Three Raman lidars were operated at the airport of Ouarzazate during SAMUM 2006. Ouarzazate (30.9°N, 6.9°W, 1133-m height a.s.l.) is located in a flat basin bounded by two almost parallel steep mountain chains of the High Atlas (more than 4000 m high) and the Anti-Atlas/Jebel Saghro (up to 2700 m high). The most powerful SAMUM lidar is the six-wavelength aerosol lidar backscatter extinction lidar-ratio temperature humidity profiling apparatus (BERTHA) of the Leibniz Institute for Tropospheric Research, IfT (Althausen et al., 2000). The Meteorological Institute of the Munich University (MIM) operated two lidars. Portable lidar system (POLIS) is a small lidar, transmitting linearly polarized laser pulses at 355 nm and can either be used as a Raman lidar or, alternatively, as a polarization lidar (Heese et al., 2002; Freudenthaler et al., 2008). Multiwavelength lidar system (MULIS) is a polarization-Raman lidar transmitting laser pulses at 355, 532 and 1064 nm wavelength (Wiegner et al., 1995; Freudenthaler et al., 2008). During SAMUM, Raman signals were only measured at 387 nm (nitrogen Raman signals). Main goal of MULIS are highly precise measurements of the depolarization ratio (Freudenthaler et al., 2008). The airborne lidar of DLR is a 532-nm HSRL with polarization-sensitive detection at 532 and 1064 nm (Esselborn et al., 2008; Freudenthaler et al., 2008). In the following, we briefly summarize the BERTHA characteristics.
The six-wavelength lidar transmits pulses at 355, 400, 532, 710, 800 and 1064 nm with a repetition rate of 30 Hz. Four lasers (two Nd:YAG and two Ti:Sa lasers) are employed. One Nd:YAG laser transmits pulses at 355, 532 and 1064 nm, the other Nd:YAG is used for optical pumping of the two Ti:Sa lasers. One of the Ti:Sa lasers transmits pulses at 710 nm and the other, pulses at 400 and 800 nm. The six laser beams are aligned onto one optical axis. A scanning unit outside the container permits measurements from 90° to −90° zenith angle (i.e. from west to east during SAMUM). Most observations are done at a zenith angle of 45° at daytime (to the west) and at 0°–5° at nighttime. Backscattered light is collected with a 0.53-m Cassegrain telescope.
A sketch of the multichannel receiver is shown in Fig. 1. The receiver unit separates the elastic backscatter signals (for the six laser wavelengths), vibrational–rotational Raman signals at 387 and 607 nm (nitrogen) and 660 nm (water vapour) and pure rotational Raman signals (Stokes and anti-Stokes lines) from oxygen and nitrogen around 532 nm (Mattis et al., 2002b; Arshinov et al., 2005). Dichroic beamsplitters, narrow-band interference filters and a double-grating monochromator (Arshinov et al., 2005) are used to separate the different signals. The grating setup for separating two temperature-sensitive Raman signals includes another 532 nm elastic backscatter channel to allow the full application of the aerosol Raman lidar method (described below) to the rotational Raman signals. This is important because the rotational Raman signals are strong enough to be detected at daytime, in the presence of a bright sky background. The vibrational Raman signals are too weak to be detected at daytime. Thus, the rotational Raman signals enable to determine dust extinction coefficients and extinction-to-backscatter ratios at daytime and nighttime, whereas the vibrational Raman signals permit us to determine these parameters at nighttime only, however, at two wavelengths (355 and 532 nm).
Figure 1. Sketch of the BERTHA receiver. The 14-channel lidar measures elastic backscatter signals (355, 400, 532, 710, 800, 1064 nm), vibrational Raman signals (387, 607, 660 nm, nitrogen N2, water vapour H2O), and rotational Raman signals around 532 nm (temperature channels, includes another 532 nm elastic backscatter channel). The number denotes the detection wavelength of the respective photomultiplier tube, abbreviations a and pc describe the kind of signal detection (analogue or photon counting), s and p (710 nm) denote channels for detection of cross and parallel polarized light, respectively.
Download figure to PowerPoint
At 710 nm, the cross- and parallel polarized components of the backscattered light are measured separately. Photomultiplier tubes (PMTs) are used as detectors. For the elastic backscatter signals (400, 532, 710, 800 nm) the analogue PMT output is pre-amplified and digitized (12 bit, 10 MHz), whereas the 355- and 1064-nm elastic backscatter signals, the Raman signals and a part of the backscattered 532-nm photons (primary wavelength of the 607 and 660 nm Raman signals) are detected with photon-counting PMTs, operated at 300 MHz.
