' s Response : The new TC-BC method and online instrument for the measurement of carbonaceous aerosols

We present the newly developed Total Carbon Analyzer (TCA08), and a new method for online speciation of carbonaceous aerosol with a high time resolution. The total carbon content is determined by flash heating of a sample collected on a quartz-fiber filter with a time base between 20 min and 24 h. The limit of detection is approximately 0.3 μgC, which 20 corresponds to a concentration of 0.3 μgC/m at a sample flow rate of 16.7 LPM and a 1-hour sampling time base. The concentration of particulate equivalent organic carbon (OC) is determined by subtracting black carbon concentration, concurrently measured optically by an Aethalometer, from the total carbon concentration measured by the TCA08. The combination of TCA08 and Aethalometer (AE33) is an easy-to-deploy and low maintenance continuous measurement technique for the high time resolution determination of equivalent organic and elemental carbon (EC) in different particulate 25 matter size fractions, which avoids pyrolytic correction and need for high purity compressed gases. The equivalence of this new online method to the standardized off-line thermo-optical OC-EC methods was evaluated during a winter field campaign at an urban background location in Ljubljana, Slovenia. The organic matter-to-organic carbon ratio obtained from the comparison with an Aerosol Chemical Speciation Monitor (ACSM) was OM/OC = 1.8, in the expected range.


Point-to-point response to Referee #1 (RC1)
Please note that page and line numbers refer to the complete manuscript and not the one in this document.
1. As mentioned on the Title, page 1 line 16, page 4 line 1 and elsewhere the authors state that this is a new or newly developed method. Nevertheless, a method bearing great similarities has been already described in the past  and references therein). The paper describes an analyser of a different manufacturer 5 that "also determines optical OC and optical EC by monitoring the laser transmission data through the quartz filter", "Total carbon (TC) is determined using the thermal-optical method, and then optical OC is deduced by subtracting optical EC from TC (optical OC =TC -optical EC)".Where optical OC and EC would be simply a different terminology given for eOC and eEC used in the current paper. How would the authors comment on the method similarities of the two studies and the suitability of the description "new" for the method ? 10 There are conceptual similarities between method mentioned above ) and TC-BC method developed in this study: the new method takes the advantage of decoupling thermal and optical method into two separate instruments, both dedicated for different measurements. With this, the new method has higher time resolution, no dead time, online loading compensation for eBC measurements and is more convenient for field 15 measurements as the thermal measurement is done without fragile quartz cross oven, high purity gases and catalyst.
The main difference are listed and described in details below: -The optical EC in semi continuous Sunset instrument is a measurement of transmittance through the filter at a wavelength of 660 nm prior to the thermal analysis, while the equivalent eEC is an equivalent BC measurement 20 with AE33 at 880 nm, then multiplied by a proportionality factor b (eEC = bBC). Contribution of light absorbing organics is higher at 660 nm. -Light source used in AE33 is non-coherent set of light emitting diodes (LEDs) with diffuser-like optics (BC6 is measured at 880 nm), while the semi continuous Sunset instrument uses diode laser at 660 nm. Distorted wave fronts of LEDs produce homogeneous signal on the exposed filter area and the transmittance signal has no 25 speckle/interference noise, making the LED light source far more convenient for filter attenuation (ATN) measurements.
-The presence of the loading effect in filter-based absorption photometers causes an ATN-dependent change in the instrumental sensitivity. The Aethalometer model AE33 performs real time compensation for this nonlinearity using patented dual spot algorithm. 30 -ATN in AE33 is determined more precisely as detector intensity signal for the measurement spot I and detector signal for the reference spot Io are measured concurrently. Any drift in LED light intensity due to temperature and other changes are compensated in real time. While the stabilization of laser source is possible but there are no publication describing this process for Sunset semi-continuous instrument. -When the attenuation reaches a certain threshold, a tape advance is induced so that measurements starts on a 35 clean spot. Significant buildup of refractory substances on the filter in the semi continuous Sunset instrument will reduce initial intensity and introduce higher signal to noise ratio at each subsequent optical analysis. Two dedicated instruments are better for this purpose.
A short paragraph was added to the manuscript (Chapter: 2.1 TC-BC method for online high time resolved OC-EC measurements, p. 4, lines 17-21: 40 -Although one can find conceptual similarities between method presented in Bauer et al., 2009 (and references therein) and TC-BC method presented in this study, the new method takes the advantage of decoupling thermal and optical method into two separate instruments, both dedicated for different measurements. With this, the new method has higher time resolution, no sampling dead time, online loading nonlinearity compensation for eBC measurements  and is more convenient for field measurements as the thermal measurement 45 is done without fragile quartz cross oven, high purity gases and catalyst.
Two references were added to the manuscript: -Bauer, J. J., Yu, X.-Y., Cary, R., Laulainen, N., and Berkowitz, C.: Characterization of the Sunset Semi-Continuous Carbon Aerosol Analyzer, J. Air Waste Manag. Assoc., 59 (7), 826-833, doi:10.3155/1047-3289.59.7.826, 2009. 50 -Drinovec, L., Gregorič, A., Zotter, P., Wolf, R., Bruns, E. A., Prévôt, A. S. H., Petit, J.-E., Favez, O., Sciare, J., Arnold, I. J., Chakrabarty, R. K., Moosmüller, H., Filep, A., and Močnik, G.: The filter-loading effect by ambient aerosols in filter absorption photometers depends on the coating of the sampled particles, Atm. Meas. Tech., 10(3), 1043-1059, doi:10.5194/amt-10-1043-2017, 2017. 2. While the terms of OC, EC and eOC, eEC are clearly defined, their use in the text overlaps and is occasionally confused. Proper terminology should be consistently used in order to avoid any misinterpretations by readers. 5 For example, the abstract mentions in lines 22-23 that this new application can result in high time resolution determination of organic and elemental carbon while in reality it provides an estimation of eOC and eEC values. Another example would be in section 3.6: eOC and eEC should be used instead of OC and EC. Also applicable in all graphs. 10 Terms OC, EC and eOC and eEC where changed throughout the manuscript using proper terminology as follows: -Page 1, line 20: The concentration of particulate equivalent organic carbon (eOC) is determined by subtracting black carbon concentration, concurrently measured optically by an Aethalometer ® , from the total carbon concentration measured by the TCA08. -Page 1, line 23: The combination of TCA08 and Aethalometer (AE33) is an easy-to-deploy and low maintenance 15 continuous measurement technique for the high time resolution determination of equivalent organic and elemental carbon (eEC) in different particulate matter size fractions, which avoids pyrolytic correction and need for high purity compressed gases. -Page 9,  -Page 16, line 15: 3.5 Comparison of online eOC measurements from TCA with offline OC filter analyses.
-Page 16, line 16: Online eOC measurements can be derived using the above EC-BC correlation plot to assign the appropriate operational value of the parameter b; the online BC data; and the online TCA data. 25 -Page 16, line 19: These results show that when using an appropriate value of b, the "TC -BC Method" yields online data for the eOC content of ambient aerosols that agree very well with conventional offline thermal analyses. -Page 16, line 23: The in-depth analysis of the relative difference between OC from 24 h filters and eOC determined by online measurement as TC-bBC shown in Fig. 9 reveals that the positive artefact can be the 30 dominant apparent source of OC for days with very low OC concentrations (< 5 μg/m 3 ) in comparison to offline 24 h filters, for which also negative artefact (desorption of VOCs) can occur. -Page 17, Figure 8. eOC on y-axis -Page 18, Figure 9. eOC and eEC on y-axis.
-Page 18, line 8: 3.6 Comparison of OM online measurements from ACSM with offline OC from filter sampling 35 and online eOC -Page 18, line 9: The data from an AE33 and TCA08 can be combined with an operational timebase of 1 hour, yielding eOC and eEC data with much greater time resolution than what can be achieved by the analysis of filter samples. -Page 20, Figure 11: eOC on x-axis. 40 -Page 20, Line 10: (b) Comparison of 3h eOC data derived as eOC = TC -bBC, to OM data measured by ACSM.
-Page 21, line 6: The diurnal variation of eOC and eEC for this urban background environment is strongly influenced by the temporal patterns of emissions from traffic and biomass burning (domestic heating) during wintertime. -Page 21, line 9: When the TCA08 is combined with an AE33Aethalometer, the TC-BC method yields eOC-eEC 45 data with much greater time resolution than that offered by the analysis of filter-based samples.

