Closure of In-Situ Measured Aerosol Backscattering and Extinction Coefficients with Lidar Accounting for Relative Humidity

Aerosol particles contribute to the climate forcing through their optical properties. Measuring these aerosol optical properties is still challenging, especially considering the hygroscopic growth of aerosol particles, which alters their optical properties. Lidar and in-situ techniques can derive a variety of aerosol optical properties, like aerosol particle light extinction, 15 backscattering, and absorption. But these techniques are subject to some limitations and uncertainties. Within this study, we compared with Mie-theory modeled aerosol optical properties with direct measurements. At dried state they were with airborne and ground-based in-situ measurements; at ambient state with lidar estimates. Also, we examined the dependence of the aerosol particle light extinction-to-backscatter ratio, also lidar ratio (LR), to relative humidity. The used model was fed with measured physicochemical aerosol properties and ambient atmospheric conditions. The model considered aerosol particles in an internal 20 core-shell mixing state with constant volume fractions of the aerosol components over the entire observed aerosol particle sizerange. The underlying set of measurements was conducted near the measurement site Melpitz, Germany, during two campaigns in summer, 2015, and winter, 2017, and represent Central European background aerosol conditions. Two airborne payloads deployed on a helicopter and a balloon provided measurements of microphysical and aerosol optical properties and were complemented by the polarization Raman lidar system Polly as well as by a holistic set of microphysical, chemical and 25 optical aerosol measurements derived at ground level. The calculated aerosol optical aerosol properties agreed within 13% (3%) with the ground-based in-situ measured aerosol optical properties at a dried state (relative humidity below 40%) in terms of scattering at 450 nm wavelength during the winter (summer) campaign. The model also represented the aerosol particle light absorption at 637 nm within 8% (18%) during the winter (summer) campaign and agreed within 13% with the airborne in-situ aerosol particle light extinction measurements during summer. During winter, in a comparatively clean case with 30 equivalent black carbon mass-concentrations of around 0.2 μg m the modeled airborne measurement-based aerosol particle light absorption, was up to 32-37% larger than the measured values during a relatively clean period. However, during a high polluted case, with an equivalent black carbon mass concentration of around 4 μg m, the modeled aerosol particle light absorption coefficient was, depending on the wavelength, 13-32% lower than the measured values. Spread and magnitude of the disagreement highlighted the importance of the aerosol mixing state used within the model, the requirement of the inclusion 35 of brown carbon, and a wavelength-dependent complex refractive index of black and brown carbon when such kind of model is used to validate aerosol particle light absorption coefficient estimates of, e.g., lidar systems. Besides dried state comparisons, ambient modeled aerosol particle light extinction, as well as aerosol particle light backscattering, were compared with lidar estimates of these measures. During summer, on average, for four of the twelve conducted measurement flights, the model calculated lower aerosol particle light extinction (up to 29% lower) as well as 40 backscattering (up to 32% lower) than derived with the lidar. In winter, the modeled aerosol particle light extinction coefficient was 17%-41% lower, the aerosol particle light backscattering coefficient 14%-42% lower than the lidar estimates. https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c © Author(s) 2021. CC BY 4.0 License.

Based on selected cases, this study presents the results of two field-experiments conducted in June 2015, and Winter, 95 2017 at the regional Central-European background measurement facility in Melpitz, located in the East of Germany. In both, a combination of airborne in-situ and remote sensing measurements, accompanied by a sophisticated set of ground-based insitu measurements, were conducted under different atmospheric conditions and aerosol load. This study aims at first to compare remote sensing measurements of σbsc(λ) and σext(λ) with calculated airborne in-situ measurement-based modeled coefficients, utilizing a closure study. Second, it gives insights on the LR enhancement, and answers the question to which extent the lidar 100 ratio depends on the ambient RH at three different wavelengths based on in-situ measurement-based optical modeling under the given aerosol conditions at the measurement site. Third, the study evaluates the capability of the used Mie-model to recreate measured σabs(λ) at different wavelengths to create a tool for the validation lidar-based σabs(λ) estimates as shown by Tsekeri et al. (2018). This study, which includes simultaneous modeling of σbsc(λ), σext(λ), and σabs(λ) in ambient and dried state based on ground-based and vertical resolved in-situ, and remote-sensing measurements, is unique in its complexity. 105 This work is structured as follows. First, an overview of the measurements site and the deployed instrumentations is given. Afterwards, details about the used optical model including a description of the applied input parameters as well as the validation with in-situ reference instrumentation are given. Subsequently, the comparison of Mie-modeled and measured aerosol optical properties is presented and discussed separated into the summer and winter experiment. This also includes with a short overview of the meteorological and aerosol conditions during the experiments. The quantification of the lidar ratio 110 enhancement with respect to RH is given for the summer case. Finally, conclusions are formulated based on the results.  Bühl et al., 2013) Melpitz Observatory comprises comprehensive instrumentation in quasi-continuous operation, for high-quality, long-term observations and can be adapted to the needs as required. An overview of the continuously operating instrumentation is presented in the following. Details about specific instrumentation additionally 130 added during the campaigns will be given within respective subsections.

