2.1.1 Site description

The PEcoRINO (Probing Ecosystem Responses Involving Notable Organics) study consisted of chemical and meteorological observations over the Chequamegon–Nicolet National Forest (CNNF) at the WLEF-TV very tall tower US-PFa Ameriflux site in Park Falls, WI (45.945^∘ N, 90.273^∘ W) (Davis et al., 2003), from 6–30 September 2020. The landscape surrounding the tower is composed of grasslands, woody wetlands, and deciduous and evergreen forests as determined from the National Land Cover Database (NLCD) 2016 (Dewitz, 2020). This location has been used for many eddy covariance (EC) studies concerning the role of surface heterogeneity in heat and carbon exchange (Desai et al., 2008, 2010, 2015; Xu et al., 2017; Bakwin et al., 1998), as well as a multi-institutional, intensive field campaign focused on the role of atmospheric boundary layer responses to scales of spatial heterogeneity in surface–atmosphere heat and water exchanges (Butterworth et al., 2021). Recently, this site has been used to investigate the exchange of O₃ and formic acid and the role of in-canopy chemistry in observed fluxes (Vermeuel et al., 2021).

2.1.2 Meteorological and O₃ measurements

For the PEcoRINO study, routine US-PFa site measurements of 10 Hz wind speed and temperature (Model K Style Probe; ATI, Inc.) at 30 m were used, along with relative humidity (HMP155, Vaisala) and solar irradiance (LI-190; LI-COR, Inc.) (Desai, 1996). Continuous 1 Hz measurements of O₃ mixing ratios using a photometric analyzer (Model 49i; Thermo Fisher) were made at a sampling height of 30 m through an inlet composed of type 1300 Synflex ($\mathrm{3}/{\mathrm{8}}^{\prime \prime }$ i.d.), drawing 30 standard liters per minute (slpm) of ambient air. The photometric analyzer was calibrated by the generation of a calibration curve every 3 d using an O₃ calibration source (Model 306 Ozone Calibration Source; 2B Technologies).

2.1.3 VOC measurements

A high-resolution proton-transfer-reaction time-of-flight mass spectrometer (HR-PTR-ToFMS) (Vocus; Aerodyne Research Inc. and Tofwerk AG) (Krechmer et al., 2018) made continuous 10 Hz measurements of VOCs at 30 m through a separate 45 m, $\mathrm{3}/{\mathrm{8}}^{\prime \prime }$ i.d. perfluoroalkoxy alkane (PFA) inlet, drawing between 25–30 slpm of ambient air in order to maintain turbulent flow in the sampling line. The Vocus subsampled from the main inlet with a 5 slpm bypass through a PFA tee located immediately in front of the Vocus capillary inlet. The sample flow into the Vocus instrument was 100 sccm with the remaining bypass flow exiting to the pump. The sample inlet was constantly heated to 40 ^∘C and was wrapped in aluminum foil throughout to avoid any potential inlet photochemistry. Attached to the front of the inlet was a PFA funnel wrapped in aluminum foil to avoid significant moisture draw during precipitation events. The focusing ion–molecule reactor (FIMR) of the Vocus was held at 1.5 mbar, and the FIMR's front and back were held at 400 and 35 V, respectively. The Vocus big segmented quadrupole (BSQ) ion guide was maintained at 215 V to allow for a higher transmission of lower-molecular-weight molecules such as methanol. Spectra with a mass range of $m/Q$ 10–504 and a resolution of $\sim \mathrm{5000}m/\mathrm{\Delta }m$ were collected, allowing for highly resolved determination of peaks in the mass spectrum. A total of 1474 peaks were integrated using the Igor-Pro-implemented (WaveMetrics, Inc.) Tofware software package (Aerodyne Research Inc. and Tofwerk AG). A three-point calibration curve using a non-methane VOC (NMVOC) standard (Apel Riemer Environmental, Inc.) and ultra-zero (UZ) air (AI UZ300, Airgas) was collected every 4 h to record dynamic in-field calibration factors for select compounds. Zeros were also performed during calibrations. Calibration standards and concentrations are presented in Table S1 in the Supplement. Calibrations were not added to the entire inlet but were rather introduced by overflowing the subsampling line. Calibration factors have been shown in lab studies to be insensitive to water content for the Vocus (Krechmer et al., 2018) and for this specific instrument (Kilgour et al., 2022). In the field the calibration factors were on average 900, 1500, 800, and 5000 cps ppbv⁻¹ for isoprene, MT (α-pinene), SQT (β-caryophyllene), and acetone, respectively, with coefficients of variation of less than 10 % across species throughout the study.

