The BAe146 aircraft operates a commercial Leosphere ALS450 backscatter lidar suitable for aerosol and thin cloud observation (Marenco et al., 2011). A description of the lidar system is provided by Chazette et al. (2012) and technical information is available in Table 1 of Marenco et al. (2014). The nadir-viewing lidar provides elastic backscatter at 355 nm and features an uncalibrated depolarisation channel, used qualitatively to distinguish depolarising layers. Data are recorded at a vertical resolution of 1.5 m and an integration time of 2 s, giving a horizontal resolution of approximately 200 m at aircraft speeds. The instrument is lightweight, has a relatively small receiver aperture of 15 cm diameter, has a 12 mJ pulse energy (20 Hz PRF) output and requires a low level of maintenance which makes it ideal for frequent operation aboard the BAe146 aircraft. However, as a consequence of aperture size and pulse energy, the signal-to-noise ratio is poorer compared to the Falcon LEANDRE Nouvelle Génération (LNG) lidar.

Initial quick-look data are provided as a range-square-corrected signal (arbitrary units), which is proportional to the total backscatter coefficient from molecules and particles at a given range, r, times the two-way transmission of light from the laser source to the range r (i.e. a function of the atmospheric optical depth), for example, as shown in Figs. 7 and 17, for which no attempt has been made to correct for attenuation by the aerosol layers. In these cases, we use the Leosphere lidar data to locate dust layers and clouds, for which the range-corrected backscattered signal is sufficient, although dust layers lower in the atmosphere may not always be evident with such a representation, due to attenuation at higher altitudes.

In a further step, aerosol extinction coefficient can be computed from the lidar range-square-corrected backscatter signal using the method described by Marenco et al. (2013), although this is labour-intensive since the method is not automated and it requires a profile-by-profile review of assumptions. Additionally, the signal-to-noise ratio for the dust laden atmosphere in the Fennec region often causes difficulties in inverting the lidar backscatter signal to extinction coefficients. This can be overcome by integrating the lidar signals as in Sodemann et al. (2015): decreasing resolution to 300 m in the vertical and a 60 s integration time, translating to extinction coefficient profiles provided at a ∼ 9 km along-track footprint at a typical ground speed of ∼ 150 m s-1. In the lowest 0–2 km layer, the uncertainty in the extinction coefficient is of the order of 100 %, but this uncertainty quickly decreases above this height, the extent of which is dependent on the ambient aerosol conditions (e.g. Marenco et al., 2014).

A very important consideration when observing aerosol particles is the efficiency of the transmission system which passes external aerosol into the aircraft cabin for collection or in situ analysis. This is highlighted in the difficulty in making accurate and reliable measurements from an aircraft platform, particularly that of coarse-mode aerosol (Wendisch et al., 2004). For objectives such as those of the Fennec programme, this is of particular importance since a significant fraction of mineral dust is in the coarse mode (Weinzierl et al., 2009). Inlet design can modify aerosol size distribution through either underestimation due to aerosol losses or overestimation due to enhancements.

The BAe146 has a specialised low turbulence inlet (LTI) which is designed to provide a characterised community inlet capable of delivering supermicron aerosol into the cabin. This is achieved by reducing turbulent flow within the tip of the inlet, reducing impaction of particles to the walls of the inlet (Wilson et al., 2004). The LTI further maintains isokinetic sampling flow using a feedback-controlled pumping system.

A Grimm Technik optical particle counter (OPC) was mounted inside the aircraft cabin behind the LTI (LTI-GRIMM), and showed that size distributions behind the LTI compare well with those from the externally mounted aircraft probes. In order to further evaluate inlet efficiency on the BAe146, Grimm OPCs were mounted behind various Rosemount inlets. This allowed for the evaluation of the size distributions passed by the standard BAe146 Rosemount inlets for the first time, from which many of the internally installed aerosol instruments draw their sample from, such as the nephelometer, particle soot absorption photometer and aerosol mass spectrometer (Trembath, 2012; Trembath et al., 2012). Significant losses and enhancements of the size distribution have been found to occur at different size ranges.

During Fennec, two TSI 3563 integrating nephelometers measuring scattering at 450, 550 and 700 nm were operated inside the aircraft cabin behind a Rosemount inlet. During Fennec 2011, the nephelometers were run in series with a BGI Very Sharp Cut Cyclone Impactor between them. The impactor has a 50 % penetration efficiency at 2.5 µm aerodynamic diameter, or around 1.5 µm geometric diameter, at a flow rate of 16.67 litres per minute (L min-1). This therefore allows the “first” nephelometer to measure scattering due to all particles passing the Rosemount inlet and the pipe work (estimated to be particles smaller than 2.5 microns, Trembath, 2012), and the “second” nephelometer to measure scattering from the fraction of particles smaller than 1.5 microns. However, due to the nephelometers being located in a series, it was difficult to account for the loss of particles between the two instruments. Therefore, during Fennec 2012 the two nephelometers were operated in parallel to avoid this problem. This was possible because a more powerful pump was used, capable of 50 L min-1, even up to altitudes of up to 9000 m. Secondly, a volume flow controller was installed to replace the mass flow meter and needle valve.

The synergy in the approach of operating a Grimm OPC behind a Rosemount inlet to measure the size distribution, and the use of the impactor to separate the sub-1.5 micron scattering from that measured as standard by the nephelometer is novel; it allows any bias in scattering and absorption due to Rosemount inlet and pipe work effects on the BAe146 to be assessed for the first time, which can lead to significant underestimation of dust absorption properties when not accounted for (Ryder et al., 2013b).

The BAe146 is well equipped to measure aerosol size distributions (for example, see Haywood et al., 2008; Johnson et al., 2012). However, the Fennec campaign was unusual amongst aerosol campaigns in the large number of instruments operated to measure particles larger than 3 µm diameter, and in the measurement of “giant-mode” particles of over 30–40 µm. Interestingly, the recent eruption of Eyjafjallajökull in Iceland has reinvigorated the interest in 1–10µm particles since volcanic ash is generally in the same size region as mineral dust and they both have similar challenges to understanding such as non-spherical morphology (Ansmann et al., 2012); therefore, there is considerable benefit to be gained from the concerted efforts surrounding the observation of volcanic ash.

Instruments measuring size distribution, and the size ranges measured, are shown in Table  and also in detail by Ryder et al. (2013b). During Fennec 2011, a total of six different instruments successfully measured size distributions between 0.15 and 300 microns diameter – namely the Passive Cavity Aerosol Spectrometer Probe (PCASP; accumulation mode), CDP (cloud droplet probe), LTI-GRIMM, SID2H and Cloud and Aerosol Spectrometer (CAS; coarse mode), and the University of Manchester CIP15 in giant mode (see Table  for explanation of acronyms). All of these are wing mounted except the LTI-GRIMM, and all are optical particle counters, making use of light-scattering techniques, except the CIP which uses imaging shadowing techniques (Knollenberg, 1970). Although the CIP15 is capable of measuring particles sizes of up to 930 µm, electrical noise only allowed measurements of up to 300 µm. During Fennec 2012, a slightly different suite of instruments was operated given certain logistical requirements and consisted of a PCASP, CDP, 2DC, SID2H, FAAM CIP15 and FAAM CIP100. Unfortunately the CIP15 suffered from electrical noise during the 2012 IOP and the data were not usable. However, the operation of other instruments such as the CDP and 2DC provide alternative measurements for this size range. Additionally, the operation of the CIP100 probe extends the measurement range up to 6200 µm.

When interpreting OPC size distribution data, it is important to note various limitations and uncertainties (e.g. Reid et al., 2003). In order to deal with several sources of uncertainty regarding OPC measurements, the instruments were calibrated and size distributions carefully processed as described in detail by Rosenberg et al. (2012). The PCASP was calibrated with PSL nanospheres with diameters from 0.4 to 3 µm and oil particles size selected by a differential mobility analyser (DMA) with diameters from 0.145 to 0.360 µm. The CDP was calibrated with glass beads, ranging from 15.9 to 49.9 µm. Smaller beads were not used due to a tendency for them to clump together; therefore, the calibration was extrapolated below this size (including over the size range influenced by the inflection in the Mie response curve). Uncertainties due to this extrapolation were included in the total uncertainty budget. Our approach is to use a rigorous methodology to assign uncertainties to the data which take account of inherent problems associated with processing OPC data. Each OPC is considered to be an instrument which directly measures the particle scattering cross section and it is calibrated in terms of this variable. Using the uncertainty in this calibration and Mie theory with an appropriate refractive index for the measured aerosol, we derive a probability density function which gives the probability of a particle of a particular size being counted in a particular OPC bin. Integrating this probability density function allows us to derive the mean diameter and effective width of each bin. This method also permits full uncertainty propagation including ambiguities caused by the nonlinear and non-monotonic Mie theory relating scattering cross-section to particle diameter. For example, there is an inflection point in this relationship in the 5 to 10 µm range, which results in larger bin size errors across this size range (e.g. see horizontal error bars in Fig. 4). Thus, we represent the degeneracy in the response curve using uncertainties in the bin widths and bin centre points without any need for arbitrary smoothing or human thresholds. Note that this method results in bin widths significantly different to those provided by the manufacturer, which if used, would have introduced artefacts in the size distributions. Finally, we highlight the regular calibration of the CDP probe during the campaign, which results in better characterised size distributions (see Rosenberg et al., 2012).

Reid et al. (2003) outline various other deficiencies in previously presented OPC results. For example, they suggest that their OPCs were not able to represent size distribution variability which they believed was occurring in reality. This was not the case during Fennec; for example, the effective diameter ranged from under 2 µm to over 20 µm (Ryder et al., 2013b), and Fig. 4 in this article clearly shows contrasting size distributions where the peaks were either narrow and centred at 10 microns diameter or broad – from 10 to 70 microns. During Fennec the OPCs were clearly responding to different ambient distributions. Reid et al. (2003) also suggest that unknown particle refractive index and shape factor has affected OPC results. Here we processed the OPC data using refractive indices spanning 1.53–0.001i to 1.53–0.003i and errors in diameter and number concentration due to this uncertainty were propagated (sensitivity tests using different real parts of the refractive index showed little impact on the final size distribution). The size distributions were produced assuming spherical particles rather than non-spherical particles, which has been shown to have a negligible impact on the resulting size distributions (Osborne et al., 2011; Veihelmann et al., 2006; Lacis and Mishchenko, 1995; Liu et al., 1992). Additionally, instruments which utilised light-scattering measurements at different scattering angle ranges (such as the CDP at 4 to 12∘, compared to the GRIMM Technik OPCs at 30 to 150∘ and 81 to 99 ∘) produced similar size distributions (Ryder et al., 2013b) suggesting that sensitivity to viewing angle during Fennec was minimal.