Ground in-situ instrumentation
In both campaigns, the PNSD was measured by a combination of a Dual Mobility Particle Size Spectrometer (D-MPSS, TROPOS-type; Birmili et al., 1999) with 10% accuracy and Aerodynamic Particle Size Spectrometer (APSS, mod. 3321, TSI Inc., Shoreview, MN, USA) with 10%-30% uncertainty depending on the size-range (Pfeifer et al., 2016). 135 A D-MPSS consist of a bipolar diffusion charger, two differential mobility analyzer (DMA; Knutson and Whitby, 1975) and two condensation particle counters (CPC; mod. 3010 and UCPC; mod. 3776, TSI Inc., Shoreview, MN, USA). The bipolar charger transforms the aerosol into a well-defined charge equilibrium, according to Fuchs (1968) and Wiedensohler et al. (1988). The TROPOS-type DMAs selects the charged aerosol particles concerning their electrical mobility, and the CPC then counts their number concentration. Overall this setup covers an aerosol particle size range of 3-800 nm in mobility 140 diameter (Dm). The PNSD is available every 20 minutes, and a scan duration is ten minutes. The final D-MPSS PNSD used in this study was derived utilizing an inversion routine (Pfeifer et al., 2014) accounting for multiple charged aerosol particles, including a diffusion loss correction based on the method of "equivalent pipe length" .
For the calculation of the optical properties with the Mie-theory, spherical particles must be assumed. Therefore, we assumed that all aerosol particles measured by the D-MPSS system used here are spherical, and the Dm is equal to the volume 145 equivalent diameter (Dv). The quality of the PNSD measurements is assured by frequent calibrations as described in Wiedensohler et al. (2018). To cover the entire size-range from 10 nm to 10 µm, the APSS PNSD extended the D-MPSS PNSD. For this purpose, the aerodynamic diameter (Daer) of the APSS was converted into Dv applying: following DeCarlo et al. (2004). Thereby ρ0 corresponds to the standard density of 1 g cm -3 , ρaer to the aerosol density, ρeff to the effective aerosol density of 1.5 g cm -3 for fine mode aerosol and already accounts for the shape of the larger aerosol particles expressed with the shape factor χ. The effective density of 1.5 g cm -3 was chosen, because with that a best overlap of the APSS and T-MPSS PNSD was achieved for the majority of merged PNSDs. Also, this effective density fits reasonably well to the findings of Tuch et al. (2000) and Poulain et al. (2014) with reported aerosol particle densities of 1.53 ± 0.31 g cm -3 and 155 1.4 g cm -3 to 1.6 g cm -3 , respectively. Although shape factor and aerosol particle density are usually size-dependent, we assumed a constant density and shape of the aerosol particles for all the measurements of the APSS. At visible wavelengths, the coarse-mode of the PNSD is less efficient than the fine-mode in terms of aerosol particle light scattering and extinction.
Hence, for aerosols dominated by accumulation mode particles, the underlying assumption is appropriate to calculate the extinction and scattering properties of the aerosol. 160 In addition to these continuously running instruments at Melpitz Observatory, a Quadrupole Aerosol Chemical Speciation Monitor (Q-ACSM, Aerodyne Res. Inc, Billerica, MA., USA; Ng et al., 2011) measured the mass concentration of non-refractory particulate matter (PM). Ammonium (NH4), sulfate (SO4), nitrate (NO3), and chlorine (Cl), as well as the organic aerosol mass, have been derived in the fine-mode regime (NR-PM1). Further details on the Q-ACSM measurements at Melpitz can be found in Poulain et al., (2020). An ion-pairing scheme (ISORROPIA II; Fountoukis and Nenes, 2007) was 165 utilized to derive the chemical compounds of the aerosol particles at 293 K and 0% RH. Furthermore, a DIGITEL DHA-80 (Walter Riemer Messtechnik e.K., Hausen/Röhn, Germany) high volume aerosol sampler collected daily the PM10 (10 denotes an aerodynamic diameter of the aerosol particles of 10 µm) aerosol particles on a quartz-fiber filter (Type MK 360, Munktell, Grycksbo, Sweden) with a total flow of 30 m 3 h -1 . Among others, Müller (1999), Gnauk et al. (2005), and Herrmann et al. (2006) provide detailed information about the aerosol sampler. The sampled quartz-fiber filter was analyzed offline and 170 allowed the determination of the total aerosol particle mass concentration (in this study we focus on PM10), water-soluble ions, and the mass of elemental carbon (EC). The EC mass concentration (mEC) was measured following the EUSAAR2 protocol (Cavalli et al., 2010), A continuously operating Multi-Angle Absorption Photometer (MAAP; Model 5012, Thermo Scientific, Waltham, MA, USA; Petzold and Schönlinner, 2004) recorded the σabs(λ) at Melpitz Observatory at a wavelength of 637 nm with an 175 uncertainty of 10% (Müller et al., 2011) to 12% (Lack et al. 2014). Several corrections were applied to the aerosol particle light absorption measurements of the MAAP. Following Müller et al. (2011), a wavelength correction factor of 1.05 was applied to all MAAP-data in this study. Furthermore, observations conducted in Melpitz by Spindler et al. (2013) andPoulain et al. (2014) have shown that the submicron aerosol regime contains 90% of the total PM10 equivalent black carbon (eBC; Petzold et al., 2013) mass concentration (meBC). Hence, on the estimated meBC data, a correction factor of 0.9 was applied to 180 match the corresponding PM1 measurements of the Q-ACSM. With mEC) and these absorption measurements, meBCwas derived using a time-dependent (t) mass absorption cross-section related to the MAAP wavelength of 637 nm (MAC(t, λ = 637 nm)) with: eBC ( , 637nm) = abs ( (hourly),637nm) ( (daily),637nm) . ( The daily average MAC(t, 637 nm) was derived by dividing the daily mEC by the daily (midnight to midnight) mean of the 185 measured σabs(637 nm): .
Following this approach, a mean daily MAC(637 nm) of 10.4 m 2 g -1 (median 10.9 m 2 g -1 ; IQR: 7.1 to 12.3 m 2 g -1 ) was derived for the period between February 1 and March 15, 2017. Recently, Yuan et al. (2020) provided MAC(870 nm) estimates for the winter campaign period of this study of 7.4 m 2 g -1 (geometric mean value, range from 7.2 to 7.9 m 2 g -1 ) which relates to a 190 MAC(637 nm) of around 10.8 m 2 g -1 (10.5 to 11.5 m 2 g -1 ) assuming an absorption Ångström exponent (AAE) of 1.2 (taken https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. from Yuan et al., 2020). Zanatta et al. (2016), also, reported a geometric mean MAC(637 nm) of 8.2 m 2 g -1 (geometric standard deviation of 1.5 m 2 g -1 ). For the period between June 1 and June 30, 2015, a mean daily MAC(637 nm) of 7.3 m 2 g -1 (median 7.2 m 2 g -1 ; IQR: 6.0 to 8.4 m 2 g -1 ) was estimated at Melpitz Observatory, which agrees with the 7.4 m 2 g -1 previously reported by Nordmann et al. (2013) and is slightly lower than the geometric mean MAC(637 nm) of 9.5 m 2 g -1 (geometric standard 195 deviation of 1.38 m 2 g -1 ) reported by Zanatta et al (2016) for the aerosol at Melpitz during summer. However, the estimates of Nordmann et al. (2013) were derived with Raman spectroscopy. Hence, the here estimated MAC(637 nm) values for summer and winter seem reasonable as well, but will be evaluated in depth later on. The specific volume fractions of each aerosol compound, fv,i, were derived based on the Q-ACSM and MAAP measurements dividing the mass of each aerosol compound with its respective density. Appendixtable 1 lists the density of each derived aerosol compound. Moteki et al. (2010) reported 200 that it is accurate within 5% to assume the density of non-graphitic carbon at 1.8 g cm -3 . Therefore, in this study a BC density of 1.8 g cm -1 is used. These measurements were completed by a Nephelometer (mod. 3563, TSI Inc., Shoreview, MN, USA), which measures the σsca(λ) at 450, 550, and 700 nm with a relative uncertainty by calibration and truncation of about 10% . The error of the Nephelometer measurements due to truncation and illumination was corrected following Anderson 205 and Ogren. (1998).
The aerosol particle hygroscopicity parameter κ, introduced by Petters and Kreidenweis (2007), represents a quantitative measure of the aerosols water uptake characteristics and depends on the chemical composition of the aerosol particles as well as their size. A Volatility Hygroscopicity-Tandem Differential Mobility Analyser (VH-TDMA), first introduced by Liu et al. (1978), measures the hygroscopic growth, and hence water uptake, of aerosol particles at a specific RH. This instrument was 210 deployed at Melpitz Observatory during the summer campaign. The VH-TDMA measured the hygroscopic growth of aerosol particles in six different size-bins (30, 50, 75, 110, 165, and 265 nm) from which the size-resolved aerosol hygroscopicity κ(Dp) was inferred. The scientific community uses a variety of VH-TDMAs, but detailed insights on the system deployed here provide Augustin-Bauditz et al. (2016). The inferred κ(Dp) allows to extrapolate the hygroscopic growth of aerosol particles to another RH. For the calculation of the hygroscopic growth of the aerosol particles under ambient conditions, we assumed 215 κ(Dp) for diameters smaller 30 nm is equal to κ(30 nm) and for diameters larger 265 nm is equal to κ(265 nm). During the winter campaign, no size-resolved direct hygroscopicity measurements were available. Therefore, the hygroscopicity of the aerosol particles encountered in the winter campaign was derived based on the parallel conducted measurements of the aerosol chemical composition utilizing the Zdanovskii, Stokes, and Robinson (ZSR;Zdanovskii, 1948;Stokes and Robinson, 1966) volume-weighted mixing rule considering the hygroscopicity parameter of every single aerosol compound κi listed in 220 Appendixtable 1. A comparison of the size-segregated κ(Dp) estimates of the VH-TDMA with bulk Q-ACSM measurements during the summer campaign has shown a 1:1 agreement with high correlation (R 2 = 0.98, fit through origin) at 165 nm. Hence, bulk Q-ACSM measurements represent the aerosol at a size of around 165 nm. However, the bulk Q-ACSM approach might over-or underestimates the hygroscopicity of aerosol particles lower or larger than 165 nm. Furthermore, Düsing et al. (2018) have conducted an optical closure experiment comparing Mie-based aerosol particle light extinction and backscatter 225 coefficients with lidar measurements, using both, κ estimates based on chemical composition and cloud condensation nuclei counter measurements at 0.2% supersaturation. In the case of the chemical composition measurements the agreement with the lidar was within 10% in terms of the aerosol particle light extinction coefficient. Hence, using κ from the bulk Q-ACSM measurements is a feasible approach.

