Relative errors of derived multi-wavelengths intensive aerosol optical properties using CAPS_SSA, Nephelometer and TAP measurements

Aerosol intensive optical properties like the Ångström exponents for aerosol light extinction, scattering and 10 absorption, or the single-scattering albedo are indicators for aerosol size distributions, chemical composition and radiative behaviour and contain also source information. The observation of these parameters requires the measurement of aerosol optical properties at multiple wavelengths which usually implies the use of several instruments. Our study aims to quantify the uncertainties of the determination of multiple-wavelengths intensive properties by an optical closure approach, using different test aerosols. In our laboratory closure study, we measured the full set of aerosol optical properties for a range of light15 absorbing aerosols with different properties, mixed externally with ammonium sulphate to generate aerosols of controlled single-scattering albedo. The investigated aerosol types were: fresh combustion soot emitted by an inverted flame soot generator (SOOT, fractal aggregates), Aquadag (AQ, spherical shape), Cabot industrial soot (BC, compact clusters), and an acrylic paint (Magic Black, MB). One focus was on the validity of the Differential Method (DM: absorption = extinction minus scattering) for the determination of Ångström exponents for different particle loads and mixtures of light-absorbing aerosol 20 with ammonium sulphate, in comparison to data obtained from single instruments. The instruments used in this study were two CAPS PMssa (Cavity Attenuated Phase Shift Single Scattering Albedo, λ = 450, 630 nm) for light extinction and scattering coefficients, one Integrating Nephelometer (λ = 450, 550, 700 nm) for light scattering coefficient and one Tricolour Absorption Photometer (TAP, λ = 467, 528, 652 nm) for filter-based light absorption coefficient measurement. Our key finding is that the coefficients of light absorption σap, scattering σsp and extinction σep from the Differential Method agree with data from single 25 reference instruments, and the slopes of regression lines equal unity within the precision error. We found, however, that the precision error for the DM exceeds 100% for σap values lower than 10-20 Mm for atmospheric relevant single scattering albedo. This increasing uncertainty with decreasing σap yields an absorption Ångström exponent (AAE) that is too uncertain for measurements in the range of atmospheric aerosol loadings. We recommend using DM only for measuring AAE values for σap > 50 Mm. Ångström exponents for scattering and extinction are reliable for extinction coefficients from 20 up to 1000 30 Mm and stay within 10% deviation from reference instruments, regardless of the chosen method. Single-scattering albedo (SSA) values for 450 nm and 630 nm wavelengths agree with values from the reference method σsp (NEPH)/σep (CAPS PMSSA) https://doi.org/10.5194/amt-2021-284 Preprint. Discussion started: 27 September 2021 c © Author(s) 2021. CC BY 4.0 License.

with less than 10% uncertainty for all instrument combinations and sampled aerosol types which fulfil the defined goals for measurement uncertainty of 10% proposed by Laj et al., 2020 for GCOS (Global Climate Observing System) applications. 35

Introduction
The precise determination of aerosol optical properties is crucial for the provision of reliable input data for chemistry transport models, climate models, and radiative forcing calculations (Myhre et al., 2013). This applies in particular to light-absorbing particles like black carbon (Petzold et al., 2013), which are produced by incomplete combustion processes and absorb visible 40 light very efficiently. Aerosol light absorbing properties are also relevant for source appointment studies and the determination of anthropogenic influences on the atmospheric aerosol (Sandradewi et al., 2008) . There are two common methods to generate aerosol light absorption data for long-term and short-term measurements, each with its own disadvantages. One method is a filter-based technique, which operates by deriving light absorbing values from the attenuation of light trough particle-loaded filter (Rosen et al., 1978). A disadvantage of all filter-based methods is linked to effects like multiple scattering inside the filter 45 matrix, shadowing of light-absorbing particles in highly loaded filters, and humidity effects (Moosmüller et al., 2009). Widely deployed filter-based light absorption measurement methods are the Particle Soot Absorption Photometer (PSAP: Bond et al., 1999) and its further development, the Tri-colour Absorption Photometer (TAP: (Ogren et al., 2017), the Aethalometer (Hansen et al., 1984), and the Multi-Angle Absorption Photometer (MAAP) . Except for the MAAP, all filterbased methods require complex correction algorithms (Collaud Coen et al., 2010;Virkkula, 2010). Another method for 50 deriving light absorption coefficients is the differential method, based on the subtraction of light scattering from light extinction coefficients. This method is commonly conducted by comparing measurements from two separate instruments which results in large precision errors particularly for lower aerosol light absorption coefficients. In laboratory studies, however, the differential method is widely used as reference technique because the applied light scattering and extinction instruments are well characterised (Bond et al., 1999;Schnaiter et al., 2005;. A significant improvement of aerosol 55 measurement capacities is achieved by the recently developed Cavity Attenuated Phase Shift particle monitor for single scattering albedo (CAPS PMSSA) (Onasch et al., 2015b) which is able to measure light extinction and scattering simultaneously and is the focus of recent studies (Perim de Faria et al., 2021;Modini et al., 2021) .
