location of the LLJ (not shown) rather than a failure to simulate an enhanced LLJ feature on this date. On the mornings of 10–12 March when no PIBAL observations were possible the LLJ was most likely even stronger. MM5 OPT experiment indicates wind speeds up to 29 m s−1 on these days. Moreover, comparison with the MODIS satellite-based “observations” of the LLJ downwind of Chicha (see section 2b) shows good agreement (Table 1). The MM5 OPT estimates in Table 1 are averaged over the approximate region occupied by the dust plume advance between two satellite overpasses, although the uncertainty in the height of the dust plume in the satellite data should be acknowledged. The spatial structure of the LLJ during strong events indicates that the jet is strongest immediately to the south of the Tibesti Mountains (Fig. 4b) where wind speeds exceed 33 m s−1, an issue considered in more detail in section 5a.
(ii) Weak synoptic forcing
Overall, MM5 accurately represents the LLJ strength at Chicha during quiescent periods. This includes the weak LLJ on the night of 1–2 March (Figs. 6a,b), the only period during BoDEx on which low-level winds at Chicha deviated from easterlies. However, the model underestimates nocturnal maximum at Chicha, most notably on 6–7 March when PIBAL data indicate a maximum LLJ strength of 22 m s−1 and MM5 indicates only 16 m s−1, although LLJ speeds are higher to the south. On these days of weak synoptic forcing the LLJ in the model becomes a rather broad and ill-defined feature and shifts away from the Bodélé region to the south (∼14.5°N, east of Lake Chad) and southeast (near 16°N, 20°E), where a downslope component associated with the western flank of the Ennedi uplands may be important (Fig. 4c). A similar southward shift is observed in the ECWMF data (not shown). This is also consistent with observations of an LLJ over N’Djamena to the south of Lake Chad reported by Gouault (1938).
3) Near-surface winds
It has been amply demonstrated in previous work (Washington and Todd 2005; Washington et al. 2006a) that variability in the Bodélé LLJ is responsible for triggering the major dust events for which the region is renowned. The frequency of these events determines variability in atmospheric dust loadings at intraseasonal and interannual time scales (Washington and Todd 2005). However, it is winds at the surface that actually raise dust. PBL processes link the LLJ and surface winds. Washington et al. (2006a) note that the diurnal cycle of surface winds and the LLJ winds are out of phase, with the former exhibiting a maximum (minimum) in the late morning (night) in contrast to the nighttime (afternoon) maximum (minimum) in the LLJ. This, they suggest, results from the decoupling of the LLJ from the surface at night owing to a strong nocturnal near-surface temperature inversion, and mixing of the PBL during the day. This observation and inference is consistent with analysis of the diurnal cycle of low-level and surface winds across the entire Sahelian sector and in particular at Niamey, Niger, during the monsoon season (Parker et al. 2005).
Thus a key requirement of the models is accurate simulation of the magnitude and phase correlation of the LLJ and surface winds. The OPT simulation has a reasonable representation of the phase of the diurnal cycle of near-surface winds, including the midmorning maximum, the afternoon “shoulder” of sustained wind speeds, and the nighttime minimum (Fig. 7a). The magnitude of the maximum is underestimated by ∼12% but more notable is the overestimation of nighttime minima at 0000 UTC by ∼100%. The day-to-day variability in near-surface wind is generally very well simulated by OPT, notably the dust events of 4, 9, and 10–12 March (Fig. 7b). Todd et al. (2007) estimate that the 2-m-height wind speed threshold for dust emission is 10 m s−1 and in the OPT experiment this threshold is appropriately exceeded during the dust event of 10–12 March. Accurate representation of the diurnal cycle ensures that the duration of the winds >10 m s−1 is about right. These results compare favorably with other model simulations of the BoDEx event. Tegen et al. (2006) found that the Lokal-Model (LM) substantially underestimated peak wind speeds such that the dust emission scheme required a recalibration. The Regional Atmospheric Modeling System (RAMS) model evaluated by Bouet et al. (2007) resolved the high wind speeds but not the diurnal cycle. It is clear that careful configuration of RCMs is necessary to provide accurate wind fields for dust emission models. The performance of other PBL schemes in resolving surface winds is considered in section 4b(3).
b. Evaluation of sensitivity of regional model simulation of the LLJ
1) Sensitivity to horizontal resolution
In terms of the mean magnitude of the LLJ speed and its diurnal cycle, the model representation is rather insensitive to model horizontal resolution. In the OPT experiment the mean wind velocity in the LLJ core at Chicha is about 14.5 m s−1 at 0600 UTC for all three model resolutions (not shown). However, the impact of horizontal resolution is exhibited close to topographic features where increased resolution provides greater detail in the wind field. Under strong synoptic forcing (e.g., 0600 UTC 10 March 2005) these differences are more pronounced and instantaneous LLJ velocities can be up to 20% higher at 9 km compared to 81-km resolution especially immediately south of the southernmost and largest Tibesti peak (Figs. 4b,d). Such finescale differences could be important in dust emission models, although over the main area of exposed diatomite in the Bodélé depression differences are minimal. Therefore a horizontal resolution of a few tens of kilometers appears adequate to resolve the LLJ feature in this case.