NDIR detectors, similarly to the one in the current application, may deteriorate in performance in long term and show a drift in their baseline. Since there is no application of an internal standard calibration or a span check, have the authors evaluated how often would an external standard calibration be required? Would there 50
be any NDIR detector related maintenance needs, e.g. source replacement, and in what frequency would that be required?
Light source life in Licor 840A NDIR detector is estimated to be 18000 hours. When light source fails the TCA instrument detects it, stops the measurements and displays Licor CO2 Error status.
Total Carbon content of the sample measured by TCA08 is a function of a CO2 difference between signal and 5 background values and thus not directly connected to absolute value of CO2 (Eq. 4). That is why the TC result is less dependent on the light source drift in the NDIR detector than if the absolute value is used in the calculations.
Additionally, the drift of the dual wavelength light source in the NDIR detector used in TCA08 is compensated online.
Concentration measurements of CO2 are based on the difference ratio in IR absorption between sample and reference signal. The CO2 sample uses an optical filter centred at wavelength of 4.26 μm corresponding to an absorption band 10 for CO2. The reference channel for CO2 has an optical filter centered at 3.95 μm, which has no absorption due to CO2.
During NDIR detector lifetime there is no need to preform internal standard calibration and span check for TC measurements, as the whole system (NDIR detector + TCA analytic chamber) can be calibrated or validated with 15 Carbon Calibration and Carbon Validation procedure for TCA08. This the great benefit of this instrument. Both procedures are described in TCA08 User Manual (TCA08, 2019). Carbon calibration of TCA08 should be done once per year or after any major maintenance or modification of the system.
A short paragraph was added to the manuscript (p. 6, lines: 11-20): 20 -Light source life in LI-840A CO2/H2O Analyser is estimated to be 18000 hours. When light source fails the TCA instrument detects it, stops the measurements and displays Licor CO2 error status. Total Carbon content of the sample measured by TCA08 is a function of a CO2 difference between signal and background values and thus not directly connected to absolute value of CO2 (Eq. 4). This is why the TC result is less dependent on the light source drift in the NDIR detector than if the absolute value is used in the calculations. During light source lifetime 25 there is no need to preform internal standard calibration and span check for NDIR detector, as the whole system (NDIR detector + TCA08 analytic chamber) can be calibrated or validated with Carbon Calibration and Carbon Validation procedure for TCA08, which is the great benefit of this instrument. Both procedures are described in TCA08 User Manual (TCA08, 2019). Carbon calibration of TCA08 should be done once per year or after any major maintenance or modification of the system. 30

The last paragraph of section 2.3 describes tests performed on the denuder efficiency but it seems that results
are not included in the paper. Page 7, line 21 also refers to TC data "(see below", which are not presented later on in the text. Related to the denuder efficiency, there is clear evidence of a positive artefact for eOC 35 concentrations below 8μg/m3, visible in Figure 8 (OC vs eOC) as well as in Figure 11 (b) (OM vs eOC). Further there are signs of a negative artefact for higher concentrations, based on the same graphs, suggesting reduced combustion efficiency. The later would be more profound if there would be an addition of second denuder monolith as suggested in page 12 lines 25-27 or if a correction for the positive artefact would be applied. Should the user then consider the use of 2 correction factors (b) related to the concentration levels measured? Or 40 would there be any other suggestion to overcome these issues?
As we noted in the first paragraph of chapter 2.3 the measurement of carbonaceous aerosols using quartz-fiber filters is always very challenging because of the possibility of positive and negative sampling artifacts which are hard to quantify. The TCA08 instrument was developed in a way, that sampling face velocity was similar to the one of high 45 volume samplers for offline analysis for easier comparison. With shorter sample time basis the positive artefact is much more pronounced and negative artefact can be neglected.
We believe that TCA08 with its denuder efficiency procedure is a great tool to investigate (1) denuder efficiency and (2) temporally variations of positive/negative artefact. In our current study (Gregorič et al., 2019(Gregorič et al., , 2020, positive and negative artefact on quartz filters depending on sample time base, face velocity, number of denuders, chemical deposition of aerosol is investigated in details. When investigating dependence on sample time base, we found out that adsorption of organic vapors in TCA08 can be described with a sum of two exponential functions with a1*(1 -5 Exp[-x/τ1]) + a2*(1 -Exp[-x/ τ 2] with a fast and slow time constants τ1 and τ2. Slow time constants are around 15-60 minutes, indicating that positive artefact prevails for sampling times up to 2 hours. Nevertheless, high VOC concentrations are usually connected to high OC concentrations as well. We believe that negative artefact due to combustion efficiency mentioned by reviewer is not the reason for lower slope. External 10 calibration of TCA shows that combustion efficiency does not reduce with higher TC, OC or EC concentrations (see Figure 1), therefore, two correction factors are not needed. Users should use appropriate number of denuders and choose appropriate sampling timebase according to denuder efficiency test and ambient TC and OC concentrations.
In our next publication guidelines and recommendations on how to asses positive/negative artefact with TCA08 and how to use this knowledge for more quality measurement data will be described. 15

Figure 1 External calibration of TCA08 -TC mass calibration range
The last paragraph in section 2.3 was changed accordingly: 20 -We developed these routines during the instrument design and performed the measurements as part of the field campaign (see Section 3). After five weeks of continuous operation with consistent TC data (see below), the measured denuder efficiency was 74%. We recommend that the denuder should be replaced or regenerated when its efficiency drops below 70% Gao et al., 2014). The Standard Operating Procedure for routine use of the TCA08 instrument recommends replacement or regeneration of the 25 denuder honeycomb element once per month. Further, in environments with high VOC concentrations, two denuder honeycombs in series are recommended (Gregorič et al., 2020).

Section 3 refers to EN 16450:2017 regarding the orthogonal regression analysis on the 31 daily measurements between the new instrument (candidate method) and 2 independent laboratories (reference method). A proper 30
application of EN16450:2017 would require a minimum of 40 valid data pairs with the further requirement of 2 candidate applications for each type testing application. EN16450:2017 further describes requirements related to the number of locations and the concentration range of data points. The use of just one candidate method limits the conclusions on performance consistency between identical instruments and restricts the candidate method uncertainty calculations. It seems that quite a significant part of section 3 discusses the comparability 35 between the two reference method applications, which is a topic thoroughly documented elsewhere in the literature (Intercomparison exercises publications are included in the reference list of the current paper). A more relevant approach would have included and compared two new instruments in parallel measurements. How would the authors comment on the approach applied and the data suitability? External calibration TCA08 We agree with reviewer that equivalence comparison was not done exactly according to standard EN 16450:2017, but it is our best attempt to do it so. As there is no standard for reference method for online measurement of OC and EC concentrations available at the time of the writing of this manuscript, we followed EN16450:2017 and choose EN 16909:2017, 2017 as the reference method. The comparison was done on the available data (31 daily filters -5 limitations of DIGITEL high volume sampler availability). Furthermore, we used only one set of instrument for the candidate method comparison, as they are both compared to the reference set of instruments after their assembly (inhouse defined requirements for successful intercomparing test are: 1. TCA08: TC concentration range up to 75.000 ng/m 3 , slope between 0.97-1.03, R 2 above 0.98 ; 2. AE33: eBC concentrations up to 30.000 ng/m 3 ; slope between 0.97-1.03, R 2 above 0.98). Following paragraph was added to the manuscript (page 9, lines: 12-21): -As there is no standard for reference method for online measurement of OC and EC concentrations available at 20 the time of the writing of this manuscript, we followed EN16450:2017 and choose EN 16909:2017 as the reference method. Nevertheless, a proper application of EN16450:2017 would require a minimum of 40 valid data pairs with the further requirement of two candidate applications for each type testing application. Additionally, the same standard further describes requirements related to the number of locations and the concentration range of data points. The results and discussion in this chapter is our best attempt of equivalence 25 comparison on the available data (31 daily filters due to the limited access to the DIGITEL high volume sampler). Furthermore, we used only one set of instrument for the candidate method comparison, as they are both compared to the reference set of instruments after their assembly as one of the tests during final inspection procedure (inhouse defined requirements for successful intercomparison test are: 1. TCA08: TC concentration range up to 75.000 ng/m3, slope between 0.95-1.05, R2 above 0.98 ; 2. AE33: eBC concentrations up to 20.000 ng/m3; slope 30 between 0.95-1.05, R2 above 0.98).
6. The uncertainty limit value of 2.00 μg/m3 mentioned in page 10 line 13 originates from calculations of PM reference methods where limit values of 30μg/m2 or more apply. Would that be directly applicable for TC method 35 and concentrations? Further section 3.3 describes method uncertainty calculations based on the NDIR detector response. Have the authors considered additional sources of uncertainty to be included in the uncertainty budget, e.g. use of denuder, zero air carrier gas, ambient temperature and pressure variations?
Carbonaceous aerosols frequently account for a large and often dominant fraction of fine particulate matter (PM2.5) mass in polluted atmospheres. Total carbon mass concentrations can contribute up to 50% of PM2.5 mass in special 5 pollution events, which justify the use of uncertainty limit value of 2.00 μg/m 3 for TC concentrations.
Temperature and pressure variations during sampling are measured with meteorological sensor and are included into calculations as the volumetric sample flow is used for concentration calculations. During analysis temperature and pressure of the analytic stream are measured within NDIR sensor and are included in CO2 concentration 10 determination.
Uncertainty due to variations in zero carrier (analytic) gas during thermal analysis was tested by measuring replicates of blank samples. This approach is also used for Limit of Detection determination, which is 0.3 μg of TC for TCA08. Again, before entering the chamber, the analytic air passes through a 10-liter buffer volume for ambient CO2 15 fluctuation averaging and a capsule filter filled with activated carbon and pleated glass fiber filter, which removes organic gases and particles from the stream.
An example of such LoD determination for TCA08-S00-00103 is shown in table below (Table 2).  semi-volatile constituents of PM (9.3.2.4). Consequently, the magnitude of the effects of adsorption of gases cannot be quantified. For the purpose of application of this European Standard the phenomenon is recognized but not considered in the uncertainty budget. -EN 16909:2017, Chapter 9.2: All sampling artefacts are inherent by convention and part of the EC and OC values according to this standard. Sampling artefacts are mainly to be expected for OC and they can be significant 5 (Chow et al. [12]).
A following sentences were added to the manuscript: -(page 11, lines 12-13): The uncertainty uRM between the reference methods for TC 10 is 0.43 μg/m 3 which is well below the limit of 2.00 μg/m 3 requested for reference methods for PM mass concentration measurements (EN 16450:2017(EN 16450: , 2017. -(page 14, lines 15-22): where LoD is the limit of detection of the TCA08 at a sample flowrate of 16.7 LPM and sample timebase of 1h. 15 In the uncertainty budget of TC measurement with the TCA08 the following sources of uncertainties were not The filters used for offline sample are 150 mm Tissuquartz 2500 Qat-Up produced by Pall Corporation. The homogeneity of samples was tested by ARSO laboratory for 5 random filters from the campaign. In the table (Table  3) below results of the homogeneity test are shown: 35 40  -IGE laboratory (Inter-laboratory comparison 2016 and 2017, contact person: Jean Luc Jaffrezo), (ACTRIS, 2016, 2017) -ARSO laboratory (inter-laboratory comparison 2018, contact person: Judita Burger), (ACTRIS, 2018) 5 8. A common practice in comparison exercises is following identical procedures on filter handling, transport and storage for all participants. In the current case filters were first analyzed by the ARSO laboratory and then shipped to IGE for further analysis. Even though the authors mention that "sampling, transport and storage of the filters was done according to the EN 16909:2017" IGE received the filters for analysis at a later period, after additional transport, handling and storage. Could that have contributed to the small uncertainty 10 observed between the reference methods? Wouldn't an approach of dividing the samples and shipping in both labs in parallel have resulted in improved comparability of the two laboratories?
We agree with the reviewer, it would be better to wait for the end of the campaign, than divide the filters and afterwards analyze them in both laboratories. Nevertheless, all measures were taken to assure that shipped samples 15 to IGE were not contaminated. The measurement campaign was conducted between 7 February and 10 March 2017 at the urban background air quality monitoring station of the Slovenian Environmental Agency (ARSO). After the cartridge with 14 sampled filters was replaced by a cartridge with fresh filters, the sampled filters were stored at petri dishes in shipped to IGE laboratory by express postal service. During shipment the petri dishes containing the filter punches were stored in cooled portable freezer box with temperatures between 5-15 o C.

