A portable fast gas chromatography-mass spectrometry workflow for quantification of biogenic volatile organic compounds with humidity‑dependent sensitivity correction Authors: Alice Claude, Carmen Kalalian, Christophe Boissard, Valérie Gros, Juliette Leymarie, Ruben Puga-Freitas Published: 2026 ABSTRACT Biogenic volatile organic compounds (BVOCs) regulate key atmospheric and ecological processes but remain challenging to quantify in the field due to reactivity, mixture complexity, and variable ambient humidity. A portable fast Gas Chromatography-Mass Spectrometry (GC–MS) workflow was developed, coupling needle-trap (NT) sampling with a compact GC–MS system to enable BVOC quantification with minimal logistics. A 15-compound gas standard (28–91 µmol mol⁻¹) was generated under controlled water vapor pressures (ea = 0.7–2.3 kPa), and two independent calibrations sessions were performed to assess reproducibility. Seven compounds (α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene) exhibited linear responses (R² > 0.85) across humidity levels. Calibration sensitivity showed a strong exponential dependence on ea, and compound-specific correction functions were derived. Concentrations estimated from portable GC–MS showed excellent agreement with Proton Transfer Reaction Mass Spectrometry (PTR–MS) reference values (R² > 0.94). Six additional VOCs were detected but not reliably quantified under 3- min GC conditions. The method provides a 15- min sampling-to-result cycle and requires only 40 mL of air. The working limit of quantification was approximately 40 ppb, making the method primarily suited to chamber or enriched‑air applications where BVOC levels exceed this threshold.•Portable NT + fast GC–MS workflow for in-field BVOC quantification.•Humidity-dependent exponential sensitivity functions used for correction.•Good agreement with PTR–MS reference concentrations (R² > 0.94). Specifications table Subject areaEnvironmental Science.More specific subject areaBiogenic volatile organic compound analysis.Name of your methodHumidity-corrected portable fast GCMS– quantification of BVOCs.Resource availabilityTorion T-9 portable GC/MS (PerkinElmer) ; Clairion Needle Trap Basic Kit (PerkinElmer). Subject areaEnvironmental Science.More specific subject areaBiogenic volatile organic compound analysis.Name of your methodHumidity-corrected portable fast GCMS– quantification of BVOCs.Resource availabilityTorion T-9 portable GC/MS (PerkinElmer) ; Clairion Needle Trap Basic Kit (PerkinElmer). Background Volatiles organic compounds (VOCs) play essential roles in atmospheric chemistry and ecological interactions [1,2]. There is a wide variety of VOCs, a first-order distinction separates anthropogenic from biogenic VOCs (BVOCs), which together influence air quality, organismal communication, and biogeochemical cycles. Owing to their high vapor pressures (typically ≥ 0.01 kPa at 20 °C) and resulting volatility, VOCs are predominantly observed in the atmosphere, but they also occur in soils [3], waters [4], and living tissues [5]. Plants are the major terrestrial source of BVOCs, exceeding 760 Tg C yr⁻¹, with isoprene and monoterpenes as dominant classes [6]. BVOCs mediate biotic interactions [7] and plant responses to abiotic stress [8], yet their structural diversity, reactivity, and spatio-temporal variability make field measurements particularly challenging [9,10]. Robust enclosure-based assessments require inert materials (glass, perfluoroalkoxy alkane (PFA), polytetrafluoroethylene (PTFE)), temperature control to avoid condensation, and oxidant scrubbing (notably ozone) to limit sampling artefacts [9,[11], [12], [13], [14]]. Online methods such as Proton Transfer Reaction (PTR) coupled to a Quadrupole Mass Spectrometer (MS) or a Time Of Flight (TOF) MS analyzer [14,15] provide rapid, highly sensitive measurements and have been deployed in the field [16], but they are logistically demanding (mass, power, air conditioning) and lack chromatographic separation, which complicates isomer-level identification, especially for monoterpenes that share nominal masses. Consequently, many studies trap VOCs on porous sorbents (e.g., Tenax-TA, Carboxen), which cover a wide range of VOC affinities [17], and analyze them offline by Gas Chromatography (GC)–MS or GC–FID (Flame Ionization Detector) after thermal or solvent desorption [15,18]. Portable instruments attempt to bridge these modes by coupling fast GC separations to compact detectors (Snyder et al., 2016) such as surface acoustic wave (SAW) sensor (zNOSE, Techmondial), metal oxide semiconductor gas sensor (New Cosmos electric), photoionisation detection (PID) systems (Alphasense, Defiant technologies FROG-5000), or