Of particular note during Fennec was the operation of shadow imaging probes, such as the CIP15 during 2011 for the measurement of particles sizes of 15 microns and above. These data are particularly valuable because unlike optical particle counter data, they do not rely the non-monotonic Mie scattering relationships to derive particle size. Both Rosenberg et al. (2012) and Ryder et al. (2013b) show that the CIP15 and CDP/SID2H size distributions agree well in the overlap zone, suggesting accurate measurements of size distributions, despite the different measurement techniques applied. This further emphasises that the reliability of the Fennec size distributions presented here.

Additionally, the PCASP and CDP agree well at their overlap zones (see Ryder et al., 2013b, and Rosenberg et al., 2012, for full details). The combination of these rigorous calibration regimes, detailed processing procedures and agreement between instruments gives good confidence in the measured size distributions, particularly when significant numbers of coarse particles are present (e.g. see Sect. ). When operated and processed with care and attention as described above, where the key uncertainties are quantified and in combination with other instrumentation, OPCs provide results which are reliable for representing volume distributions in the coarse mode.

In addition to the core pyranometers on the upper and lower parts of the aircraft fuselage measuring downwelling and upwelling shortwave irradiance respectively, a number of specialist radiometers were operated during Fennec, which in the future will allow for considerably more detailed radiative measurements and radiative closure to be performed. In the shortwave spectrum, the Spectral Hemispheric Irradiance MeasurementS (SHIMS) measured spectrally resolved up- and downwelling irradiance from 0.3 to 1.7 µm. The shortwave spectrometer (SWS) measures spectrally resolved radiances from 0.3 to 1.7 µm, using an externally mounted scanning telescope designed for viewing at particular angles. In the longwave spectrum, the Airborne Research Interferometer Evaluation System (ARIES) measured spectrally resolved radiances from 3.3 to 18 µm, at either nadir or zenith, as well as several different downward-pointing angles. Further details of SHIMS, SWS and ARIES can be found in Osborne et al. (2011). Operation of these instruments allows detailed radiative closure to be performed (e.g. Haywood et al., 2011b; Osborne et al., 2011). Further work will examine the radiative measurements made under extremely high dust loadings when very large particles were present.

Due to the scientific objectives of the Fennec programme, the ability of the aircraft to make robust observations of atmospheric turbulence was of paramount importance. Three-dimensional wind vectors are generated using a five-port radome-mounted turbulence probe at the aircraft nose which provides angle of attack (AOA) measurements. These are combined with pitot tube measurements of air speed and position information from a GPS inertial navigation unit to generated ground referenced wind vectors at 32 Hz (Petersen and Renfrew, 2009). A known linear dependence between the vertical component and aircraft pitch results in additional post-processing. This is likely the result of uncertainties in the calibration of the turbulence or pitot probes. Some of the parameters (static pressure and airspeed required for the processing) are generated through the on-board aircraft computer; this is calibrated in situ annually as part of the maintenance schedule, using a pressure calibrator. Airspeed is calibrated similarly. The radome transducers are calibrated at a calibration laboratory annually, or as determined by inspection of the data for drifts or other artefacts. The inertial navigation unit (INU) alignment is assessed annually with a physical survey for pitch, roll, and heading. AOA and angle of sideslip (AOSS) calibrations derive from AOA/AOSS flight manoeuvres that were carried out when the facility was commissioned, as they are physically dependent on the radome mounting. These were subsequently validated to confirm this. The AOA/AOSS are further corrected using yawing orbits, where further corrections are introduced to these quantities. True airspeed is corrected using reverse-heading manoeuvres, where the correction minimises the difference in derived upward/downward wind measurements.

The concentration and properties of cloud condensation nuclei (CCN) were measured using a commercial dual column continuous flow stream-wise thermal gradient instrument (Droplet Measurement Technologies, Boulder, Co). The principles of its design are outlined in Roberts and Nenes (2005), Lance et al. (2006) and Rose et al. (2008). Ambient air is drawn into a pair of temperature-controlled columns where it encounters a particle free sheath flow which is humidified to near-saturation. A thermal gradient exists along each of the columns, meaning that supersaturation occurs as the samples flows through the columns. Activated aerosol forms droplets which increase in size depending upon their hygroscopicity. The instrument is configured to provide a pair of supersaturations at any time and has a supersaturation range nominally between 0.07 and 2 %. The residence time within the humidified zone is sufficient for these activated droplets to grow to diameters larger than 1 µm; all particles with a diameter below this threshold are judged to be unactivated interstitial particles. An optical particle counter at the base of each column estimates the size distribution of the droplets (0.75–10 µm across 20 size bins).

In order to ensure stable volumetric flow to the CCN instrument, vital for robust measurements across altitude ranges encountered by airborne platforms, it draws air from a reduced pressure buffer volume which is connected to a modified Rosemount 102E inlet (Trembath, 2012). In addition to the CCN, a condensation particle counter, CPC (modified 3786 UCPC, Quant Technologies) also samples from this plenum to allow the total concentration of particles (2.5 nm–3 µm) to be determined.

The synoptic situation during the short Fennec pilot campaign during 5–8 April 2011 generated numerous dust emission events characteristic of springtime dust events over the Sahara. On 1 and 2 April, a high pressure ridge over Algeria–Libya sector drove a strong northeasterly harmattan surge over the central–eastern Sahara activating multiple dust sources in Algeria, Libya, Niger and Chad created a large dust plume of advected dust southwestward over northern Mali, southern Algeria by 3 April. Further westward transport of this plume into the Fennec aircraft operations zone was prevented by strong northeasterly circulation around an intense cut-off low on 3–4 April (feature A in Fig. 3a). This low tracked northwards from western Algeria to Morocco over this period and was accompanied by strong cyclonic near-surface winds with pronounced dust emission along primary and secondary cold fronts penetrating southeastward over southern Morocco and northern Mauritania on 4 April. Fennec flight b589 was able to observe this dust feature and the accompanying cold surge. Subsequent flights on 5–8 April (see Table ) observed the interaction of the cold maritime intrusion with dusty Saharan air, after which the dust was transported towards Portugal (Preissler et al., 2011).

During this IOP most of the F20 and BAe146 flights were conducted over northern Mauritania and northern Mali. In terms of the large-scale structure of the atmosphere during June 2011 in this region, a clear distinction can be made between a “maritime phase” from around 2 to 12 June and a “heat low phase” from around 13 to 30 June (see Todd et al., 2013, for full details). These phases essentially determine conditions across the entire central–western Sahara. These maritime and heat low phases are broadly congruent with the “east” and “west” and phases, respectively, of the intraseasonal SHL mode of variability described by Chauvin et al. (2010). During the maritime (“heat low east”) phase the upper-level pattern exhibited a trough centred over the Iberian Peninsula extending southwards over the northern extremity of northern Africa (feature A in Fig. 3b). In addition, at low levels the SHL remained relatively stationary in an anomalously eastward location centred at ∼ 15∘ E (feature B in Fig. 3b), similar to the mean state for May, and the Azores High ridged towards the coast of northwest Africa. These conditions combined to drive anomalous westerlies throughout the troposphere over northwest Africa creating a strong northwesterly inflow of maritime air over much of the Fennec flight domain (feature C in Fig. 3b), with the ITD displaced southward (not shown). As such, the Sahara is effectively “ventilated” by cool advection from the Atlantic sector restricting the heat low to the central–eastern Sahara. Accordingly, Fennec observations at both supersites (not shown) indicate that the SABL during the maritime phase is anomalously cool and dry with shallow daytime convective boundary layer development (Marsham et al., 2013b; Todd et al., 2013) and generally cloud free conditions. Aerosol loading was low due to the relative absence over the Fennec flight domain of the two dominant dust-generating processes, namely cold pools from moist convective systems, favoured within the southerly monsoon flow (ITD “bulge”) on the eastern flank of the SHL, and enhanced northeasterly harmattan winds around the western flank of the SHL trough. As a consequence, these two dust-generating activities were largely restricted to the central Sahara with the eastward-displaced SHL.

Subsequently, during the latter heat low (west) phase anomalous positive geopotential heights dominated over Iberia and the extremity of northwest Africa (feature A in Fig. 3c), associated with the passage of three upper-level ridges. At lower levels, the SHL exhibited an abrupt westward displacement to ∼ 5–10∘ W (feature B in Fig. 3c) in two distinct intraseasonal pulses. These conditions combined to drive anomalous mid- and upper-level easterly flow, with easterlies at lower levels around the SHL, evident over the western Saharan sector (feature C in Fig. 3c) and Fennec flight domain. Fennec ground-based observations indicate the SABL during the heat low phase of June 2011 to be substantially hotter with deeper afternoon convective boundary layer (CBL) development and cases of almost “pure” well-mixed near-dry adiabatic profiles from the surface to the top of the Saharan residual layer (SRL) at ∼ 5 km height. Dust aerosol loadings are substantially higher over the western Sahara region and Fennec flight domain during the heat low phase associated with enhanced mesoscale convective activity and strong easterlies around the heat low and African easterly wave troughs. Shallow convective clouds often developed in the late afternoon in the relatively moist upper SRL.

Flight planning to meet Fennec science objectives was largely determined by synoptic meteorology, as well as logistical constraints. As such, the science objectives of specific flights (Tables , and ) are geared to the prevailing meteorology described above. Overall, flights during the maritime phase (Falcon only) were able to sample substantial dust emission events over northern Mauritania (F13, F18). During the heat low phase, certain flights were able to measure dust/meteorological processes associated with both northeasterly low-level jet-related emissions (e.g. b600/601/602, b610, b614) and mesoscale convective system (MCS) cold pool events originating over central Mali (b604) and also the Atlas Mountains to the north (b605 and F22/F23). Flights to survey the SABL were able to measure the pronounced evolution in the structure of the PBL over this transition from the maritime (e.g. F14–F17) to heat low phases (e.g. b607/b608, F24/F25), representing the intraseasonal variability and seasonal evolution of the Saharan atmosphere.