Ground-based remote sensing 230
In addition to the in-situ measurements on the ground, in both campaigns a Lidar system was used to determine σbsc (λ) and σext(λ). This system was Polly XT , a 3+2+1 wavelengths Raman polarization lidar system, in the first version introduced by Althausen et al. (2009). The Polly XT version in this study was introduced by Engelmann et al. (2016) and did operate with https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. three channels for aerosol particle light backscattering and two for aerosol particle light extinction. During the summer campaign a near-field channel at 532 nm was available. After the summer campaign, Polly XT was updated and equipped with 235 an additional near-field channel at 355 nm and therefore available during the winter-campaign. Vertical profiles of these aerosol properties were available each 30 s with a vertical resolution 7.5 m. The geometry of emitted laser and far field-ofview (FOV) leads to a partial overlap below 800 m altitude, which is known as the overlap height and can be determined experimentally (see Wandinger and Ansmann, 2002). Below 800 m, an overlap correction was applied to the lidar data (see Ansmann, 2002). The standard far FOV is 1 mrad and the near FOV is 2.2 mrad 240 . The automated data evaluation routines and quality check control are presented in detail in Baars et al. (2016). An intercomparison campaign presented by Wandinger et al. (2016) including different EARLINET (European Aerosol Research LIdar NETwork) instruments, including the system within this study (see Lidar system named le02 therein) has shown a maximum deviation of less than 10%. Hence, we assume a 10% measurement uncertainty of the σbsc(λ)

measurements. 245
During daytime, the signal-to-noise ratio in the Raman-channels is too weak due to solar radiation to provide robust Raman σext(λ). Therefore, in this and other studies, the σbsc(λ) have been converted to σext(λ)by means of the extinction-tobackscatter ratio, also known as lidar ratio (LR, in sr), with: LR is an aerosol intensive property. 250 In the past, several studies, investigated the LR of different aerosol types with ground-based lidar systems (Haarig et al., 2016, Mattis et al., 2004, and Ansmann et al., 2010 with an airborne lidar system by Groß et al. (2013). Cattrall et al. (2005) estimated LRs at 550 nm and 1020 nm wavelength based on retrievals of direct sky radiance and solar transmittance measurements. Tao et al. (2008) and Lu et al. (2011) determined the LR with a synergistic approach combining space-borne and ground-based lidar. Düsing et al. (2018) provide LR based on airborne in-situ measurements estimated with 255 Mie-theory. All these investigations clearly show that the LR is highly dependent on the predominant aerosol types. Müller et al. (2007) and Mattis et al. (2004) provided an overview of the LR for different aerosol types. Mattis et al. (2004Mattis et al. ( ) provided long-term (2000Mattis et al. ( -2003 estimates of the LR for central European haze (anthropogenic aerosol particles) of 58 (±12) sr for 355 nm, 53 (±11) sr for 532 nm, and 45 (±15) sr for 1064 nm wavelength, respectively. In this study, the measured σbsc(λ) was transformed into σext(λ) with these estimates. The uncertainties of the estimates of Mattis et al. (2004) and the measurements 260 uncertainties of the lidar system were accounted in the derived σext(λ).
Additionally, a sky spectral radiometer (mod. CE318, Cimel Electronique, 75011 Paris, France) was deployed during both intensive periods of both campaigns as part of the AERONET observations. This pointed sun radiometer derived the AOD at several wavelengths, and Holben et al. (1998) provide detailed insights on the working principle of this instrument. It was used to cross-check the lidar retrievals in terms of validation of the integrated σext(λ) profiles with the AERONET AOD. 265 With a combination of both, the lidar and the sun-photometer, profiles of σabs(λ) can be estimated using the Generalized Aerosol Retrieval from Radiometer and Lidar Combined data algorithm (GARRLiC; Lopatin et al., 2013). But AOD at 404 nm of 0.4 and more are needed for this purpose, thus we could not apply it for our study.

Airborne in-situ measurements during summer The Airborne Cloud and Turbulence Observation System 270
During the intensive period of the summer campaign, a set of state-of-the-art instruments, installed on the airborne platform ACTOS (Siebert et al., 2006), determined microphysical and aerosol optical properties. ACTOS was designed as an external cargo under a helicopter with a 150 m long aerial rope and was operated maximum ascend and descend speeds of 6 m s -1 . Ambient RH and temperature (T) were recorded as well and were averaged to a temporal resolution of 1 Hz. A data https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. link was established between ACTOS and a receiver station installed on the helicopter and allowed the scientist on board of 275 the helicopter a real-time data observation to adjust flight height and track.

Aerosol sampling on ACTOS 295
On ACTOS, a custom-made silica-bead based diffusion dryer dried the air sample to ensure an aerosol humidity below 40% following the recommendations of . A TROPOS-built MPSS determined the PNSD with a temporal resolution of two minutes covering a size range of 8 nm to 230 nm. This temporal resolution translates into a vertical spatial resolution of several 100 m depending on the ascent/descent speed of the helicopter. Like the D-MPSS on the ground, this MPSS included a bipolar charger (here mod. 3077A, TSI Inc., Shoreview, MN, USA) containing radioactive Kr-85, a 300 TROPOS-type DMA (Hauke-type, short) and a condensation particle counter (CPC; mod. 3762A, TSI Inc., Shoreview, MN, USA) with a lower cut-off diameter (Dp,50%; the CPC detects 50% of the aerosol particles with this diameter) of around 8 nm and counting accuracy of 10%. An optical particle size spectrometer (OPSS; here mod. skyOPC 1.129, GRIMM Grimm Aerosol Technik, Ainring, Germany) recorded the optical equivalent PNSD covering an aerosol particle size range of 350 nm https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. to 2.8 µm (optical diameter) with a temporal resolution of 1 Hz. The corresponding two-minute averaged OPSS PNSD 305 extended the MPSS PNSD. The detailed geometry of the optical cell inside the instrument is unknown. Hence, a correction regarding the complex aerosol refractive index (n = nr + ini) could not be applied to the data set. The upper cut-off of the inlet system is estimated at around 2 µm following Kulkarni et al. (2011). The PNSD has been corrected concerning aspirational and diffusional losses following Kulkarni et al. (2011) and  using the method of the "equivalent pipe length". 310

315
The quality of the airborne in-situ measurements was checked by comparing average of the lowermost 200 m with the ground in-situ measurements (see Figure 2). The intercomparisons revealed a distinct underestimation of the aerosol particle number concentration above 800 nm in optical diameter (see Figure 2). This underestimation is caused presumably due to a mixture of losses within the system which cannot be addressed appropriately and the here missing refractive index 320 correction of the OPSS which would shift the OPSS PNSD more to larger particle diameters (see Alas et al., 2019). Since, the in-situ instrumentation at ground is quality-assured, the ground-based measurements is the reference and was utilized to correct the airborne measurements. Therefore, above 800 nm, the airborne in-situ PNSD recorded by the OPSS was replaced and extended with a height-corrected PNSD measured on the ground at Melpitz Observatory establishing a non-fixed altitudecorrection factor fh. The altitude-correction factor fh(h, scan) was calculated according Eq. (6): 325 Where NOPSS,<200 m is the mean aerosol number concentration derived with the OPSS in the lowermost 200 m. NOPSS(h, scan) is the mean aerosol particle number concentration detected by the OPSS during the corresponding scan-time of the MPSS at a given altitude h (NOPSS(h)). Advantageously, this method accounts for uncertainties introduced due to differences in the complex refractive index of the calibration aerosol and the prevalent aerosol and accounts for the upper cut-off limit of the 330 inlet-system. first filter. A photodetector detects the intensity of light of the given wavelength behind these filter spots. All raw data have been recorded on a 1 Hz time-resolution. At default the STAP estimates σabs(λ) based on 60 s running averages of the measured intensities. At this averaging period, the measurement uncertainty is estimated to 0.2 Mm -1 . Based on differential light attenuation measurements between two time-steps, the STAP calculates the σabs(λ). Filter-loading and the enhancement of 340 absorption due to multiple-scattering within the filter-material have been corrected following Ogren (2010) and Bond et al. (1999). These corrections include the real-time estimated filter-transmission dependent loading correction factor: where the transmission τ is defined as the ratio of the intensity I(t) measured at time t and the blank-filter intensity I0 = I(t0).
Due to the limited computational power of the internal chip onboard of the STAP σext(λ) was recalculated at 1 Hz time 345 resolution during the postprocessing with larger precision. Since the STAP was in an early developing state we faced issues concerning the implemented analog-to-digital converter and data of the STAP sampled during summer is not presented in this study.