A GC system designed for online atmospheric analysis (ARI GC; Aerodyne Research Inc.) was used to separate isomers of reactive compounds (e.g., MT) by coupling it to the Vocus to create a GC-ToF-MS system. A previous version of the instrument is described in detail in Claflin et al. (2021) but will be briefly described here. In the GC-ToF-MS system, sample air passes through a multi-stage thermal desorption preconcentration (TDPC) system (Aerodyne Research, Inc.) to collect and focus analyte species from ambient air before separation on the chromatographic column. The GC sample flow rate is controlled via a mass flow controller (MFC) and is held at 100 sccm for 10 min, resulting in a 1 L ambient sample per GC cycle. Before collection onto the TDPC, sample gases passed through a sodium sulfite (Na₂SO₃) oxidant trap to remove reactive gases, such as ozone, to reduce sampling artifacts that can occur at high mixing ratios (Helmig, 1997). After passing through the oxidant trap, the sample is then collected onto a multi-bed sorbent tube (Tenax TA/Graphitized Carbon/Carboxen 1000, Markes International), which is then forward-purged with zero gas for 2 min to reduce the level of trapped water. After the post-collection purge, the sample is then transferred to a multi-bed, narrow bore cold trap (Tenax TA/Carbopack X/Carboxen 1003, Markes International) for focusing before injection onto the GC column. Both the sample collection and focusing are conducted at sub-ambient (20 ^∘C; optimized to avoid condensation) temperatures through the use of a Peltier thermoelectric cooler. After focusing, the flow is then injected onto a GC column, which then undergoes a programmed temperature ramp from 35–225 ^∘C. The column used in this study resolves non- to mid-polarity VOCs including hydrocarbons, oxygenates, and some nitrogen- and sulfur-containing compounds (MXT-624, Restek). For this study, the ARI GC was used to resolve C₅–C₁₂ hydrocarbons, DMS, and some oxygen-containing VOCs. For the majority of this work, the total chromatograph times were 10 min, which allowed for the full resolution of all MT species at this site. A subset of 14 sets of chromatograms was collected with a 20 min chromatograph length to also speciate SQTs and larger isomers, although this was not used for routine analysis, as it was too time-demanding. Chromatogram peak areas were fitted using the Igor-implemented TERN software v2.2.9 (Aerodyne Research Inc.) (Isaacman-VanWertz et al., 2017, 2022), and the resulting values, in units of cts s per extraction where cts are signal counts, were multiplied by the ToF extraction rate (24.4 kHz) to calculate quantifiable cts.

In the field, the operation was divided between a 10 Hz ambient collection solely through the HR-ToF-MS and collection via the GC-ToF-MS system, herein referred to as real-time (RT)-Vocus and GC-Vocus sampling, respectively. The GC-Vocus collection routine is described in the Supplement .

3.1.1 Meteorology

The CNNF canopy experienced a variety of meteorological and physical (e.g., leaf stage, leaf area) conditions during the sampling period (Fig. S4). For example, there was a wide range of observed daytime maxima (7.9–26 ^∘C) in ambient temperature (Fig. 1a). Wind speed (Fig. 1b) generally peaked in the late afternoon (13:00–16:00 CDT) with an average daytime value of 2.3 m s⁻¹ (average daytime maximum of 2.8 m s⁻¹) and a range in daytime maxima of 1.9–6 m s⁻¹. Winds primarily originated from the west (Fig. 1c) with large, abrupt changes in wind direction (WD) concurrent with low wind speeds, indicative of periods of a stable boundary layer. There were many precipitation events throughout the study (Fig. 1d), with 24 September onward experiencing many rainy, misty, and cloudy days. Also shown in Fig. 1d is the LAI product from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on board the NASA Terra satellite for the pixel over the WLEF-TV site (Savtchenko et al., 2004). Throughout September, we also observed a decrease in LAI from 4 to 0.8 m² m⁻², indicating loss of leaves or declining leaf greenness throughout the month, both of which may be crucial in controlling the exchange of BVOCs via surface area required for emissions and/or deposition and uptake. Measurements of PPFD (Fig. 1e) indicate an attenuation of solar radiation in the last week of the study, suggesting increased cloud cover at the site during that time.