Table 3 provides with a wide range of b factors and the recommendation right above is: "the determination needs to be performed for each location and with filters sampled over the time-period of interest". 25
Considering that the b factor is location and season specific, could the authors suggest a typical coverage period range with OC/EC offline analysis in parallel? How often would the b factor have to be re-evaluated per location?
The proportionality parameter b is an effective value with a local and regional component. Usually, the local contribution to concentrations is dominant and the local BC and EC contributions dominate the relationship. The 30 differences in b values presented in Table 3 show, that there is a big variation between different rural/regional background sites, and also between the urban sites. This is the reason why similar offline-to-online intercomparison is recommended for every new background site or site with strong mixture of local and regional contribution. The time period of intercomparison should cover seasonal variations in b values, for example 2-3 weeks each season. The re-evaluation intercomparison campaign for the certain location should be done if significant changes in the BC 35 emission inventory is expected (traffic or wood burning restrictions, etc.). For sites with dominant traffic contribution, where the b factor mostly depends on the properties of the vehicle in the fleet, the intercomparison measurements will result in similar b values unless a significant fleet change occurs.
Following sentences were added to the manuscript: 40 -(page 15., lines 11-18) The proportionality parameter b (Eq. 3) is compared with values taken from the literature in Table 3. These values depend on the location, the nature of the aerosol, and the thermal protocol used for analysis. The value of 0.44 which we determined in this study for an urban background site is slightly lower than values for other urban and 45 urban background sites using EUSAAR 2 thermal protocol, and considerably lower than the values for rural sites.
The proportionality parameter b is an effective value that features a local and a regional contribution of BC and EC. Usually, the local contribution to concentrations is dominant and the local BC and EC contributions dominate the relationship. The differences in b values presented in Table 3 show, that there is a big variation between different rural/regional background sites, and also between the urban sites. This is the reason why similar 50 offline-to-online intercomparison is recommended for every new background site or site with strong mixture of local and regional contribution.  Table  3. Further the slope from the OM -eOC comparison of ( Figure 11,(b)) is 1.82, and would have been even higher 10 when considering the high negative intercept. Following the ranges provided by the literature in page 17, lines 14-15, these slopes fit better a rural site rather than an urban environment. This comes in contradiction with the characteristics of the selected site which is influenced from traffic emissions. If the estimated b factor was within the literature range that would have further resulted in a lower OM -eOC slope and would fit the literature range better. How would the authors comment on these observed differences compared to the literature? 15 There is a big variation between different rural/regional background sites, and also between the urban sites. These differences show that a site and season specific factor b needs to be assessed. The obtained OM/OC of 1.82 indicates on mixture of traffic and biomass emissions which agrees with other studies done in Ljubljana (Gjerek et al., 2018;Ogrin et al., 2016). Ljubljana is located in a subalpine basin surrounded by hills. Large forest areas provide a cheap 20 heating source, which represents a government promoted alternative to the use of fossil fuel. While the measurement of black carbon in Ljubljana apportioned to traffic show great variability in concentrations among measurement sites, the contribution of biomass burning is spatially distributed much more homogeneously across wider area in Ljubljana. In winter 2013/14 campaign at 2 urban background measuring sites, study showed averaged value of BC from traffic of 2.5 ± 1.8 μg/m 3 , while from biomass burning contribution of 1 ± 0.7 μg/m 3 (Ogrin et al., .  Similary, the comparison of the offline analysis for EC on filter by sampling date (Figure 3) or by date of the analysis at the ARSO laboratory ( Figure 4) does not show any significant descripancies. At the moment, the temporal variations of descripancy between eEC and offline EC values is unknown.

Following Graph 7, it seems that in the first half of the campaign BC is overestimated while on the second half underestimated, compared to EC. Would there be any interpretations for this observation? Further, around the
.

Figure 4 Comparison of offline EC analysis between ARSO and IGE laboratories sorted by the date of analysis by ARSO laboratory.
10 12. Page 15, lines 19-21, suggest that the intercept due to the eOC positive artefact can be neglected based on the comparability with the OC intercept of the two independent laboratory measurements. Nevertheless, as discussed earlier in the text, the eOC intercept is systematic and attributed to the denuder performance. It should 15 also be noted that the eOC concentrations are compared to the average values between the two laboratories. It would be more appropriate if the artefacts observed for a new instrument would not be overlooked but rather investigated further and be dealt with. The use of two candidate method analyzers and an extended data set would be required for in-depth analysis. The above comes also in contradiction with the conclusion of page 20, line 12: "the correlation analysis showed very high agreement between eOC and eEC to the EC and OC". It seems 20 that there is room for improvement in agreement once the artefact issues are resolved.
We agree with the reviewer. We are aware of the importance of OC positive/negative artefact when comparing online OC/EC instruments to analysis of the 24h filters. In our current study (Gregorič et al., 2020), positive and negative artefact on quartz filters depending on sample time base, face velocity, number of denuders, chemical deposition of 25 aerosol is investigated in details in order to quantify the magnitude of these effects. In this publication guidelines and recommendations on how to asses positive/negative artefact with TCA08 and how to use this knowledge for more quality measurement data will be described. Again, positive artefact and denuder efficiency is recognized by standards EN 12341:2014 and EN 16909:2017: but not included in the uncertainty budget.
Please refer to points 4, 5, and 6 for details. 5
-The amount of OC converted into PC during the analysis depends on many factors, including the amount and type of organic compounds, the sources of air pollution, temperature steps in the analysis, the residence time at 10 each temperature step, and the presence of certain inorganic constituents (Yu et al., 2002). Page 15, line 4: It is not clear on which comparison the authors refer.
-The EC data determined by offline OC/EC analysis used in the comparison depends greatly on the thermal protocol used . Page 17, line 21: Fig 10(b) to Fig 11(b). 15 -The negative offset in the regression model with intercept ( Fig. 11 (b)) again reveals the pronounced positive sampling artefact due to adsorption of organics on quartz fiber filters for short sampling times in TCA08 method. Page 18: Figure 11 (a) misses the dashed trendline. Please note that page and line numbers refer to the complete manuscript and not the one in this document.