2) Sensitivity to vertical resolution
The key features of the model-simulated LLJ, specifically the intensity, height, mean diurnal cycle, and day-to-day variability are largely insensitive to the vertical resolution in the lowest 1000 m as defined in our model experiments under the OPT configuration. Even with 24 vertical layers and only 3 in the lowest 1000 m the LLJ is well reproduced (not shown). However, near-surface winds are much more sensitive to vertical resolution. The downward mixing of the LLJ after sunrise evident in Figs. 3a and 3b is not apparent in the model runs with <30 levels and as such the diurnal cycle of surface winds is heavily damped (see eta with 23 levels in Fig. 7a). In the OPT configuration, to resolve the diurnal phase anticorrelation between near-surface winds and the LLJ requires high vertical resolution in the lowest few hundred meters and the lowest model layer to be close to the surface (<20 m) (Fig. 7a). This is crucial for dust emission models.
3) Sensitivity to PBL, land surface, radiation schemes, and surface roughness
In all model configurations a mean LLJ feature is present, with a core near 960 hPa and a marked diurnal cycle, and a spatial structure similar to that of the OPT experiment (Fig. 3a). However, the intensity of the feature is strongly sensitive to the PBL scheme used. As described in section 4a(1) the simulated LLJ intensity is closest to observations using the eta PBL scheme from which an underestimation of the nocturnal peak LLJ wind speeds of 15% is noted. The mean nocturnal peak LLJ speed as a proportion of that simulated using the eta PBL scheme is 48% for MRF, 52% for BLR, 70% for BT, 76% for PX, and 80% for GS. The underestimation of LLJ wind speeds using these PBL schemes is most pronounced under weak synoptic forcing. Under strong synoptic forcing the effects of physical channeling are more important.
Importantly for dust emission models the PX and eta produce peak daytime surface wind estimates close to observations (Fig. 7a). All the other PBL schemes (including the RegCM3 simulation) underestimate daytime maximum in surface winds (Fig. 7a). The nighttime minima are well simulated by all schemes except eta and PX, which overestimate. All schemes except GS represent well the phase of the diurnal cycle in surface winds but only eta and BT have notable evidence of the afternoon “shoulder” of sustained wind speeds. The phase of the surface winds from the GS leads observations by 3 h. Because surface winds are sensitive to surface roughness (further reducing the surface roughness in the model would increase surface winds for all schemes) the diurnal ratio of surface winds may be a more appropriate metric of comparison. All PBL schemes underestimate this quantity by about a factor of 2. It should be noted that this disparity between observed and modeled surface winds could be partly a result of comparison of area-averaged model output and point observations. However, Zhang and Zheng (2004) note a similar underestimation of the diurnal ratio in surface winds over the summertime continental United States in a study that used area-averaged observations.
Explaining differences in the performance of the PBL schemes is problematic due to the absence of detailed observational information on the surface energy budget and the vertical structure of the PBL in this study. As such we cannot verify model simulations of surface/PBL vertical fluxes of heat and momentum. A radiosonde mounted on a meteorological kite was destroyed when the kite platform failed on the first ascent during BoDEx. From our results there is some indication that overall the PBL schemes based on the more physically sound turbulent kinetic energy (TKE) closure of turbulence (eta, BT, and GS) provide a more realistic estimation of LLJ speeds than those based on K-theory (BLR, MRF) as might be expected. We do not find that nonlocal turbulence closure provides better performance than local closure. The schemes with nonlocal closure (BLR, MRF, PX) produce much deeper daytime PBLs than the other schemes due to their parameterization of large eddies, but this does not result in improved wind speed accuracy. Regarding the OPT configuration we may speculate that the tendency for the Eta scheme to underestimate (overestimate) the nocturnal LLJ (surface winds) may be connected. This condition may be indicative of excessive downward mixing of momentum between the LLJ layer and the surface associated with an underestimation of the decoupling of the two layers. All PBL schemes evaluated here use a local K-theory to determine vertical fluxes under stable conditions that are characteristic of all nighttime periods in this study, and downward transport of horizontal momentum from the LLJ to the surface is determined by the eddy exchange coefficients K. We may infer that K values are too high in the eta scheme and that lower values might reduce both the nocturnal overestimation of surface wind and underestimation of LLJ strength. It is possible that this is related to difficulties the models may have in resolving the near-surface temperature gradients, notably the inversion, which may be shallower and the magnitude larger than that simulated by the models, a condition noted in other semiarid regions by, for example, Zhong and Fast (2003) and Hanna and Yang (2001). In this condition the model will overestimate the vertical heat and momentum fluxes. The finding that the eta scheme performs best in this context should not be seen as a generalized statement. Other similar intercomparisons have found very different results (e.g., Berg and Zhong 2005; Bright and Mullen 2002; Zhang and Zheng 2004).