Figure 1Meteorology at the measurement site: (a) temperature, (b) wind speed, (c) wind direction, (d) precipitation (black line) and LAI (grey dots), and (e) photosynthetic photon flux density (PPFD).

Download

3.1.2 Mixing ratios of ambient chemical species

We first examine the impact of physical and meteorological changes that occurred during this period on the mixing ratios of reactive terpenes (MT, SQT, and isoprene), DMS, and O₃ (Fig. 2). Uncertainties for all mixing ratios are presented as shaded regions and were calculated by propagating the uncertainty from the fraction lost to the inlet (Fig. S1), accuracy in calibration standards and mass flow controllers, and the standard deviation (1σ) in field calibrations. This produced average uncertainties of 11.1 %, 29.0 %, 12.0 %, and 8.0 % for ΣMT, ΣSQT, DMS, and O₃, respectively. The average isoprene uncertainty was 17.1 % during the day and 35.8 % at night and is further discussed in this section.

Figure 2Mixing ratios of (a) ΣMT, (b) isoprene, (c) ΣSQT, (d) DMS, and (e) O₃ from 6–30 September 2020 with shaded uncertainties. Leaf senescence generally began around 16 September, and leaf abscission began around 21 September.

Download

Figure 2a shows the time series of summed MT (ΣMT) (black line) as detected through the MH⁺ ion ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}^{+}$ ($m/Q$ 137.1325). Concentrations of ΣMT peaked in the evening due to late-afternoon emissions and buildup thereafter due to reduced vertical mixing and oxidative removal. We observe an increase in ΣMT concentrations following 21 September, most likely due to senescing leaves, as we will discuss in Sect. 4.2. From on-site visual assessment, senescence defined by changes in deciduous leaf color (and thus the end of the growing season) in this region generally began around 16 September, and leaf abscission began around 21 September (Fig. S4). In addition, MODIS LAI decreased by more than half (from 4 to 1.5 m² m⁻²) by 21 September, indicating a large portion of the region's leaf area losing greenness. Since this region is a mixed forest with species of varying lengths of developmental cycles, these assessments are approximations based on a few studied trees and may not be reflective of individual species that undergo mid- or late-autumn senescence. Following the beginning of leaf abscission of deciduous trees, higher concentrations of MT were observed. Prior to 21 September, peak daily mixing ratios were regularly below 0.5 ppbv but were above 1.0 ppbv following 21 September, peaking at 1.4 ppbv. Average concentration diel profiles for all species in Fig. 2 for periods before and after 21 September are presented in Fig. S5 to highlight these changes. The campaign-average [ΣMT] was 0.26 ppbv, which is close to autumn measurements of [ΣMT] at the SMEAR II station in a boreal coniferous forest in Hyytiälä, Finland (0.25 ppbv), where the latter would be expected to have a higher density of MT-emitting species (Hakola et al., 2003).

Data show that the speciation of MT changed throughout September (Fig. 2a). The colored regions in Fig. 2a show fractions of the major identified MT isomers using the GC-Vocus. Prior to senescence, α-pinene initially comprised on average ∼40 % of the total emissions, ∼32 % of the total MT emissions during senescence, and decreased to ∼20 % during abscission. β-pinene showed a slight increase in MT fraction throughout the month, increasing from 44 % to ∼57 % as leaves moved from the mature to abscission stages, respectively. Similarly, camphene also showed an increase in relative proportion, shifting from mature (16 %) to senescent (23 %) stages. Speciation from the GC-Vocus is described more in Sect. 3.2.