Intensive aerosol parameters like the Single Scattering Albedo (SSA) or Ångström exponents are often not directly measured, but calculated from multiple instrument datasets, which could lead to an increase in errors and uncertainties concerning this 60 parameter. The importance of reliable intensive parameters is undisputable, especially, when the use of them is required for an experiment or sensitive climate related modelling. The Ångström exponents are widely used to adjust extensive parameters to a desired wavelength (Ångström, 1929); Foster et al. (2019) for instrument comparison and more importantly for aerosol characterisation (Russell et al., 2010) like the refraction index calculation of mineral dust (Petzold et al., 2009) or black carbon (Kim et al., 2015), or for source identification of mineral dust (Formenti et al., 2011). The scattering Ångström exponent (SAE) 65 is size-dependent and therefore, used as an indication of the size distribution of aerosols in the investigated medium. The SAE https://doi.org/10.5194/amt-2021-284 Preprint. Discussion started: 27 September 2021 c Author(s) 2021. CC BY 4.0 License. value of 4 indicates either a gaseous medium or a medium with nanometer-sized particles, whereas a value of 0 indicates coarse particles (Kokhanovsky, 2008). The absorption Ångström exponents (AAE) depends on the chemical composition of the aerosol. A value of 1 indicates an aerosol which absorbs light strongly across the entire visible spectral range and is composed of nanometer-sized spheres (Berry and Percival, 1986). This behaviour is characteristic for fresh soot or black carbon fractal 70 agglomerates (Kirchstetter and Thatcher, 2012;Xu et al., 2015). AAE values higher than unity indicate the presence of brown carbon (Kim et al., 2015) or mineral dust (Formenti et al., 2011), both of which are characterised by a stronger absorption in the blue and ultraviolet compared to the red spectral range. The extinction Ångström exponent (EAE) is often used for aerosol classification by remote sensing methods such as Lidar and depends on size distribution and chemical composition (Kaskaoutis et al., 2007;Veselovskii et al., 2016). Combining those exponents in a cluster plot is a reliable method for classify aerosol 75 sources (Russell, 2010). The SSA of an aerosol is the key parameter for its direct and semi direct impact on the climate (Penner, 2001). It describes the ratio of scattering to total extinction of a medium. The value of 1 indicates that light extinction relies exclusively on light scattering. In contrast, low SSA values indicate an aerosol with a large fraction of light-absorbing components, which may cause heating of the atmosphere. The intensive parameters are only available through multipleinstrument approaches at different wavelengths which calls for a detailed analysis of measurement uncertainties. Our study 80 contributes to this topic with a detailed optical closure study in which we deploy standard and advanced instrumentation for measuring aerosol optical properties and sample mixtures of light absorbing and scattering aerosol to assess method uncertainties and precision errors.