In OPT, sensitivity of winds to Z0 is more apparent at the surface rather than at the LLJ height. With the default value of Z0 = 0.1 m the daytime maximum in surface winds are underestimated by ∼60% (20%) (Fig. 7b) such that the winds never exceed the 10 m s−1 threshold for dust emission. With Z0 = 0.0001 m peak surface winds are overestimated by ∼20% (not shown) such that the best representation is obtained with Z0 = 0.001 m in the OPT configuration. With respect to land surface schemes the Noah land surface model (LSM) when used with the eta PBL scheme results in a slightly reduced LLJ intensity. This is likely to result from lower near-surface temperatures south of 16°N with the Noah LSM, which reduces the large-scale north–south temperature gradient. The LLJ strength is insensitive to the radiation schemes used. The RegCM3 simulation resulted in a mean LLJ with velocity roughly in the middle of the range of the MM5 simulations with various PBL schemes (not shown), but the 10-m-height winds have a diurnal out of phase with observations with a peak at 0600 UTC (Fig. 7a). Finally, it is important to note that underestimation of the LLJ velocity and surface winds may also result from errors in the lateral boundary conditions in which the RCMs are nested. As noted earlier the NCEP reanalysis underestimates the magnitude of the LLJ by ∼60% and displaces the jet about 3° latitude to far to the south.
5. Mechanisms driving the LLJ
a. The effect of orographic channeling
The effect of orographic channeling is most acutely expressed during conditions of strong synoptic forcing when the northeasterlies pass through the Tibesti–Ennedi gap. In the 27-km-resolution OPT experiment at 0600 UTC on 10 March maximum LLJ speeds of 34 m s−1 occur at 18.5°N, 18°E, immediately south of the largest peak in the Tibesti range (∼3415 m), where the channel width between the Tibesti and Ennedi Mountains is at its minimum (Figs. 4b,d). Figure 4e suggests that the orography of the Tibesti mountains creates a split in the LLJ north and south of the mountains with considerable acceleration to the south. The LLJ is asymmetric across the Tibesti/Ennedi gap with strongest winds closer to the Tibesti side (Fig. 4e). Given the high elevation of the Tibesti, it is very likely that the flow is blocked and forced to accelerate around the topographic barrier. On the Ennedi side of the gap the lower elevations mean higher Froude numbers and thus less tendency for blocking. The existence of shallow low-level cold air to the north of the Tibesti, potentially related to evaporational cooling even farther north [see Fig. 5c, section 4a(2), and Knippertz and Fink 2006], enhances the vertical stability and thus the tendency for blocking. After passing the gap this cold air is accelerated downslope into the Bodélé depression by the pressure gradient resulting from the density difference between the cold air to the northeast and the warm air on the southwestern side of the Tibesti.
Comparison of the LLJ in NO-OROG, in which all orography in the region is reduced to 200 m, with that in the otherwise identical OPT further reveals the role of orography in driving the LLJ. In NO-OROG a mean LLJ feature with a marked diurnal cycle is maintained at the location of Chicha but relative to OPT the mean maximum LLJ speed is reduced by ∼12%, and the jet core is about 20 hPa closer to the surface (Figs. 3a,d). The mean LLJ in NO-OROG appears as a broad feature centered near 12°N and is clearly displaced from its focus over the Bodélé depression to the south (not shown). During days of strong synoptic forcing, when the influence of channeling between the Tibesti and Ennedi Mountains is greatest, the LLJ appears in NO-OROG as a rather broad feature over the north of the domain and peak wind speeds are reduced by up to 40% over much of the domain in the NO-OROG simulation (Figs. 4b,f). During conditions of weak synoptic forcing the jet is displaced south to about 13°N and exists as a zonally broad feature over much of the Sahelian zone (not shown). These results indicate that the nocturnal LLJ exists in the study domain irrespective of orography, but that many of its key features, notably the intense core region focused onto the Bodélé, are a result of orographic channeling. In a similar study on the Turkana LLJ, located between the Ethiopian and the East African highlands, Indeje et al. (2001) found a peak wind speed reduction of about 60% after the removal of orography.
b. The diurnal cycle
A pronounced diurnal cycle in lower level winds has long been documented in the topical North African sector (e.g., Farquharson 1939). Indeed, Parker et al. (2005) demonstrate the importance of the nocturnal low-level flow for meridional moisture transport in the West African monsoon system. The strong diurnal cycle of the LLJ modeled and observed at Chicha (Fig. 3) is likely to be related to the inertial oscillation mechanism (Blackadar 1957), whereby diurnally varying eddy viscosity and therefore frictional force can lead to an oscillating LLJ. Nocturnal radiative cooling leads to a decoupling of the air above the developing inversion from surface friction, which creates an imbalance between the pressure gradient and the Coriolis force in this layer. This initiates an oscillation around the geostrophic wind resulting in supergeostrophic maximum wind speeds some hours after sunset (Blackadar 1957; Hoxit 1975). The oscillation has a period of 2π/f, where f is the Coriolis parameter. During the day the PBL is deep and turbulent mixing of momentum increases friction and reduces the wind to subgeostrophic speeds.