ion trap MS systems (Perkin Elmer Torion ® T-9, FLIR Griffin™ G510, Hapsite® ER), enabling in-field identification and quantification. The PerkinElmer Torion® T-9 has been successfully used for rapid volatile profiling [[19], [20], [21]], but its quantitative performance for VOCs, particularly across contrasting humidity, requires validation. Here, a portable Torion® T-9 workflow (coupled with needle-trap sampling) is compared with PTR-MS, and a humidity-corrected calibration is proposed to enable quantitative field analysis in potentially BVOC-rich matrices (e.g., plants, human breath). Method details Calibration gas and humidity control For sensitivity determination and calibration-factor estimation, a multicomponent VOC standard (15 compounds at nominal 1 µmol mol⁻¹; SIAD, Bergamo, Italy, see Supplementary Table S1) was supplied to a Gas Calibration Unit (GCU; Ionicon Analytik GmbH, Innsbruck, Austria). The standard was diluted with zero air generated by the GCU’s built-in catalyzer to yield target concentrations of 28, 38, 48, 70, and 91 µmol mol⁻¹. Water-vapour pressure (ea) was set to approximately 0.7, 1.2, 1.9, 2.2, and 2.3 kPa using the GCU dew-point generator. The GCU’s mass-flow controllers and dew-point control were used to regulate both concentration and ea. The outlet flow was 150 mL min⁻¹. Two independent calibrations were performed on separate days. Between the two calibration sessions, the Torion® T‑9 GC–MS system was fully shut down and restarted, allowing assessment of day‑to‑day reproducibility and the robustness of the calibration procedure. In Fig. 1, Fig. 2, Fig. 3, these appear as the “first” and “second” calibrations, respectively.Fig. 1Calibration curves across humidity and days for seven target VOCs. Peak area versus standard concentration for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene at multiple water vapor pressures ea (0.7, 1.2, 1.9, 2.2, 2.3 kPa) and across two independent calibration days. Points are means of duplicate NT samples. Symbols denote calibration day (° = first; △ = second), and shading encodes ea as indicated in the legend. Dotted lines are linear fits (computed per day and ea condition). Fits were typically R² > 0.85 with intercepts not significantly different from zero.Fig 1 dummy alt textFig. 2Exponential dependence of calibration sensitivity on water vapor pressure for seven VOCs. Calibration sensitivity (slope of peak area vs. concentration from Figure 1) as a function of partial water vapor pressure, ea (kPa), for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene. Each point represents one sensitivity estimate obtained at a given ea and calibration day. Symbols denote calibration day (° = first; △ = second). Red curves show exponential fits of sensitivity versus ea (parameterized by a and b; Table 1), with R² > 0.9 for all seven compounds.Fig 2 dummy alt textFig. 3Calculated vs. true concentrations for seven VOCs using humidity-corrected calibration. Calculated concentrations for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene plotted against the true standard concentrations (ppb). Concentrations were obtained from portable GC-MS peak areas using the humidity-corrected calibration functions (compound-specific parameters in Table 1, see Eq.1–2). Each point represents the mean of duplicate NT samples acquired at a given partial water vapor pressure ea and calibration day. Symbols denote calibration day (° = first; △ = second) and shading encodes ea (0.7–2.35 kPa). Blue lines show linear least-squares regression between estimated and true concentrations, and the shaded bands represent the 95% confidence intervals of the regressions. The regression equation (slope) and R² for each compound are reported directly on the panels.Fig 3 dummy alt text Calibration curves across humidity and days for seven target VOCs. Peak area versus standard concentration for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene at multiple water vapor pressures ea (0.7, 1.2, 1.9, 2.2, 2.3 kPa) and across two independent calibration days. Points are means of duplicate NT samples. Symbols denote calibration day (° = first; △ = second), and shading encodes ea as indicated in the legend. Dotted lines are linear fits (computed per day and ea condition). Fits were typically R² > 0.85 with intercepts not significantly different from zero. Exponential dependence of calibration sensitivity on water vapor pressure for seven VOCs. Calibration sensitivity (slope of peak area vs. concentration from Figure 1) as a function of partial water vapor pressure, ea (kPa), for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene. Each point represents one sensitivity estimate obtained at a given