Unlike the equivalent period of June 2011 Fennec IOP 2012 period 1–17 June there was no clear projection of the circulation onto the east–west heat low mode of Chauvin et al. (2010). As such, the period was characterised by a relatively stationary SHL centred close to the triple point of Algeria, Niger and Mali, further west than during the first half of June 2011. However, relatively subtle synoptic-scale variations strongly influenced the circulation over the western Saharan sector and the Fennec flight domain. First, during the early part of June 2012 (1 to 9), a weak upper-level trough extended south towards the coast of Morocco (feature A, Fig. 3d) and a heat low extension was established over far western Algeria (feature B, Fig. 3d) driving a strong northwesterly maritime flow over the Fennec domain (feature C, Fig. 3d). As with the maritime phase of IOP 2011, this led to the characteristic maritime conditions of a cool, dry SABL with shallow CBL daytime development and relatively cloud- and aerosol-free conditions over almost all the domain. This maritime flow weakened after the 10 June and a heat low extension west into northwest Mali from 14 to 17 June (not shown) established more characteristic heat low SABL conditions over the eastern Fennec flight domain. Specifically, a strong northeasterly low-level flow around the western flank of the SHL trough favourable to dust emissions and a northern extension of monsoon flow to the east over Mali developed. MCS activity increased as the maritime flow weakened after 8 June and substantial cold pool events were observed in the monsoon flow over southern Mauritania on this day (see ITD “bulge” feature D in Fig. 3d) and over southern Mali on 12 and 14 June (not shown).

Fennec 2012 flights targeted specific features of the evolving Saharan atmosphere, including surveys of the maritime flow in the early period (b699/700), aged dust from MCS cold pools to the south of the flight domain sampled over the ocean (b702–3) and southern Mauritania (b704), boundary layer heat fluxes close to edge of the SHL (b705), the SHL tongue and low-level jet (LLJ) dust emission (b706–8) and dust uplift and radiative processes (b708–9).

During the Fennec Pilot campaign in April 2011, seven flights were performed (Table , Fig. 2a). b589 was an initial shakedown flight to test operational logistics, and was conducted at high altitude only, but overflew a dust front which was observed with the lidar and dropsondes. b590 (morning) and b591 (afternoon) were the first flights performing in situ measurements, and sampled maritime inflow over Mauritania, which was overlaid by dust layers at higher altitudes. b592 took place 2 days later on 7 April (note b592 was actually two separate flights, one in the morning and one in the afternoon) and sampled the diurnal evolution of the recovering SABL (Saharan boundary layer) following the retreat of marine air. b593 continued the sampling of the recovering SABL, but over a different surface albedo. b594 was a science transit return of the BAe146 to the UK, sampling dust transported northwards by a low pressure system over Morocco.

June 2011 was the main flying period of Fennec, when both the Falcon and the BAe146 conducted missions over the Sahara. Eleven flights were performed with the F20 during the period 2–16 June (Fig. 2b, Table ). The first four flights (F09–F12) were designed to sample the dust outflow from the continent, over the coastal Atlantic, though almost no dust was sampled during F10. The subsequent seven flights were conducted over the continent, with two flights (F13 and F18) dedicated to the study of the morning dust uplift over alluvial sources of northern Mauritania in connection with the decay of the low-level jet. The flights were part of the RAIN4DUST project funded by the EUropean Facility for Airborne Research, EUFAR (Schepanski et al., 2013), designed to examine alluvial deposits as a dust source. Four flights were conducted along the exact same track (F14, F16, F17 and F19) to document evolution of the thermodynamics, the dynamics and the composition of the SABL over north central Mauritania in response to an approaching Saharan heat low (SHL), which was migrating westward during that period (see Sect. ). Flight F15 was conducted to document the SABL over northern Mauritania together with a dust plume transported from Algeria and associated with a Mediterranean wind surge.

The first three flights performed by the BAe146 on 17 and 18 June were a set of missions designed to investigate very strong low-level winds over northern Mali (b600, b601 and b602). During these flights, some of the largest particles encountered during Fennec were measured (see Sect. 4.1.1), and elevated dust concentrations were seen at altitudes beneath 1km, although vertical mixing played a role in the afternoon. The Falcon also flew on 17 June (F20) with a mission dedicated to the documentation of the SABL over northern Mauritania and northern Mali, west of an approaching African easterly wave, and of the structure of the dust plume associated with a Mediterranean wind surge.

Flight b603 was a calibration flight performed over the Canary Islands at high altitudes under clear skies for the radiation instruments. Flight b604 was a LADUNEX EUFAR flight sampling dust which had been uplifted more than 24 h previously by an MCS and associated haboob over Mali, and then transported over Mauritania by prevailing winds (Sodemann et al., 2015), retaining giant-mode dust particles despite large transport distances (Ryder et al., 2013a). The BAe146 crossed the dust front at high and low altitudes for in situ and remote sensing measurements. F21 consisted of a long rectilinear flight across northern Mauritania and northern Mali to survey the SABL as well as document the dust uplift in the region of the intertropical discontinuity (ITD, i.e. the near-surface convergence zone between the monsoon and the harmattan flow) to the south of the SHL, over Mali.

On 21 June both the Falcon and BAe146 performed two flights each (b605, b606, F22 and F23). On the preceding day, convection over the Atlas Mountains initiated a dust front which propagated southwards over Mauritania by 21 June, with aged dust overlying it. During the day the layers mixed together. Both aircraft missions aimed to sample this dust and diurnal mixing (see Sect. ). On 22 June, again, both aircraft missions were in the morning and afternoon (b607, b608, F24 and F25). The missions were aimed at sampling the SHL and therefore flight tracks extended well into Mali (Fig. 2). Lidar, dropsondes and radiation instrumentation were used to sample the spatial and diurnal evolution of the SHL (see Engelstaedter et al., 2015). On 23 June, F26 performed a mission dedicated to the study of the morning dust uplift over alluvial sources of northern Mauritania in connection with the decay of the low-level jet (RAIN4DUST project).

From 24 June onwards, dust conditions were generally more well-mixed vertically with less fresh dust being sampled. Flight b609 on 24 June sampled dust and cumulus developing on the top of the dust layers (see Sect. ). Flight b610 sampled the low-level jet and dust uplift mechanisms over eastern Mauritania. b611 flew over the Zouerate ground supersite – see Sect.  for a comparison of in situ measurements to sunphotometer retrievals. The purpose of the b612 and b613 missions on 26 June were to achieve radiative closure and measure heat fluxes over the desert. Both were performed under clear sky conditions with a series of stacked runs, under low dust loadings. Flight b614 sampled dust uplift and the low-level jet early in the morning. Flight b615 on 28 June was the return transit to the UK, and included radiation calibration manoeuvres.

Since the initial flying period during Fennec 2012 was initially dominated by Atlantic inflow, with dust being observed at the confluence of this and Saharan air (see Sect. ), most of the earlier flights aimed to sample this boundary (Table ). b698 was a science transit from the UK to Fuerteventura, during which calibration manoeuvres for radiation instruments were performed. b699 and b700 were a pair of flights on 6 and 8 June which sampled the gradient of Atlantic inflow and its eastern boundary at high and low levels over northern Mali and northern Mauritania. b701 and b702 were similar flights, but here the edge of the Atlantic inflow was contingent with the ITD, and larger dust loadings were sampled over central and southern Mauritania. Following b702, the BAe146 landed at Dakar, and then returned to Fuerteventura over the Atlantic (b703) sampling continental dust outflow. Flight b704 sampled Atlantic inflow and the ITD again, this time measuring the highest submicron aerosol optical depths (AODs) of Fennec, 3.4 at 550 nm, over southern Mauritania. b705 on 12 June was performed around midday to measure Saharan heat fluxes over a stable pressure gradient.

The b706 and b707 pair of flights examined dust uplift over the Mauritania–Mali border, with exactly the same track, and uplift began to happen under stronger winds during b707. b708 was designed to measure dust uplift by the LLJ over Mali under clear sky conditions so that the radiative impact of the dust could also be measured. This flight saw the highest scattering measurements on the nephelometer during the campaign (see Sect. ), from dust at very low altitudes. By contrast, b709 on 17 June sampled dust which had been transported into the SHL and was well-mixed vertically up to 6 km. This flight aimed to sample the pressure structure of the SHL and also perform radiative closure. b710 flew over the Zouerate ground supersite as part of SAVEX in order to compare AERONET (AERosol Robotic NETwork) retrievals and aircraft measurements of dust. Finally, b711 was a science transit return to the UK.

During Fennec 2011 six different instruments were used to measure size distribution, as described in Sect. 2, covering the size range of 0.1 to 300 µm diameter. Of these, the PCASP, CDP and CIP operated consistently during the whole campaign (see Rosenberg et al., 2012, for details of calibration and errors). Very large particles were measured during Fennec 2011, with the effective diameter of the full size distribution ranging from 2.3 to 19.4 µm (Ryder et al., 2013b). Examples of different types of size distribution are shown in Fig. 4. The solid lines show measurements from flight b600 at around 700 m above ground level, under aerosol optical depths greater than 3.0 at 550 nm when the dust was being actively uplifted by strong winds and was encountered beneath 1 km above ground level. These were some of the largest particles encountered during Fennec 2011, and the size distribution shows a strong coarse and giant mode present with a broad peak in volume concentration from around 10 to 60 µm. Contrastingly, b612 (dashed lines) shows more aged dust (24–48 h based on satellite imagery) which was well-mixed within the SABL up to 5 km, with optical depths at 550 nm of around 0.6. Here there are fewer particles across all sizes upwards of 0.5 µm compared to b600, and the peak volume concentration is now at 10 µm diameter, reflecting a shift to lower number concentrations and fewer coarse particles as dust is mixed vertically through the entire SABL, and larger particles are deposited during transport as well as dispersion decreasing the total number concentration. Ryder et al. (2013a) examined the effects of vertical mixing and transport on dust properties further. Interestingly, at sizes smaller than 0.5 µm there are more particles in the case of b612, which gives the size distribution a flatter shape than b600. This may be due to different dust sources, soil types or uplift wind speeds acting.