Airborne in-situ measurements during winter
During MelCol-Winter, the tethered balloon system BELUGA (Balloon-bornE modular Utility for profilinG the lower Atmosphere, Egerer et al., 2019) carried a set of payloads, which determined meteorological conditions, including ambient T and RH, as well as microphysical and aerosol optical properties. The 90 m³ helium-filled balloon was attached on a 2 km long 355 tether (3 mm Dyneema®), an electric winch allowed profiling with a climb and sink-rate of 1 to 3 m s -1 .
A temperature-insolated container included the same STAP also deployed during the summer campaigndetermined σabs(λ). An OPSS (mod. 3330, TSI Inc., Shoreview, MN, USA) was sampling the PNSD in a range of 0.3 to 10 µm in 16 size bins every 10 seconds. The OPSS PNSD was corrected in terms of the complex aerosol refractive index. Here, a complex aerosol refractive index of 1.54 + i0 was used since this resulted in OPSS PNSD with a good overlap to the MPSS PNSD. The 360 imaginary part of the complex aerosol refractive index was forced to 0 because it leads to a significant overestimation of the coarse mode in the PNSD when the imaginary part of the complex aerosol refractive index is above 0 (see Alas et al., 2019).
Note, that this complex aerosol refractive index is not the refractive index used in the Mie model.
The missing size-range of the PNSD, here all particles smaller than 0.3 µm in optical diameter, was extended with the altitude-corrected average ground-based PNSD of the corresponding flight period analogue to the summer campaign. Here, 365 the variable altitude correction factor fh from Eq.(6) was for each OPSS PNSD the ratio of the aerosol particle number concentration detected by OPSS within the lowermost 50 m (NOPSS,<50 m) and the aerosol particle number concentration detected by the OPSS at an altitude h (NOPSS(h)). Particles larger than 800 nm have not been replaced by the PNSD measurements at ground since the refractive index correction was applied to the OPSS data.
Varying wind-speeds during the campaign changed the inclination of the aerosol inlet accordingly. Therefore, we do 370 not account for the varying upper cut-off of the inlet. However, calculations following Kulkarni et al. (2010) with an inclination angle of 90° show that 50% of 10 µm aerosol particles with a density of 2 g cm -3 are aspirated by the inlet at a wind-speed of around 0.8 m s -1 . Diffusional losses at the OPSS size-range are negligible. The aerosol was dried with a silica-bead based dryer similar to the one on ACTOS to dampen sudden changes in the RH of the aerosol stream, which can have significant influences on the filter-based absorption measurements of the STAP as shown, for instance, by Düsing et al. (2019). They estimated a 375 deviation of around 10.08 (± 0.12) Mm −1 % -1 s (10.08 Mm -1 per unit change of RH (in %) per second), which is significant, especially under clean conditions. An RH sensor (model HYT939, B+B Thermo-Technik GmbH, Donaueschingen, Germany) sensor recorded the RH of the sampled air downstream of the drier. particles are spherical. The Mie-model applied here fulfilled three major tasks. First, it was tested to what extent it can reproduce measured σabs(λ) with the given constraints. Second, it was compared to lidar-based σbsc(λ) and σext(λ) based on airborne in-situ measurements accounting the ambient RH. Third, it derived LR(λ) at ambient aerosol conditions to examine the LR-RH dependence.
Regarding the mixing state of the aerosol, three different approaches are considered in the scientific community: 1) external mixture, in which each aerosol compound is represented by its own PNSD, 2) internally homogeneous mixture, with homogeneously mixed aerosol compounds within the aerosol particles, and 3) the internal core-shell mixture, in which a core 395 of a specific compound, like sea salt or light-absorbing carbon, is surrounded by a shell of, e.g., organics or inorganic salts.
Regarding internally mixed aerosols, Ma et al. (2012) have shown that for the aged aerosol conditions at Melpitz, the coreshell mixing model usually is the better representation of the internally mixed approaches to estimate the aerosol optical properties. Rose et al. (2006) furthermore have shown that the number fraction of externally mixed soot aerosol particles at 80 nm diameter is rather low in Melpitz, indicating a majority of internally mixed aerosol particles at this size-range. The study 400 of Yuan et al. (2020), conducted at Melpitz observatory, has shown coating thicknesses of several tens of nm of BC cores with a diameter of about 200 nm estimated for February 2017. Based on these findings, the core-shell internal mixture model was utilized in this study to calculate the aerosol optical properties for both campaigns. We assumed that the aerosol particles consist of a non-water-soluble core of light-absorbing carbon and a shell of water-soluble, non-absorbing material. However, it must be mentioned that in general the mixing of aerosol particles is rather complex and a more sophisticated approach would 405 be to consider mixtures of aerosol particle populations. For instance, a mixture could be a combination of homogeneously mixed aerosol particles containing no BC, and aerosol particles containing a light absorbing BC core surrounded by a shell of inorganic salts, organic material, or something else. However, the number fraction of both populations would remain unclear.
Similar to Düsing et al. (2018), the Mie-model used the aerosol particle diameter and number concentration, extracted from the dried-state PNSD, the aerosol particle core diameter, and the complex refractive index of the aerosol particle core 410 and shell as input parameters to derive the aerosol particle optical properties in the dried state. The aerosol particle core diameter Dc was calculated with: where fv,eBC is the volume fraction of eBC and was assumed to be constant over the entire size-range. The volume fraction of the eBC particles was estimated as described in Section 2.1.1. Due to a lack of airborne chemical composition measurements, 415 we assumed that the chemical composition derived on ground was representative for the planetary boundary layer in both campaigns.
Within the model, an additional optional module calculated the aerosol optical properties in the ambient state. This module required additional information about the aerosol and environment, like its hygroscopicity parameter κ, and the ambient temperature T and RH. At first, the module simulated the hygroscopic growth of the aerosol particles utilizing the semi-empirical parameterization of Petters and Kreidenweis (2007). For this, the in Sect. 2.1.1 introduced κ-estimates from the ground in-situ measurements were utilized. In a second step, it estimated the volume fraction of water of each aerosol particle based on these hygroscopic growth simulations.
Following Ma et al. (2014) and references therein, the complex refractive index of water-soluble compounds was set to be 1.53 + 1e-6i, with a 0.5% uncertainty of the real part and 0% of the imaginary part, respectively. The water-insoluble 425 light-absorbing (eBC) compounds were estimated to have a wavelength-independent complex refractive index of 1.75 + 0.55i, with a 4% and 6.6% uncertainty, respectively. This approach leads to inaccuracies especially for calculating σabs(λ) since the complex aerosol refractive index depends on the wavelength. Bond and Bergstrom (2006), e.g., recommended a complex refractive index of BC at 550 nm of 1.95 + 0.79i at 550 nm whereas Moteki et al. (2010) reported values of 2.26 + 1.26i at 1064 nm. 430 Also, only BC was considered, whereas brown carbon (BrC), usually organic material and hence part of the particle shell, was not. But, BrC is especially effective in light absorption at lower wavelengths, whereas the contribution of BC to σabs(λ)decreases towards lower wavelengths. A brief discussion of the spectrally resolved Mie-based σabs(λ) follows in Sect.
4.2.1. Hale and Querry (1973) provided the complex refractive index of water (liquid; 25°C). Following this publication, 435 the mean (± standard deviation) of the real part of the complex refractive index of water is 1.33 (± 0.0043) in the range from 0.3 to 1.0 µm wavelength. The imaginary part is negligibly small (4.5e-7) in this wavelength range. Hence, the complex refractive index of water was set to 1.33 + 0i with an assumed real part uncertainty of 0.5%. At ambient state, the complex refractive index of the aerosol particle shell was derived based on the volume weighted ZSR mixing rule of the complex refractive index of the water-soluble components and the additionally added water. Although the sampled aerosol was dried, 440 it always contained a small amount of residual water, which is negligible for the hygroscopic growth calculations. In the Miemodel, each estimate of the aerosol optical properties was derived with a Monte-Carlo approach with n = 50 runs. Bevor each run, the input parameters were varied according to their uncertainty with a Gaussian normal-distribution or an uniformdistribution when the Gaussian normal-distribution creates physically unreasonable input parameters, e.g., a negative volume fraction of eBC, or negative ambient RH. Table 1 summarizes the input parameters of the Mie-model with the uncertainties 445 and the underlying distribution for the variation within the Monte-Carlo approach.  the CAPS (see Figure 4). Considering the correlation with the ground-based in-situ measurements of σsca(450 nm), the model agrees within 3% during the summer campaign (underestimation, Figure 3a)) and within 13% (overestimation, Figure 3b)) during the winter period. Based on the correlation in Figure 3, the Mie-model reproduced the σabs(λ) derived with the MAAP 455 at 637 nm within 8% (Figure 3b)) during winter, and within 18% (Figure 3a)) during the summer period overestimating the measured σabs(λ) in both cases. In the summer case, two distinct clusters in the σabs(λ), one above and one below the fitting line, indicating different aerosol types and that the model constraints might represented the prevalent aerosol type of lower cluster better since the data points are close the 1:1 line. The aerosol represented by the lower cluster was prevalent at Melpitz from 13 June 2015 on and 465 the comparison of the modeled and measured σext(λ) (σsca(λ)) has shown an agreement within 4% (2%). Therefore, the mixing approach within the model is a good representation of the aerosol the intensive period of the measurement campaign in summer between 15 June and 28 June 2015. However, the model utilized rough assumptions to represent the aerosol. Besides the assumption of a wavelengthindependent complex aerosol refractive index, the assumption of a constant volume fraction of eBC resulted in an 470 underestimation of the BC content in the smaller aerosol particles and led to an overestimation in the larger aerosol particles, because BC usually is largely found in the aerosol accumulation and Aitken mode (Bond et al., 2013) with a mass peak at around 250 nm of BC core diameter. Also, the coating thickness of same-sized soot cores is not constant and the size of BC cores covers only a certain size-range as shown by Ditas et al. (2018). No size-resolved BC mass concentration measurements have been available during the summer campaign, and would also be limited to a certain size-range. Therefore, the 475 implementation of a constant eBC volume fraction within an optical model is a handy approach and is often used in other studies (e.g., Düsing et al., 2018.