Experimental Design
In this study, we combined the use of different instruments with various aerosol types. In order to minimize instrument and measurement errors, a couple of preparations were necessary. For example, we ensured that the aerosol production is operated using constant volumetric air flow. Also Ammonium-sulphate concentrations used where not changed during the experiment. 90 Otherwise, this would change the particle size and thus the size distribution, leading to a less-well defined aerosol. Every measurement was done under ambient conditions in the lab. This was monitored by the internal pressure and temperature sensors of the nephelometer.  The aerosol production was controlled by multiple Mass Flow Controllers (MFC, Bronkhorst High-Tech B.V., Ruurlo, Netherlands). A Labview based program controlled the complete measurement system and recorded centrally all data from the individual instruments. Downstream the production the aerosol was injected in a mixing chamber assuring homogenous mixing. The mixing chamber is attached to the aerosol supply line. Several instruments are connected to the central aerosol 100 supply line where the individual instruments are connected to using an iso-axial orientated and isokinetic operated nozzle located in the centreline of the supply line. A Grimm optical particle size spectrometer (SKY-OPC, model 1.129, Grimm Aerosol GmbH & Co. KG, Ainring, Germany) was used to characterize and monitor the resulting size distribution. The particle scattering coefficient σsp was measured with a integrating multi wavelength nephelometer (Model 3563, TSI Inc., Shoreview, MN, USA) (Bodhaine et al., 1991) and by a integrating sphere used in the CAPS PMSSA monitor (CAPS PM_SSA, Aerodyne 105 Research Inc., Billerica, MA, USA; Onasch et al. (2015)). For the particle light absorption coefficient σap we used the small sized TAP (Brechtel Inc., Hayward, CA, USA) based on the well-known Particle Soot Absorption PSAP and the Continuous Light Absorption Photometer (CLAP) developed by NOAA (Ogren et al., 2017). The particle light extinction coefficient σep was directly measured with the phase shift channel of the CAPS PMSSA monitor.
All tubes and connections after the nebulizer were made of stainless steel or conductive silicone tubing to reduce particle loss 110 by electrostatic forces. The humidity rarely exceeded 7%, which was an additional parameter measured by the nephelometer, so deliquescence effects were avoided. Because all instruments were connected to one central aerosol supply line. It was necessary to reduce the air flow towards the nephelometer from 20 l/min to 2.2 l/min due to the flow limits of the aerosol production. The flow-range of the other instruments, span from 0.6 l/min to 3 l/min. Due to the reduced air flow of the nephelometer also the time resolution of the nephelometer was reduced due to the longer flushing time of around 10 minutes. 115 https://doi.org/10.5194/amt-2021-284 Preprint. Discussion started: 27 September 2021 c Author(s) 2021. CC BY 4.0 License.

Corrections and calibrations 120
The CAPS PMSSA instrument extinction channel was calibrated with polystyrene latex beads (PSL) particles as reference and Mie theory using BHMIE Python code derived from Bohren & Hoffman (1983). Additionally, the 450 nm wavelength CAPS PMSSA was calibrated with CO2 for additionally validating the same factor and the calibration was applied to the nephelometer (Anderson and Ogren, 1998;Modini et al., 2021). The scattering channel of the CAPS PMSSA using the integrating sphere method was internally adjusted to the extinction channel using ammonium sulphate as a light-scattering aerosol assuming a 125 single scattering albedo of 1. A truncation error correction was not necessary regarding the size of the aerosols used (Onasch et al., 2015a) since highest amount of aerosols were smaller than 200 nm in diameter size. The CAPS PMSSA has a drifting shift of the base line as long the system is heating up, which apparently stabilized after 30 min of operation (Faria et al., 2019).
The nephelometer (NEPH) correction for light absorbing aerosols was calculated according to (Massoli et al., 2009). Because of the reduced air flow, the nephelometer needed at least 15 minutes to reach a stable plateau after changing aerosol production 130 settings. After that, a new Filter Spot for the TAP was selected, to minimize transmission uncertainties increases by loaded filters. Data inversion for the Nephelometer was done by correction of truncation effects which alterned the data of additionally 5% maximally. These corrections were either made using the approach proposed by Anderson and Ogren (1998)  Corrections of the TAP data were made according to (Virkkula, 2010). A new filter spot was selected for each measurement. where σep is the extinction coefficient, σsp the light scattering coefficient and σap the coefficient for light absorption by particles.
Solving equation 1 for it is possible to derive the absorption coefficient by combining CAPS_SSA and nephelometer measurements for comparison. In the following this will be called Differential Method (DM). 150 To calculate the Single Scattering Albedo (SSA), the particle light scattering must be divided by the particle light extinction: The Ångström exponents AE are calculated from: 155 By solving Eq. 3 for ( 1) and assuming a valid Ångström exponent the resulting equation (3a) is used for wavelength adjustments ( 1) = ( 2) • ( 1 2 ) − Eq. (3a) 160 For the particle coefficient σxp the corresponding σsp, σep or σap could be put into calculations (Eq. 3) to obtain the absorption Ångström exponent (AAE), extinction Ångström exponent (EAE) and scattering Ångström exponent (SAE) accordingly.