For the first time on the FAAM BAe-146, all size-resolved particle measurements were made with high temporal resolution (≥ 10 Hz) allowing their correlation with the vertical wind speed, and therefore permitting measurements of the eddy covariance particle flux. This technique has been previously employed to derive heat, momentum and moisture fluxes from FAAM BAe-146 data (Petersen and Renfrew, 2009). During Fennec we were able to resolve particle flux both in terms of eddy length scales and particle diameter. During flights b600, b601 and b602, upward particle fluxes were observed associated with synoptic-scale winds in Algeria and northern Mali. Upward particle fluxes were also observed during flight b604 again associated with synoptic-scale winds in this area. In general it has been found that particles above 10 µm diameter dominate the mass flux and in some cases particles above 100 µm diameter make a significant contribution. Full details are provided in Rosenberg et al. (2014).

To date, information on the mineralogical composition of coarse mineral dust can only be obtained by post-field analysis of filter samples. Mineralogical composition is a fundamental property for determining the impacts of mineral dust on climate. Mineralogical composition controls the complex refractive index, determining the radiation interactions in the shortwave and longwave spectrum (relevant to the direct radiative effect); the water uptake capability, determining the cloud and ice nuclei activation efficiency (relevant to the indirect radiative effect); the solubility in water, controlling the capability of deposited mineral dust to be assimilated by the marine phytoplankton; and the surface reactivity relevant to interactions with the gas phase (Formenti et al., 2011a; Scheuvens et al., 2013).

The mineralogical composition of mineral dust is obtained by X-ray diffraction (XRD) (Caquineau et al., 2002). Nonetheless, this technique is not always applicable to aircraft samples because of limited sampling times yielding light loadings which are incompatible with the detection limits of this analytical technique. Typically, about 800 µg of total dust mass are needed for analysis (Caquineau et al., 1997). For this order of magnitude, at least 1 h sampling at high volume is required (∼ 50 L min-1) for low to moderate atmospheric concentrations (< 200 µg m-3) and at least half an hour for concentrations of the order of 200 µg m-3 and above.

Alternatively, useful indications on the mineralogical composition of mineral dust can be obtained by examining the concentrations of typical trace elements such as Al, Si, Fe, Ti, Ca, K, Mg and Na, which can be obtained by X-ray fluorescence techniques which have typical detection limits of 10 µg or less across a filter sample (Formenti et al., 2011b). In particular, the inter-elemental ratios provide indications of the origin of mineral dust. Typically, Al is used as a unique tracer as aluminosilicates dominate the dust mass. However, the Fe / Ca ratio has also proven useful for tracing the origin of the dust plumes (Kandler et al., 2007; Formenti et al., 2011a, 2014; Scheuvens et al., 2013).

Ninety-three samples are available in total from the Fennec 2011 and 2012 campaigns from the BAe146 (55 and 38 for each field phase, respectively). Samples were collected in the Saharan boundary layer at altitudes ranging between 350 and 2700 m a.s.l. The total dust concentrations, estimated as the sum of oxides of Na, Mg, Al, Si, K, Ca, P, Fe, and Ti, varied between 22 and 4012 µg m-3.

The analysis of dust uplift potential (DUP, Sect. ) restricted to the filter sampling legs suggests that the Fennec 2011 was characterized exclusively by emissions from Saharan sources in Algeria, Western Sahara and Mauritania, with the exception of samples from b604 where dust had been uplifted by MCS outflow over Mali and transported by a large-scale haboob (Sodemann et al., 2015). However, during the Fennec 2012 period, additional emissions of Sahelian dust from convective activity in Mali constituted a much larger proportion of the samples. This contrast is a result of the dominant heat low west phase during the latter half of Fennec 2011 driving anomalous northeasterlies over western Algeria (Fig. 3c, Sect. ) compared to a northern extension of the monsoon flow over Mali during Fennec 2012 with increased MCS activity (Sect. ).

The elemental composition is consistent with the DUPs indications for those source regions. This is shown in Fig. 5a, where the Fe / Ca and the Si / Al ratios obtained for the Fennec 2011 and Fennec 2012 samples are compared to those measured during the AMMA, DABEX, DODO and GERBILS campaigns summarised in Formenti et al. (2014). As a consequence, and with the exception of samples collected during flights b699 and b700 when dust originated from the sources in the Algeria, Western Sahara and Mauritania areas, samples collected during Fennec 2012 had a lower Ca and Mg percent content with respect to Fennec 2011, reflecting the absence of calcium carbonates (calcite and dolomite) in Sahelian soils (Journet et al., 2014).

Likewise, there is a clear difference between the measured single scattering albedo (SSA) at 550 nm during Fennec 2011 and Fennec 2012 (Fig. 5b). Even when excluding the outlier corresponding to a pollution plume encountered during flight b710 at Zouerate during Fennec 2012, when the single scattering albedo value averaged over the filter collection run was 0.91 (±0.02), the mean single scattering albedo value for the Fennec 2012 period is lower than that for Fennec 2011 (0.94 ± 0.01 and 0.97 ± 0.01, respectively).

Future work will investigate the possible link between the changes in composition and optical properties during the 2011 and 2012 periods. This will also involve taking into account the particle size distribution as a function of origin and of the age of the sampled air masses.

It is possible to use in situ measurements of scattering and absorption by the nephelometer and PSAP on the BAe146, respectively, to calculate extinction profiles and hence AOD. Measurements are restricted firstly by the altitudes flown by the aircraft, which are usually between above the aerosol layer and as close to the surface as is safe and permissible. Depending on visibility, this varied between around 50 m and 1 km during Fennec. Secondly, the measurements are restricted by the aircraft inlets, which do not sample particles larger than around 2 µm (Ryder et al., 2013b). The former has been accounted for by assuming that the aerosol profile is constant beneath the minimum aircraft altitude to the ground, while the latter is not accounted for and therefore the AODs presented here represent only extinction from the submicron size distribution, and are therefore an underestimate. Scattering and absorption measurements are corrected as described in Ryder et al. (2013b).

AODs from Fennec 2011 and 2012 are shown in Fig. 6, with circles representing 2011 and diamonds 2012. AODs ranged from 0.2 to 3.6 at 550 nm. Of particular interest were a few heavy dust events which the aircraft sampled, including b600, b601 and b602 on 17 and 18 June 2011 in northern Mali (orange, red and green circles), during which very large dust particles were measured and dust fluxes have been calculated (as described in Sect. ). Secondly, flights b707 (blue and green diamonds on Mali–Mauritania border) and b708 (orange diamonds in northern Mali) in 2012 sampled very high dust loadings, the first with very low-altitude fresh dust (see Sect. ), and the second with well-mixed dust to above 5 km, both under clear skies (i.e. no cloud). These flights will make excellent radiation closure case studies. Thirdly, we draw the reader's attention to the large number of profiles over the ocean between the land and Fuerteventura. The vertically resolved changes in particle size and optical properties between fresh, aged and oceanic profiles are examined by Ryder et al. (2013a), who find a significant reduction in particle size, number and associated changes in optical properties for dust measured over the ocean.

Saharan clouds have the potential to be significantly different than other continental mid-latitude clouds due to the abundance of dust, which can act as ice nuclei (IN) and giant cloud condensation nuclei (GCCN), and the fact that the hot dry boundary layer prevents precipitation from reaching the surface. Flight b609 on 24 June 2011 investigated a convective system in northern Mauritania. According to analyses from the Met Office operational Africa Limited Area Model, an overnight monsoon surge associated with an easterly wave brought moist southerlies as far as 24∘ N at 8∘ W. Over the course of the day, a linear convective feature formed, extending from 18.5∘ N to link with a system over the Atlas Mountains at 30∘ N. Dusty cold pool outflows, which affected supersite 2 (BBM), were visible in SEVIRI (Spinning Enhanced Visible Infra-Red Imager) satellite imagery from at least 18:00 to 23:00 UTC. Flight b609 consisted of an overflight of the system and a series of north–south aligned legs at 8.0∘ W between 23.8 and 25.8∘ N on the eastern flank of the convective system from 12:42 to 15:36 UTC. The run locations were restricted by operational constraints.

Figure 7 shows the flight pattern and measurements. The flight path (thick black lines) consisted of an initial high-level leg, followed by a descent to minimum altitude and then three legs beneath the clouds, each increasing in altitude to 4500m, just below the cloud base (5400 to 5800 m). Once above the cloud base, a series of short legs were performed targeting three cloud cells, with the aircraft finally ascending through the cloud tops at 8000 m. Cloud droplet concentration is shown in Fig. 7 on top of the aircraft track, appearing red when the aircraft was in clouds. Range-corrected Leosphere lidar backscatter signal is shown measured during the highest-altitude aircraft leg and also beneath the aircraft descent where available, since the signal is strongly attenuated by the clouds along the high-level leg. Here we solely use the lidar measurements to describe the presence and structure of clouds present, not the vertical distribution of dust, due to the strong attenuation of the lidar signal by the clouds.

The initial lidar observations indicated that cloud tops ranged from 6.1 km to above the aircraft altitude of 8.75 km, equivalent to approximately -11 to -28 ∘C (based on the profile measured during the descent). It was observed visually from the aircraft cockpit that the cloud tops had no observable anvil cirrus outflow. During the descent to low levels, the aircraft passed through one isolated cloud at 24.18∘ N. Lidar observations of this cell 13 min prior to the intersection provided a cloud top height of 6.65 km, which is estimated to be at -15.0 ± 0.2 ∘C. The lidar data showed no links to, or particle flow between, any other clouds. Particle images recorded by the CIP showed that this cloud consisted of pristine hexagonal plates. Freezing at this warm temperature is uncommon even for clouds in the vicinity of a source of IN (Kanitz et al., 2011; Ansmann et al., 2008; Sassen et al., 2003; Raymond and Blyth, 1989). An explanation could be the very high dust concentrations acting as IN in the heart of the Sahara.

The descent to 500 m provided a measurement of the aerosol input into the cloud. At the surface particle concentrations above 0.13 µm diameter measured by the PCASP and CDP ranged between 60 and 80 cm-3 south of 25.33∘ N. North of this point the concentrations were 200 cm-3. Note that most of this concentration is measured by the PCASP and therefore does not show up on the number concentration scale in Fig. 7. As the aircraft climbed to the cloud base, the aerosol concentration fell to 40 cm-3, although the number of particles above 4 µm diameter rose from 0.05 to 0.15 cm-3.