485
However, considering the airborne in-situ correlation, the model agrees to measured σext(630 nm) within 13% (slope = 1.13 with R 2 = 0.98; p = 0) averaged over all available data points of all conducted flights. But, the modeled σext(630 nm) overestimate the measured one especially on June 25 (light blue data points). Excluding that day from the correlation, the model would overestimate the measured σext(630 nm) by 2.2% (R 2 = 0.98), which is within the measurement uncertainty of the CAPS. Note that for the airborne in-situ correlation, the underlying airborne PNSD used in the Mie-model 490 was not corrected for diffusional and aspirational loss, because both systems were sampling through the same inlet system. In winter, the altitude corrected PNSD measured at ground which was used to replace of the missing aerosol particle size range (up to 300 nm) was, however, corrected for the diffusional losses inside the tubing. Diffusional losses inside the tubing of the balloon platform lower the in-situ measured σabs(λ). Therefore, the in-situ measured σabs(λ) would have been smaller than modeled ones by default. To which extent, however, remains unclear. 495 Nevertheless, the agreement of both approaches, Mie modeling and in-situ measurements, at ground and airborne implies that the model constraints provide a good representation of the "real" aerosol properties, at least in the dried state with the limitation of a MAC(637 nm) applied to all considered wavelengths.    Panel 5 of Figure 6 displays the spectrally resolved modeled LRMie(λ) and the LR(λ) with the given uncertainty-range reported by Mattis et al. (2004). In the lowermost 1200 m LRMie(λ) was relatively constant and the RH did increase from ground to 1200 m from around 50% to 70%. The impact of the RH on the LR(λ) was small due to small hygroscopic growth of the aerosol particles in this RH range. Under these conditions, the mean LRMie(λ) was 54 sr at 355 nm and 532 nm, respectively. This mean LRMie(λ) is in the range of reported LR(λ) for urban haze aerosol reported by Müller et al. (2007) and Mattis et al. 575 (2004) and is reasonable considering also the LR(532 nm) of polluted dust aerosol of 60 sr reported by Omar et al. (2009. The anthropogenic influence (urban, polluted) is indicated by a larger meBC compared to June 17, and 28 (see Figure 5). The mean    Table 2). The model calculated significantly lower (42.9% to 35.9%) σext(λ) in the ambient state than derived with the lidar using the LR(λ) of Mattis et al. (2004). 590 We assume that the LRs for urban haze aerosol reported by Mattis et al. (2004) might not apply to that day. The the LRMie(λ) displayed in the fifth panel to σbsc,lid(λ), the slope of the linear fit of modeled and the lidar-based σext(λ) was much closer to 1 and the agreement was within 12.9% (underestimation of 7% at 1064 nm, 7.9% at 532 nm, 5.2% at 532 nm near-600 field channel, and 12.9% at 355 nm).
Averaged over all four investigated flights, the Mie-model calculated lower optical coefficients than derived by the lidar. Table 2   near-field channel, and 9.2 (±3.6)% lower at 1064 nm; the modeled σext(λ)was 25.2 (±2.1)% lower at 355 nm, 13.6 (±2.9)% at 532 nm, 12.9 (±3.4)% at 532 nm near-field channel, and 28.9 (±3.9)% lower at 1064 nm. Ferrero et al. (2019) have shown that unaccounted dust has a significant impact on the modeling of σbsc(λ). Their Mie-calculations have been 72% to 39% lower than the corresponding lidar measurements without considering dust. After considering the 45% of unaccounted PM10 mass as dust, the modeled results agreed with the lidar measurements (37% overestimation at 355 nm, and within 7% at 532 nm and 615 1064 nm) and increased the intensity of the scattered light at 180° significantly. In this study we do not consider dust or any other crustal material within the chemical composition. Hence, the missing dust and crustal material could explain the underestimation of the Mie-model.
Another reason could be an underestimation of the aerosol hygroscopicity and hence an underestimation of the aerosol particle growth resulting in a lower simulated extinction and backscatter cross-section of the aerosol particles in ambient state. 620 As stated by Wu et al. (2013) evaporation of NH4NO3 within the VH-TDMA system can occur and therefore the hygroscopicity is underestimated compared to size-segregated hygroscopicity estimates based on chemical composition measurements. Also, as shown by Rosati et al (2016b), the variation in temperature and RH can have an influence on the apportionment of ammonium nitrate which has a κ of 0.68 (see Appendixtable 1). A lower temperature at higher altitudes results in less evaporation and thereby to a larger volume fraction of ammonium nitrate and hence to a larger hygroscopicity in that altitude. 625 Furthermore, De Leeuw and Lamberts (1986) have showed that σbsc(λ) is sensitive to a) the refractive index and b) covered size-range. At a size-constant imaginary part of 0.05 the variation in σbsc(λ) for a real part of 1.4 to 1.6 is almost one order of magnitude. At a real part of 1.56, they have shown that increasing the imaginary part from 10 -3 to 10 -1 decreases σbsc(λ) by one to two orders of magnitude. Since the imaginary part is mainly driven by the BC content within the aerosol, an overestimation of the BC mass would result into a larger imaginary part of the refractive index and hence to a σbsc(λ) which 630 would be too small. Also, they stated, extending the covered aerosol particle diameters to more than 32 µm significantly increases both extinction as well as backscatter. They also showed that σext(λ)is in general less sensitive to the imaginary part https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. complex refractive index compared to σbsc(λ). However, the real part is important and the aerosol particle light extinction increases with increasing real part. Thereby, the increase is larger the smaller the wavelength is. Hence, a) non-captured aerosol particles larger than the observed size-range could led to larger σbsc(λ) and σext(λ), and b) the constant complex aerosol 635 refractive index over all wavelengths and for all particle sizes could also had an influence on the results. However, the bulk chemical composition approach has shown good agreements with the in-situ scattering measurements on groundat least at 450 nm. A wavelength-dependent complex refractive index of the aerosol components could improve the agreement.
Furthermore, the approach of correcting the airborne PNSD with the OPSS-based altitude correction factor fh might underestimates dN/dlogDp in higher altitudes which would result into lower modeled optical coefficients than observed with 640 the lidar. Ma et al. (2012) has already shown, that a mixture of fully externally and internally core-shell mixed aerosol containing light absorbing carbon is a better representation to derive the hemispheric aerosol particle light backscattering coefficients (HBF) and they reported a mass fraction of fully externally mixed light absorbing carbon of 0.51 (±0.21) for in the North China Plain for July 12 to August 14, 2009. With fixed refractive indices of the aerosol components (1.8 + 0.54i for 645 light absorbing carbon and for the less absorbing components 1.55 + 1e-7i) and constant volume fractions for the whole observed particle size range, they have shown that the core-shell approach overestimates the measured HBF at 450 nm by around 10% and underestimates the measured HBF by about 5% at 700 nm wavelength. Although HBF is not σbsc(λ), these results show that the constant mixing approach in this study might led to biases in the modeled aerosol optical coefficients.