In RCM simulations over the Bodélé, mean temperature profiles show clear evidence of a pronounced nocturnal inversion of ∼9 K at Chicha below 950 hPa (Fig. 8). The LLJ is located near the top of this layer and increases with the strength of the inversion in the course of the night (Fig. 3a). Turbulent mixing triggered by surface heating removes the inversion by 0900 UTC and leads to a peak in near-surface wind speed (Fig. 7a) as the LLJ is mixed through the PBL. By 1200 UTC the lowest 200 hPa of the atmosphere has become well mixed and exhibits an almost dry-adiabatic lapse rate (Fig. 8). Comparison of the total winds with the geostrophic component at Chicha in a hodograph shows characteristics consistent with an inertial oscillation (Fig. 9a). At night, winds are supergeostrophic (by ∼5.4 m s−1 at 0600 UTC) while they are subgeostrophic (by ∼6.5 m s−1 at 1200 UTC) during the day. At 18°N the period of a pure inertial oscillation is approximately 34 h. This implies that turbulent mixing after sunrise interrupts the oscillation before a full half-period is completed and the maximum possible wind value is reached, thereby effectively locking the observed LLJ to the cycle of night and day. On weak synoptic forcing days the location of nocturnal LLJ maximum is displaced considerably to the south of the Bodélé (Fig. 4c) and is roughly collocated with the zone of maximum nocturnal temperature inversion (not shown). On these days wind hodographs throughout the broad zone of the LLJ show patterns (not shown) similar to the mean hodograph in Fig. 9a.
Generally, a marked diurnal cycle is maintained during strong synoptic forcing events even though the daytime minimum LLJ speeds are considerably higher (8–12 m s−1) than on days of weak synoptic forcing (Fig. 6b). The circular hodograph pattern in wind speed/direction is apparent but less clear than that found for the BoDEx period mean (Fig. 9b). The fact that the simulated northeasterly flow is strongly ageostrophic is likely to be related to the dynamics of the acceleration of the cold air described by Knippertz and Fink (2006). We infer that under conditions of strong synoptic forcing nighttime radiative cooling and cold air advection lead to pronounced near-surface cooling (Fig. 5c). However, mechanical turbulence associated with the stronger nighttime winds creates a shallow but well mixed dry adiabatic layer between the surface and 970 hPa, such that the inversion layer is elevated to lie between 957 and 880 hPa (see the plot for 0600 UTC on 10 March in Fig. 8). The magnitude of this inversion is similar to that on other days (Fig. 8) resulting in nocturnal decoupling of the LLJ from near-surface layers and consequent diurnal cycle as in other days. However, the LLJ is elevated slightly (Fig. 6b) as a result of the elevated inversion layer.
In addition, orography may influence the diurnal cycle of the LLJ through elevated heating and cooling over mountain regions. The model simulations suggest that the Tibesti massif to the north of the LLJ core does act as an elevated heat low during the daytime, as shown by local extremes in 850-hPa geopotential height and temperature (Fig. 10). This may reduce the north–south pressure gradient and consequently the strength of the northeasterly winds over this region during the day. From consideration of the magnitude of the diurnal cycle across the wider study domain (Fig. 11) it is apparent that the phase of the diurnal cycle in LLJ is similar across the region but the magnitude is most pronounced between 16° and 20°E, south of the Tibesti, the difference between 0600 and 1500 UTC being 8.9 m s−1 compared to 4.8 m s−1 averaged over 12°–16°E. Assuming a linear dependence of frictional effects on wind speed, however, the increased diurnal cycle south of the Tibesti is likely to reflect the higher mean wind speeds (Fig. 11) associated with channeling, rather than any influence of elevated heating/cooling. It should be noted that the MM5 simulations all use a desert surface albedo of 0.25. In reality the exposed diatomite has an albedo of ∼0.4 and the Tibesti ∼0.15 (Fig. 1a), such that the thermal contrast will be greater than that simulated in the model. The diurnal ratio at Chicha still reaches ∼2.7 in NO-OROG (compared to ∼3 in OPT) suggesting that nighttime frictional decoupling plays a dominant role over heating/cooling of the elevated surface. Further, comparison of the effect of orography on LLJ strength under weak and strong synoptic forcing can highlight this. In the former case the nighttime LLJ maximum in NO-OROG is actually quite similar to that in OPT, while in the latter underestimation reaches 40% (Figs. 4b,f). This suggests that the effect of orography is primarily expressed through the gap channeling with thermal forcing being a secondary process. Therefore, to the first order the diurnal cycle appears to be a result of diurnally varying eddy viscosity in line with LLJ features observed in numerous other locations (see May 1995; Stensrud 1996; and references therein).