During ascent back towards the cloud base, sporadic ice precipitation was observed by the CIP probe from altitudes of 4.4 km (4 ∘C) and graupel was observed impacting the aircraft. Clouds were encountered at 5.75 km (-8 ∘C) although the cloud base could have been slightly lower (minimum of 5.4 km or -5 ∘C). It is of note that cloud base may have been too cold for the Hallett–Mossop ice multiplication process which occurs around -6 ∘C. No columnar ice crystals typically produced by this process were observed. Near the cloud base, the cloud was found to be in mixed phase with the droplet number peaking at 250 cm-3 coincident with the peak updraught speed of 10 m s-1. This measured droplet concentration was found to be significantly higher than the aerosol concentration reported by the PCASP and CDP below the cloud base: the shortfall in CCN must have been made up of particles smaller than the PCASP detection limit. Twohy et al. (2009) showed that dust with zero hygroscopicity, κ, is entirely activated in cloud by a 10 m s-1 updraught and, because of its large size, it can form the majority of the CCN population over other hygroscopic particles when they have a small but non-zero κ (Koehler et al., 2009). It is therefore likely that dust particles were acting as CCN or GCCN in this case. Higher in the cloud, there is evidence of liquid water in updraught regions, and near the cloud top a population of homogeneously nucleated bullet rosettes were observed. No cirrus or precipitating particles were observed above the cloud top.

These measurements have shown that dust is likely acting as a CCN and as an IN at temperatures of -15 ∘C. Sampling of clouds earlier in their evolution would provide further limits on the effectiveness of dust as an IN. The lack of the Hallett–Mossop process in these clouds makes them a useful case for assessing IN concentrations and the extreme size of the dust particles may provide tests on the impact of GCCN.

For a non-precipitating cloud we expect that equivalent potential temperature, θe, and total water concentration (condensed plus vapour) are conserved and hence any point in cloud should lie on a mixing line or in a mixing region of these parameters (Paluch, 1979; Blyth et al., 1988). Here the cloud is precipitating meaning that total water concentration is no longer conserved but these variables are still useful in diagnosing the transport and mixing processes (Fig. 8).

Much of the sampled in-cloud air had higher water content, greater than 5 g kg-1, but similarly high ranges of θe compared to boundary layer air (Fig. 8). These are inconsistent with the clouds being a simple mixture of boundary-layer and entrained air. Out of cloud and above-cloud-base air had some regions consistent with simple mixing, some in a similar moist warm region to the in-cloud air, but also some regions with low moisture content less than 1 g kg-1 and high θe, similar to boundary layer air. Profiles of water vapour mixing ratio (WVMR) (Fig. 7) show that in the boundary layer (below 2500–5000 m, varying from profile to profile) WVMR increases with altitude. Similar behaviour was also seen in the mean WVMR profile at Fennec supersite 1 (BBM) between 15:00 and 18:00 UTC, the time of maximum cloudiness (Marsham et al., 2013b). This is again inconsistent with simple mixing of a growing boundary layer. We hypothesise that in this low shear environment, precipitation is evaporating in the boundary layer air but is not able to arrest the updraught, allowing water to be recycled and concentrated in the cloud. High θe air rises in the boundary layer and receives extra water from evaporating precipitation, such that when it enters the cloud base it has more moisture than its environment. In cloud air parcels either precipitate adding to the recycled moisture reservoir before being detrained as dry, high θe air, or they do not precipitate and are instead detrained as moist, high θe air. We also expect dust and aerosol to be affected by this recycling process. Precipitation accumulates CCN and upon total evaporation releases them as a single aggregate particle. The increase in large dust particle concentration below the cloud base is qualitatively consistent with this expectation. This concentrating of moisture and dust in the boundary layer top and the modification of the dust size distribution has implications for long-range transport of these atmospheric constituents. To our knowledge, these are the first observations of such a mechanism increasing the moisture content within the SABL mid-levels.

Heterogeneous uptake of photochemical species leads to changes in the gas-phase composition of the atmosphere, affecting the global ozone budget (Bauer et al., 2004). Previous campaigns have observed ozone depletion during high dust loadings (de Reus et al., 2000, 2005). These have also been investigated through modelling (Bian and Zender, 2003) and laboratory studies (Chang et al., 2005; Hanisch and Crowley, 2003). There is still some debate as to whether the removal of ozone is due to heterogeneous chemistry on the surface of the dust or to a feature associated with a change in air mass between high and low dust loadings. The alkalinity of mineral dust has been shown to enhance the uptake of gases on the surface (Grassian, 2002). Bauer et al. (2004) propose that the coarse-mode of mineral dust could be important for heterogeneous uptake, whilst Chang et al. (2005) found that there was no mass accommodation limitation to the rate of ozone uptake coefficients, concluding that freshly emitted Saharan dust is potentially a significant route of ozone loss. Hanisch and Crowley (2003) discussed that mineral dust surface sites could be deactivated by the extended presence of ozone. Ultimately, the change in the surface of mineral dust may have repercussions for subsequent aerosol–cloud interactions and modify the cloud nucleating properties of the mineral dust. A number of cases observed during the Fennec campaigns were investigated Brooke (2014).

Fennec flight profiles provided the opportunity to sample very recently lofted mineral dust which would not have undergone significant atmospheric “processing” and thus provided a good opportunity to investigate heterogeneous dust/ozone interactions. These observations of decreased ozone concentrations correspond with increased mineral dust surface area associated with elevated dust concentrations. Figure 9 presents box and whisker diagrams of mineral dust mean surface area correlated with ozone mass mixing ratios observed during b707, where dust uplift was encountered at the far eastern section of the flight track in northern Mali (orange line in Fig. 2d). The red central line of the box and whisker denotes the median, the edges of the box are the 25th and 75th percentiles and the whiskers extend to the most extreme data points. Mean surface areas of 0.15 to 0.35 µm2cm-3 (roughly count median diameters of 0.22 to 0.33 µm) correspond to ozone mass mixing ratio of 49–52 ppb. As the mean dust surface area increases to 0.45 to 0.75 µm2cm-3 (count median diameters of 0.38 to 0.49), the ozone mass mixing ratio decreases to 41–44 ppb. The spread in ozone concentrations at mean surface areas of 0.45 µm2cm-3 is associated with crossing into a harmattan airflow.

These in situ observations suggest that increased mineral dust surface area associated with fresh dust uplift and a large coarse-mode contribution to the size distribution act as a route for the reduced ozone concentrations. However, from the analysis presented here, it is not possible to unequivocally conclude if the air mass initially contained lower ozone concentrations and that the mineral dust was subsequently uplifted, or if mineral dust uplift could have contributed to the reduced ozone concentrations observed. There is scope within the Fennec data set to further investigate air mass source regions, potentially with Lagrangian study methods.

Aircraft data can play an important role in validating satellite-based retrievals of AOD, covering a more extensive spatial area than that which is viewed from fixed ground-based measurements. Particularly useful in this regard are active remote sensing observations from lidar, since they can sample the full depth of the atmosphere below the aircraft instantaneously (i.e. a physical vertical profile by the aircraft is not required) and can provide vertically resolved information.

In Fig. 10, middle panel, we show an example of the level of agreement seen between three different co-located measures of AOD, one provided at 532 nm by the LNG lidar on the F20, one from MODIS Aqua, derived using the Deep Blue algorithm collection 5.1 (Hsu et al., 2004) and one from the SEVIRI instrument on Meteosat-9 (Brindley and Russell, 2009; Banks and Brindley, 2013), all at a wavelength of 550 nm. Here we focus on an afternoon flight (F23, see Fig. 2c) made by the Falcon on a track leading across to northern Mali from northern Mauritania on the 21 June 2011.

The satellite observations are co-located spatially with the lidar by averaging the satellite pixels within 25 km of each lidar pixel. Temporally, the Aqua satellite overpass time is always within 90 min of the aircraft observations, with a minimum time difference of 37 min. For SEVIRI, we take advantage of the improved temporal sampling available from geostationary orbit such that each lidar observation is within 30 min of the corresponding satellite retrieval. The lower panel in the figure shows the vertical extinction coefficient derived from the lidar observations, while the top coloured band illustrates the colouring of the standard “desert-dust” red–green–blue (RGB) composite (Lensky and Rosenfeld, 2008) extracted from SEVIRI along the flight track.

Looking at the middle panel, the longitudinal behaviour of the AOD derived from all three instruments is generally in good agreement although SEVIRI tends to show consistently higher AODs than those derived from the lidar and from MODIS. The MODIS retrievals contain more data gaps as a result of various data quality tests: both the lidar and SEVIRI retrievals and the RGB composites suggest that these tests may be slightly too severe as there is no clear evidence of a break in the aerosol layer or the presence of cloud. The intense pink colour of the composite at the western edge of the track would suggest the largest dust loadings are located here, associated with a thick dust plume at an altitude of ∼ 3 km and another distinct layer observable at ∼ 5.5 km seen in the lidar profile (which may have originated from Mali on 19 June). By the eastern end of the track, the AODs measured by MODIS, SEVIRI and the lidar are slightly smaller than the values seen at the western end, the dust is much more uniformly spread throughout the lowest ca. 5 km of the atmosphere, and the intensity of the RGB signal is somewhat reduced.

Further work has explored co-located aircraft and satellite data in more detail, utilising a more extensive suite of satellite instruments (such as the Multi-angle Imaging SpectroRadiometer (MISR) instrument on Terra and the IASI instrument on the METOP satellites (Banks et al., 2013), and between the BAe146 in situ measurements and spaceborne lidar CALIOP (Pappas et al., 2015). In the former study, the differences between retrievals were investigated, including an evaluation of the sensitivity of the retrievals to variations in dust loading, to atmospheric conditions (such as column water vapour), to surface features (such as albedo) and to aerosol height. As diagnosed by Banks et al. (2013), when the dust loadings are high, the SEVIRI retrievals appear most capable of retrieving the appropriate AODs, whereas the other retrievals are biased low. On the other hand, the SEVIRI retrievals are most sensitive to meteorological conditions, especially column moisture, under high levels of which the SEVIRI-retrieved AODs are biased high; conversely, the MODIS Deep Blue and MISR aerosol retrievals appear to be relatively insensitive to such factors. The aircraft data will be of substantial benefit to interpreting the “desert-dust” RGB imagery.

Considering the wide application of size distributions from AERONET retrievals such as for aerosol models and climate forcing assessments (e.g. García et al., 2012; Kinne et al., 2003), it is important to validate AERONET retrievals where possible with field observations. Moreover, some discrepancies have been found between retrieved size distributions using the AERONET algorithm (Dubovik and King, 2000; Dubovik et al., 2006) and the same size distributions derived with the Sky Radiation (SKYRAD) algorithm (Nakajima et al., 1996), as described in Campanelli et al. (2012) and Estellés et al. (2012b). The SAVEX project aims to explore these discrepancies, and its creation was motivated by studies such as Estellés et al. (2012a, b), where differences between different sunphotometer retrieval algorithms are examined.

