1. Introduction

Nitrogen dioxide (NO₂) is one of the most important constituents of the Earth's planetary boundary layer, with a strong influence on both air quality and climate. It plays a crucial role in the formation of tropospheric ozone [1] and aerosols [2] which directly affect the Earth's radiation budget. Moreover, it is also toxic to humans at higher concentrations causing short- and long-term health problems [3]. In order to regulate and monitor NO₂ levels, the European Environment Agency has set a limit value for the annual average concentration of NO₂ of 40 µg/m³. Additionally, concentrations should not exceed 200 µg/m³ more than 18 times in the same time period [4]. Although total emissions of nitrogen oxides decreased by 48% between 2005 and 2020, approximately 49,000 premature deaths in 2020 were attributed to exposure to nitrogen dioxide levels above the WHO guidelines from 2021 [5]. The annual average limit of 10 µg/m³ set by the World Health Organization (WHO) was exceeded in 2022 by all 35 reporting countries in Europe, while nine reporting countries even reported values above the European Union (EU) annual limit [6].

The largest anthropogenic contributors to nitrogen dioxide production are the combustion of fossil fuels in conventional vehicles and the heating of buildings. Therefore, it is critical to closely monitor urban NO₂ levels, considering the complex factors that influence its abundance. The atmospheric lifetime of NO₂ depends on solar radiation-induced photolysis, resulting in seasonal, daily, hourly and more transient variations. Mapping NO₂ concentrations in cities with high NO₂ levels, focusing on annual and seasonal variations as well as changes caused by weather conditions or traffic volume, would help to implement tailored strategies to reduce NO₂ levels while considering urban development.

At present, nitrogen dioxide concentrations are typically measured using either various alterations of differential optical absorption spectroscopy (DOAS) [7,8], cavity ring-down spectroscopy [9] or chemiluminescence [10]. Chemiluminescence provides a high sensitivity (e.g., 50 ppt [10]) but typically has a response time of at least one minute. Several other methods with a high temporal resolution and a high sensitivity are employed to measure nitrogen dioxide at a single point in space. For example, laser-induced fluorescence reaches sensitivity limits of 5-10 ppt in 60 s [11].

Methods that use an optical cavity to enhance absorption typically provide a lower detection limit with a one-second acquisition time. A prominent method is cavity ring-down spectroscopy, which achieves a detection limit of 100 pptv [12], 30 ppt/1 s [13] and 14.5 pptv/10 s [14]. Cavity enhanced (CE) DOAS achieved a detection limit of 0.4 ppbv with an integration time of 0.5 s in NO₂ concentration measurements [8]. In addition, cavity attenuated phase shift instruments were demonstrated with a sensitivity 0.3 ppb in 10 seconds [15]. However, it is important to notice that the CE experiments are not conducted with an open path. Instead, the sample is collected from a single point in the field and then directed into the cavity.

DOAS-based methods can detect small concentrations over long absorption paths in the atmosphere. Detection limits of 0.1 ppb were reported almost 45 years ago [16]. Since then, multiple different iterations [17], such as MAX-DOAS [7,18] have been developed and successfully used in field experiments over kilometer-long paths [19].

In addition, NO₂ measurement instruments based on the combination of cavity enhancement and DOAS have been developed. Iterative CE-DOAS can achieve a detection limit of 0.4 ppbv with an integration time of 0.5 s [8].

An alternative technique that only recently has advanced from exclusively laboratory applications to field-deployed atmospheric measurements is dual-comb spectroscopy (DCS). This method merges the advantages of conventional Fourier transform spectroscopy (FTS), high spectral resolution and broad spectral range, with the high accuracy of laser frequency combs and short measurement times (up to 10⁶ shorter than conventional FTS, see, e.g., [20]) for environmental monitoring. Currently, DCS is successfully employed for atmospheric sensing in the near- and mid-infrared (IR) [21–23], allowing real-time monitoring of gas concentrations. Long interaction path lengths of several kilometers, as they are typically used for field sensing in the IR region, result in strong absorption signals and with that in advanced sensitivity. A long interaction path length can also be advantageous to monitor large areas, resulting in path-integrated concentration values. The performance of environmental monitoring in the IR spectral region allows the detection of multiple relevant greenhouse gases such as CO₂, H₂O, and CH₄. Other gases of high atmospheric relevance, such as NO₂ and ozone (O₃), exhibit strongly absorbing electronic transitions in the visible and ultraviolet wavelength range. Expanding DCS into the visible and ultraviolet wavelength range will enable the detection of these important trace gases, providing the mentioned fourfold advantage of high spectral resolution, broad spectral coverage, short measurement times and high accuracy [24].