Figure 2b shows mixing ratios of isoprene, as determined by the ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ ion ($m/Q$ 69.06988), throughout the month of September. The ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ signal required correction due to the presence of n-aldehyde fragments in the ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ chromatogram. Although some of the n-aldehyde contribution in the chromatogram was determined to be from reactions of ozone with the system sorbent tubes due to unconditioned Na₂SO₃ used in the oxidant trap (Sect. 3.2), n-aldehydes were also observed in RT-Vocus measurements, and the known fragmentation of these n-aldehydes to ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ required a correction. Corrections were performed by taking advantage of the consistency in signal ratios of fragment ions to parent ions (M⁺) across the GC- and RT-Vocus (as in α-pinene, Fig. S6). To correct the ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ for n-aldehydes, the peak area ratio of ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ to M⁺ of heptanal, octanal, and nonanal were multiplied by the M⁺ RT-Vocus signal to get the corresponding n-aldehyde ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ signal (Fig. S7). The sum of these n-aldehyde ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ signals was then subtracted from the total ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ signal to get an “isoprene-only” signal, which was then calibrated for isoprene from in-field calibrations. The contribution of n-aldehydes made up 36 % (148 ppt correction) and 59 % (140 ppt correction) of the daytime and nighttime ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ signal, respectively. S3 describes this correction and associated uncertainties in more detail with uncertainties calculated from accuracies in calibrant standards and mass flow controllers, as well as 1σ uncertainty of isoprene calibration factors and the GC peak area ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ to M⁺ ratio. The daily peak in isoprene concentrations was variable, ranging from 0.13–1.1 ppbv, and the campaign average was 0.16 ppbv, a value between year-averaged measurements of isoprene in a northern temperate forest in MI in 2001 and 2002 (0.1 and 0.5 ppbv, respectively) (Karl et al., 2003). The two forests may not serve as direct comparisons, but comparison to the MI forest range does show the high interannual variability of isoprene in mixed northern temperate forests. Figure 2c shows the mixing ratios of ΣSQT detected at ${\mathrm{C}}_{\mathrm{15}}{\mathrm{H}}_{\mathrm{25}}^{+}$ ($m/Q$ 205.1951) and calibrated for β-caryophyllene. There is no clear diurnal cycle in ΣSQT, and the campaign average [ΣSQT] was 7.2 pptv, a value over a factor of 6 lower than late-summer observations in a primarily coniferous forest (∼44 pptv) where mixing ratios are expected to be higher and may serve as an upper bound (Bouvier-Brown et al., 2009).

Figure 3Time series of monoterpene oxide concentrations: (a) C₁₀H₁₆O calibrated for camphor, (b) C₉H₁₄O calibrated for nopinone, (c) C₁₀H₁₄O calibrated for thymol, and (d) C₁₀H₁₆O₂ and (e) C₁₀H₁₆O₃ calibrated for cis-pinonic acid. Measurement uncertainties are presented as shaded regions.

Download

Observations of DMS at CNNF, detected as C₂H₇S⁺ ($m/Q$ 63.0263) in the Vocus, are presented in Fig. 2d. The diel profile in DMS is consistent, displaying an evening maximum around 21:00 CDT and a minimum in the early morning (∼ 05:00–07:00 CDT). This profile of evening buildup is indicative of a compound that has a late-afternoon source that extends into the evening. The short lifetime of DMS in the early morning suggests removal due to boundary layer mixing or advection, since the lifetime of DMS against OH is too long to account for this loss (∼1 d). DMS at CNNF is low, with an average mixing ratio of 7.7 pptv for the entire observation period. There was no dependence of [DMS] on leaf stage or LAI, suggesting that DMS may not be sourced from plants or are from plants that did not show a change in LAI. Vertical mixing ratio profiles of terrestrial DMS have been recorded at an Amazon Forest between September 2010–January 2011, with mixing ratios < 160 pptv. In that study, there was a clear enhancement of [DMS] in the late afternoon and at warmer temperatures, and there was a strong nocturnal accumulation within the canopy and closer to the forest floor, indicative of light-independent soil emissions (Brown et al., 2015). Temperate coniferous ecosystems can have DMS sourced from trees. Vertical distributions of DMS in a loblolly pine forest near Atlanta, Georgia, also showed enhanced [DMS] closer to the forest floor and at night (∼12 pptv) compared to the day (∼4 pptv) (Berresheim and Vulcan, 1992), with abundances similar in magnitude to this study. The authors of the Georgia study attribute this distinction to reduced photooxidation at night and concluded that DMS emissions were from the pine trees. However, soil emissions, although highly dependent on microorganisms in the soil, have been proven to provide a small source in other ecosystems (Goldan et al., 1987; Banwart and Bremner, 1975; Yang, 1996) and can also explain the magnitude of observed mixing ratios at the site. Without leaf-level or soil chamber measurements of DMS we cannot definitively state whether DMS comes from the soils or trees.