AERONET CIMEL sunphotometers were installed and operated at the two supersites of Zouerate (western Mauritania) and Bordj Badji Mokhtar (BBM, Algeria) as part of the Fennec programme. As part of the SAVEX project, sunphotometers were also installed and operated at several different sites on Tenerife during June 2012 with the intention of overflying the instruments during dust events. However, overflights were not performed at Tenerife due a lack of dust outflow in this location during the campaign. The aircraft range from Fuerteventura did not permit overflights at BBM. Therefore, overflights as close as possible to the Zouerate station under dusty conditions were performed during 2011 (b611, 25 June) and 2012 (b710, 18 June, SAVEX flight).

During these flights, profiles and stacked legs were performed to measure in situ aerosol properties and radiative measurements to allow for radiative closure of the column above the ground site. Radiative flux measurements were also made at the ground site. Here we present some measurements from b611 in 2011. Dust sampled during this flight was around 19 to 43 h old, originating from Algeria (Ryder et al., 2013b), with AERONET AODs at 440 nm from 0.8 to 0.94, and was relatively well mixed in the SABL up to around 5.5 km, although extinction coefficient measurements from the aircraft approximately doubled beneath 2.5 km. Similar measurements are available from flight b710, although for that flight, layers of anthropogenic pollution were detected between dust layers, thus making comparisons between platforms more complicated, and are not shown here.

Figure 11 shows a comparison of the size distributions measured by the BAe146 compared to AERONET retrievals on 25 June 2011. The in situ aircraft measurements were taken over a vertical profile close to Zouerate on 25 June 2011 between 8 km and 80 m a.g.l. from 15:58 to 16:27 UTC. In situ size distribution measurements shown in Fig. 11 are therefore shown as the median, and the 10th and 90th percentiles between 80 m and 5.5 km.

Sunphotometer retrievals of size distribution from almucantar scans are not present during much of the day due to cloud cover over Zouerate. Nevertheless, several retrievals are available during the morning (dark blue), one during the flight (black) and two from 18:06 and 18:30 after the aircraft had left the region (light blue). Size distribution retrievals shown are those directly available from AERONET (L1.5, V2) and converted to dV/dlog(D) to match the aircraft measurements, and adjusted to measurements in cm-3 assuming the dust layer is distributed evenly above 5.5 km. Further work will examine measurements from aircraft legs at different altitudes, and different ways of representing a column-average measurement from the aircraft measurements, such as extinction-weighted averaging.

The median aircraft measurements show a peak volume concentration at 12 µm, while the AERONET retrievals show peaks between 3 and 6 µm. This is consistent with previous aircraft–AERONET comparisons finding larger particles measured by aircraft (Reid et al., 2003; Müller et al., 2012, 2010b; McConnell et al., 2008). However, only one retrieval shows a peak volume concentration at 13 µm which appears to agree much more closely with the shape of the size distribution from the aircraft measurements. Satellite images show a small convective cloud developing close to, but not over, Zouerate around this time. It is possible that small-scale downdraughts produced some freshly uplifted dust which may have resulted in different size distribution retrievals. However, we cannot reject the possibility that optically thin cirrus cloud contamination affected the quality of this inversion (although it is not visible in satellite imagery), which would bias the size distribution towards larger sizes, and we note that the retrieval error is around double for this particular time compared to the others shown. At sizes smaller than 3 µm, differences in volume concentration are substantial between AERONET and the aircraft, with AERONET reporting more particles. Further work will explore possible factors causing this difference.

Rather few coarse particles were seen during b611 (towards the end of the campaign) relative to the rest of Fennec, perhaps due to the aged nature of the dust which meant that the largest particles had already been deposited. This is reflected by the absence of particles larger than 16 µm in the median, and the absence of particles larger than 45 µm in the 90th percentile (see the one CIP data point for the 90th percentile), though particles of these sizes were measured, but the standard deviation was very large, as shown by the large error bars on the median above sizes of 16 µm.

Existing publications show contrasting examples of agreement and disagreement between airborne, ground-based and AERONET size distributions, and there has been much debate over the causes. Reid et al. (2003) provide an overview of many commonly used sizing techniques and their limitations. These authors find that aerodynamic measurement methods and sunphotometer inversions tend to produce mass median diameters (MMDs) of around 3 µm diameter, while published OPC measurements at that time produced volume median diameters (VMDs) of the order of 8–13 µm. Limitations of OPCs, as described in Sect. , are principally uncertainties due to assumptions of refractive index, particle shape and the Mie response curve, the latter leading to a sizing ambiguity in the range of 5–10 µm. As outlined in Sect. , we consider uncertainty due to particle shape to be small (as evidenced by similar size distributions resulting from OPCs measuring over different scattering angles), and we provide error bars to account for the remaining uncertainties. Aerodynamic measurement systems, such as aerodynamic particle sizers (APS) and cascade impactors rely on the particle dynamic shape factor, which varies with dust particle shape, causing uncertainties in the size distribution and may undersize particles by 25 % for dynamic shape factors of 1.2. These instruments are also impacted by cut-offs of larger particle sizes imposed by inlets. Open path OPC instruments such as the CDP and CIP do not suffer from inlet effects, but do have uncertainties in their measurement volume. Cascade impactors can also be affected by particles bouncing off substrates. Thus, each measurement technique has its own advantages and disadvantages. Reid et al. (2003) found for dust aerosol at Puerto Rico that the AERONET and APS size distributions agreed well with MMD at around 3.5 µm, while OPC size distributions produced a VMD of 9 µm. Reid et al. (2003) conclude that OPC data are most likely to have the largest biases based principally on the response function and uncertainty/variability in particle refractive index and shape. However, no attempt was made by Reid et al. (2003) to determine uncertainties in the size distributions due to the response function or the uncertain refractive index. This error analysis has been rigorously performed here, and is represented in the error bars in Figs. 4 and 11. Additionally, data from the CIP, which uses light-shadowing techniques rather than light-scattering techniques as with the OPCs, further increase confidence in the shape of the Fennec size distribution presented by the CDP data.

Reid et al. (2006) compared aircraft OPC measurements to surface-based APS observations in an environment dominated by sea-salt aerosol, and drew similar conclusions to Reid et al. (2003). AERONET size distribution inversions were also found to compare favourably to APS surface measurements. However, in this sea-salt environment, aerosols are not likely to reach such large sizes as was observed during Fennec – the precise size range which poses challenges for sunphotometer retrievals, and additionally the inlet to the APS had a cut point of around 12 µm in this case. Finally, Reid et al. (2008) report observations of dust from the United Arab Emirates in 2004. The authors found that AERONET and APS size distributions agreed well, although here the inlet cut point to the APS was around 10 µm, thus excluding measurements in the exact size range challenging to AERONET retrievals.

Contrastingly, other previous work (Müller et al., 2012, 2010b; McConnell et al., 2008) has found relative disagreement between aircraft and AERONET size distribution retrievals for dust, finding that AERONET retrievals significantly undersize dust. In some of these cases detailed radiative closure has been achieved, validating OPC observations when a reasonable coarse mode was sampled, both for dust (Osborne et al., 2011; Müller et al., 2010b, a) and for volcanic ash (Turnbull et al., 2012; Newman et al., 2012). Thus, despite the contrasting conclusions concerning AERONET size distributions, it is important to repeat these closure flights in dusty environments.

Additionally, AERONET size distribution retrievals are subject to their own set of limitations and associated errors. Firstly, the maximum diameter extends only to 30 µm and the tails of the size distributions are constrained to very small values (Hashimoto et al., 2012), and encounter large errors (Dubovik and King, 2000) which are dependent on the particle size. As noted in Estellés et al. (2012b), for the diameter interval 0.2 to 14 µm, the retrieval errors do not exceed 10 % in the maxima but could increase up to 35 % in the minima. Outside this intermediate range the errors increase, rising up to 80–100 % or higher for diameters less than 0.2 µm and greater than 14 µm (Dubovik et al., 2002).

Unfortunately the flights during Fennec when large particles were strongly evident did not take place close to AERONET sites, due to the remoteness of the flight locations. Ryder et al. (2013b) find that particle sizes are larger close to dust sources in remote locations, and Ryder et al. (2013a) show that giant particles (d > 37.5 µm) are a feature of freshly uplifted dust events, and some long-range transported cases. This should be a caution for using AERONET retrievals as a basis for dust size distributions over the central Sahara, particularly since they only extend to 30 µm diameter and the tails of the size distributions are constrained to very small values (Hashimoto et al., 2012). Further studies will examine aircraft and sun-photometer data from both 25 June 2011 and 18 June 2012, in terms of in situ aircraft measurements, airborne and ground-based radiation measurements, and using both the AERONET and SKYRAD retrievals for the inversion of sun-photometer radiances.

It is relevant for several areas of dust measurement analysis to identify the sources of dust sampled during research flights (e.g. Sect. 4.1.2). Lagrangian backward trajectory calculations with the FLEXPART model (Stohl et al., 2005) have been initiated in “tropospheric curtains” run along the track of each research flight to investigate the sources of the dust sampled. For this a large number of virtual air parcels (1000) were released at 30 s intervals in a vertical column between the surface and a pressure of 200 hPa along the flight tracks. Each parcel was tracked for 3 days backward in time using ECMWF analysis winds at a 1 × 1∘ horizontal grid spacing. We utilise the metric of dust uplift potential (DUP), defined as fU3(1+Ut/U)(1-Ut2/U2), with f being the desert and bare soil fraction, the wind velocity U, and the threshold velocity Ut= 6.5 m s-1 (Marsham et al., 2011). Despite being a simplified representation of likely dust uplift (e.g. variations in soil moisture are neglected, and dust uplift may not be linear with threshold velocity (Kok et al., 2014), DUP is a useful indicator of where likely uplift occurred and is relatively easily computed. DUP was calculated along the 3-day back-trajectories for locations where the tracked air parcels were within the boundary layer. DUP values were gridded on a 0.25 × 0.25 ∘grid and integrated over time. The DUP thus calculated for the tropospheric column at the aircraft location characterises the air mass as measured by the onboard lidars when the BAe146 and Falcon were flying at high altitudes. During lower flight legs this analysis allows for the interpretation of in situ dust measurements with respect to their mobilisation conditions and source regions.