Introduction: The intro part mainly discussed the definitions of different carbon fractions and various
protocols for measuring the carbon fractions. However, some of the content is repetitive (e.g. it was discussed 5 in Page 2 Line 26 to Page 3 Line 2 that a few factors will influence the OC-EC split while similar points were mentioned again in Page 3 Lines 9-27). The authors also listed out three protocols that have been widely used in different regions of the world (Page 2 Lines 16-24). Since in the following sections the online data were compared with the offline data obtained from the EUSAAR2 protocol, readers might expect to see more discussions on the specific differences among the protocols (EUSAAR2 vs. IMPROVE_A, EUSAAR2 vs. 10 NIOSH).
We believe that discussion on the specific difference among protocols is beyond the scope of this paper and was very well described in Karanasiou et al. (2015). We focus on operational definition aspects and harmonization between different experimental approaches. 15 Following sentence was added to the manuscript (page 2, lines 24-25): -This protocol has recently became part of the European standard for the determination of OC-EC in PM2. 5 samples (EN 16909:2017(EN 16909: , 2017. Detailed discussion on the specific difference among protocols can be found elsewhere (Cavalli et al., 2010;Karanasiou et al., 2015). 20

Section 2.2: How is the performance of external calibration (using sucrose or KHP or other chemicals) by TCA08? What is the maximum carbon concentration tested?
External calibration of TCA08 was performed with punches of ambient filters with known TC content. This is done to simplify the calibration procedure on the field and to achieve better calibration accuracy (ambient filter includes 25 mixture of different EC and OC fractions). Carbon calibration constant is defined with the slope between measured integrated pulse of CO2 by TCA08 and known TC mass content of the filter punch ( Figure 5). The maximum carbon content tested in the TCA08 was up to 100 μg of TC. Integrated signal CO 2 (PPM) External calibration of TCA08 3. Section 3.3: What is the "carbon calibration factor"? How is it derived? What's the loading effect compensation algorithm used for treating AE33 data in this study? Will different algorithms introduce uncertainties?
Carbon calibration factor mentioned in paragraph in section 3.3 is carbon calibration constant Ccarb from Eq. 4. It is 5 defined as the slope between TC mass on the punch of ambient filter with known TC content and the integrated value of the CO2 signal measured by the NDIR detector ( Figure 5).
Carbon calibration factor was changed to carbon calibration constant (page 14, line 5): -The uncertainty uTCA associated with the TC data from the TCA08 includes individual uncertainty sources of the 10 carbon calibration constant Ccarb; the uncertainty of the analytic flow measurement; and the uncertainty of the signal and blank CO2 peak measurement (Eq. 4).
Loading effect compensation algorithm used for treating AE33 data in this study is an real time "dual spot" loading compensation (Drinovec et al., 2015. The inlet air stream of the AE33 is split, and the sample is collected on 15 two filter spots concurrently. The flow through each of the two spots is different, so the loading rates on the respective sample spots are different. Different loading rates cause the accumulation of the sample to be different between the two spots, resulting in a different magnitude of filter loading effect (FLE) between the spots. Measurement of FLE enables the compensation of the datausing the parametrization described in Drinovec et al. (2015), the compensation parameter k can be derived. We believe this method is the most appropriate one as it measures loading effect with the 20 same time resolution as measuring black carbon concentrations and it does not make any assumptions. Using different compensation algorithms will introduce uncertainties based on the assumptions they use.
A following sentence was added to the manuscript (page 14, line 26): - Figure 7 shows the regression of the off-line thermo-optical analysis of samples for EC (from the ARSO and IGE  25 laboratories, using the EUSAAR_2 protocol) with the 24-hour averaged BC (Aethalometer data) obtained during the field campaign period. An AE33 integrated "dual spot" real-time loading compensation algorithm was used for BC data treatment (Drinovec et al., 2015).  (Table 3) that the relationship between BC and EC is location dependent. If the aerosol composition at a certain location has a very clear temporal variation pattern, the parameter b could be sensitive to the sampling time period as well. Since the PM monitoring networks usually adopt the filter-based sampling approach followed by offline laboratory analysis and the historical dataset was very likely obtained from offline measurement, do the authors suggest 35 that every time the online TC-BC system is deployed to one sampling location, the online-offline comparison needs to be conducted to derive the b value so that the measured data can be compared to other dataset?
We agree with the reviewer. The sampling time period for offline filters can affect the b factor as well. We followed a common practice for sampling period; sampling start time was at 00.00 am and sampling stop time was at23.55 pm 40 each day. During 5-minute idle period, the sampler automatically stored sampled filter and replaced it with the new one.
The proportionality parameter b is an effective value with a local and a regional component. Usually, the local contribution concentrations are dominant and the local BC and EC contributions dominate the relationship. The 45 differences in b values presented in Table 3 show, that there is a big variation between different rural/regional background sites, and also between the urban sites. This is the reason why similar offline-to-online intercomparison is recommended for every new background site or site with strong mixture of local and regional contribution. The time period of intercomparison should cover seasonal variations in b values, for example 2-3 weeks each season. The re-evaluation intercomparison campaign for the certain location should be done if significant changes in the BC 50 emission inventory is expected (traffic or wood burning restrictions, etc.). For sites with dominant traffic contribution, where the b factor mostly depends on the properties of the vehicle in the fleet, the intercomparison measurements will result in similar b values unless a significant fleet change occurs.
Following sentences were added to the manuscript: 5 -(page 8, lines [35][36][37] The TCA08 was operated on a 1-hour time-base, sampling PM2.5 fraction at 16.7 LPM; co-located with a Model measurements of the OM-to-OC ratio. The ACSM, equipped with a PM1 aerodynamic lens, was sampling through a PM1 sharp cut cyclone (SCC 1.197, BGI Inc.) at a flow rate of 3 LPM yielding a particle cut off diameter of roughly 3 μm. Furthermore, the sample was driven through a Nafion dryer, upstream the instrument inlet, keeping the sample relative humidity below 40% throughout the campaign. The chemical composition dependent collection efficiency of the instrument was determined according to Middlebrook et al., 2012. Due to variability 20 in the ACSM time-base, we gathered the data into 3h averages. All of the instruments were checked regularly and operated without interruption throughout the campaign. No data were selectively removed from the results presented in the following.
-(page 15., lines 11-18) 25 The proportionality parameter b (Eq. 3) is compared with values taken from the literature in Table 3. These values depend on the location, the nature of the aerosol, and the thermal protocol used for analysis. The value of 0.44 which we determined in this study for an urban background site is slightly lower than values for other urban and urban background sites using EUSAAR 2 thermal protocol, and considerably lower than the values for rural sites.
The proportionality parameter b is an effective value that features a local and a regional contribution of BC and 30 EC. Usually, the local contribution to concentrations is dominant and the local BC and EC contributions dominate the relationship. The differences in b values presented in Table 3 show, that there is a big variation between different rural/regional background sites, and also between the urban sites. This is the reason why similar offline-to-online intercomparison is recommended for every new background site or site with strong mixture of local and regional contribution.