6. Summary and conclusions
Mineral dust is an important component of the global climate system but one that is relatively poorly constrained in model simulations. The LLJ and surface winds of the Bodélé depression in northern Chad are responsible for the emission and transport of vast quantities of mineral dust aerosols over West Africa, the tropical Atlantic, and beyond. A requirement of model simulations of wind fields, which drive dust emission–transport models, is therefore an accurate representation of these parameters including their diurnal evolution. In this paper we evaluate the ability of regional models and global analyses and reanalyses to represent this key dust-generating atmospheric feature. We utilize observational data obtained during the Bodélé Dust Experiment (BoDEx 2005) for comparison. In answer to the questions posed in the introduction to this paper we can conclude as follows:
(i) How well do the RCM simulations of winds over the Bodélé match the observations? The vertical structure of the LLJ, its magnitude, mean diurnal cycle, and day-to-day variability are very well represented in the OPT configuration of MM5. Representation of the LLJ magnitude is strongly sensitive to the model PBL scheme. The magnitude of surface winds is strongly influenced by surface roughness and PBL scheme but can be well simulated when the model is carefully configured. All PBL schemes tend to underestimate the magnitude of the diurnal cycle in surface winds, which suggests that further development of these schemes for these particular conditions is necessary.
(ii) How does RCM model resolution influence the agreement with the observed? The representation of the intensity and structure of the LLJ is relatively insensitive to model resolution below the 81 km of the first RCM nest used here, except that fine detail close to orographic features is enhanced at the highest resolution. This could have important implications for dust emission models where dust sources are close to orographic features but seems not to be crucial in the case of the Bodélé where the erodible material is at least 100 km from the topographic features. The LLJ is quite insensitive to model vertical resolution. This is not the case for the surface winds due to the need to resolve the lowest few tens of meters where energy and momentum fluxes between the surface and the PBL result in the pronounced phase shift between the two layers.
(iii) What model resolution is necessary to realistically simulate the behavior of the Bodélé LLJ and surface winds? We conclude that models require at least a 1° resolution for an adequate representation of the time mean field of the LLJ. However, to resolve the fine detail of the diurnal cycle and the precise positioning of the LLJ a model resolution of a few tens of kilometers is necessary with output fields at least every 3 h. High vertical resolution in the lowest ∼100 m is necessary to resolve the near-surface winds adequately for dust emission simulation. Such detailed information is important given (a) the relatively small area of diatomite sediment from which dust is emitted and (b) the role of the diurnal cycle in creating the observed pulses of dust emission. RCM simulations with high horizontal resolution of 27 km or finer provide crucial additional detail on the LLJ location and intensity in comparison to coarse-resolution NCEP reanalysis and even ECMWF fields. Thus, the resolution of current general circulation models (GCMs) does not appear adequate to simulate many key features of the Bodélé LLJ.
The second aim of this paper is to diagnose the causes of the LLJ, specifically the role of orography and the contribution of diurnally varying eddy viscosity. We conclude that the latter process is the main driver of the pronounced diurnal cycle although elevated heating/cooling over the Tibesti may be responsible for a smaller component. However, orography is most clearly expressed through the effects of acceleration through the Tibesti–Ennedi gap under conditions of strong synoptic forcing. During these strong LLJ events the great Bodélé dust plumes occur, which are a dramatic and uniquely frequent feature of the regional climate. There is a close relationship between the orography and the spatial structure of the RCM-simulated LLJ. RCM sensitivity experiments with no orography demonstrate that orography acts to (i) focus the jet onto the Bodélé depression and (ii) accelerate the LLJ by up to 40% under conditions of strong synoptic forcing and by ∼12% in the time mean average over the BoDEx period. Downslope acceleration of cold air from the north might further enhance low-level wind speeds during these events. Under weak synoptic forcing the LLJ is evident as a much broader feature considerably displaced to the south and southeast. In summary the Bodélé LLJ may be seen as a particularly well-defined part of a wider low-level wind speed maximum extending across much of the Sahelian zone of Niger, Chad, and Sudan. This is largely a nocturnal feature whose diurnal cycle is driven largely by the inertial oscillation mechanism. Local orography acts to focus the jet onto the Bodélé depression. Our results suggest that dust models will benefit from high-resolution winds provided by suitably configured RCMs or high-resolution GCMs but that further research into PBL parameterization is required to improve the representation of near-surface winds in particular. We should add that the conclusions here on the relative performance of the various model configurations, especially the PBL schemes, should not be generalized to other regions.
Acknowledgments
The authors are grateful to many people whose help ensured that BoDEx 2005 took place. BoDEx 2005 was supported by the Gilchrist Educational Trust with the Gilchrist Fieldwork Award, administered by the Royal Geographical Society (with the Institute of British Geographers). We are greatly indebted to Mahamat Abrerahman Troumba for logistical support in Chad. We also thank Ilan Koren for provision of the satellite estimates’ wind speeds. The British Atmospheric Data Centre (BADC) provided the ERA-40 and ECMWF data. Thanks are also due to Bob Bornstein for very helpful discussions and to the referees for their comments.