Figure 12 shows the composite of the DUP from (a) all the Fennec 2011 Falcon flights, (b) Fennec 2011 BAe146 flights and (c) Fennec 2012 BAe146 flights. The areas contributing to the sampled air masses, which experienced strong winds that would be associated with dust uplift for dust source regions (i.e. high DUP areas), were mostly located in a NE–SW oriented swath extending from central Algeria to northern Mali and Mauritania during 2011. This dominant pattern is related to the inflow into the Saharan heat low, as shown by the 925 hPa winds in Fig. 3c over southwest Algeria. DUP locations from 2012 suggest more southerly dust sources, from southern Mauritania, stretching to the Mali–Algeria–Niger triple point and along the Mali–Algeria border towards southern Libya. This is consistent with additional convective activity in Mali driving emissions which were more Sahelian-dominated during 2012 (Sect. ).

Individual flights exhibit additional sources and substantial variability (see Supplement for DUP maps for individual flights). For example, dust from more southerly sources in Mali and Mauritania was intercepted during flights b600–602, b604–b606, b608, b611 and b614. Dust from northern Niger was sampled during flight b607. Note that the connection of dust filter samples to Fig. 12 is not immediate, because only the DUP for the selected legs corresponding to the filter sampling duration and position are considered in that case (see Sect. 4.1.2). We note that DUP from events associated with convective downdraughts such as haboobs may not be accurately represented due to the ECMWF analyses not fully capturing these events (Marsham et al., 2011). For example, this is the case for b604, where a large MCS generated a haboob over Mali, which subsequently travelled towards Mauritania (Sodemann et al., 2015). Therefore, in situations where it is possible for dust to be uplifted by events associated with convection, back trajectories and more generally operational meteorological analysis and forecast data should not be used in isolation to determine dust source regions. For example, a combined analysis of SEVIRI RGB satellite imagery and Lagrangian methods can be used to ensure consistency with observations (e.g. Ryder et al., 2013b).

One particularly notable flight was b600 during the morning of 17 June 2011, under which the highest dust loadings observed during Fennec 2011 and very large particles were measured. This was followed by flight b601 in the afternoon, and b602 the following morning in the same region. At this time, the SHL was centred on the Mali–Algeria–Niger triple point, producing strong low-level northeasterlies through Algeria to northern Mali, which were particularly pronounced on the morning of the 17 (b600, Fig. 13c, d, e). A region of lighter winds in Mauritania was associated with moisture remaining from the monsoon flow. Flights b600 to b602 were aimed at sampling these air masses, travelling out at high levels to descend into the strong winds in northern Mali and returning northwestwards at low levels into the moister air mass (Fig. 13a, b). In situ aircraft profile measurements are shown in Fig. 14.

Forecasts showed a pronounced decrease in the strong 925 hPa winds in northern Mali from 06:00 to 09:00 UTC, with a corresponding increase in 10 m winds, consistent with the downward mixing of momentum from the nocturnal LLJ around the SHL, likely deflected around the Hoggar Mountains (Birch et al., 2012). The existence of a LLJ is confirmed by the observation from the b600 descent into Mali (Fig. 14, black) of a wind maximum of 16.7 m s-1 at a pressure height of 1700 m (1400 m a.g.l.), located above the growing turbulent moist and dusty CBL found below 1400 m a.g.l.. The dust number and mass concentrations below 1400 m were the highest observed during the Fennec 2011 campaign with particularly large particles observed during b600 and b601; the size distribution during the initial part of the horizontal run in the dusty CBL following the profile descent of b600 can be seen in Fig. 4, with particles present up to nearly 300 µm. The high dust concentrations are consistent with the very high extinction measurements from the nephelometer and PSAP of over 1250 Mm-1 in both profile descents (Fig. 14). By the time of the profile descent of b601 at approximately 17:00 UTC, the dust had been mixed up into a CBL that reached 3.7 km (Fig. 14, red), with no remaining LLJ. The upwards vertical mixing of the dust resulted in the “pinkness” in the SEVIRI images (Fig. 13a, b) becoming more pronounced by the time of the second flight (the RGB product is sensitive to dust altitude; Brindley et al., 2012). Flight b601 then travelled back under the moist convection developing over Mauritania, with some precipitation observed falling onto the aircraft, but no extensive cold pool outflows at the aircraft altitude at this time.

To the authors' knowledge, this is the first airborne observation of dust size distributions (including the presence of coarse and giant particles) measured under uplift conditions caused by the breakdown of the Saharan nocturnal LLJ. Flights b706, b707 and b708 (Sect. ) from 2012 also collected in situ measurements of dust under LLJ breakdown conditions, thus providing scope for further analysis.

Recent studies have shown that haboobs (dust fronts occurring at the leading edge of cold pools emanating from convective storms) are a significant source of dust over the Sahara and Sahel (Flamant et al., 2007; Knippertz et al., 2007; Schepanski et al., 2009; Tulet et al., 2010). For example, Marsham et al. (2008b, 2013b) and Allen et al. (2013) show that haboobs cause around 50 % of dust uplift in the summertime Sahara, contributing to the seasonal cycle in dustiness. Radiosonde observations show that the transport of cold moist air in haboobs was a major cause of global model forecast bias at the Fennec BBM supersite in June 2011 (Garcia-Carreras et al., 2013), consistent with the role of haboobs diagnosed from convection-permitting simulations (Marsham et al., 2013a).

On 21 June 2011, aircraft measurements were taken over and through an aged haboob emanating from convection over the Atlas Mountains in Morocco (Kocha et al., 2013). The cold pool passed over dust sources and uplifted large quantities of dust. The haboob was observed over the central Sahara over northern Mauritania and northern Mali in the morning with the LNG lidar on the Falcon 20 during flight F22 (see Fig. 2b).

The haboob appears as the layer characterized by large extinction coefficient values at pressure heights beneath 1.5 km (Fig. 15a). The aerosol optical thickness (AOT) derived from the lidar extinction coefficient profiles reached an average of 1 around 09:00 UTC. At the same time, the BAe146 flew through the haboob to directly sample its characteristics during flight b605. In situ measurements from the BAe146 show that the dust concentration and observed extinction in the cold pool air increased by a factor of around three compared to its environment. The number of large particles of size around 10 µm increased to 0.1 cm-3 µm-1 (not shown). The properties of the dust sampled during this event also had a significant impact on the radiative fluxes within the haboob. For instance, the downward shortwave flux measured by the BAe146 decreased by 100 Wm-2 when entering the dusty cold pool (Fig. 15b).

In the afternoon, both aircraft sampled the growth of the SABL again (flights F23 and b606) as the haboob was mixed into the Saharan residual layer above. An unambiguous influence of the haboob composition and thermodynamics was observed on the development of the SABL (Kocha et al., 2013). Simulations with and without dust are being used to investigate role of the haboob on the dynamics/thermodynamics on the development of the SABL over the central Sahara.

Several flights were performed during Fennec to use in situ aircraft aerosol measurements and radiative measurements to potentially achieve radiative closure and examine the radiative properties of dust. Flight b708 on 16 June 2012 aimed to observe freshly uplifted dust at the time of downwards mixing of strong LLJ winds to the surface which was forecast to uplift dust over the Mali–Mauritania border. Additionally, since clouds were absent, the flight aimed to attain radiative closure measurements since the dust loadings were high but with very low-altitude dust, with AODs at 550nm of 0.54 and 1.92 measured during the two aircraft profiles by the nephelometer and the PSAP.

Figure 16b shows information from the aircraft profiles: extinction calculated from corrected scattering and absorption measurements is shown for the descent (black) and ascent (red) in Mali. During this flight, the aircraft flew a high-level leg at 7.5 km for radiative measurements, followed by a profile down to minimum safe altitude, which was around 100 m above ground level (a.g.l.) initially (see black line in Fig. 16a). During the descent the aircraft entered the dust layer at around 900 m. At this time the dust was not visible in the SEVIRI RGB desert dust imagery, despite an AOD of 0.54, likely because the RGB imagery is sensitive to dust altitude (Brindley et al., 2012). Absence of a “pink” signal in the SEVIRI RGB imagery during active dust uplift such as occurred during this flight would have major implications for dust source maps that have previously been created based on this imagery (e.g. Schepanski et al., 2007). Following the descent, the aircraft flew a low-level leg. Figure 16a shows the extinction as a function of longitude. As the aircraft flew eastwards, the amount of dust increased until visibility was so poor that the aircraft had to ascend to 400 m a.g.l. Despite this, extinction continued to increase to the east, with a maximum of 5500 Mm-1, the highest value ever observed from the FAAM nephelometer and PSAP.

At the end of the low-level leg, the aircraft ascended (red line in Fig. 16b). The dashed lines in Fig. 16b mark potential temperature and show inversions at the height of the rapid increases in dust extinction. This is one example of many during Fennec where the dust was encountered in a low layer, which was gradually mixed upwards during the day as the SABL grew. The red line in Fig. 16a shows the measured downwelling shortwave irradiance (SWD) during the low-level run. Note that during the legs (around 30 min), the solar zenith angle decreased in a way that SWD would be expected to increase with increasing longitude. Instead, during the western portion of the leg, SWD decreased with increasing extinction (dust above the aircraft). During the eastern portion of the leg there is a notable drop in SWD of around 150 Wm-2 at around -5.7 W at the same time as the peak in extinction. This flight, as well as b709 in the SHL where dust was well mixed vertically up to 5 km, will be further used to examine the radiative effect of dust over the Sahara under different dust conditions, using the spectral radiation instruments SHIMS, ARIES and SWS on the BAe146 in conjunction with radiative transfer models and satellite observations.