Introduction 30
Carbonaceous aerosols frequently account for a large and often dominant fraction of fine particulate matter (PM2.5) mass in polluted atmospheres. They are extremely diverse (Gelencsér, 2004;Karanasiou et al., 2015) and they directly impact air quality, visibility, cloud formation and properties, the planetary radiation balance, and public health (Pöschl, 2005). The carbonaceous fractions can be described as black (BC) or elemental (EC) carbon, and organic matter (OM). OM is made up of many different molecular structures and includes not only particulate organic carbon, but also hydrogen, oxygen, nitrogen, and 35 sulfur (Brown et al., 2013;Crenn et al., 2015). The amount of carbon that can be found in carbonaceous aerosols is called total carbon (TC), which is commonly categorized into fractions of organic carbon (OC) and elemental carbon (EC). OC can be directly emitted to the atmosphere in particulate form as primary organic matter by combustion and biogenic processes, or it can have a secondary origin from gas-to-particle conversion of (semi)volatile organic compounds in the atmosphere to aerosols after oxidation and condensation/nucleation (Hallquist et al., 2009). EC, on the other hand, is a mixture of graphite-40 like carbonaceous matter and is exclusively of primary origin and emitted by the incomplete combustion of carbonaceous fuels (Fuzzi et al., 2006;Karanasiou et al., 2015;Xu et al., 2015).
The first thermo-optical method for OC and EC determination was developed in 1982 by Huntzicker et al. (Huntzicker et al., 1982;Malissa et al., 1972). In thermo-optical methods, the carbonaceous aerosol deposited on the quartz filter is thermally desorbed according to a prescribed temperature protocol, first in an inert atmosphere (helium) and then in an oxidizing atmosphere (2% oxygen, 98% helium) (Cavalli et al., 2010). EC is thermally refractive and does not volatilize in an inert 5 atmosphere below ~700°C and can be combusted by oxygen at temperatures above 340°C Petzold et al., 2013;Schmid et al., 2001). Ideally, the OC fraction would desorb in the inert stage of the analysis, while EC would desorb and combust in the high temperature oxidizing stage of the analysis. Nevertheless, thermally unstable organic compounds pyrolyze (char) in the inert atmosphere to form pyrolytic carbon (PC), which combusts in the He+O2 gas stream in a manner similar to the original EC (Cavalli et al., 2010;Karanasiou et al., 2015;Schmid et al., 2001). The PC that is formed during 10 analysis, if not properly accounted for, would be incorrectly reported as EC. To account for this, illumination by a laser beam is used to monitor the optical properties of the filter during the analysis by measuring reflectance or transmittance (Chow et al., 1993). Because PC absorbs light, light transmission and reflectance signals decrease during the inert stage of the analysis when the PC is created; and increase again in the oxidizing stage as the remaining carbonaceous material is burnt off the filter.
The time when the reflectance or transmittance signal values meet the pre-pyrolysis value is called the OC-EC split point. 15 The three most commonly used thermal protocols are IMPROVE_A, NIOSH 5040 and EUSAAR2. The IMPROVE protocol using light reflectance for correction was designed to be applied to the Interagency Monitoring of Protected Visual Environments network in USA by Chow et al. (Chow et al., 1993). The NIOSH protocol using light transmittance was developed for the analysis of the carbonaceous fraction of particulate diesel exhaust based on the U.S. National Institute of 20 Occupational Safety and Health method 5040. In 2010, the thermal-optical analysis protocol EUSAAR2 was developed for European regional background sites. In order to improve the accuracy of the OC-EC split of this protocol, lower temperature steps in the inert stage of the analysis and longer residence times are used to achieve reduction of PC and more complete evolution of OC (Cavalli et al., 2010). This protocol has recently became part of the European standard for the determination of OC-EC in PM2.5 samples (EN 16909:2017(EN 16909: , 2017. Detailed discussion on the specific difference among protocols can be 25 found elsewhere (Cavalli et al., 2010;Karanasiou et al., 2015).
The charring of organic material during thermal analysis is an important uncertainty of the thermo-optical methods. tTe The amount of OC converted into PC during the analysis depends on many factors, including the amount and type of organic compounds, the sources of air pollution, temperature steps in the analysis, the residence time at each temperature step, and the 30 presence of certain inorganic constituents (Yu et al., 2002). When correcting for PC, thermal-optical methods make two important assumptions: (1) PC created by charring during the helium stage of the analysis is more easily oxidized and will evolve before the original EC (2) The specific light attenuation cross section of PC ( PC ) is similar to that of the original EC on the filter ( EC ). 35 However, PC and original EC combust concurrently in the oxidizing stage of the analysis. Moreover, PC can evolve even prematurely in the inert atmosphere depending on the thermal protocol used for the analysis, especially in the presence of oxygen donor substances in the sample (Sciare et al., 2003). Additionally, PC and EC have been shown to have significantly different values of (Bhagawan et al., 2015a;Cavalli et al., 2010;Chen et al., 2014;Karanasiou et al., 2015;Subramanian et al., 2004). The PC is mostly affected by the composition of its organic precursors, aerosol type and duration of sampling. For 40 this reason, the magnitude of the uncertainty of the OC-EC split point varies from one aerosol sample to another. Overall, the uncertainty derived from an incorrect determination of the OC-EC split is a function of the following parameters : -Aerosol type: the amount of PC converted from OC in the sample and its properties.
-Interference from other aerosol components: Carbonate carbon, Metal metal Oxidesoxides, Inorganic inorganic salts, Brown brown carbon.
-Thermal protocol used for analysis. 5 Because OC is the larger and often the dominant fraction of TC, the uncertainty from an incorrect OC-EC split point has a greater effect on the EC value. However, TC is a measurement of all evolved carbon, irrespective of the possible conversion of the fractions or the sample properties. Hence the TC determination is not influenced by the amount of PC formed during analysis or the thermal protocol used, and is therefore independent of the parameters mentioned above. 10 Thermal and optical methods refer to different properties of carbonaceous aerosol and specific attention needs to be paid to use appropriate terminology when inter-comparing carbonaceous analysis techniques using different measurement methods (Petzold et al., 2013). Measurements of optical attenuation or absorption are converted to mass concentration of black carbon (BC) using an externally determined mass attenuation/absorption cross-sectionthe resulting quantity is called equivalent 15 black carbon (eBC, Petzold et al., 2013). The thermo-optical and optical measurements share more than the optical pyrolysis determination during the inert phase of the heating in a thermal-optical analyzer. The definition of eBC is tied to the thermal determination of the sample carbon contentthe sample optical attenuation was compared to its thermally determined carbon content, both analyses performed after Soxhlet extraction (to remove non-soluble carbon), obtaining the BC mass attenuation cross-section independent of a specific thermal protocol (Gundel et al., 1984). 20 It was shown that the soluble carbon fraction did not absorb significantly, as the attenuation for the extracted samples decreased by no more than 7% compared to the non-extracted ones. While the insoluble fraction is not identical to the thermally refractive one, the relationship between the optically determined BC and the thermo-optically determined EC can be determined by analyzing samples obtained at the same site during the same period. Differences in thermal protocols, giving (systematically) 25 different EC values (Bae et al., 2009;Karanasiou et al., 2015), will result in different EC-to-BC regression slopes. At the same time, differences in the sample composition (and the sources of the aerosols) will influence the OC-EC split point, resulting in evolution of the less refractive part of EC in the inert phase and the more refractive part of OC in the oxidizing phase . Sample composition and sources also impact the sample optical properties, especially at shorter wavelengths (Sandradewi et al., 2008;Zotter et al., 2017). All of these factors affect the relationship between EC and BC. 30 Carbonaceous aerosols are the major, dominant component of the mass of suspended particles in polluted atmospheres.
Accurate, continuous and high time resolved data are needed in order to assess the severity of the problem and to identify and investigate the main sources which require attention; and to quantitate the improvements following the application of controls and regulations. The new TC-BC method presented in this study is an easy-to-deploy and low maintenance continuous 35 measurement technique for the high time resolution determination of organic and elemental carbon in different PM fractions (PM10, PM2.5 and PM1). It can be used for routine air quality monitoring applications, field work and laboratory research.
For example, high-time resolution data from the TC-BC method in combination with different size selective inlets can be used for quality control in aerosol mass spectrometry through comparison of differently derived oxygen to carbon (O-C) and organic aerosol to organic carbon (OA-OC) ratios (Pieber et al., 2016). In this study, the new online TC-BC method was tested during 40 a field campaign from 7 February to 10 March 2017 at an urban background air quality monitoring station of the Slovenian Environmental Agency (ARSO). High time resolved data of TC and BC were compared to EUSAAR2 OC-EC analysis of PM2.5 filter samples that were collected in parallel with a high volume sampler; and to organic aerosol mass measured by ACSM with a PM1 aerodynamic lens. The equivalence of the new online TC-BC method to the standardized off-line thermooptical OC-EC methods is evaluated through analysis of regression models of the various compared methods.

TC-BC method for online high time resolved OC-EC measurements 5
In this study we present the newly developed TC-BC method, which combines an optical method for measuring mass equivalent black carbon (eBC) by the AE33 Aethalometer (Drinovec et al., 2015;Hansen et al., 1984), and a thermal method where eEC = b·eBC (3) is equivalent to elemental carbon (EC) and the determined proportionality parameter b is region/site specific but also depends to a large extent on the thermal protocol used to determine the EC fraction with a conventional OC-EC method. We call this 15 determined parameter 'equivalent elemental carbon' (eEC) since the measurement method is an optical one, and its result is converted to an equivalent concentration of elemental carbon, following the terminology logic of Petzold et al. (2013).
Although one can find conceptual similarities between method presented in Bauer et al., 2009 (and references therein) and TC-BC method presented in this study, the new method takes the advantage of decoupling thermal and optical method into 20 two separate instruments, both dedicated for different measurements. With this, the new method has higher time resolution, no sampling dead time, online loading nonlinearity compensation for eBC measurements  and is more convenient for field measurements as the thermal measurement is done without fragile quartz cross oven, high purity gases and catalyst.