Test Aerosol Generation
For every day of the experiments the solutions of Aquadag (AQ, Aqueous Deflocculated Acheson Graphite; Acheson Industries, Inc., Port Huron, MI, USA), Cabot Black (BC) or the Acrylic Paint Magic Black (MB) were prepared by ultra-175 sonication first, before nebulization in a Constant Output Atomizer (Model 3076, TSI Inc.). The resulting size for these aerosols, as well as of the atomized ammonium sulphate, depended on the concentration put into solution and the air flow rates. In order to vary the aerosol concentration with minimized sizes distribution chances, the mixture was controlled by a MFC-determined active extractive flow after the dehydration tube. The inverted flame soot generator (Argonaut Scientific Corporation, Edmonton, AB, Canada) was operated with a pre-determined propane to oxidation air ratio so that the flame 180 produced a stable and low organic carbon soot. It has previously been shown that at least 30 min were necessary to reach stable aerosol concentrations (Bischof et al., 2019;Kazemimanesh et al., 2018)  With these sets of different aerosol types and shapes, the behaviour of instrument measurement is investigated. The results of the intercomparison of Aquadag is expected to be best described by Mie theory, since its spherically shape and therefore applied correction schemes to the instruments apply best, since calibration is done by ideal PSL spheres (polystyrene latex beads), which were treated the same as all other aerosol solution samples and their size was approved by DMA and OPC.
Fractal agglomerates could have multiple internal scattering effects. Spherical shapes and several optical properties are 190 determined of the primary particle, this is expected to differ the most in intercomparison approaches (Barber and Wang, 1978;Moosmüller et al., 2009).
3.21 ± 0.08 2.03 ± 0.38 1.43 ± 0.65 0.52 ± 0.10 1.10 ± 0.10 Table 3 shows the aerosol types used along with the measured size parameters and their calculated intensive parameters. The size distribution was measured beforehand with the combination of a Differential Mobility Analyzer (DMA 5.400, Grimm Aerosol Technik GmbH Co & KG Germany) and Condensation Particle Counter (CPC 5.411, Grimm Aerosol Technik) system 200 in a sequential mode of operation. For internal calibration of the CAPS integrating sphere channel-measuring the light scattering coefficient-AS particles were used as purely scattering substance. By assuming a SSA of 1.0 the CAPS PMSSA Extinction channel is used as calibration reference.
The Ångström exponents for the pure substances are in typical ranges for these types of aerosols and size distributions reported in the literature. For example, the SAE decreases from a value of 3,22 for 40nm AS particles which is close to the SAE value 205 of 4 for air molecules with increasing particle diameter. Thus, the SAE drops to 0,76 for 130 nm AQ particles. As expected by Eq. 3a the SSA increases with shorter wavelength (Bohren and Huffman, 1983). The AAE for fractal combustion soot is close to 1 as reported by e.g. (Török, 2018) for the mini-CAST soot generator.
The errors reported are either the instruments uncertainties or are calculated from error propagation. The light extinction channel of the CAPS instrument has an uncertainty of 5% and precision of 2% and a scattering uncertainty of 8% and 2% 210 precision respectively (Onasch et al., 2015). The TAP has an uncertainty of around 8%, with a precision of 4% ( (Müller et al., 2014;Ogren et al., 2017), while the nephelometer has an uncertainty of less than 10% and a precision of about 3% ( (Anderson and Ogren, 1998)  In order to give a brief overview of the test aerosol size distributions reported by the DMA and CPC system as a function of the electric mobility size diameter in nm, Figure 2 provides the size distributions of the different aerosol types normalized to 1000 particle counts (N) per cubic centimetre. 220

Measurements
In a first step, the extensive parameters must be validated for all instrument combinations to ensure the reliability of the intensive parameters derived from them. Aquadag is well-known for its physical properties and it is easy to handle by 225 nebulising. We have selected AQ as it is commonly used as a reference material for instrument comparisons (Foster et al., 2019) for all the viewgraphs. The results for the other aerosol types are added in the associated tables 6-9. Respective data points are given as averages of at least 100 seconds of stable aerosol production.