Figure 2e shows half-hourly averaged mixing ratios for O₃ as measured by the photometric analyzer. The average [O₃] for the whole study was 23 ppbv, and the day with the highest measured [O₃] was 23 September 2020, a day that also experienced the peak in ambient temperature (26 ^∘C). There was a period of sustained [O₃] maintaining >20 ppbv from 19–25 September, which correlates with periods of higher temperature in that time range.

Figure 4Observed fluxes of (a) ΣMT, (b) C₁₀H₁₆O, (c) isoprene, and (d) ΣSQT from 6–30 September 2020 with shaded uncertainties. Also presented are diel profiles of (e) ΣMT, (f) C₁₀H₁₆O, (g) isoprene, and (h) ΣSQT. Diel profile shaded regions are 95 % confidence intervals.

Download

Monoterpene oxides exhibited enhancements in ambient mixing ratio during the seasonal transition. Mixing ratios of the monoterpene oxide C₁₀H₁₆O, detected as C₁₀H₁₇O⁺ ($m/Q$ 153.1638) and calibrated for as camphor, exhibited a similar behavior as the time series of [ΣMT] (Fig. 3a). Following 21 September, mixing ratios are enhanced on average by over a factor of 2 compared to the period before 21 September. This suggests that C₁₀H₁₆O and ΣMT follow the same mechanisms of primary emissions and/or that C₁₀H₁₆O is an MT oxidation product. This type of observation from a mixed temperate forest has been published before: observations from a mixed forest in New England show that summertime emissions of C₁₀H₁₆O were closely related to those of ΣMT, although emissions were negligible by September (McKinney et al., 2011). A recent study using a Vocus PTR-ToF-MS system at the coniferous Landes Forest found the diel profile of MT to be consistent with C₁₀H₁₇O⁺, with an average evening [C₁₀H₁₆O]:[ΣMT] of 0.03, compared to 0.08 in this study. The authors suggested that C₁₀H₁₆O was partially sourced from directly emitted camphor or an MT oxidation product (Li et al., 2021), although at the CNNF site there are few species that directly emit camphor and those that do (north white cedar, white spruce) emit camphor in low amounts (Helmig et al., 1999). Figure 3 provides concentrations of MTOs measured with the RT-Vocus at a collection of other ions that were determined to be lightly oxidized products of MT oxidation (C₉H₁₅O⁺, C₁₀H₁₅O⁺, ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{2}}^{+}$, ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{3}}^{+}$). Since there was no in-field calibration for monoterpene oxides, we approximate uncertainties for C₉H₁₅O⁺, C₁₀H₁₅O⁺, and C₁₀H₁₇O⁺ to have the same relative uncertainty as ΣMT and ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{2}}^{+}$ and ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{3}}^{+}$ to have the same relative uncertainty as ΣSQT based on assumptions of volatility. The C₉H₁₅O⁺ ion ($m/Q$ 139.1117) is commonly assigned as nopinone, one of the main products formed during β-pinene ozonolysis (Atkinson and Arey, 2003; Lee et al., 2006a). In this study the ion was calibrated for nopinone and the identity confirmed by GC observations (Sect. 3.2). Nopinone had the highest recorded concentration among the observed monoterpene oxides. Other common assignments of observed monoterpene oxides are the major α-pinene ozonolysis product pinonaldehyde (C₁₀H₁₆O₂), which can appear as a parent ion ($m/Q$ 169.1223; ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{2}}^{+}$) or as a fragment ($m/Q$ 151.1118; C₁₀H₁₅O⁺), and pinonic acid ($m/Q$ 185.1172; ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{3}}^{+}$), a minor product of α- and β-pinene ozonolysis (Atkinson and Arey, 2003). The C₁₀H₁₅O⁺ ion was identified by GC measurements to be primarily thymol, but ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{2}}^{+}$ and ${\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{O}}_{\mathrm{3}}^{+}$ could not be identified due to retention times outside of our collection window. All monoterpene oxides in Fig. 3 show an enhancement in concentrations following 21 September, suggesting either a similar physical mechanism in the enhancement of emissions or increases due to the oxidation of concurrently increased MT. Concentration diel profiles for these monoterpene oxides for periods before and after 21 September are presented in Fig. S8 to highlight changes due to the seasonal transition.