REFERENCES
Berg, L. K., and S. Zhong, 2005: Sensitivity of MM5-simulated boundary layer characterizations to turbulence parameterizations. J. Appl. Meteor., 44 , 1467–1483.
Blackadar, A. K., 1957: Boundary layer wind maxima and their significance for the growth of nocturnal inversions. Bull. Amer. Meteor. Soc., 38 , 282–290.
Bouet, C., G. Cautenet, R. Washington, M. C. Todd, B. Laurent, G. Marticorena, and G. Bergametti, 2007: Mesoscale modeling of aeolian dust emission during the BoDEx 2005 experiment. Geophys. Res. Lett., 34 .L07812, doi:10.1029/2006GL029184.
Bright, D. R., and S. L. Mullen, 2002: The sensitivity of the numerical simulation of the southwest monsoon boundary layer to the choice of PBL turbulence parameterization in MM5. Wea. Forecasting, 17 , 99–113.
Brooks, N., and M. Legrand, 2000: Dust variability over northern Africa and rainfall in the Sahel. Linking Climate Change to Land Surface Change, S. McLaren and D. Kniveton, Eds., Kluwer Academic, 1–25.
Drake, N., and C. S. Bristow, 2006: Shorelines in the Sahara: Geomorphological evidence for an enhanced monsoon from palaeolake Megachad. Holocene, 16 , 901–911.
Dudhia, J., 1993: A nonhydrostatic version of the Penn State/NCAR Mesoscale model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121 , 1493–1513.
Egger, J., and Coauthors, 2005: Diurnal circulation of the Bolivian Altiplano. Part I: Observations. Mon. Wea. Rev., 133 , 911–924.
Farquharson, J. S., 1939: The diurnal variation of wind over tropical Africa. Quart. J. Roy. Meteor. Soc., 65 , 165–180.
Gasse, F., 2002: Diatom-inferred salinity and carbonate oxygen isotopes in Holocene waterbodies of the western Sahara and Sahel (Africa). Quat. Sci. Rev., 21 , 737–767.
Giles, J., 2005: The dustiest place on Earth. Nature, 434 , 816–819.
Ginoux, P., M. Chin, I. Tegen, J. Prospero, B. Holben, O. Dubovik, and S. J. Lin, 2001: Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res., 106 , 20255–20273.
Giorgi, F., G. Bates, and S. Nieman, 1993a: The multi-year surface climatology of a regional atmospheric model over the western United States. J. Climate, 6 , 75–95.
Giorgi, F., M. R. Marinucci, and G. T. Bates, 1993b: Development of a second-generation regional climate model (RegCM2): Boundary layer and radiative transfer processes. Mon. Wea. Rev., 121 , 2794–2813.
Gouault, J., 1938: Vents en altitude Fort Lamy (Tchad). Ann. Phys. Globe France d’Outre-Mer, 5 , 70–91.
Goudie, A. S., and N. J. Middleton, 1992: The changing frequency of dust storms through time. Climatic Change, 20 , 197–225.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 122 pp.
Hanna, S. R., and R. Yang, 2001: Evaluations of mesoscale models’ simulations of near-surface winds, temperature gradients, and mixing depths. J. Appl. Meteor., 40 , 1095–1104.
Herman, J. R., P. K. Bhartia, O. Torres, C. Hsu, C. Seftor, and E. Celarier, 1997: Global distribution of UV-absorbing aerosols from Nimbus 7/Toms data. J. Geophys. Res., 102 , 16911–16922.
Herrmann, L., K. Stahr, and R. Jahn, 1999: The importance of source region identification and their properties for soil-derived dust: The case of Harmattan dust sources for eastern West Africa. Contrib. Atmos. Phys., 72 , 141–150.
Hoxit, L. R., 1975: Diurnal variations in planetary boundary layer winds over land. Bound.-Layer Meteor., 8 , 21–38.
Indeje, M., F. H. M. Semazzi, L. Xie, and L. J. Ogallo, 2001: Mechanistic model simulations of the East African climate using NCAR Regional Climate Model: Influence of large-scale orography on the Turkana low-level jet. J. Climate, 14 , 2710–2724.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Knippertz, P., and A. H. Fink, 2006: Synoptic and dynamic aspects of an extreme springtime Saharan dust outbreak. Quart. J. Roy. Meteor. Soc., 132 , 1153–1177.
Koren, I., and Y. J. Kaufman, 2004: Direct wind measurements of Saharan dust events from Terra and Aqua satellites. Geophys. Res. Lett., 31 .L06122, doi:10.1029/2003GL019338.