RH dependence of the LR(λ) 650
Based on the four measurement flights during the summer campaign, the LR(λ) dependence on the RH have been examined. The winter cases have been excluded in this analysis because the underlying measurements were, although basically based on airborne in-situ measurements, different in a) the underlying hygroscopicity estimates, and b) the measured aerosol particle number size distribution.
The fifth panel of Figure 6 and Figure 8 displays the Mie-based ambient state LR(λ) at the given wavelengths (dots 655 with error bars) and the reference LR(λ) of Mattis et al. (2004), represented by the color-coded vertical lines with the given uncertainty range marked as dashed lines around these. The mean LR(λ) of flight 26a calculated with the Mie-model in the ambient state was 64.1 (±14.1) sr at 355 nm, 61.7 (±10.9) sr, and 36.2 (±8.0) sr at 1064 nm which is 10.5% larger, 16.4% larger and 19.6% lower than the corresponding LR(λ) reported by Mattis et al. (2004) but in the given range. The vertical structure of LRMie(λ) did follow the trend of the RH. 660 Previous studies reported a significant influence of the RH on the aerosol optical properties often expressed with an enhancement factor. Zieger et al. (2013), e.g., presented the aerosol particle light scattering enhancement for different European sites, Skupin et al. (2016) published a four-year-long study on the impact of the RH on the aerosol particle light extinction for Central European aerosol, and Haarig et al. (2017) showed the backscatter and extinction enhancement for marine aerosol. Ackermann (1998) investigated the dependence of the LR(λ) on RH for different aerosol types with a numerical simulation, 665 but has not presented a LR(λ) enhancement factor and the underlying PNSD were solely based climatology data and not based on actual measurements like within this study. Following the approach of Hänel (1980) the RH-and wavelength-dependent enhancement factor of the LR(λ), fLR (RH, λ), is expressed with: where fLR,dry is equal to fLR(RH = 0, λ), the LR(λ) enhancement factor at 0% RH and is forced through 1. γ(λ) denotes the 670 wavelength dependent fitting exponent. https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License.

translated into the lidar ratio enhancement factor is displayed as solid lines.
The estimated fLR(RH, λ) for the four investigated measurement flights (17b, 26a, 28a, 28b) is displayed in Figure 9 and Table 3 shows the corresponding fitting parameters with the standard errors of the fit. Note that the "dried state" LR(λ) was calculated for aerosol with some residue water, because the sampled aerosol was never completely dry. The RH measured 680 after the dryer was at most 48.3% on flight 20150617b and reached a maximum of 35.8% on the other days. In the Mie-model the aerosol particles in dried state were treated as completely dry. However, the growth in size of the aerosol particles at this RH level is small (around 10%) and the bias on the LR(λ) enhancement estimates should be negligible small.
The LR(λ) enhancement factor shows a clear dependence on the ambient RH with an expected enhancement factor of around one at low RH. The observed trend follows the results reported by Ackermann (1998) (solid lines in Figure 9) for 685 continental aerosol but with larger quantities especially at larger RH. Also, the aerosol sampled in this study resulted in a LR(λ) enhancement factor of up to 3.7 at 532 nm and up to 2.4 (2.2) at 1064 nm (355 nm) at 93.7% RH. The power series representation of Ackermann (1998)  A predominant aerosol type at that day, which was different to the other shown days, is assumed to be the reason of a different LR(λ) enhancement factor behavior. γ(532 nm) is significant larger than γ(355 nm) and γ(1064 nm), respectively. The data-points sampled under ambient conditions of 60% to 80% RH are overrepresented in the fit. Furthermore, Mie calculations (settings: fv,eBC = 0.03, κ = 0.3, 695 T = 20°C, core-shell mixture), conducted on the basis of the PNSD measured at Melpitz Observatory during June 26, 2015, have shown that in this RH range the LR(532 nm) gets more enhanced than the LR(1064 nm) or LR(355 nm) and might be a typical feature of the predominant aerosol or results from the model constraints. Similarly, in the results of Ackermann (1998) the LR-to-RH dependence for continental aerosol was not following the exponential curve perfectly. Also, LR(λ) for marine aerosol is more enhanced at this RH range as reported by Ackermann (1998). The fit for 532 nm at this RH range, therefore, 700 might was over-weighted which might led to an overestimation of γ(532 nm). Also, at 355 nm Ackermann (1998)  The results are opposed to the findings of Takamura and Sasano (1987), showing a negative correlation of LR(λ) and RH at 355 nm and a small dependence of the LR(λ) on the RH at larger wavelengths. This might be caused by their different 705 analysis approach since Takamura and Sasano (1987) used PNSDs inferred from angular light scattering measurements of a polar Nephelometer including more uncertainty-increasing processing steps. Also, their Mie calculations were based on PNSD estimates at different RH levels with assumed homogeneously mixed aerosol particles with an effective complex refractive index at ambient state. Contrary, our investigations based on hygroscopic growth simulations and a core-shell mixing approach.
Furthermore, the limited covered size-range of the aerosol particle hygroscopicity might introduces some bias in our results 710 since the κ(Dp) estimates above 265 nm are maybe too large or too small, which would have an impact on the Mie-model results, especially on σbsc, which is more sensitive to the complex aerosol refractive index than σext(λ).
Nevertheless, the presented results provide good first estimates of the RH-induced LR(λ) enhancement factor based on in-situ measured PNSD for the observed RH range. Although Ackermann (1998), already, has shown the LR-to-RH dependence for three different aerosol types (marine, continental, dessert dust), future research should collect more data to 715 provide fLR(RH, λ) with the corresponding γ(λ) estimates including a separation into different aerosol types.
Future research should investigate the impact of the mixing-state and hygroscopic growth factor representation within the Mie-model on the lidar ratio enhancement factor as well.