The TCA08 Total Carbon Analyzer
The TCA08 Total Carbon Analyzer instrument uses a thermal method for total carbon (TC) determination. The instrument contains two parallel flow channels with two analytical chambers, which alternate between sample collection and thermal analysis. While one channel is collecting its sample for the next time-base period, the other channel is analyzing the sample collected during the previous period. This sequential feature offers the great advantage of a continuous measurement of TC. 30 Fig. 1 (a) shows the TCA08 flow diagram, controlled by a system of valves which alternate the two channels to the common elements of pump, CO2 analyzer, etc. The instrument collects the sample of atmospheric aerosols on a central spot area of 4.9 cm² of a 47-mm diameter quartz fiber filter enclosed in a small stainless-steel chamber ( Fig. 1 (b)), at a controlled sampling flow rate of 16.7 LPM, i.e. 1 m³ per hour, provided by a closed-loop-stabilized internal pump. The sampling time may be preset from 20 minutes to 24 hours. A 1-hour time-base was used in the studies reported here. 35 At the end of the collection period, the sample flow is switched from one channel to the other. A different configuration of valves provides a small analytical flow of 0.5 LPM of ambient air through the quartz-fiber filter and then to the CO2 detector.
Before entering the chamber, the analytic air passes through a 10-liter buffer volume for ambient CO2 fluctuation averaging and a capsule filter filled with activated carbon and pleated glass fiber filter, which removes organic gases and particles from 40 the stream. High-power electrical elements above and below the quartz filter heat the sample almost instantaneously to 940°C, efficiently combusting carbonaceous compounds into CO2. Since the amount of CO2 produced is large compared to the internal volume of the system, this creates a pulse of CO2 in the analytical air stream of short duration but well-defined amplitude over the baseline.

10
This has the very great advantage that filtered ambient air may be used as the analytical carrier gas, after temporal stabilization in the internal buffer volume to remove any rapid ambient fluctuations. This feature facilitates the field deployment of the TCA08 instrument, as it does not require compressed (carrier) gas for the analysis. The carrier gas concentration of CO2 is measured before and after the combustion step and fit using a polynomial function to create the baseline. The increase in CO2 concentration above baseline is measured and integrated to give the Total Carbon content of the sample (mTC): 15 where Ccarb is a carbon calibration constant determined by a calibration with punches of ambient filters with known TC content; 2 − 1 is the combustion duration of heating 1; ( ) is the analytical air flowrate during combustion; and The CO2 sensor used in TCA08 is the LI-840A CO2/H2O Analyser (LICOR, Inc., 2016). It is an absolute, non-dispersive 5 infrared gas analyser based upon a single path, dual wavelength and thermostatically controlled infrared detection system. Concentration measurements of CO2 and H2O are based on the difference ratio in IR absorption between sample and reference signal. The CO2 sample uses an optical filter centred at wavelength of 4.26 μm (reference at 3.95 μm), while for H2O at 2.595 μm (reference at 2.35 μm). The concentration measurement of CO2 is pressure compensated and corrected for spectral crosssensitivity of water molecules with an uncertainty less than 1 ppm (at 370 ppm and 1 second signal filtering). 10 Light source life in LI-840A CO2/H2O Analyser is estimated to be 18000 hours. When light source fails the TCA instrument detects it, stops the measurements and displays Licor CO2 error status. Total Carbon content of the sample measured by TCA08 is a function of a CO2 difference between signal and background values and thus not directly connected to absolute value of CO2 (Eq. 4). This is why the TC result is less dependent on the light source drift in the NDIR detector than if the absolute 15 value is used in the calculations. During light source lifetime there is no need to preform internal standard calibration and span check for NDIR detector, as the whole system (NDIR detector + TCA08 analytic chamber) can be calibrated or validated with Carbon Calibration and Carbon Validation procedure for TCA08, which is the great benefit of this instrument. Both procedures are described in TCA08 User Manual (TCA08, 2019). Carbon calibration of TCA08 should be done once per year or after any major maintenance or modification of the system. 20

Positive and negative sampling artifacts in the TCA08 Total Carbon Analyzer
The measurement of carbonaceous aerosols using quartz-fiber filters is challenging because of the possibility of positive and negative sampling artifacts (Cheng et al., 2009;Kirchstetter et al., 2001;Subramanian et al., 2004;Watson et al., 2008). The adsorption of organic vapors (Volatile Organic Compounds, VOCs) onto quartz-fiber filters during aerosol sampling causes OC concentrations to be over-reported, while volatilization of the collected aerosols from the filter results in the loss of OC. 25 These sampling artifacts have been estimated to range between +50% for adsorption (Arhami et al., 2006a;Kirchstetter et al., 2001) to -80% for volatilization (Modey, 2001). In the European standard (EN 12341:2014(EN 12341: , 2014 this phenomenon is acknowledged but not considered in the uncertainty budget, as its magnitude cannot be quantified precisely. However, different studies of positive and negative sampling artifacts have shown that the magnitude depends on the sampling face velocity, sampling duration, filter substrate, pre-firing of filters, ambient temperature, and location with its characteristic 30 aerosol type Mader, 2003;Subramanian et al., 2004;Turpin et al., 2000). For comparison purposes, table 1 shows a comparison of sample flow, sample face velocity, sample time-base and filter media for the two different filter based instruments used in this study; Digitel Sample DHA-80 (DIGITEL Elektronik, 2012) and the TCA08. Different studies have noted that adsorption tends to be the dominant artifact at low-volume ambient sampling and shorter sample time-bases.
Consequently, we expect that volatilization effects will be small for the conditions used in the TCA08 instrument (McDow 35 and Huntzicker, 1990;Subramanian et al., 2004;Turpin et al., 2000)  Different approaches have been used to minimize the adsorption artifact and to quantify its magnitude: such as the "two filters" 5 approach (quartz behind quartz, QBQ; quartz behind Teflon, QBT); the "slicing filters" approach; regression intercept approach; and the use of denuders (Eatough et al., 1999;Watson et al., 2008). For routine measurements in monitoring networks, a VOC denuder appears to be the most practical and realistic approach (Cavalli et al., 2016;Watson et al., 2009).
Such denuders trap gaseous carbonaceous species, which would otherwise be adsorbed by quartz fiber filters and measured as a positive sampling artifact. The denuder adsorbs organic gases by diffusion to its wall surfaces, while the aerosols remain 10 suspended in the sample stream and are unaffected. The TCA08 instrument uses a honeycomb charcoal denuder to remove gas-phase OC with high efficiency at the sampling flow rate of 16.7 LPM. Residence time for one denuder monolith in the TCA08 is 175 ms. Honeycomb denuders have a high density of channels and offer a large active surface area in a compact size (Mader et al., 2001). Additionally, solid charcoal material does not deteriorate under the influence of humidity, which is an advantage compared to denuders fabricated with carbon impregnated strips (Cavalli et al., 2016). 15 Depending on location and the concentration of organic gases, some VOCs can still penetrate through the denuder and be adsorbed by the quartz-fiber filter matrix (denuder breakthrough, Arhami et al., 2006b;Zhang et al., 2013). Denuder breakthrough occurs when the time for trapping VOCs is longer than the residence time. During the sampling the actual capacity of the denuder slowly decreases, as the denuder surfaces become occupied by adsorbed VOC, leading to increased 20 times to trap all VOC. Longer residence times are needed in such occasions (2 or more denuder monoliths). To account for this artefact, the TCA08 instrument incorporates a test procedure which can be used to determine the on-site efficiency of the VOC denuder and denuder breakthrough value on site. This (QBQ) approach integrates an in-line filter in the sample inlet stream to remove filterable aerosols. The denuder is then installed in the flow stream passing to Channel 1, while Channel 2 receives the un-denuded stream (Fig. 3). 25 Figure 3: TCA08 setup when (a) sampling and (b) performing denuder efficiency test. Note that the tubing length is identical in both setups. This permits the test to be performed at a permanent installation without disturbing the inlet plumbing.
The denuder efficiency ED is determined by comparing the TC results in chamber 1 and chamber 2 as where TCF+D,n is n-th Total Carbon content measured in chamber 1, where air sample stream goes through filter above divider and denuder and TCF,n is n-th Total Carbon content measured in chamber 2, where air sample stream goes only through filter above divider. Constant gaseous OC concentration approximation through n measurements is used for calculation. TCF+D,n also represents denuder breakthrough value. 10 We developed these routines during the instrument design and performed the measurements as part of the field campaign (see Section 3). After five weeks of continuous operation with consistent TC data (see below), the measured denuder efficiency was 74%. We recommend that the denuder should be replaced or regenerated when its efficiency drops below 70% Bhagawan et al., 2015b;Gao et al., 2014). The Standard Operating Procedure for routine use of the TCA08 15 instrument recommends replacement or regeneration of the denuder honeycomb element once per month. Further, in environments with high VOC concentrations, two denuder honeycombs in series are recommended (Gregorič et al., 20192020).

Field testing measurement campaign
The TCA08 instrument was evaluated during a field measurement campaign at an urban background site in Ljubljana, Slovenia. Ljubljana is a city of ~350,000 inhabitants located at the southern edge of a geographic basin. In wintertime, it is 20 characterized by poor ventilation and frequent temperature inversions. Air quality in Ljubljana is influenced mostly by traffic and also by the combustion of biomass for household heating, both within the city and in surrounding areas (Ogrin et al., 2016).