Combining lidar observations and dropsonde-derived atmospheric profiles allows for a detailed analysis of the spatial and vertical structure of the atmosphere as well as the boundary layer processes that control the emission, vertical mixing and transport of mineral dust. Flights b607 and b608 were part of an extensive survey of the troposphere in the SHL region with the aim (a) to characterise the spatial variability the SHL, CBL, monsoon inflow and dust distribution in the central Sahara, (b) to analyse how these features change throughout the day and (c) to assess the processes that control these features and dust dynamics. Both flights followed a straight track crossing from northern Mauritania into Mali in the morning of 22 June 2011 (see Fig. 2c for b607 flight track; the afternoon flight b608 overlies b607). The aircraft flew the track twice, once in the morning (b607) and once in the afternoon (b608), allowing the evolution of the atmosphere over time to be studied. Dropsonde measurements were obtained during the outward and return flight at fixed locations. Dropsonde data were interpolated to reference times at each location thereby creating a snapshot of the state of the atmosphere in the study region at the reference times. Engelstaedter et al. (2015) analysed the observed SHL characteristics and evaluated the performance of the UK Met Office limited area model for Africa (Africa-LAM). They identified two moisture transport pathways, one curving around the SHL core in the north (especially pronounced in a morning near-surface layer), and the other going towards the northeast within the roughly 2 km deep monsoon surge. The deep afternoon CBL simulated by the Africa-LAM in the monsoon surge region (more than twice as deep as observations) suggests a significant model error due to moisture being vertically mixed into northeasterly flow above about 2 km.

As an example of the combination of observations from different instruments, Fig. 17 shows Leosphere lidar and dropsonde-derived data from BAe146 flight b607.

The range-corrected lidar signal (see Sect.  for more detail on the lidar measurements) is shown here as coloured blocks and has a vertical resolution of 45 m and an integration time of 1 min. It is used here as an indicator for the presence of dust and clouds in the atmosphere but limitations apply. For instance, attenuation of the laser beam when it passes through an elevated dust layer can limit the lidar's ability to detect dust at lower levels. Dropsonde observations allow for the identification of atmospheric structures such as the top of the CBL and SRL as well as temperature inversions. The CBL depth was determined by locating the altitude in the sonde profile (from the surface upwards) where the potential air temperature (θ) first reaches 0.3 ∘C above the value at 150 m above the surface. In cases where θ increased monotonically from the surface up to 150 m, the surface θ value was used as a reference. The top of the SRL was determined manually where possible by identifying a sharp decrease in water vapour mixing ratio coinciding with a sharp increase in θ. The resulting CBL and SRL tops were linked by solid lines in Fig. 17 in order to illustrate spatiotemporal changes of these features. The depth of air temperature inversions, defined as an increase in air temperature with altitude, are indicated as grey boxes along the vertical sonde tracks together with the inversion strength in ∘C km-1 (Fig. 17).

Dropsonde-derived near-surface winds ranging between 11 and 17 m s-1 observed during the b607 outgoing flight (not shown) led to local dust emissions observed by the lidar at about 7.3∘ W (also seen in lidar depolarisation data, not shown) that were prevented from upward mixing by a low-level temperature inversion (Fig. 17a). At that time in the morning, the CBL was still relatively shallow (mostly < 1 km deep), the top of the SRL varied between about 4.3 and 5.5 km a.m.s.l. and an aged dust layer of varying intensity could be identified close to the SRL top. Cloud development was identified west of 11∘ W in the lidar data. In the time that passed between the outgoing and return flight, surface emissions ceased and the CBL expanded to about 4.5 km a.m.s.l. (B4, Fig. 17b) as a result of increasing near-surface temperatures. East of about 7.5∘ W, the CBL was prevented from growing deep by temperature inversions and the influence of monsoon flow (not shown). Clouds continued to develop west of about 10.5∘ W. CBL growth rates can be calculated for each dropsonde location based on the two dropsonde profiles. The SRL top showed little change compared to the outward leg apart from at B4 where the SRL was consumed completely by the fast-growing CBL (Fig. 17). It should be noted that SEVIRI imagery did not show any dust presence along the flight tracks on this day, suggesting that the lidar dust signal represents background dust levels – some dust is almost always present over northern Africa at this time of year (Israelevich et al., 2003).

As part of this SHL survey, the Falcon 20 took measurements at the same time as the BAe146 but on a more southern track (flight F24 in Fig. 2b). The analysis of the combined aircraft data showed that the SHL had an elongated shape with a NE–SW orientation. Moisture from the monsoon inflow was transported around the SHL at low levels in the morning. These unique measurements allow for the first time to challenge climate models in the SHL region and to understand the processes that control the observed temporal and spatial variability.

The Saharan atmospheric boundary layer (SABL) is probably the deepest on Earth, often reaching 5–6 km, and is crucial in controlling the vertical redistribution and transport of dust, moisture, heat and momentum fluxes in the Sahara (Cuesta et al., 2009). Before Fennec, aircraft observations and radiosondes (Cuesta et al., 2009; Messager et al., 2010; Marsham et al., 2013b) showed the persistence of a deep near-neutral Saharan residual layer (SRL) over large areas of the Sahara throughout the day, with only a very small temperature inversion separating the SRL from the CBL below. Flamant et al. (2007) and Messager et al. (2010) showed that the SRL may have a maximum humidity mixing ratio at its upper levels, and that small errors in model representation of this humidity can have substantial consequences in terms of relative humidity, cloud cover and, therefore, radiation. This unusual structure of the SABL means that relatively small perturbations to CBL temperature (e.g. from a surface albedo anomaly) are expected to have significant impacts on vertical mixing and perhaps induce circulations that may affect the CBL in neighbouring regions. There was evidence of such effects in observations from the CBL (Marsham et al., 2008a) and in modelling studies (Birch et al., 2012; Huang et al., 2010), but observations of impacts on the SRL were lacking. Observations from Fennec BBM supersite 1 showed that the CBL tends only to reach 5 or 6 km between 15:00 and 18:00 UTC (Marsham et al., 2013b). Fennec flights have provided new insights into the vertical structure of and mixing within the SABL (see Garcia-Carreras et al., 2015, including schematic (their Fig. 14) and SABL mesoscale circulations, below).

During Fennec, aircraft lidar and in situ observations were used to better understand the vertical stratification and transport mechanisms within the SABL, as well as its temporal and spatial variability (Garcia-Carreras et al., 2015). In order to sample the turbulent vertical structure of the SABL during Fennec, stacked legs were performed at different heights, determined from inspecting dropsonde profiles launched at both ends of the leg before descending. Each run was at least 10 times the SABL depth (≥ 60 km) and took place between 13:00 and 15:00 LT, when sensible heating was maximum. Heat fluxes were computed from the stacked legs, as well as the ascents and descents, taking advantage of the shallow angle of the aircraft profiles. These indicate that entrainment fluxes are very weak, as a result of detrainment at the CBL top. This is a result of the weak temperature inversion, and high vertical velocity of overshooting parcels, which are characteristic of the SABL, and can explain the slow development of the CBL despite the strong surface heating. Lidar measurements from high-level runs also showed that the boundary layer depth can vary by up to 100 % over distances of a few kilometres due to turbulent processes alone, so that any given dropsonde profile may not be representative of the whole run.

Figure 18 shows an example from a flight where small variations in heating from an albedo anomaly appear to be generating mesoscale circulations within the SABL. Figure 18 shows the vertical extinction coefficient at 532 nm retrieved with the LNG lidar on 20 June 2011 (14:05–14:46 UTC, flight F21, see Fig. 2b) from the Falcon flying southeastward in Mauritania, with water vapour mixing ratio (WVMR) and wind profiles from four dropsondes overplotted. The lidar transect highlights a number of important boundary layer (BL) processes encountered during the Fennec campaign, showing variability from the turbulent to the synoptic scales, as described below.

At the synoptic scale, there is a temperature and humidity gradient across the transect, with warmer and drier conditions in the northwest (by ∼ 5 K and 7 g kg-1), leading to a deeper CBL compared to the southeast (4 km at 24∘ N compared with 2 km at 21∘ N). The monsoon flow at night reached approximately 20∘ N along the flight track, bringing in cool moist air into the southern end of the transect (from UK Met Office analysis, not shown), which was then redistributed vertically as the CBL grew during the day. The more spatially homogeneous residual layer, on the other hand, reflects the conditions from the day before; the monsoon front on the night of 19 June was considerably further south, leading to a deep CBL throughout the transect. There is substantial variability in the SABL depth and structure superimposed on the synoptic gradient. Variability at the turbulent eddy scale can be observed in the northern end of the transect (24.5–25.2∘ N), with changes in the depth of the well-mixed aerosol layer (and so the CBL) of ∼ 1km over short horizontal distances (5–10 km, ∼ 0.05 to 0.1 ∘), consistent with idealised simulations and lidar measurements described in Garcia-Carreras et al. (2015).

At the mesoscale, there is a region with cloud and deeper BLs at the boundary between the warm, dry conditions in the northwest and the moister conditions in the southeast (21.4–22.5∘ N), with an orange plume reaching 6 km at 22.2∘ N. Satellite imagery shows that the clouds observed by the lidar are part of a band of clouds coincident with a negative albedo anomaly of around 0.2 that is just west of the flight track at 21.6∘ N (red line, Fig. 18). The surface hotspot leads to a local increase in the CBL depth, cloud formation and an upward transport of dust. The impact of another smaller hotspot can be observed at 22.8∘ N. Easterly winds in the SRL in the southeast lead to the air mass overriding the deeper CBL in the northwest, potentially contributing to the cloud formation. The 3 g kg-1 contour in Fig. 18 has been drawn using the dropsonde data and the lidar-inferred aerosol distribution and suggests that the deeper CBL around 22∘ N acts to transport water vapour and dust directly to the top of the SRL, where it spreads laterally, capping the adjacent CBL and leading to weak maxima in water vapour mixing ratios at the top of the SRL in the three eastern dropsondes. This supports the hypotheses of Marsham et al. (2008a) and Messager et al. (2010) of mesoscale variability in the SABL and its role in the transport of CBL air into the SRL, with implications for the long-range transport of dust.

Approximately 15 pollutant plumes were observed on the BAe146 in the upper troposphere (6 to 8.5 km altitude) above the Sahara desert during the Fennec campaign in June 2011. Using Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model trajectory analysis and MODIS satellite fire products, four source regions were identified for these pollutant plumes: flaring from oil fields in Algeria and biomass burning in the southern USA, Venezuela and western Africa. The pollutant plumes displayed high concentrations of ozone and sub-micron particles, with differing characteristics from each source region. Values for the single scattering albedo ranged from 0.57 to 0.99 and for the Angstrom exponent from -0.85 to 2.44 for individual plumes. If the HYSPLIT trajectory calculations are robust (uncertain due the substantial errors identified in the vertical wind fields), it is believed this is the first aircraft measurement of flaring from oil fields and merits the attention of further research, planned for the forthcoming DACCIWA field campaign in southern west Africa.