Extensive Parameters 230
The two CAPS_SSA monitors used measure the extinction coefficient of particles directly with a small precision error of around 2% (Modini et al., 2021) for 450nm and 630nm wavelength. In Figure 3 we show scatter plots of these direct Here mixtures of nebulized Aquadag particles and ammonium sulphate particles are used as a proxy for the mixing ratio the SSA is shown as colour code. The extinction coefficients align the 1:1 line within 10% in a broad range of the extinction coefficient for 450 and 630 nm wavelength as well as for SSA of the mixtures ranging from 0.3 close to 1. This shows that the 245 instruments are not sensitive to the SSA of the particle type used for both wavelengths of interest.
As the next extensive parameter, the scattering coefficient at 450 and 630 nm wavelengths are compared using scatterplots for different techniques in Figure 4. Here we use the Nephelometer and the integrating sphere channel of the CAPS_SSA instrument capable of measuring the scattering coefficient directly. In addition we calculated the scattering coefficients using 250 a Differential Method (DM) solving Eq.(1) for the scattering coefficient by subtracting the absorption coefficient measured by the TAP from the extinction coefficient measured by CAPS_SSA, The nephelometer is used as reference because it has well proven correction functions for light absorption particles, as described in Section 2.2.1.  There is neither a trend visible of the mixture ratio with ammonium sulphate, which the SSA is the indicator for, nor a strong shift for high or low volumetric cross-section values. This is true for both examined wavelengths of 630 nm and 450 nm.
Overall, it is visible, that some data points scatter more roughly on the 1:1 line, which is true for mostly pure Aquadag aerosol, where the TAP contributes the biggest uncertainties due to higher values and deviates from high agreement, which was 265 approved by the Reno Study  and although visible in our measurements. When the 1σ precision errors are tripled, it is still undistinguishable from the 1:1 line.
As a last extensive parameter, we focused on the particle light absorption coefficient. This is the most complicated to measure, as for filter-based methods a bunch of correction schemes must be applied. Using a differential method e.g.  In Figure 5, the light absorption values for wavelengths of 450 nm and 630 nm are depicted. To compare instruments, the overall uncertainty is often estimated to be 30% (Bond et al., 1999)In this work we stay within a 20% deviation for this 280 parameter. Most data points correlated for both the σap (CAPS,CAPS) and σap(TAP) reference, without any mixing ratio dependence. When the σap(CAPS,CAPS)is compared to σap(CAPS,NEPH), the values agree within the uncertainty errors. The high Pearson correlation (r > 0.95) coefficients in Table 4 indicate that the correlation is highly linear and reveals a stable behaviour of the instrument measurements characteristics. The slopes are all close to unity within the expected errors ranges.
Thus, the extensive parameters can be trusted for instrument comparison especially for the light scattering and light extinction 290 information. The slopes reported for light absorption coefficients are with 0,92 ±0.07 and 1.04 ±0.08 below the expected error from literature. Higher values influence linear regression slopes, for which the filter methods are drifting to lower values respective to intercomparison instruments . We provide further regression analysis for all other aerosol types individually in Tables 7-9. An excellent agreement (r=0.99) is shown for σsp measurements of the nephelometer and the CAPS PMSSA scattering channel. Thus, the CAPS PMSSA gives reliable scattering coefficient measurements for aerosol mixtures and 295 could be considered as a substitute for the nephelometer and delivers reliable SSA measured simultaneously in one volume of the same instrument.
Instead of using regression analysis, where outliers and/or high values are dominating the slope of the regression-a more robust statistical analysis of the ensemble averaged instrumental ratios (σap (instrument #1) / σap (instrument #2) will be shown in the following section. For the table 5table 6 the ratios are calculated using averaged 1Hz measurement data. The average 300 intervals are adapted for constant conditions, waiting 15 minutes until the production and nephelometer were settled/relaxed lasting for about 5 minutes until the next mixture was setup in the sample line restarting the procedure. Filtering these instrument ratios for σap < 10 Mm -1 the relative frequency distribution shows almost no modal value. Filtering 310 the data for σap > 10 Mm -1 about 80% of these data are within the range of 0.8-1.2 .  -Redoing this analysis for 450 nm wavelength the light extinction and scattering of smaller particles increases compared to the 315 values at 630nm wavelength. As a result, this increase also the errors associated with the differential method. As demonstrated in Table 6, only the ratio σap (TAP) / σap (CAPS, NEPH) for spherical particles deviate less from unity, with over 50% of the data being within the range of 0.8-1.2. Still all ensemble averages are close to 1 but with an associated error of up to ±1.7 these values are not significant, which means, that the ratios scatter widely with no clear modal value.