ea and calibration day. Symbols denote calibration day (° = first; △ = second). Red curves show exponential fits of sensitivity versus ea (parameterized by a and b; Table 1), with R² > 0.9 for all seven compounds. Calculated vs. true concentrations for seven VOCs using humidity-corrected calibration. Calculated concentrations for α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene plotted against the true standard concentrations (ppb). Concentrations were obtained from portable GC-MS peak areas using the humidity-corrected calibration functions (compound-specific parameters in Table 1, see Eq.1–2). Each point represents the mean of duplicate NT samples acquired at a given partial water vapor pressure ea and calibration day. Symbols denote calibration day (° = first; △ = second) and shading encodes ea (0.7–2.35 kPa). Blue lines show linear least-squares regression between estimated and true concentrations, and the shaded bands represent the 95% confidence intervals of the regressions. The regression equation (slope) and R² for each compound are reported directly on the panels. Sampling of volatile organic compounds The calibration gas was sampled directly at the GCU outlet. The line was split three ways as close as possible to the GCU using PTFE T-connectors. Orthogonal branches were routed either to PTR-MS (online comparator) or to the Clarion NT-pump interface (PerkinElmer, France) for offline, portable GC–MS analysis. Online sampling to PTR-MS was performed at 80 mL min⁻¹. For offline sampling, Custodion® needle traps (NTs) packed with a tri-bed sorbent (Tenax TA, Carboxen 1016, Carboxen 1003) were used. Two NTs were connected in parallel to the NT-pump and gas was sampled for 10 min at 4 mL min⁻¹ per NT (see Supplementary Figure. S1). Sampling commenced once the PTR-MS signal (cps) was stable. Analysis of volatile organic compounds The protocol used for PTR-MS is described elsewhere [22] and used as a reference analytical method during calibration. Offline analyses were performed on a Torion® T-9 portable GC–MS. NTs were thermally desorbed in the GC injector. Separations used a 3- min method with a 2 °C s⁻¹ ramp from 50 °C to 290 °C (initial and final temperatures 50 °C and 290 °C). Detection employed the instrument’s toroidal ion-trap mass spectrometer. Compound identification was carried out in Chromion® software (version 2.0.4.0; PerkinElmer Inc., Waltham, MA, USA). Targeted ions and retention-time (RT) windows were:•α-Pinene: m/z 93, 92, 91, 79 (relative intensities 100, 40, 38, 28); RT 46–52 s.•Benzene: m/z 78 (relative intensities 100); RT 20–26 s.•Chlorobenzene: m/z 112, 114, 77 (relative intensities 100, 39, 25); RT 37–43 s.•1,2-Dichlorobenzene: m/z 146, 148 (relative intensities 100, 64); RT 55–61 s.•Isoprene: m/z 67, 68, 53 (relative intensities 100, 55, 55); RT 10–16 s.•Toluene: m/z 91, 92 (relative intensities 100, 64); RT 28–36 s.•o-Xylene: m/z 91, 106, 105 (relative intensities 100, 63, 29); RT 42–48 s s. •α-Pinene: m/z 93, 92, 91, 79 (relative intensities 100, 40, 38, 28); RT 46–52 s. α-Pinene: m/z 93, 92, 91, 79 (relative intensities 100, 40, 38, 28); RT 46–52 s. •Benzene: m/z 78 (relative intensities 100); RT 20–26 s. Benzene: m/z 78 (relative intensities 100); RT 20–26 s. •Chlorobenzene: m/z 112, 114, 77 (relative intensities 100, 39, 25); RT 37–43 s. Chlorobenzene: m/z 112, 114, 77 (relative intensities 100, 39, 25); RT 37–43 s. •1,2-Dichlorobenzene: m/z 146, 148 (relative intensities 100, 64); RT 55–61 s. 1,2-Dichlorobenzene: m/z 146, 148 (relative intensities 100, 64); RT 55–61 s. •Isoprene: m/z 67, 68, 53 (relative intensities 100, 55, 55); RT 10–16 s. Isoprene: m/z 67, 68, 53 (relative intensities 100, 55, 55); RT 10–16 s. •Toluene: m/z 91, 92 (relative intensities 100, 64); RT 28–36 s. Toluene: m/z 91, 92 (relative intensities 100, 64); RT 28–36 s. •o-Xylene: m/z 91, 106, 105 (relative intensities 100, 63, 29); RT 42–48 s s. o-Xylene: m/z 91, 106, 105 (relative intensities 100, 63, 29); RT 42–48 s s. Quantification used the integrated peak area of the reconstructed-ion chromatogram (RIC) built from the targeted ions within the respective RT windows. Calibration curves obtained by PTR-MS for the same seven VOCs across the four humidity levels are provided in Supplementary Figure S2. PTR-MS sensitivity remained effectively independent of humidity across the tested range (Supplementary Figure S3), which confirms its suitability as a stable reference instrument for cross-calibration. Statistical analysis All statistical analyses were performed in R (version 4.4.0 [23]). Calibration relationships between peak area and gas-standard concentration were evaluated using ordinary least-squares (OLS) linear