In this work, we apply DCS for field monitoring with all its advantages, for the first time to our knowledge in the visible spectral region. The short sampling times, typically on the µs scale, for single-trace measurements are one of the key advantages of DCS in field applications, as environmental disturbances are mitigated. Averaging over several individual data sets maintains high sensitivity while reducing the influence of atmospheric turbulences. The use of a laser allows active sensing with around-the-clock measurements, independent of natural light sources. In addition, DCS with its coherent laser radiation and its multiplex advantage requires only low power (µW levels), reducing the sensitivity to link losses caused by beam divergence and atmospheric turbulence. Different scattering processes (e.g., Rayleigh and Mie scattering) along the absorption path lead to broadband losses, which can be accounted for by a low order polynomial baseline correction [25]. The wavelength range from 514 nm to 520 nm was specifically chosen for the presented experiments to avoid any sharp water absorption lines which would mask any trace gas absorption.

2. Experimental setup

Figure 1(a) depicts the experimental set-up. The system was designed to be compact and mobile for measurements in different locations to permit flexible beam paths. Therefore, the optical part is placed on a vibration damped, rolling cart of size 120 cm x 75 cm allowing easy transport. An enclosure, built around the optical components, passively mitigates air and temperature fluctuations for different measurement conditions.

 

Fig. 1. (a) Schematic of the mobile DCS set-up. Laser A is operated at a stabilized repetition rate of 80 MHz – δ with a detuning of δ = 4 Hz while Laser B is stabilized at 80 MHz. The lasers are separately frequency-doubled via second harmonic generation (SHG) in LBO crystals and afterwards superimposed with a beam combiner (BS1). There are three interferograms recorded by fast photodiodes and an oscilloscope card (DAQ), respectively: a reference (PD_Ref), an iodine calibration (PD_Iod) and the field measurement trace (PD_Field). For the field experiment, we use a single-mode fiber to guide the laser to an external collimator for flexible sampling paths. (b) Backside of the reflector array during a field campaign on a humid day. In this configuration the beam travels three times back and forth (first beam on the right of the photograph) along the path under investigation. See text for more details.

Download Full Size | PDF

Two commercially available ytterbium-doped fiber frequency combs with a center wavelength of 1035 nm and a FWHM of 13 nm drive two separate second harmonic generation (SHG) stages based on LBO crystals. These convert the frequency comb into the visible wavelength range covering the interval from 514 nm to 520 nm. The resulting green laser radiation enables the detection of distinctive absorption features of NO₂ and, in conjunction with the absence of absorption features of other atmospheric gases, provides an effective tool for determining the absolute NO₂ concentration.

The repetition rates of Laser A and B are stabilized to 80 MHz - 4 Hz and 80 MHz, respectively. The carrier envelope offset (CEO) frequency is not stabilized to minimize the technological complexity of the system. The two laser output beams are spatially superimposed by a beam combiner (BS1). One output after the first beam combiner is split again (BS2), to determine the incident laser spectrum (PD_Ref) and for wavelength calibration (PD_Iod), both recorded with fast photodiodes. An iodine reference with a distinctive sawtooth absorption profile serves for wavelength calibration. The second output of BS1 is used for the field measurement and is coupled into a single-mode fiber, guiding the laser onto an off-axis parabolic mirror, with a reflective focal length of 5 cm for the collimation of the beam to a 1/e² diameter of 14 mm, serving as the sender port. The visible beam exits the building and is directed to a mobile reflector array, placed at various locations on the university campus at distances ranging from 50 m to 90 m from the sender port. The average power of the sender port is less than 1 mW in all measurements, qualifying the dual-comb spectrometer for laser class 2, where no safety equipment is required for the outdoor part of the setup.