The raw signals are stored with range and time resolutions of 7.5–60 m and 10–30 s, respectively. All raw signal profiles are corrected for deadtime effects and for sky background noise before the data are further processed to obtain the dust optical properties.
The profiles of the volume extinction coefficient of the particles at 355 and 532 nm are calculated from the measured profiles of the Raman signals. In the case of the rotational Raman signals, the sum of the two signals is used. Signal averaging of 30–120 min and vertical smoothing of the signal profiles with window lengths from 300 (lower heights) to 1200–2400 m (upper part of optically dense dust layers) was necessary to reduce the statistical errors of the extinction coefficients to values of 5%–25%. Systematic uncertainties caused by the removal of Rayleigh-scattering and air–density effects from the backscatter signals are on the order of 5%–10%.
It is interesting to note that in the case of the airborne HSRL, extinction profiling is based on measurements of almost pure Rayleigh backscatter signals. Rayleigh backscattering is more than two orders of magnitude larger than vibrational Raman backscattering and thus permits day and night operations with high temporal resolution of few seconds to few minutes. Signal smoothing of several hundreds of meters is, however, also necessary here to reduce the influence of signal noise (Esselborn et al., 2008).
The profiles of the volume backscatter coefficients at 355 and 532 nm are determined from the profile of the ratio of the elastic backscatter signal (at 355 or 532 nm) to the corresponding nitrogen Raman signal (387 or 607 nm). In a similar way, the 1064 nm signal and the 607 nm nitrogen Raman signal are combined to calculate the dust backscatter coefficient at 1064 nm. The use of a photon-counting PMT at 1064 nm enables us to accurately measure Rayleigh backscattering at this wavelength in the aerosol-free upper troposphere and lower stratosphere. The signal calibration height was typically set to 7–9 km height above ground level (a.g.l.), and a total-to-Rayleigh backscatter ratio of 1.05 was assumed in the data analysis. Calibration uncertainties are estimated to be less than 10% at 1064 nm.
As mentioned, these retrievals can only be applied to the nighttime signals. The advantage of using signal ratios is that laser-beam receiver-field-of-view (RFOV) overlap effects (as discussed below) widely cancel out so that the profile of the backscatter coefficient (indicating the dust layering) is trustworthy to low heights above the lidar. Furthermore, input parameters such as the lidar ratio that is used to correct for attenuation effects are not needed. Similarly, the aerosol and Rayleigh signals at 532 nm are used in the case of the HSRL to calculate the backscatter coefficients (Esselborn et al., 2008). Once the profiles of the extinction and the backscatter coefficients are available, the vertical profile of the extinction-to-backscatter ratio is obtained in addition.
Alternatively, the elastic backscatter signals alone can be used to compute the volume backscatter-coefficient profile by applying the Fernald method (Fernald, 1984) and to estimate, in a subsequent step, the corresponding extinction coefficients from the backscatter coefficients. In the Fernald method, the extinction-to-backscatter ratio is an input parameter. The advantage of the method is that extinction and backscatter profiles can be determined with high temporal and vertical resolution at nighttime as well as at daytime.
The relative statistical error of the backscatter coefficients is generally of the order of 5%–10%. Systematic effects (calibration, Rayleigh scattering correction) are estimated to be about 5%–10% in the desert environment. The extinction coefficients estimated from the backscatter coefficients are also obtained with comparably high accuracy of 10%–20%, because the dust lidar ratio (main input parameter) is well known from the Raman and HSRL observations and does not introduce large uncertainties in this homogeneous environment. In polluted continental areas, where the relative contributions of maritime particles, fresh and aged urban haze and smoke particles to the observed aerosol are unknown and may vary strongly with time and height, the uncertainty in the assumed lidar ratio profile and in the extinction estimation from backscatter profiles is usually very high.
The extinction measurements with ground-based lidars are biased by the incomplete overlap of the laser beam with the RFOV in the near range which is about 100–200, 300–700 and about 2000–3000 m in the case of POLIS, MULIS and BERTHA (most channels), respectively. The RFOV of BERTHA is 0.8 mrad for most of the channels and only 0.1–0.15 mrad for the grating monochromator (rotational Raman channels and corresponding 532 nm elastic backscatter channel). The full laser-beam RFOV overlap is reached in a distance of about 4000-5000 m from the lidar in the case of the rotational Raman signals.