Again filtering only for σap >10 Mm-1 the methods agree well with significant ratios σap (TAP) / σap (CAPS, NEPH) =1.08 320 ± 0.33 for BC. The best instrumental ratio with 1.01 ± 0.13 is shown for AQ in Table 6 at 450 nm wavelength.   Table 7. Linear regression analysis of attenuation coefficients using Cabot Black and ammonium sulphate mixtures are shown.
Presenting: the slope (m), Pearson (r) and y-axis intersection (b) for different instruments combinations.  For Magic Black the light absorption measurements using the DM method for 450nm shows the highest difference compared 365 to the TAP measurement with a regression slope of 0.21 ± 0.14. The reason could be different absorption behaviour for in-situ to filter measurements and no clear indications of the particle shape could be made.

Single scattering Albedo (SSA) 370
The Single Scattering Albedo (SSA) as an important climate parameter is used for instrument validation in this section. To obtain this parameter, different methods are shown in Table 10. Each method excludes at least one instrument from the calculation, thus, instrument intercomparison is possible.   shows reasonable results within a +-10% error band. Spotting for dark colours in the colourcode -Low σap values are only seen for SSA >0.6 as expected reflecting that there are just fewer particles of Aquadag in the aerosol mixture. General, the load of 395 absorbing particles seems not to influence the accuracy of the method, except for high absorption coefficients over 50 Mm -1 .
Here the TAP shows a nonlinear response which is visible in Figure 7 as offset of 0.1 higher than the SSA reference calculated using Eq. (10). Like in the previous section, the ensemble average (average of instrument-to-instrument measurement ratios) was calculated 400 to show a robust measure for the overall agreement of this parameter. In Table 10 the SSA values for all aerosol types are summarized. The nephelometer and CAPS extinction was used again as reference. The highest deviation is visible with combustion soot for TAP related data. The deviations of the reported mean from 1 are less than the relative uncertainties which range around 0.09.

Ångström exponents: EAE
In this section we will now focus on the next important and climate model- When directly comparing EAE(TAP, NEPH) to EAE (CAPS) the EAE values agree within 10% deviation. Again, the best correlation is visible with Aquadag mixture particles. For EAE(CAPS) > 2.5 the EAE(TAP, NEPH) tents to underestimate the EAE. Nevertheless EAE(NEPH, TAP) shows the highest values close 3.21 which corresponds to EAE values reported in Table   3 for the pure AS particles which are small in size of about 40nm. The instrument-to-instrument ration of SAE were calculated for each particle type. Most data points show a ratio of close to 1 not biased by σap or by the SSA shown in and Figure 11.
Looking for the instrument-to-instrument SAE ratios for the different absorbing species individually in Table 12 only soot  475 shows an instrument-to-instrument ratio of about 1.43+-0.61 which is statistical not significant different from 1. When comparing the SAE dataset obtained by using Nephelometer and CAPS is measurements in Figure 12 and Table 12,  480 Aquadag shows the best instrument-to-instrument ratio of 0.99 +-0.15. A small nonlinearity for SAE values higher 3.0 begins to deviate from the 1:1 line but stays within 15% deviation as already seen for EAE. Here again NEPH shows higher SAE values compared to CAPS by a factor 0.9. This factor corresponds as the observed factor for the EAE values and is linked to nephelometer measurements for fine AS particles. Since the Nephelometer correction is calculated based on the scattering angstrom exponent, which contains a vague size distribution information, it could fail to give correct values for aerosol 485 mixtures and for different sizes.