3.1.3 Eddy covariance fluxes of BVOCs

The effect of the seasonal transition on forest–atmosphere exchange of BVOCs is shown in Fig. 4, which presents the quality-controlled fluxes of ΣMT, C₁₀H₁₆O, ΣSQT, and isoprene. F_ΣMT (Fig. 4a) regularly exhibited daytime maxima less than 2.5×10¹⁰ molec cm⁻² s⁻¹ prior to 21 September (Fig. 4a). Following 21 September, emissions were enhanced (maximum 9.3×10¹⁰ molec cm⁻² s⁻¹) and were in agreement with observed mixing ratios during the same time period. There was no strong increase in temperature during this period (Fig. 1a), indicating that factors other than leaf temperature control emissions of ΣMT following leaf senescence. While there is a source area shift for F_ΣMT from the west half to primarily southwest for pre- and post-21 September, respectively, it is unclear if this shift caused emission enhancements, since both footprints overlap according to flux footprint prediction (FFP) parameterizations (Kljun et al., 2015) (Fig. S9). The diurnal profile of F_ΣMT (Fig. 4e) shows that emissions follow a temperature profile, with emissions peaking in the late afternoon (13:00–15:00 CDT). The only other measurement of F_ΣMT in this region was in July 1993 with values ranging from 0.46–9.1×10¹⁰ molec cm⁻² s⁻¹ based on leaf-level measurements and estimates of area foliage densities (Geron et al., 1994; Isebrands et al., 1999). Based on a synthesis of MT speciation in the US, a high-end estimate of F_ΣMT in northern WI would be 1.2×10¹¹ molec cm⁻² s⁻¹ (Geron et al., 2000).

The fluxes of ΣMT and C₁₀H₁₆O are closely related, with $F_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}}$ enhanced by nearly a factor of 2.5, on average, following 21 September (Fig. 4b). A regression of $F_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}}$ and F_ΣMT provides an r² of 0.83 and a slope of 0.041 ($F_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}}:F_{\mathrm{\Sigma }\mathrm{MT}})$, showing that the two processes are correlated (Fig. S10). Branch enclosure measurements of ponderosa pine trees at Manitou Forest in CO, USA, from 21 August–4 September 2008 showed consistent emission ratios of C₁₀H₁₆O:MT of ∼0.1 (Kim et al., 2010), suggesting direct emissions of C₁₀H₁₆O. The New England canopy-scale flux study over a mixed temperate forest recorded a summertime C₁₀H₁₆O:MT of ∼0.03 (McKinney et al., 2011). This highlights the observed range in this ratio potentially due to ecosystem differences or in-canopy loss of C₁₀H₁₆O. However, in the McKinney et al. (2011) study, $F_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}}$ and  F_ΣMT were not correlated (r²=0.18).

F_isoprene exhibited high day-to-day variability (Fig. 4c; daytime maxima: 0.067–1.9×10¹¹ molec cm⁻² s⁻¹) and was not enhanced after 21 September, consistent with presumed cessation of in situ leaf synthesis. The variability in F_isoprene was partially controlled by ambient temperature, as observed by F_isoprene enhancement on warmer days (e.g., 6 September; 22–23 September) and suppression on colder days (e.g., 8–9 September). The diurnal cycle of F_isoprene (Fig. 4g) peaked with air temperature and was low or zero outside of daylight hours, implying that parameterizations of isoprene emissions based on sunlight and temperature are appropriate during this season. Since there was no measurable flux from the parent masses of heptanal, octanal, and nonanal, we are confident that there is minimal to no added error from corrections to the isoprene signal, since the aldehyde signals do not vary with w and therefore should not contribute to ${\mathrm{C}}_{\mathrm{5}}{\mathrm{H}}_{\mathrm{9}}^{+}$ flux. Previous area-averaged fluxes in this region from the July 1993 study, where leaf temperatures reached 35 ^∘C and caused high emissions, were 2.8×10¹¹ molec cm⁻² s⁻¹, providing an upper bound for the observations here.