MelCol-winter
Data representing another season with different atmospheric conditions was collected and evaluated for the winter of 2017.
Exemplarily, the data of two measurement days within winter 2017 is discussed in the following.

Aerosol Particle Light Absorption
During winter, two balloon launches during different levels of pollutions were conducted. This part focuses on the evaluation 745 of the model with airborne in-situ measurements in a dried state. The corresponding atmospheric conditions are shown. The findings provide insights to, e.g., evaluate σabs(λ) derived from lidar with similar setups.

775
The profiles of the Mie-modeled and measured σabs(λ) in dried state conducted on February 9 and March 9, 2017, are shown in the last panel of Figure 11 and Figure 12. The linear fit and the corresponding fittings are displayed in Figure 13, Figure 14, fitting parameters are given in Table 4.
(10) 785 The AAESTAP(624 nm, 450 nm) was 1.67 ± 0.14 on average within the lowermost 700 m on February 9, and was slightly larger than the daily mean AAEAE33(660,450 nm) of 1.49 (±0.08 standard deviation of mean) derived from parallel conducted, spectrally resolved, σabs(λ)measurements of an Aethalometer at Melpitz (model AE33; Magee Scientific, Magee Scientific, Berkeley, CA, USA). For March 9, 2017, we could not compare the AAE since the AE33 was stopping its measurements on As shown in Figure 3b), in the winter period, the Mie-model simulated on average around 8% larger σabs(637 nm) than measured by the MAAP. For the airborne measurements, the assumptions within the Mie-model to derive σabs(λ)in the 795 dried state led to a 31.8 (±1.5%), 24.7 (±1.7%) and 13.2 (±1.7%) underestimation at 450 nm, 525 nm, and 624 nm respectively on February 9. On March 9, 2017, a 32-37% overestimation of the airborne measured σabs(λ)was observed (see Figure 13, Figure 14; corresponding profiles in Figure 11 and Figure 12). This indicates a spectral dependence.
At ground, the Mie-simulation based on the aerosol microphysical measurements calculated a σabs,Mie(630 nm) on small as a result of probably too small mEC measurements. However, we considered EC as eBC, which could have led to some bias in the MAC(637 nm) estimate as well. In particular, on February 9, a MAC(637 nm) of 10.9 m 2 g -1 was derived, on March 805 9, a small MAC(637 nm) of 6.6 m 2 g -1 . The time-series of the MAC(637 nm) estimates are displayed in Appendixfigure 1. Zanatta et al. (2018) and Yuan et al (2020), e.g., have shown that the mixing of BC is an important parameter influencing directly the value of the MAC(λ). They reported MAC(λ) for pure externally mixed BC aerosol particles. For Melpitz, during the winter period of this study and applying an AAE of 1, the MAC(870 nm) of 5.8 m 2 g -1 reported by Yuan et al. (2020) translates into 7.9 m 2 g -1 at 637 nm. With an AAE of 1, modeled MAC(550 nm) for pure BC particles reported by 810 Zanatta et al. (2018). translate into very small 3.5 m 2 g -1 to 5.7 m 2 g -1 at 637 nm depending on the particle size. Nevertheless, the MAC(637 nm) on February 9, coincided with the estimates of Yuan et al. (2020). Therefore, on February 9, 2017, σabs,Mie(624 nm) and σabs,STAP(624 nm) agree reasonably well within 13.2% since a MAC estimated at 637 nm represents 624 nm reasonably well.
The core-shell mixing representation within the model was not applicable to the aerosol on March 9, because a 815 MAC(637 nm) in the range of the estimates of Yuan et al. (2020) and Zanatta et al. (2018) indicates external mixture rather than an internal core-shell mixture. The larger MAC(637 nm) on February 9, on the other, hand suggest a good representation of the mixing state of the prevalent aerosol.
The spectral dependence of the over-and underestimation for both days can be explained with the AAE. Within the lowermost 700 m above ground, a median AAEMie(624 nm, 450 nm) of 0.94 was found; on February 9, and of 1.05 on March 820 9, respectively. The corresponding median AAESTAP(624 nm, 450 nm) of 1.64 on February 9, and of 1.20 on March 9, clearly indicated a significant amount of BrC aerosol particles according to Zhang et al (2020). The AAE of BC is near unity at visible and near-infrared wavelengths (e.g., Kirchstetter and Thatcher, 2012) but also can go as high as 1.6 when BC is coated with transparent material as stated by Cappa and Lack (2010). The values of AAEMie(624 m, 450 nm) of around 1 agree with these findings. AAESTAP on both days, and AAEAE33 on February 9 indicated the presence of BrC. BrC contributes less to the 825 absorption at near-infrared wavelengths with increasing contribution to the aerosol particle light absorption towards UV wavelengths (e.g. Kim et al., 2020 andSun et al., 2007). The daily mean volume fraction of organic material detected by the Q-ACSM on February 9 was 45.1% peaking at around 50% during the flight time. On March 9 during flight time, a volume fraction of 34.4% was found with values as small as 17% in the morning hours. The small volume fraction (March 9) had less of an impact on the Mie-model and led to the smaller spectral dependence of the overestimation. The larger volume fraction 830 on February 9, on the other hand, indicated a large content of BrC and hence a larger spectral dependence of the deviation.
To summarize, for March 9, it is more likely that a combination of the aerosol mixing representation within the model as well as the possibly too small MAC(637 nm) led to the overestimation by the model rather than the missing BrC. For February 9, the agreement within 13.2% at 624 nm indicated that the MAC(637 nm) represented the prevalent aerosol within a satisfying range, the missing BrC content within the model, however, resulted into a larger spread in the underestimation. 835 The mixing approach within the model seemed to have better represented the aerosol present on February 9.
In conclusion, that future studies should a) consider the mixing state of the aerosol or at least include this in the uncertainty analysis, and b) should include BrC with a spectral resolved MAC(λ).