The mobile reflector set-up is mounted on a stable tripod optimized for field measurements in geodesy (see Fig. 1(b)). The reflector array consists of up to five collocated protected silver mirrors in standard mirror mounts (compare to Fig. 1(b)). A similar array is used on the sender/receiver side with a maximum of four mirrors in order to achieve a path length of up to 900 m in a multi-pass geometry. In total, for all nine reflections, the loss of the signal is approximately 35%. The receiver port consists of a 250 mm aperture lens and a fast photodiode (PD_Field). The folded path geometry enables a high spatial resolution of 90 m and a high detection sensitivity due to the long path length of almost a kilometer. Comparable field-deployed systems for nitrogen dioxide detection have a lower spatial resolution of several kilometers (e.g., [7]).

The superposition of the frequency comb outputs causes an interference pattern that down-converts the optical signal to the radio frequency domain, by the scaling factor m = f_rep,1/δ. This allows the detection of time-dependent optical transmission signals with fast photodiodes. A more detailed description of the dual-comb spectroscopy principle can be found in the reviews [26,27]. The three photodiodes implemented in the presented setup detect signals with a bandwidth of 150 MHz and have a rise time of 2.3 ns for the incident laser beam (PD_Ref), for the wavelength calibration (PD_Iod) and for the atmospheric field measurement (PD_Field), respectively. They are connected to a 16 bit, 250 MSa/s digitizer card, allowing simultaneous detection of all three interference patterns in the time domain. The Fourier transformation of these simultaneously measured interferograms results in corresponding spectra in the radio frequency domain, which are up-converted to optical frequencies, depending on the repetition rate detuning. To increase the sensitivity and the signal-to-noise ratio, averaging over multiple recorded spectra is performed, which requires a post-processing routine.

Instabilities resulting from the unlocked CEO frequencies of the two laser sources and atmospheric turbulence, are corrected in post-processing. To compensate for fast jitter [28] and slow thermal drifts [29], the recorded spectra are shifted twice, in analogy to [30]. The first shift compensates for large drifts by shifting the spectrum to its correct position according to the center frequency determined by a Gaussian curve fit of the spectrum. This correction is applied in the Hilbert transformed time domain, by multiplying the term exp(-j × 2π×Δf × t) onto the Hilbert transform of the measured time trace. A second shift is applied to compensate for small jitter, mathematically equivalent to the first shift. This step utilizes iodine absorption features as reference. Iodine provides a distinctive sawtooth-like absorption pattern from a single interference pattern due to its strong absorption in the visible domain, see Fig. 2(a). The cross-correlation between the measured iodine spectra and literature curves is calculated to find the frequency shift for this second correction. A relative absorption of 15% is achieved by passing through our iodine sample cell twice. The two shifts are applied equally to all three simultaneously recorded channels. In addition, the iodine reference provides an absolute frequency calibration in the optical frequency domain (e.g., see [31]). In this final step, the calculated iodine transmission is shifted and stretched along the frequency axis to match the literature iodine transmission signal. This calibrates the wavelength axis for the nitrogen dioxide spectrum. Afterwards, the spectra are averaged to calculate the concentration of nitrogen dioxide. Figure 2(b) compares an averaged NO₂ transmission spectrum with a best fit literature transmission spectrum, calculated for 41 ppb NO₂ with a 450 m long absorption path. The measured spectrum is baseline-corrected with a seventh-order polynomial and then vertically shifted to match the literature transmission for a better comparison.

 

Fig. 2. Typical DCS transmission spectra compared with literature transmission calculated from absorption cross sections. The DCS measurements averaged over 172 spectra from single interference events (green lines) are compared with literature (black lines). (a) The normalized, averaged and smoothed iodine features are used as wavelength calibration and the nitrogen dioxide absorption is shifted according to literature [31]. (b) The baseline corrected transmission is shown after subtracting a seventh order polynomial trend (calculated for the interval 513.5 nm – 519.0 nm) for better visibility of the absorption features. The literature transmission curve (black line) is calculated from the absorption cross section [33] via the Beer–Lambert law with an interaction path length of 450 m of the experiment and a NO₂ concentration of 41 ppb. This value could be determined from the amplitude of the measurement transmission spectrum. The offset of the transmission spectrum has been adjusted to match the literature curve. The absorption feature used to calculate the concentration of nitrogen dioxide is at 578 THz (518.7 nm, red solid circle). A secondary feature, which is observed around 583 THz (514.2 nm, red dotted circle), is used for determining the uncertainty of the concentration.