Figure 4d presents, to our knowledge, the first canopy-scale fluxes of SQTs in a mixed temperate forest. Similar to F_isoprene and F_ΣMT, F_ΣSQT also demonstrated a diel temperature dependence, although the day-to-day variability was not as pronounced (Fig. 4d). In addition, the time series of F_ΣSQT was not dependent on leaf stage, and we did not observe an enhancement in emissions post-21 September. Daily maxima of F_ΣSQT ranged from 0.028–1.8×10⁹ molec cm⁻² s⁻¹. No measurements of F_ΣSQT have been performed near this site for comparison, although branch enclosure measurements of summertime north temperate pine suggest canopy-scale SQT emissions up to as much as 2.5×10¹⁰ (Holzke et al., 2006), providing an upper bound nearly 2 orders of magnitude larger than observations. Depending on the chemical lifetime of the dominant species in observed SQTs, the magnitude and profile of F_ΣSQT can be influenced by in-canopy ozonolysis. A study measuring above- and within-canopy ambient concentrations of SQTs in the Amazon showed that 46 %–61 % of SQTs by mass undergo in-canopy ozonolysis (Jardine et al., 2011), and a multi-layer gas dry deposition model using observations from the SMEAR II station showed that ∼70 % of SQTs are removed within the canopy due to chemical oxidation. We estimate the impact of within-canopy ozonolysis on F_ΣSQT in Sect. 4.1.

The error bars presented in Fig. 4a–d were determined through calculations of flux LoD (Sect. 2.3). The campaign average uncertainties for MT, C₁₀H₁₆O, isoprene, and SQT were 25 %, 33 %, 27 %, and 37 %, respectively. Table 1 provides a summary of observed mixing ratios and fluxes throughout the study.

Table 1Summary of mixing ratios (MR) and fluxes (F) of select compounds during the PEcoRINO study.

Download Print Version | Download XLSX

4.2.1 Changes in ambient oxidation chemistry

Reduction in gas-phase oxidant concentrations would reduce the rate of in-canopy oxidation and increase the net detected flux. To estimate how a change in in-canopy oxidation would impact measured F_MT, we rearrange Eq. (4) to get the fraction of F_MT,net to F_MT,emitted, which can be used to assess in-canopy loss under different oxidant loadings:

Since O₃ increases after 21 September, the concentration of OH would need to decrease after 21 September for this to increase the net emitted MT as observed. Using observed O₃ concentrations and a high estimate of 1×10⁷ and 5×10⁶ molec cm⁻³ OH before and after 21 September, respectively, and a τ_canopy of 5 min, the ratio of $\frac{F_{\mathrm{MT},\mathrm{net}}}{F_{\mathrm{MT},\mathrm{emitted}}}$ would achieve a 1.1× increase from 0.79 to 0.89 before and after 21 September, a modest increase. Since the parameterized F_MT (Fig. 9a) is a factor of 1.5 higher in the period before 21 September relative to after, this change due to oxidation is small. Still, a reduction in OH from attenuated solar radiation in the latter parts of the month (Fig. 1e) has a small impact on sustained [ΣMT]. The approach used here is a simple parameterization and not a replacement for more comprehensive 1-D vertical models such as the Canopy Atmospheric Chemistry Emission model (Bryan et al., 2012) where changes in concentration with time are explicitly treated. However, since we use conservative estimates of τ_oxidation and a fixed τ_canopy, we are directly comparing scenarios where in-canopy oxidation would have the strongest impact.