regression, with the intercept constrained to zero, as expected for calibration data where a concentration of zero should yield a peak area of zero. Each regression was performed separately for each compound, humidity level, and calibration day. The slope of each linear fit was taken as the compound-specific sensitivity at a given partial water vapor pressure ea. The dependence of sensitivity on ea was modelled using non-linear least-squares (NLS) with an exponential function of the form: y = a exp (bx), where y is sensitivity, x the partial water vapor pressure ea and a and b are compound-specific fit parameters. Goodness of fit for both linear and non-linear models was assessed using the coefficient of determination (R²) and inspection of residuals. Concentrations predicted from portable GC-MS were compared with PTR-MS reference concentrations using linear least-squares regression with the intercept constrained to zero, and agreement was quantified using the resulting R² values. All regressions were performed independently for each compound, and no data transformations other than those implied by the calibration functions were applied. Method validation Quantification performance across the calibration mixture Among the 15 VOCs (including both biogenic and non-biogenic compounds) present in the calibration gas mixture (Supplementary Table S1), seven compounds showed calibration behavior suitable for reliable quantification with the portable GC–MS system (Figure 1). These included α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene and o-xylene. For all seven targets, calibration curves were linear across 28–91 µmol mol⁻¹, with R² > 0.85 and intercepts not significantly different from zero. This performance was reproducible across the two independent calibration days (first vs second) and across humidity levels (Figure 1). Using the slope of each linear calibration at each humidity (Fig. 1) as the sensitivity metric, a clear exponential dependence of sensitivity on partial water vapor pressure ea (kPa, Fig. 2) was observed for the seven compounds (R² > 0.90; Table 1). This relationship was modelled using an exponential function:(1)Sensitivity=aeb×ea And then estimated concentrations from Reconstructed Ion Chromatogram (RIC) peak areas (A) via:(2)[VOC]=Aaeb×ea Where [VOC] is the gas-phase mixing ratio (ppb), A is the integrated peak area of the reconstructed-ion chromatogram (RIC) within the compound-specific RT window, ea is the partial water-vapour pressure (kPa), and a,b are the compound-specific parameters reported in Table 1. At a given ea, sensitivity remained compound-dependent (e.g., α-pinene tended to exhibit higher sensitivity than isoprene), consistent with their differing ionization and chromatographic responses under the fast-GC/MS conditions used. Detected but non-quantifiable compounds In addition to the seven reliably quantified compounds, the portable system consistently detected six other VOCs present in the standard mixture: acetone, 1,2,4-trichlorobenzene, 2-butanone (methyl ethyl ketone), acenaphthylene, phenol, and acetic acid. However, their calibration curves showed either non-linear responses, poor reproducibility, or strong humidity-dependent variability, leading to unreliable quantification under the tested conditions. Unreliable calibration for these compounds primarily reflected chromatographic and ionization limitations. In particular, several showed partial co-elution with adjacent peaks during the 3- min fast-GC separation, exhibited low signal-to-noise ratios close to the instrument’s limit of quantification, or displayed humidity-dependent ionization behavior typical of oxygenated and polar VOCs. Given these limitations, these six compounds were excluded from quantitative analysis but remain fully detectable, indicating potential for quantification after method optimization (e.g., extended GC runtime, specific SIM ion selection, or compound-tailored humidity correction). Comparison with PTR-MS For the seven VOCs that met the quantitative criteria (α-pinene, benzene, chlorobenzene, 1,2-dichlorobenzene, isoprene, toluene, and o-xylene) concentrations computed from portable GC–MS peak areas using the humidity-dependent sensitivity equations showed excellent agreement with the PTR-MS reference across the tested range of partial water vapor pressure. Specifically, the relationship between estimated concentrations (portable GC–MS) and the true gas concentrations (PTR-MS) was highly linear with R² > 0.94 for all seven compounds (Figure 3), with slopes between 0.95 and 1.02, indicating no systematic bias (i.e., no over- or under-estimation), confirming that the humidity-corrected calibration is broadly applicable