Koren, I., Y. J. Kaufman, R. Washington, M. C. Todd, Y. Rudich, V. J. Martins, and D. Rosenfeld, 2006: The Bodélé depression—A single spot in the Sahara that provides most of the mineral dust to the Amazon forest. Environ. Res. Lett., 1 , 1–5.
Marengo, J. A., M. W. Douglas, and P. L. S. Dias, 2002: The South American low-level jet east of the Andes during the 1999 LBA-TRMM and LBA-WET AMC campaign. J. Geophys. Res., 107 .8079, doi:10.1029/2001JD001188.
May, P. T., 1995: The Australian nocturnal jet and durnal variations of boundary-layer winds over Mt. Isa in north-eastern Australia. Quart. J. Roy. Meteor. Soc., 121 , 987–1003.
Oke, T. R., 1987: Boundary Layer Climates. Methuen, 372 pp.
Paegle, J., 1998: A comparative review of South American low-level jets. Meteorologica, 23 , 73–81.
Parker, D. J., and Coauthors, 2005: The diurnal cycle of the west African monsoon circulation. Quart. J. Roy. Meteor. Soc., 131 , 2839–2860.
Penner, J. E., and Coauthors, 2001: Aerosols, their direct and indirect effects. Climate Change 2001: The Scientific Basis, J. T. Houghton et al., Eds, Cambridge University Press, 289–416.
Pinker, R. T., G. Pandithurai, B. N. Holben, O. Dubovik, and T. O. Aro, 2001: A dust outbreak episode in sub-Sahel West Africa. J. Geophys. Res., 106 , 22923–22930.
Prospero, J. M., P. Ginoux, O. Torres, S. E. Nicholson, and T. E. Gill, 2002: Environmental characterization of global sources of atmospheric soil dust identified with the Nimbus 7 Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product. Rev. Geophys., 40 .1002, doi:10.1029/2000RG000095.
Ramaswamy, V., and Coauthors, 2001: Radiative forcing of climate change. Climate Change 2001: The Scientific Basis, J. T. Houghton et al., Eds, Cambridge University Press, 350–416.
Rosenfeld, D., 2006: Atmosphere—Aerosols, clouds, and climate. Science, 312 , 1323–1324.
Stensrud, D. J., 1996: Importance of low-level jets to climate: A review. J. Climate, 9 , 1698–1711.
Swap, R., M. Garstang, S. Greco, R. Talbot, and P. Kållberg, 1992: Saharan dust in the Amazon basin. Tellus, 44B , 133–149.
Tegen, I., S. P. Harrison, K. Kohfeld, and I. C. Prentice, 2002: Impact of vegetation and preferential source areas on global dust aerosol: Results from a model study. J. Geophys. Res., 107 .4576, doi:10.1029/2001JD000963.
Tegen, I., B. Heinold, J. Helmert, M. C. Todd, R. Washington, and O. Dubovik, 2006: Modeling soil dust aerosol in the Bodèlè depression during the BODEX campaign. Atmos. Chem. Phys., 6 , 4345–4359.
Todd, M. C., R. Washington, V. Martins, O. Dubovik, G. Lizcano, S. Mbainayel, and S. Engelstaedter, 2007: Mineral dust emission from the Bodèlè Depression, Chad during BoDEx 2005. J. Geophys. Res., 112 .D06207, doi:10.1029/2006JD007170.
Uppala, S. M., and Coauthors, 2006: The ERA-40 Re-analysis. Quart. J. Roy. Meteor. Soc., 131 , 2961–3012.
Washington, R., and M. C. Todd, 2005: Atmospheric controls on mineral dust emission from the Bodélé depression, Chad: Intraseasonal to interannual variability and the role of the low level jet. Geophys. Res. Lett., 32 .L17701, doi:10.1029/2005GL023597.
Washington, R., M. C. Todd, N. Middleton, and A. S. Goudie, 2003: Dust-storm source areas determined by the total ozone monitoring spectrometer and surface observations. Ann. Assoc. Amer. Geogr., 93 , 297–313.
Washington, R., M. C. Todd, S. Engelstaedter, S. M’Bainayel, and F. Mitchell, 2006a: Dust and the low level circulation over the Bodélé depression, Chad: Observations from BoDEx 2005. J. Geophys. Res., 111 .D03201, doi:10.1029/2005JD006502.
Washington, R., and Coauthors, 2006b: Links between topography, wind, deflation, lakes and dust: The case of the Bodélé depression, Chad. Geophys. Res. Lett., 33 .L09401, doi:10.1029/2006GL025827.
Zender, C. S., H. Bian, and D. Newman, 2003: The mineral Dust Entrainment and Deposition (DEAD) model: Description and 1990s dust climatology. J. Geophys. Res., 108 .4416, doi:10.1029/2002JD002775.
Zhang, D-L., and W-Z. Zheng, 2004: Diurnal cycles of surface winds and temperatures as simulated by five boundary layer parameterizations. J. Appl. Meteor., 43 , 157–169.
