| 22 Sep 2026
Reliable VOC Monitoring in Industrial Coastal Areas: Uncertainty Quantification and Variability in the Berre Basin
Abstract. Continuous measurements of 23 volatile organic compounds (VOCs) were deployed at three sites surrounding the lake of Berre industrial basin (southern France). This dataset was used to characterize VOC variability in a complex industrial environment and to establish a robust methodological framework for long-term monitoring. The sites, located near major Seveso-classified industrial platforms, exhibit distinct chemical signatures driven by contrasting source influences, mesoscale circulations, and topographic effects. The monitoring strategy relied on dual online GC-FID systems and was supported by an extensive quality assurance and quality control (QA/QC) program, including routine calibration, assessments of linearity, repeatability, blank correction, and environmental influences. A comprehensive uncertainty calculation scheme was developed by combining elements of the NF EN 14662-3 Benzene standard with the ACTRIS network’s standard operating procedure (GAW report No.281, WMO, 2023) and implemented within the Incert'R R-based tool to enable automated uncertainty estimation for each VOC concentration. The resulting expanded uncertainties, calculated for a reference concentration of 5 µg m-3, ranged from 9 % to 25 % depending on the compound. Except for 1,2-dichloroethane (25 %), all VOCs exhibited uncertainties below 20 %, while benzene showed an uncertainty of 9 %, well below the 25 % limit required by the benzene standard. Overall, these uncertainty levels confirm the reliability of the dataset for interpreting concentration patterns across the three sites. Having established this, the analysis of VOC concentrations reveals pronounced spatial and temporal contrasts across the sites: Berre-l'Étang is strongly affected by land-sea breeze recirculation, leading to recurrent morning and evening peaks and exceptionally high values during industrial maintenance periods, while Lavéra shows frequent episodic enhancements dominated by chlorinated and light hydrocarbon compounds. Port-de-Bouc records lower and less structured concentrations, consistent with more intermittent plume impacts. These findings highlight the importance of combining high time-resolution measurements with harmonized QA/QC procedures and systematic uncertainty estimation to reliably detect high variability of VOC concentration in industrial regions. The resulting dataset provides a rigorous basis for subsequent source-apportionment analyses and for improving emission inventories and atmospheric modelling in one of Europe's most industrially impacted coastal areas.
General comments:
This manuscript " Reliable VOC Monitoring in Industrial Coastal Areas: Uncertainty Quantification and Variability in the Berre Basin" presents an extensive automatic uncertainty calculation to strength the QA/QC procedure at three different monitoring stations located in the vicinity of industrial activities at the lake of Berre with different wind regimes. In this study, 23 VOCs are monitored with broad boiling point range and affinity to sorb to surface making challenging to access for each compound/instrument/site their uncertainties. A tool has been developed for this purpose to calculate these uncertainties and can be further applied at other (industrial) monitoring sites. Additionally, the instruments used in this study are gaining interest in VOC monitoring community, thus the results presented here can support further development. VOC fingerprints for the main peak events at the different stations are also described. The monitored VOCs are summarized in Table 2, nevertheless, a better overview of the different stations could have been given in Fig. 3. The discussion on BTEX could also include further discussion on T:B ratios.
The manuscript deserves publication given the extensive work done to estimate the uncertainties. These validated datasets allow further studies such as source apportionment analyses at these sites. However, several points outlined below require further clarification and detail
Specific comments:
L191: I would suggest adding a list of the 23 compounds selected for this study in the SI or refer to table 2.
L196: Material of the sampling line?
L220: "The base sensitivity is calibrated annually" – are previous works using these FID analyzers available to show that the BS remains stable throughout the year?
L230: "a canister with 48 VOCs (full details in SI) and the permeation tubes present in the AirmoCal", please refer to the corresponding figure/table in the SI.
L261: In the section "Impact of sample filter and sampling line", are the 6 measurements considered for the calculation or is the first measurement automatically discarded?
Table 1: Are the linearity results also available for the other VOCs and for the other stations?
L285-287: "(i) partial breakthrough of the preconcentration at higher loadings, […] response at elevated concentrations" Are the breakthrough tests available? Additionally, it would be interesting to indicate the upper end of the working range as it is relevant in Table 2 with the high concentrations observed for example at Martigues-Lavéra for isopentane, 1,2-DCE. Do you observe carryover/memory effect after measuring high concentrations, for example after a contamination event? If so, how do you process the data?
Table 2: Either use DL or LoD throughout the manuscript. As for some of the compounds the contamination events drive the maximum concentration, are the data normally distributed? If not the case, using median values would be probably more adapted.
L370: Concerning the temperature sensitivity, the station temperature (variability) should be mentioned earlier in the material and methods section, e.g., in 2.1..
Figure 2: It is mentioned that the uncertainties vary from instrument to instrument. How does it look like at the other stations? Are the results based on 5 measurements?
Figure 3: Total concentration of the 23 VOCs is indicated. That would be interesting to highlight the contributions of the main compounds/VOC classes/industrial activities throughout the year. Are the data aggregated hourly? Do you observe a strong difference in the concentration ranges between weekdays and weekend? The Fig. S5 is more informative, especially with the comparison between the different stations.
L436-437: " These pronounced hourly peaks are strongly modulated by the multidirectional wind regime characteristic" Due to the different wind regimes at the stations, polar rose plots would be easier to visualize/highlight the results.
L520-521: "using different reference years does not bias the interpretation" I agree with the approach but the seasonal variation, as most of the summer and autumn data are missing at Port-de-Bouc is also considered when the first and last semester are compared.
L541: " Aromatic VOCs, in particular BTEX, were selected for focused analysis in this study". This would be interesting to discuss the T/B ratio as they should be characteristic of industrial emissions.
Technical corrections:
L174 – not need to give the abbreviation for NMHCs as already indicated in the introduction
L179: indicate the meaning of "LoD" abbreviation
L273: "each concentration tested (4, 21, 38, and 0 µg m⁻³)" for the last number, which concentration did you use?
L293: I would suggest using "Limit of detection" instead of detection limit (DL).
L309: Citation should be included differently concerning the Ambient air quality reference.
L374: Use K-1 if referring to the temperature
L549: " surpassing the EU annual limit", indicate here the limit concentration, especially since the new EU air quality directives of 2024 target 3.4 ug/m3 for benzene.