30
The TCA08 was operated on a 1-hour time-base, sampling PM2.5 fraction at 16.7 LPM; co-located with a Model AE33 Aethalometer measuring Black Carbon aerosols in PM2.5 on a 1-minute time-base at 5 LPM. At the same location, 24-hour PM2.5 filter samples were collected in parallel with a Digitel high volume sampler for OC-EC offline analysis at two different laboratories; the Slovenian Environmental Agency (ARSO, Ljubljana, Slovenia), and IGE (Grenoble, France) both using the Sunset offline OC-EC analyzer with the EUSAAR_2 thermal protocol. Sampling start time was at 00.00 am and sampling stop 35 time was at 23.55 pm each day. During 5-minute idle period, the sampler automatically stored sampled filter and replaced it with a new one. Additionally, non-refractory organic matter (OM) measurements were also performed during the campaign with an ACSM (Aerodyne, Billerica, MA; Ng et al., 2011) on a 29-30 min time-base to derive high-time resolution measurements of the OM-to-OC ratio. The ACSM, equipped with a PM1 aerodynamic lens, was sampling through a PM1 sharp cut cyclone (SCC 1.197, BGI Inc.) at a flow rate of 3 LPM yielding a particle cut off diameter of roughly 3 μm. Furthermore, 40 the sample was driven through a Nafion dryer, upstream the instrument inlet, keeping the sample relative humidity below 40% throughout the campaign. The chemical composition dependent collection efficiency of the instrument was determined according to Middlebrook et al., 2012. Due to variability in the ACSM time-base, we gathered the data into 3h averages. All of the instruments were checked regularly and operated without interruption throughout the campaign. No data were selectively removed from the results presented in the following.
3 Results and discussions 5 Table 2  to the number of locations and the concentration range of data points. The results and discussion in this chapter is our best attempt of equivalence comparison on the available data (31 daily filters due to the limited access to the DIGITEL high volume sampler). Furthermore, we used only one set of instrument for the candidate method comparison, as both instruments are compared to the reference set of instruments after their assembly as one of the tests during final inspection procedure (in-house defined requirements for successful intercomparison between new and reference set of instruments are: 1. TCA08: TC 20 concentration range up to 75.000 ng/m 3 , slope between 0.95-1.05, R 2 above 0.98 ; 2. AE33: eBC concentrations up to 20.000 ng/m 3 ; slope between 0.95-1.05, R 2 above 0.98).    These results show that the off-line analyses of filter samples collected during the field campaign were consistent between the two external laboratories, both for the Total Carbon content of the samples, as well as for the partitioning into EC and OC components. The uncertainty uRM between the reference methods for TC 10

Orthogonal regression results
is 0.43 μg/m 3 which is well below the limit of 2.00 μg/m 3 requested for reference methods for PM mass concentration measurements (EN 16450:2017(EN 16450: , 2017. However, the difference in slope for OC and consequently for TC is around 10%, with a negative intercept value of around -0.80 μg/m 3 for OC and TC (using linear orthogonal regression model with intercept) which can indicate differences in instrument calibration, suboptimal performance of one of the instruments (featuring artefacts) 15 or inadequate filter sample handling. The filter samples were first measured in ARSO laboratory, and then shipped to IGE laboratory. Sampling, transport and storage of the filters were done according to the 16909:2017 standard (EN 16909:2017(EN 16909: , 2017. These uncertainties and the regression slope are consistent with the results of the inter-laboratory comparisons conducted in 20 the ACTRIS (Aerosol, Clouds and Trace Gases Research Infrastructure) framework, where TC repeatability (intra-laboratory measurement comparison) and reproducibility (inter-laboratory measurement comparison) were reported to be in the range of 2% -6% and 3% -13%, respectively (ACTRIS, 2016(ACTRIS, , 2017(ACTRIS, , 2018. For EC/TC, the ACTRIS exercises gave much larger reproducibility percentages, so, while there seems to be here a systematic (about 10%) difference between the two laboratory analyses, the difference is within the range expected for the OC-EC determination. The OC-EC determination is quality 25 controlled in the comparison exercise in which the Slovenian laboratory was participating. The 10% difference in TC is larger than the reproducibility and repeatability of urban background samples analyzed in this exercise , and the difference is smaller for EC (ACTRIS, 2016) . This leads us to conclude that while the differences between the laboratories can be large, the 10% difference between two laboratories using the same thermal protocol and sample protocols according to the applicable standard (EN 16909:2017(EN 16909: , 2017 is not unusual (Panteliadis et al., 2015). 30 To reduce the uncertainty of OC-EC data in further analysis, an average of TC, OC and EC measurements on filters from both laboratories is used and reported in Table 2 where 1 ≤ i ≤ 31 represent each 24 h filter during the measurement campaign.
10 Figure 5 shows a time series comparison of the 1-hour and 24-hour average TCA08 data, together with the offline analyses results for TC analysis of filter samples defined by Eq. 7. Gaps in the TCA08 measurement data are due to regular maintenance and quality control procedures (quartz filter change procedure, denuder efficiency test, etc) 15

20
These results show that on-line operation of the new TCA08 instrument with its simplified analysis method agrees very well with TC data measured by off-line thermo-optical analyses of filters. Figure 6 shows the comparison of these two datasets. The fact that these slopes are close to unity for both regression models, shows that the of TCA instrument using no catalyst and filtered ambient air as the carrier gas during analysis, has as high a combustion efficiency as the conventional offline OC-EC analyser. The intercept of 0.99 ± 0.15 μg/m 3 may indicate a positive sampling artefact as described in Chapter 2.3. The 15 positive sampling artifact attributed to VOC adsorption is more pronounced for the TCA method compared to offline filter analysis due to the difference in the sampling time, since both methods use similar face velocity (Table 1). VOC adsorption is most pronounced at 1 h sampling time and saturates in a few hours (Gregorič et al., 2019); with a 24 h sampling time, the VOC contribution is small. Over a period of 24 hours, VOCs adsorbed onto the filter can during cooler parts of the day may be desorbed during warmer parts of the day, reducing their contribution to the OC result. The contribution of positive and negative 20 artefacts for the 24 h filters is hard to estimate, while for short sample time base the positive artefact prevails and can be described with a saturation curve. Therefore, the measured offset can be accounted for by denuder breakthrough, which was measured and confirmed by the denuder efficiency test. The delta analysis between TC analysis done on 24h offline filters and online TC with 1 h time resolutions confirms this phenomenon, especially for the days with lower total carbon concentrations (lower than 5 μg/m 3 ), where the relative difference between both methods can reach 25-50 % (Figure 9). To achieve a lower 25 offset in comparison to OC-EC measurements based on 24 h filters for the sampling sites with lower concentrations of TC, two denuder monoliths or a longer sampling time base should be used. .

TCA08 method uncertainty
The uncertainty of TC data from conventional OC-EC analyzers is determined by the uncertainty of the volume of injected gaseous standard at the end of each analysis; the uncertainty of the external calibration standard; and the uncertainty of the CO2 and flow measurements during analysis (EN 16909:2017(EN 16909: , 2017. The uncertainty uTCA associated with the TC data from the TCA08 includes individual uncertainty sources of the carbon calibration factor constant Ccarb; the uncertainty of the analytic 5 flow measurement; and the uncertainty of the signal and blank CO2 peak measurement (Eq. 4). To calculate the measurement uncertainty of data from the TCA08, the CO2 signal measured by the NDIR detector is approximated with a box function, with its integral value the same as of the measured CO2 signal function (Fig. 3). The height of the CO2 box function is a linear function of TC mass collected on the filter. The relative uncertainties of Ccarb and analytic flow are determined to be 5% and 2%, respectively, while the absolute uncertainty of CO2 measurement is approximately 1 ppm. The uTC for a representative 10 range of concentrations of TC in air, using a 1h timebase and sampling at 16.7 LPM, is estimated to be where LoD is the limit of detection of the TCA08 at a sample flowrate of 16.7 LPM and sample timebase of 1h. In the 15 uncertainty budget of TC measurement with the TCA08 the following sources of uncertainties were not included: (1) Temperature and pressure variations in the sample flow as they are measured by meteorological sensor and included in TC of these effects cannot be quantified precisely, they are not considered in the uncertainty budget. Figure 7 shows the regression of the off-line thermo-optical analysis of samples for EC (from the ARSO and IGE laboratories, using the EUSAAR_2 protocol) with the 24-hour averaged BC (Aethalometer data) obtained during the field campaign period. 25

Comparison of on-line BC measurements with off-line EC filter analyses
An AE33 integrated "dual spot" real-time loading compensation algorithm was used for BC data treatment (Drinovec et al., 2015). The Pearson correlation coefficients of 0.87 and 0.88 are very similar for each of the regression models (with/without intercept). The linear relationship between EC and BC is described by slope s when using orthogonal regression model without The proportionality parameter b (Eq. 3) is compared with values taken from the literature in Table 3. These values depend on the location, the nature of the aerosol, and the thermal protocol used for analysis. The value of 0.44 which we determined in this study for an urban background site is slightly lower than values for other urban and urban background sites using EUSAAR 10 2 thermal protocol, and considerably lower than the values for rural sites. The proportionality parameter b is an effective value that features a local and a regional contribution of BC and EC. Usually, the local contribution to concentrations is dominant and the local BC and EC contributions dominate the relationship. The differences in b values presented in Table 3 show, that there is a big variation between different rural/regional background sites, and also between the urban sites. This is the reason why similar offline-to-online intercomparison is recommended for every new background site or site with strong mixture of 15 local and regional contribution.