and that needle-trap sampling coupled to a portable fast GC–MS provides robust quantitative performance for targeted BVOCs under field-relevant humidity conditions. Limitations The method described here was validated under specific experimental conditions, which define its current scope of applicability. First, although needle-trap (NT) sampling combined with portable fast GC–MS provided robust quantification for several VOCs, the limit of quantification (LOQ) of approximately 40 ppb may not be sufficient for environments where BVOC concentrations routinely fall below this threshold [24]. Increasing sampling time may lower the LOQ, but breakthrough volumes were not fully characterized beyond 20 min, and breakthrough may depend on humidity and compound volatility. Second, the method was validated only within a partial water-vapor pressure range of 0.7–2.3 kPa. Under very dry air (< 0.7 kPa), the LOQ increased and method performance may degrade. Such dry conditions are uncommon in most BVOC field environments, where surrounding biomass typically implies non-negligible ambient humidity, but they can occur in specific settings (e.g., arid sites, cold and very dry air masses, or air-conditioned laboratory streams). At very high humidity although, no artefacts were observed under the conditions tested, the risk of condensation in sampling lines or NTs remains a potential limitation, particularly in cooler environments. Third, the fast GC–MS method used a 3- min separation, which limits chromatographic resolution. Under these conditions, some compounds showed partial co-elution, lower signal-to-noise ratios, or humidity-dependent ionization effects, preventing reliable quantification. These issues may be mitigated by adjusting the temperature ramp or extending the runtime. Although the sorbents used in the needle-trap (Tenax TA, Carboxen 1003, Carboxen 1016) are hydrophobic, strong polarity or very high volatility may still influence trapping efficiency for certain analytes. Finally, the method was evaluated using controlled standard mixtures and PTR-MS as a reference. Additional validation may be required when applying the workflow to complex field matrices (e.g., plant emissions with rapid temporal variability, anthropogenic mixtures, or highly reactive BVOCs) and under extreme temperature or humidity conditions not explored in this study Table 1.Table 1Parameters of the humidity-dependent sensitivity function for seven VOCs. Sensitivity (slope of peak area vs. concentration; Figure 1) was fitted to Eq. 1 with partial water-vapor pressure ea (kPa). The table lists a and b (fit parameters) and R² (goodness of fit). These parameters are used to convert reconstructed-ion chromatogram (RIC) peak areas A to concentrations according to Eq.2.Table 1 dummy alt textParametersCompoundabR²(+)-α-Pinene3.8291.0160.96Benzene3.5710.5530.90Chlorobenzene2.3620.7880.921,2-Dichlorobenzene1.6650.9690.94Isoprene0.7731.0040.93Toluene4.020.6950.91o-Xylene3.4560.8540.97 Table 1Parameters of the humidity-dependent sensitivity function for seven VOCs. Sensitivity (slope of peak area vs. concentration; Figure 1) was fitted to Eq. 1 with partial water-vapor pressure ea (kPa). The table lists a and b (fit parameters) and R² (goodness of fit). These parameters are used to convert reconstructed-ion chromatogram (RIC) peak areas A to concentrations according to Eq.2.Table 1 dummy alt textParametersCompoundabR²(+)-α-Pinene3.8291.0160.96Benzene3.5710.5530.90Chlorobenzene2.3620.7880.921,2-Dichlorobenzene1.6650.9690.94Isoprene0.7731.0040.93Toluene4.020.6950.91o-Xylene3.4560.8540.97 Parameters of the humidity-dependent sensitivity function for seven VOCs. Sensitivity (slope of peak area vs. concentration; Figure 1) was fitted to Eq. 1 with partial water-vapor pressure ea (kPa). The table lists a and b (fit parameters) and R² (goodness of fit). These parameters are used to convert reconstructed-ion chromatogram (RIC) peak areas A to concentrations according to Eq.2. Ethics statements None. CRediT author statement Alice Claude: Investigation, Writing – original draft, Writing – review & editing. Carmen Kalalian: Conceptualization, Investigation, Writing – review & editing. Christophe Boissard: Conceptualization, Investigation, Supervision, Writing – review & editing. Valérie Gros: Conceptualization, Supervision, Writing – review & editing. Juliette Leymarie: Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review & editing. Ruben Puga- Freitas: Conceptualization, Formal analysis, Investigation, Methodology, Supervision, Visualization, Writing – original draft, Writing – review & editing. Related research article None. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.