Zhang, J., and S. A. Christopher, 2003: Longwave radiative forcing of Saharan dust aerosols estimated from MODIS, MISR, and CERES observations on Terra. Geophys. Res. Lett., 30 .2188, doi:10.1029/2003GL018479.
Zhong, S., and J. Fast, 2003: An evaluation of the MM5, RAMS and Meso-Eta models at sub-kilometer resolution using VTMX field campaign data in the Salt Lake Valley. Mon. Wea. Rev., 131 , 1301–1322.
(a) MODIS channel 4 surface reflectance (8-day composite 27 Feb–6 Mar 2005), (b) MODIS true color composite (channels 3, 4, and 1) image for 1245 UTC 4 Mar 2005, and (c) as in (b) but for 0950 UTC 11 Mar 2005. The location of the BoDEx 2005 field site at Chicha (16.9°N, 18.5°E) is indicated in (a).
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Climatological mean circulation over the study region (1979–2001). (a) Mean winter (October–March) sea level pressure (hPa) from NCEP reanalysis. (b) Mean winter 925-hPa wind speed (m s−1, shaded) and wind vectors from ERA-40 reanalysis. Dashed line represents 1000-m surface elevation contour. (c) Vertical profile of monthly mean wind speed (m s−1) at the grid cell over Chicha (17°N, 18.5°E) from ERA-40. (d) As in (b) but for NCEP reanalysis data.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Vertical profile of the mean diurnal cycle of wind speed (m s−1) at Chicha (17°N, 18.5°E) over 28 Feb–9 Mar (coincident with PIBAL data) from (a) MM5 OPT experiment, (b) PIBAL data, (c) ECMWF operational analyses, and (d) MM5 simulation NO-OROG with reduced topography. All MM5 data are for 27-km resolution. All times are UTC. Local time is 1 h ahead of UTC.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
MM5-simulated wind speed (m s−1) over the model domain. (a) Mean 960-hPa wind speed over the BoDEx period from MM5 OPT experiment at 27-km resolution; (b) as in (a) but for 0600 UTC 10 Mar 2005; (c) as in (a) but averaged over 1–3 and 6–8 Mar 2005; (d) as in (b) but for 9-km resolution, where location of domain is indicated by box in (b); (e) latitude–height profile of wind speed at 19°E from MM5 experiment OPT at 0600 UTC 10 Mar 2005; and (f) as in (b) but for NO-OROG experiment. In each case areas where surface elevation is above a given pressure level are masked out.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Regional circulation for selected days during BoDEx 2005. (a) Mean 925-hPa geopotential height (m) and vector winds for dust-free days 1–3 and 6–8 Mar. (b) As in (a) but for dust event of 10–12 Mar. Data are from NCEP reanalysis. Unit vector wind speed is shown in m s−1. (c) The 925-hPa temperature anomalies (K, shaded) and 925-hPa geopotential height (gpm, contours) at 0600 UTC for 10 Mar 2005 (with respect to the 0600 UTC mean 3–17 Mar 2005) from ECMWF operational analyses. Note that the grayscale for temperature anomalies is the same for positive and negative values, but that negative (positive) anomalies have dotted (solid) contours.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Time–height section of wind speed (m s−1) at Chicha (17°N, 18.5°E) from (a) PIBAL observations and (b) MM5 OPT with 27-km resolution.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Near-surface wind speeds (m s−1) at Chicha from observations and various model experiments. (a) Mean diurnal cycle over BoDEx period; (b) time series over BoDEx period where thick line is observations, thin line is MM5 OPT, and dotted line is MM5 OPT but with Z0 = 0.1 m. All model winds refer to 3-m height except where noted in legend.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Vertical air temperature profile (K) at Chicha (17°N, 18.5°E) from MM5 experiment OPT with 27-km resolution at 1500 UTC averaged over BoDEx period (solid line), at 0600 UTC averaged over BoDEx period (dashed line), and at 0600 UTC 10 Mar 2005 (thick dotted line). Dry adiabats are shown as thin dotted lines.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
The 940-hPa wind hodograph at Chicha from the MM5 OPT experiment at 27-km resolution showing zonal and meridional components of wind (m s−1) for (a) mean diurnal cycle over BoDEx period and (b) diurnal cycle over 10–11 Mar 2005. Crosses are total wind; circles are geostrophic winds. Times (UTC) given only for total winds.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Mean 850-hPa geopotential height (shaded, m) and temperature (contours, K) at 1200 UTC averaged over BoDEx period from MM5 experiment OPT at 27-km resolution.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Lon transect at 18°N of mean diurnal cycle of 940-hPa wind speed (0600 UTC minus 1500 UTC in m s−1 solid line) and daily mean wind speed (m s−1, dotted line) from MM5 experiment OPT.
Citation: Journal of Climate 21, 5; 10.1175/2007JCLI1766.1
Comparison of satellite-derived wind speeds downwind of the Bodélé depression with 960-hPa wind speed as simulated by the MM5 OPT experiment.