4.2.2 Contributions from soil and leaf litter

A potential explanation for the high observed F_ΣMT and [ΣMT] may be enhanced emissions from the forest floor. Throughout autumn, the forest floor can be a significant contributor to VOC flux from the decomposition of leaf litter (Isidorov et al., 2010; Greenberg et al., 2012) or from microbial activity in exposed soils (Mäki et al., 2019). Aaltonen et al. (2011) observed forest floor BVOC emissions in a boreal coniferous forest peaking in early summer and autumn with emissions of MT averaging 6.2×10⁸ molec cm⁻² s⁻¹ and containing primarily α-pinene, camphene, and Δ³-carene with a negligible contribution from isoprene and SQT. Hellén et al. (2006) recorded forest floor emissions from the same site, noting that the highest MT emissions occurred in the spring and autumn with values up to 4.6×10¹⁰ molec cm⁻² s⁻¹ composed of α-pinene, camphene, Δ³-carene, limonene, and β-pinene. However, this behavior has been observed primarily in boreal forests when needleleaf litter is high and may not be reflective of the CNNF at the observed stage in the needleleaf cycle during the PEcoRINO study. Bare soils can serve as a source of MT in temperate regions, although that contribution ($\sim \mathrm{1.0}\times {\mathrm{10}}^{\mathrm{7}}$ molec cm⁻² s⁻¹) is generally negligible (Trowbridge et al., 2020) compared to foliar and leaf litter emissions. Based on these studies we assume that decaying leaf litter and soils may provide a small contribution to the early-autumn rise in F_ΣMT and to the change in MT speciation during the summer to autumn transition. However, without soil and floor observations we cannot quantitatively conclude their impact on observed F_ΣMT. This study shows a need for measurements of forest floor emissions in temperate regions or in mixed forests that can confirm the magnitude and speciation of these emissions.

4.2.3 Leaf degradation or enhanced reactive carbon synthesis

A third explanation for elevated MT emissions is the leaf senescence process itself. During stages of senescence, changes in the biomechanical properties of the epidermis cuticle make diffusion of certain compounds easier through a degraded epidermal layer, which could lead to increased emissions of BVOCs, although this has been primarily recorded for hydrophilic species (Mozaffar et al., 2018). Degradation of plant structural components can also generate leaf wounds during senescence, which may enhance MT emissions if the lamina is wounded (Portillo-Estrada and Niinemets, 2018). In addition, the degradation of cells provokes desiccation of the lamina, driving the emission of volatile compounds stored within the cytosol or in specialized reservoir organs (Portillo-Estrada et al., 2020). These two processes have been witnessed in trees of the genus Populus, including aspen, which make up 25 % of the acreage of CNNF as of 1996 (Haugen et al., 1998).

A study measuring VOC emissions via the eddy covariance method over a poplar plantation of 12 different genotypes observed a peak in the emissions of 25 VOCs at the beginning of September and the onset of senescence (Portillo-Estrada et al., 2020), with increases in OVOCs of 1–2 orders of magnitude and modest increases in MT. The authors also observed a cessation of isoprene flux beginning with the onset of leaf senescence. This agrees with observations at CNNF with increases in F_ΣMT and $F_{{\mathrm{C}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}\mathrm{O}}$ coincident with sharp drops in F_isoprene near to 21 September. Still, if aspens within the flux footprint were controlling BVOC emissions it would be expected that other OVOCs such as methanol or acetone would present an enhancement in emissions (Portillo-Estrada et al., 2020). However, observed fluxes of methanol and acetone post-21 September exhibited a net sink (Fig. S19). It is possible that the net flux of those OVOCs was controlled by other enhanced routes of deposition, such as uptake to water films or biotic processes (Laffineur et al., 2012; Fulgham et al., 2020), making it difficult to assess changes in the gross source of methanol and acetone.

A final but most speculative enhancement route related to the senescence process may be due to the increased synthesis of, and the need to mitigate, reactive oxygen species (ROS) within leaves. During senescence, ROS are generated in response to stress, which depletes antioxidants and damages cells (Jajic et al., 2015). To reduce ROS, plants may increase synthesis of reactive carbon (Affek and Yakir, 2002). Monoterpenes serve as ROS scavengers during senescence (Chen and Cao, 2008), and production of MT in response to increased oxidative stress has been shown in the leaves of evergreen oak (Quercus ilex) (Loreto et al., 2004). Although there are no studies confirming enhanced MT synthesis in senescing plants at this site, there is the possibility for this biochemical pathway to be upregulated during this phenological stage, and future studies aimed at assessing plant reactive carbon synthesis in response to senescence and oxidative stress would be invaluable.