5
Uncertainties associated with the reported Aethalometer BC mass concentrations incorporate the uncertainty in flow calibration, the uncertainty in the attenuation measurement and the uncertainty in the conversion of the attenuation coefficient to mass concentrations -constant mass attenuation cross-section approximation ( Gundel et al., 1984;Hansen, 2007, Drinovec et al., 2015, Healy et al., 2017. The overall estimated uncertainty for reported BC mass concentrations is approximately 25% (World Meteorological Organization and Global Atmosphere Watch, 2016). The EC data determined by 10 offline OC/EC analysis used in the comparison depends greatly on the thermal protocol used . In addition, the uncertainty can be determined using the procedure described in the standard EN16909:2017. The uncertainty we use has been taken as the laboratory-to-laboratory variability of 10%.

Comparison of online eOC measurements from TCA with offline OC filter analyses 15
Online eOC measurements can be derived using the above EC-BC correlation plot to assign the appropriate operational value of the parameter b; the online BC data; and the online TCA data. Figure 8 shows the correlation between online OCTC-BC and offline OC derived from the 24-hour filter samples analyzed with a thermo-optical OC-EC analyzer. These results show that when using an appropriate value of b, the "TC -BC Method" yields online data for the eOC content of ambient aerosols that agree very well with conventional offline thermal analyses. The offset i = 1.33 ± 0.18 μg/m 3 lies in the same range as that 20 determined by TC correlation analysis, which confirms that organic carbon is the origin of the offset in the correlation plots in Figs. 6 and 8. The in-depth analysis of the relative difference between OC from 24 h filters and eOC determined by online measurement as TC-bBC shown in Fig. 9 reveals that the positive artefact can be the dominant apparent source of OC for days with very low OC concentrations (< 5 μg/m 3 ) in comparison to offline 24 h filters, for which also negative artefact (desorption of VOCs) can occur. Longer sample time base or usage of two denuder monoliths in TCA08 would decrease the offset in such comparisons. Nevertheless, as the offset lies in the same range as that determined by the inter-laboratory comparison of offline filter analyses (Table 2, Fig 4.), we can assume that it can be neglected and the regression without intercept can be used 5 for intercomparison,

3.6 Comparison of OM online measurements from ACSM with offline OC from filter sampling and online eOCTC-BC
The data from an AE33 and TCA08 can be combined with an operational timebase of 1 hour, yielding eOC and eEC data with much greater time resolution than what can be achieved by the analysis of filter samples. In order to assess the high-time 10 resolution performance of this on-line technique, comparison of BC (from AE33) and TC (from TCA08) together with OM analyzed by ACSM is shown in Fig. 10. Due to variability in ACSM timings, the data was gathered into 3h averages. The chemical composition dependent collection efficiency of the Q-ACSM was calculated according to Middlebrook et al., 2012. Ambient organic-mass-to-organic-carbon ratio (OM/OC) in organic aerosol (OA) in organic aerosol (OA) is an important 5 parameter to investigate OA chemical composition. OM/OC can vary widely depending on the sources, monitoring location, season and meteorology. The lower ambient OM/OC ratios are consistent with fresh aerosol emission from traffic, while the higher values are usually observed for aged ambient oxygenated OA (Chirico et al., 2010).
The slopes s of the regressions without intercept represent average OM/OC values measured during this campaign (Fig. 11). 10 The ratios determined from comparison of daily averages of OM measurements to OC from offline filters ( Fig. 11 (a)) and to eOC from TC-bBC method ( Fig. 11 (b)) are 1.79 and 1.82 respectively. The ratio lies on the higher end of OM to OC range determined for urban environments which is 1.4 to 1.8, while for the rural sites it varies from 1.7 to 2.3 3 (Aiken et al., 2008;Gilardoni et al., 2006;Sun et al., 2009;Turpin and Lim, 2001). This is consistent with other studies in similar urban environments with close proximity of the sampling site to fresh vehicle emissions and additional contribution of biomass 15 burning (Brown et al., 2013;Turpin and Lim, 2001;Xing et al., 2013). The sampling site used in this study is mainly influenced by fresh emissions from traffic with a regionally homogeneous contribution of biomass burning for household heating (Ogrin et al., 2016). The in-depth source apportionment analysis of OA and high time resolution of OM/OC ratio from this campaign will be discussed in a different study.

20
The negative offset in the regression model with intercept ( Fig. 11 (b)) again reveals the pronounced positive sampling artefact due to adsorption of organics on quartz fiber filters for short sampling times in TCA08 method. This is not the case of the nonfilter based ACSM measurement of organic aerosol mass. The influence of such sampling artefact is noticeable only during conditions with low atmospheric loading of particulate organic aerosols. Again, the installation of two denuders monoliths or increased sample time base for TCA08 is recommended in such environment in order to minimize the influence of these 25 sampling artefacts.

Diurnal profiles of high-time resolution measurements of eOC, eEC, and eEC/TC ratio
The coupling of TCA08 and Aethalometer instruments offers new opportunities to investigate the short-term variability of carbonaceous aerosols, and the factors that control their atmospheric concentrations such as source variability and/or atmospheric (dynamic/photochemical) processes. For this purpose, diurnal profiles of organic carbon and elemental carbon concentrations were calculated for each hour of the day (Fig. 12 (a)), separately grouped for working days (Monday 5 to Friday) and for weekends (Saturday and Sunday). The diurnal variation of eOC and eEC for this urban background environment is strongly influenced by the temporal patterns of emissions from traffic and biomass burning (domestic heating) during wintertime. Two traffic peaks can be observed for working days in OC and EC concentrations; the first one observed during morning rush hours (between 6:00 and 10:00 LT) and the second in the afternoon, after 16:00 LT.
Between the two peaks, (e.g. between 10:00 and 16:00), OC and EC concentrations decrease due to atmospheric dilution 10 in the increasing mixing height of the planetary boundary layer (Ogrin et al., 2016). During the weekend the morning traffic peak disappears, while the evening one remains present. Peaks in average eEC to average TC ratio are concomitant with the eEC peaks which is aligned with the EC-rich pattern of traffic emissions (Fig 12 (b)). Average eOC and eEC values during the measurement camping campaign were 7.3 ± 4.9 μg/m 3 and 1.3 ± 1.3 μg/m 3 , respectively, which is

Summary
We present the newly developed Total Carbon Analyzer model TCA08, which offers measurement of the concentrations of total aerosol carbon continuously with high time resolution as rapid as 20 min. Two parallel flow channels provide continuous 5 operation: while one channel analyzes, the other collects the next sample. Thermal analysis by flash-heating of the sample collected on a quartz fiber filter efficiently converts all the particulate carbon to CO2. The increase in CO2 concentration above baseline in a flow of analytic air is measured by an integrated NDIR detector. When the TCA08 is combined with an AE33Aethalometer, the TC-BC method yields eOC-eEC data with much greater time resolution than that offered by the analysis of filter-based samples. In this study, we show results from these instruments combined on an operational timebase 10 of 1 hour and compare them to conventional 24h filter measurements of EC and OC, and high-time-resolution measurements of organic aerosols with an ACSM. The correlation analysis showed very high agreement between for eOC = TC-bBC and eEC= bBC derived by the TC-BC method, to OC-EC analysis using EUSAAR2 thermal protocol on 24h filters and OM from ACSM. The value of the calibration proportionality parameter b can be derived for the desired OC-EC thermal protocol to obtain high time resolution eOC and eEC data. 15 These two instruments are automatic, rugged, and designed for unattended operation in field monitoring situations.
Measurements can be done in different PM size fractions (PM1, PM2.5, PM10) The combined data may be analyzed to examine repetitive diurnal patterns, reflecting both anthropogenic inputs of carbonaceous aerosols to the atmosphere; production of secondary aerosols; as well as atmospheric processing and dispersion into mixing layers of varying depth. Additional analyses 20 can compare these results between workdays and weekends, seeking patterns of human activity that may reflect changes in traffic or industrial emissions. Studies such as this, requiring large numbers of closely-spaced data points, are greatly facilitated by online instruments.

Acknowledgments
This work was partly funded by the EUROSTARS grant E!8296 TC-BC.

Code/Data availability
The data used in this publication is available upon request to the corresponding author (martin.rigler@aerosol.eu). 30

Author contribution
MR, LD, and GM designed the study, MR, GL, LD, GM, and IS performed and analysed TC, BC and OM online measurements. MR, GL, LD, GM, AV, AP, and AH were involved in the new instrument development. JJ, JS, JB, IK, and JT preformed OC/EC measurements on offline filters. All authors contributed to the scientific discussion. 35

Conflicts of interest
At the time of the research, M. Rigler, L. Drinovec, G. Lavrič, and G.Močnik were also employed by the manufacturer of the Aethalometer and Total Carbon instruments. Other authors declare no conflict of interest. The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision 40 to publish the results.