Ångström exponent: AAE
The absorption Ångström exponent depends entirely on the absorbing particle type and should not differ when the light absorbing particle is mixed with non-light absorbing particles. This independency of adding AS to the mixture was observed 490 for the filter based instrument TAP, visible as x-axes in the scatter plot Figure 13, but when the AAE was calculated using measurements of the in situ instruments Nephelometer and CAPS, the AAE deviates more with increasing SSA and lowering The reason for this is the high relative precision error associated with low absorption particle loads for AAE (CAPS, NEPH), 500 which we had crosschecked by calculating the variation of the AAE by varying the input variables by their possible max errors, showing the same results. Pure Aquadag particles are made visible by an open circle and it is visible, that those does not stray far from the 1:1 line in Figure 13 including AAE vales for Aquadag. For the pure substance a higher particle load could be used and no other negative interfering non absorbing aerosols influence the measurements.  In Figure 14 showing the Method-to-method ratios AAE(CAPS, NEPH) / EAE (TAP) there is a strong dependency as function 510 of σap(TAP) and SSA(CAPS,NEPH)_630 visible. Lowering the absorption coefficients below 100 Mm -1 or a SSA higher than 0.5, the AAE begins to differ strongly and up to tends up to triple the AAE value calculated from TAP coefficients only. As long as for laboratory studies, these high particle concentrations could be archived, but are rarely present in atmospheric conditions EAE(CAPS, NEPH) method is not applicable for atmospheric measurements 515  Table 12. Here an ensemble average and the associated variance was considered as a good reference. The instrument to instrument ratios for Ångström exponents for light extinction and Ångström exponents for scattering correspond within 10% deviation. The most prominent exception is again freshly produced combustion soot. For light absorption, a large deviation for the AAE ratios value is associated with 525 weak absorption coefficients of the mixtures used. Therefore, the AAE shows the biggest differences within the instrument to instrument ratio analysis.

530
A major goal of this study was to determine the errors associated with instrumental uncertainties of intensive optical aerosol parameters such as single scattering albedo and Ångström exponents. Basis was an instrument intercomparison study of widely used measurement techniques that are suitable for long-term observations. The methods used agreed the most for a mixture of the spherical-shaped colloidal graphite (Aquadag) as light-absorbing and ammonium sulphate as light-scattering aerosol component. Results for this mixture have low uncertainties and agree within 10% deviation between the methods for single 535 scattering albedo, extinction Ångström exponent and scattering Ångström exponent. Laj et al.,2020 recently stated requirements for GCOS (Global Climate Observing System) applications. Here he proposed uncertainties lower than the 20% measurement uncertainty for single scattering albedo measurements for attributing and detecting changes to a climate feedback.
The uncertainties and deviations shown in this work are with 8-10% measurement uncertainty fulfil the required limit. Overall, we were able to show study that extensive parameters agree within the limits of uncertainty for the individual instruments. For 540 spherical particles, we achieved the highest correlations for each light extinction, scattering and absorption coefficients. For fractal-like particles, the correlation for light absorption between the in-situ and filter method weakens but stays within instrument uncertainty ranges. Uncertainties increase for intensive parameters, especially for parameters obtained with the differential method that calculates light absorption as the difference between light extinction and light scattering. In addition, extinction Ångström exponents, scattering Ångström exponents, and single scattering albedo were not as much affected by the 545 uncertainties associated with the differential method used for σap compared to the absorption Ångström exponent. Using the differential method, AAE was rarely within typical physical values for the differential method. Low single scattering albedo values (<0.5) and, more importantly, high particle loads of at least 50 Mm -1 are necessary to reach satisfactory uncertainty levels. Freshly-generated combustion soot differs the most, with results disagreeing up to 30% between filter-based absorption https://doi.org/10.5194/amt-2021-284 Preprint. Discussion started: 27 September 2021 c Author(s) 2021. CC BY 4.0 License. coefficient data and in-situ methods. This is due to the combined effects of small flickers of the inverted flame generator during 550 the experiment, the overall filter correction schemes, and the physical behaviour of agglomerates. The single scattering albedo for 630 nm wavelength could be determined within 10% deviation between the instrument combinations of CAPS, TAP and Integrating Nephelometer, but tends to differ by at least 0.1 for light absorption coefficients of over 50 Mm -1 . A similar accuracy could be achieved with a wavelength of 450 nm, for which a 15% deviation between the instrument combinations must be considered. Even with the strong deviation within absorption values, the intensive parameters for the scattering and 555 extinction Ångström exponent stay within 10% deviation, regardless which instrument combination is used for calculation.
With this approach the intensive aerosol properties showed a high rate of agreement between different instrument sets for the determination of these properties for techniques used for long-term measurements, except for the absorption angstrom exponent. As an additional result, we can present that for stable aerosol production, the internal scattering coefficient measurement by the CAPS PMSSA agrees with the integrating Nephelometer within 10% deviation and therefore could be 560 substitute the TSI Nephelometer 3563 for light scattering measurements which is not produced any longer.