Aerosol particle light backscattering and extinction coefficient
Besides the in-depth view on the σabs(λ), also a comparison of the lidar estimates of the σbsc(λ) and σext(λ)was conducted and 840 is shown below.  The σbsc(λ) and σext(λ)are displayed in panels three and four of Figure 11 and Figure 12 for February 9, and March 9, 2017. The Mie-modeled coefficients are represented by dots with the three times standard deviation of the mean of the  calculation, the lidar estimates as lines with the corresponding color.
Panel one and two of Figure 13 and Figure 14 display the correlation of the modeled and measured σbsc(λ) and σext(λ) shown in Figure 11 and Figure 12 (panel three and four in each), correspondingly. The linear fit estimates, the corresponding standard error of fit and correlation coefficients are given in Table 4. Note that the shown fit of Figure 13 (Figure 14) is forced through the coordinate origin which artificially enhances the coefficient of determination R 2 . The fits have been forced through zero 855 since a) the range of the values of the observed optical coefficients was small and b) because both model and measurements rely on the present aerosol and if no aerosol is prevalent both, model and observation, should be zero. Therefore, results of R² should be considered with care.
For February 9, over all considered wavelengths, and field-of-view configurations of the lidar, the model results agreed within 21.2% to 37.8% (21.2% at 1064 nm to 37.8% at 523 nm; R 2 close to 1 in all cases) with the measured σbsc(λ). 860 The modeled σext(λ) were up to 30.5 (±1.8)% (at 1064 nm) lower than those derived based on the lidar measurements with a mean underestimation of 18.3 (±0.8)%. The approach of correcting the lower aerosol particles with the altitude correction factor might underestimated the aerosol particle number concentration of particles up to 300 nm. In Mie-theory, particles with about the same size of the incoming radiation wavelength are most efficient in scattering. In the study of Virkkula et al. (2011), aerosol particles in the range of 100-1000 nm contributed most to the aerosol particle light scattering at 550 nm. Therefore, at 865 355 nm an artificial under-sampling of the aerosol particles up to 300 nm in diameter induced by the altitude correction factor could have led to an underestimation in the modeled aerosol particle light scattering and thus extinction. Also, the Mie-model, as well as the correction of the OPSS, did not consider aspherical particles which could have led to a bias induced by the PNSD. Also, the wavelength-independent complex aerosol refractive index and probably, at this time present, non-captured, huge particles, as discussed already in the summer part, could explain some of the deviations. However, all modeled σext(λ) 870 were within the range of the aerosol particle light extinction coefficients calculated with the minimum and maximum LR(λ) provided by Mattis et al. (2004).
The fifth panel of Figure 11 shows the LR(λ) with the range-bars indicating the minimum and maximum value of the result of the ambient state Mie modeling. Like in the summer cases, a clear connection between the increase of the LR(λ) and the increase of the RH was significant: with increasing RH the LR(λ) increased. Overall, the average LR(λ) in the shown profile 875 was 63.8 sr at 355 nm, 69.0 sr at 532 nm, and 37.6 sr at 1064 nm, which was in the range of the LR(λ) reported by Mattis et al. (2004) except for the LR(532 nm) at 532 nm which was 7.8% larger than the maximum reported LR(532 nm). However, these and σsca(λ), overestimating the in-situ measured σabs(λ). However, the hygroscopic growth and the refractive index of the aerosol particles, estimated by their chemical composition, might have been inaccurate. Nevertheless, most of the modeled σext(λ) matched with the lidar estimates within the range of the LR(λ) estimates of Mattis et al. (2004). Except above 450 m 890 altitude and 355 nm wavelength, where the modeled σext(λ) was significant smaller than the lidar estimates, which indicated an underestimation of the aerosol particle number concentration at this altitude and size-range due to an inaccurate altitude correction factor of the PNSD.    To summarize, the Mie-model reproduced a σext(λ) at ambient state closer to the lidar estimates at the more polluted case, whereas the in the clean case the underestimation was larger. In the case of σext(λ), no spectral trend was observed in terms of agreement indicating a bias induced by the PNSD rather than the by the complex aerosol refractive index. At 1064 nm, also, the Mie-model results were closest to the measured σbsc(λ). That might be a hint, that the correction approach of utilizing an altitude correction factor for the ground in-situ PNSD measurements was not able to reproduce the PNSD aloft of Melpitz, 910 at least in the lower size-ranges. Equivalent to the summer cases, also the findings of De Leeuw and Lamberts (1986) and Ferrero et al. (2019) may provide some explanation for the observed results. However, both, modeling and lidar estimates, underlay uncertainties so that not only the modeled results could have been too small, also the lidar estimates could have been too large, especially in the extinction where the LR(λ) is subject to a large uncertainty range. The underlaying reasons are speculative and many parameters within the model can be varied. For σbsc(λ) and σext(λ), we do not suspect that the missing 915 BrC within the model would result into significant different results. However, considering the limitations of the measurements setup, e.g., the limited covered size-range and no vertical resolved chemical composition measurements, the results are promising.

Summary and Conclusion
This study presented the comparison of lidar estimates of σbsc(λ) and σext(λ) with airborne in-situ measurement-based modeled 920 ones and examines the effect of the RH to the aerosol particle light extinction-to-backscatter ratio. Also, it evaluated modeled σabs(λ) with airborne measured ones in a dried state to determine whether the presented model can be utilized to evaluate lidarbased aerosol particle light absorption estimates. For this purpose, the results of two field campaigns carried out near Melpitz conducted in the summer of 2015 and February/March, 2017, covering different states of aerosol load, were utilized. In the two campaigns, two different airborne systems were deployed to carry out aerosol in-situ measurements complemented by a 925 https://doi.org/10.5194/acp-2021-21 Preprint. Discussion started: 4 March 2021 c Author(s) 2021. CC BY 4.0 License. set of state-of-the art ground-based in-situ instrumentation as well as by a polarization Raman-lidar system directly measuring the aerosol particle light backscattering coefficient at three wavelengths. In this study a height-constant LR(λ) was utilized to derive aerosol particle light extinction profiles from aerosol particle light backscattering profiles derived by the lidar system.
The in-situ measurements were used to calculate aerosol optical properties using Mie-theory. A core-shell mixture of the aerosol particles was assumed. The chemical composition of the aerosol particles measured on the ground was set constant 930 over all particle sizes and was assumed to be representative for all altitudes above ground. The model validation under dry conditions confirmed the underlying assumptions with modeled values matching the in-situ measurements within 18%. An additional module of the Mie-model calculated the aerosol optical properties in ambient state utilizing a hygroscopic growth simulation based on Kappa-Köhler theory. In both campaigns the airborne-based PNSD was extended with height-extrapolated ground-based in-situ PNSD measurements. 935 Mie-model results and lidar measurements were compared with each other. In the summer case, the Mie-model calculated aerosol optical coefficients up to 32% lower than the lidar estimates, in the winter campaign they have been up to 42% lower. In both, the summer and winter campaign, a spectral dependence in the slope of the linear fit of the modeled and measured σbsc(λ) was observed, whereas in σext(λ) not. This agrees with previous studies who have shown that σext(λ) (major fraction is σsca(λ)) is less sensitive to the complex aerosol refractive index than σbsc(λ) and is more driven by the PNSD. The 940 results were promising, since the σbsc(λ) especially requires a very precise determination of the aerosol state in terms of PNSD and chemical composition (refractive index and mixing state).
In the winter campaign, the Mie-model result was directly compared to the filter-based airborne in-situ σabs(λ) measurements. In the more polluted case, the Mie-model derived up to 32% lower σabs(λ) with the best agreement at 624 nm wavelength and a showed a distinct spectral dependence of the agreement. In the cleaner case, the Mie-model calculated up to 945 37% larger σabs(λ) with a small spectral dependence. The results indicated that the mixing-state of the aerosol, the wavelengthdependent complex refractive index of the aerosol compounds, as well as the BrC content, must be accurately represented by the model to match the measured σbsc(λ) within a narrow uncertainty-range.
Utilizing a height-constant LR(λ) is widely applied to determine σext(λ) from σbsc(λ)and the modeled LR(λ) shown here are in the range of LR(λ) estimates presented by previous studies for different aerosol types. In both campaigns, the Mie-model 950 ambient state calculations, however, revealed a dependence of the LR(λ) to the ambient RH and resulted in a RH and wavelength-dependent LR(λ) enhancement factor ( , ) = ( = 0, ) × (1 − ) − ( ) , with ( = 0, ) forced through 1. Estimates of γ(λ) were derived based on the summer campaign data-set.
In conclusion: a) Conducting closure studies of optical aerosol properties requires a precise determination of the aerosol mixing state, 955 its composition, the inclusion of BrC, and the application of a wavelength-dependent complex refractive index. b) Airborne in-situ measurements of, e.g. the aerosol chemical composition including the BrC content, would provide improvements in such studies and would allow to validate lidar-based σabs(λ). c) A wide range of aerosol particle sizes was covered within this study. However, the modeled σbsc(λ) was lower than the measured one. A much further extension of the observed aerosol particle size-range beyond 10 µm would ensure 960 that this parameter would not cause such an underestimation based on the finding of the De Leeuw and Lamberts (1987). d) Knowing the connection between RH and the LR(λ), the LR(λ) enhancement factor would be a useful tool to estimate the LR(λ) at ambient state, when the dry state LR(λ) is known. Also, it allows to calculate back the LR(λ) in dry state, when the LR(λ) is directly measured in ambient state and a RH profile is known, e.g. by radio soundings. 965 e) However, long-term measurements must be conducted to verify the LR(λ) enhancement estimates for various aerosoltypes as well as different seasons. Data set and source codes underlying this work can be requested via email to the corresponding author.

Authors contribution.
The authors SD, BW, AA, and HB were responsible for the conceptualization of the study. Data curation, investigation, and 985 the development of the methodology was done by SD. Further, for the study needed, data was provided by CD (