Fortunately, the overlap effect can partly be corrected as shown in Fig. 2. During clear, almost dust-free nights, the overlap profile of the lidar can be determined by applying the procedure suggested by Wandinger and Ansmann (2002). Correction of the overlap effect by means of the measured overlap profile enables us to obtain trustworthy extinction values down to 700–1000 m at optimum conditions with a stable overlap profile over days. However, this was not the case during SAMUM. Because of permanent air conditioning problems in the BERTHA lidar container (with four lasers and many data acquisition systems producing a lot of heat), the overlap characteristics changed significantly from day to day in this desert environment. As a consequence, BERTHA signals measured at heights below 2 km a.g.l. were often not reliable enough to be further used in the extinction, backscatter and lidar ratio retrievals.
Figure 2. (a) Lidar overlap function as a function of height above ground (a.g.l.), computed from the 355 and 387 nm signals (solid line) and from the 532 and 607 nm signals (dashed line) measured during a clear night (11 May) and dust extinction coefficients computed from 387 nm (b) and 607 nm Raman signals (c) observed on 15 May. The overlap effect is ignored in the case of the dashed lines and considered in the case of the solid lines in (b) and (c).
Download figure to PowerPoint
A trustworthy correction of the overlap was usually possible down to 150 m height a.g.l. in the case of POLIS and to 350 m height a.g.l. in the case of MULIS. No overlap problems occur for a downward looking lidar aboard an aircraft flying at 8–12 km height a.s.l.. The full overlap is reached well above the dust layers.
Finally, another error source has to be discussed which may play a critical role in desert environments (Mattis et al., 2002a). Our polarization-dependent calibration measurements at Leipzig, Germany, revealed that the efficiency factor η of most detection channels of BERTHA is polarization sensitive. η describes the transmission of light from the primary mirror of the telescope to the photomultiplier. For most channels, however, this polarization-dependent effect is small and can be ignored in the retrieval. The exception is the 532 nm pc channel in Fig. 1.
The efficiency factor η∥ for the parallel-polarized component (parallel with respect to the plane of polarization of the emitted linearly polarized laser light) is about 5–7 times smaller than the related cross-polarized efficiency factor η⊥. To properly account for this effect, we follow the procedure suggested by Mattis et al. (2002a). The consequence of the effect is illustrated in Fig. 3a. Without correction, the computed 532 nm backscatter coefficient in the dust layer is, by a factor of two, larger than the true one (MULIS data). The measured signal strengths are corrected by dividing the 532 nm signals P532(z) by K(z, λ) given by
K(z, λ) for height z and wavelength λ of the 532-nm elastic backscatter channel (532, pc, Fig. 1) is shown in Fig. 3c. D=η⊥/η∥ is obtained from the calibration measurement at Leipzig. For the observation shown in Fig. 3, D= 6.76. The required volume depolarization ratio δ(z) at 532 nm, the ratio of the cross polarized to the parallel polarized total (Rayleigh + particle) backscatter coefficient, is provided by MULIS (Freudenthaler et al., 2008). The volume depolarization ratio for the measurement on 3 June 2006, is shown in Fig. 3b. Remaining uncertainties in the corrections are estimated to be of the order of a few percent. Profiles of the volume depolarization ratio for all available measurement days were required from the MULIS observations to obtain a trustworthy 532-nm backscatter data set. MULIS was operated in a distance of about 10 m from BERTHA.
Figure 3. Influence of the polarization-sensitive detection efficiency of the 532 nm elastic backscatter channel (532, pc, Fig. 1) on the retrieval of the dust backscatter coefficient. (a) The dashed curve shows the uncorrected case, the thick solid curve is obtained from the corrected 532 nm signals. For comparison, the profile of the dust backscatter coefficient computed from MULIS data is shown (thin solid line). In the signal correction, the profile of the 532 nm volume depolarization ratio measured with MULIS, shown in (b), is required to calculate K(z) (eq. 1), shown in (c). For correction, the 532 nm signals (as function of height z) are divided by K(z), then the backscatter coefficients are calculated from the corrected signals.
Download figure to PowerPoint
It should be mentioned that the other two 532 nm channels of BERTHA do not show this polarization-dependent effect and can thus be used to check the quality of the correction. This strong polarization effect on our main 532 nm signal was not visible during foregoing campaigns such as LACE 98 (Wandinger et al., 2002) and the Indian Ocean Experiment (INDOEX, Franke et al., 2003). This may partly be caused by the fact that anthropogenic aerosols which produce very low depolarization ratios prevailed during these campaigns. However, during these campaigns, the polarization effect was also not visible in cirrus clouds. The most reasonable explanation is therefore that one of the mirrors that directs the light to the 532 nm PMT degraded significantly during the last 6 years, with the result of an increased sensitivity of this channel to the state of polarization of the incoming light.