Download Full Size | PDF

The baseline correction of the field measurement path is performed by a seventh order polynomial fit [32], in the interval 513.5 nm – 519.0 nm, to focus on the strong absorption features of nitrogen dioxide. The optimum order of seven of the polynomial fit has been chosen after two independent tests of the sensitivity of the baseline removal. In the first test, various types of baselines were removed from a field experiment, ranging from a sixth to a ninth order polynomial. The calculated concentration 1σ standard deviation for different baselines was found to be less than 3%. In the second test, a baseline correction was simulated using a measured background but a simulated absorption signal derived from literature, using the NO₂ absorption cross-section for the given experimental parameters (41 ppb at 450 m) [33]. The recovered signal was compared to the input, revealing a concentration difference of 7%.

The concentration of nitrogen dioxide is determined from the amplitude of the absorption resonance at 578 THz (518.7 nm, red solid circle in Fig. 2(b)) using the measured transmission spectrum (green line in Fig. 2(b)). A corresponding literature-based reference is calculated from the absolute NO₂ absorption cross section [33]. For a quality check, the accordance of the two curves at the feature located at 583.2 THz (514.0 nm, red dotted circle in Fig. 2(b)) has been analyzed. The agreement is within 96%. The absorption of NO₂ across the observed region originates from overlapping features of the X²A₁ → B²B₁ and X²A₁ → A²B₂ bands [34]. The transmission T is calculated using the equation T = I_T/I_Ref= exp(-αl), where I_Ref is the initial intensity, I_T is the transmitted intensity, α is the absorption coefficient and l is the path length. The absorption coefficient can be determined from the literature's wavelength-dependent absorption cross-section for a given concentration. The secondary feature, observed at 583.2 THz is used as a check for the post processing routine. Between the two main features several other absorption peaks can be observed (e.g., at 579 THz and 580 THz), but are not used in the absorbance calculation due to their relatively low signal-to-noise ratio. The drop in transmission in Fig. 2(b) around 583.7 THz (513.6 nm) is caused by the limited laser bandwidth resulting in a poor signal-to-noise ratio beyond 583.7 THz.

To determine the round-trip path length, the time delay between the reference beam path, having a negligible length, and the field beam path giving the full interaction length of one round trip is used. The RMS envelope is calculated for both channels and then converted to distance using the speed of light. The concentration is then calculated from the absorption using the Beer-Lambert law at a given distance, absorption depth and a known absorption cross section from literature [33].

3. Experimental results

Since March 2022, we have regularly determined the daytime peak concentrations of NO₂ in the Graz atmosphere by measuring at the TU Graz campus “Neue Technik” located close to the city center. The data presented in Fig. 3(a) show the peaks of the 30-minute averaged peak concentration values from 200 to 400 individual time traces recorded on random days. The signal-to-noise ratio and the number of averages is used to calculate the measurement uncertainty, shown as bars. While rain and fog increase the likelihood of signal reduction or even temporary link loss, the short acquisition time mitigates the influence on the spectral information. Capturing multiple interference patterns with more rigorous post selection can compensate for any unwanted effect of weather conditions. The first three measurements shown were taken in the evening hours (6 p.m.) to investigate the rush hour traffic, while the data recorded afterwards was measured in the morning (9 a.m.). The data sets in Fig. 3(a) clearly show the general behavior of NO₂ concentration peaking during the morning and evening traffic.

 

Fig. 3. (a) NO₂ peak concentration values at different days in 2022 and 2023. Dual-comb spectroscopy data (DCS, green bars) is compared with published data of the Environment Agency Austria (EAA) [35], measured at a fixed monitoring site 1 km away from our location (outside of the map, red star), and with chemiluminescence measurements at our location – Petersgasse 16 (red circle). The data of the orange shaded part was measured in a courtyard with a total absorption path length of 450 m. Afterwards, the measurement path was chosen closer to a street, with the light blue background having an interaction length of 500 m, while the dark blue shaded area indicates the third variation with a round-trip length of 900 m. The DCS data is in excellent agreement with the chemiluminescence concentration and shows high correlation with the official data. (b) Map of the different measurement sites and paths from OpenStreetMap [36]. The orange path corresponds to the orange shaded background from Fig. 3(a). The blue arrows correspond to the blue backgrounds from Fig. 3(a) with different total interaction path lengths.

Download Full Size | PDF

Initially, an interaction path of 450 m was chosen within a courtyard (orange shaded background) for simplified measurement conditions in the set-up phase. These first three measurements are carried out for an hour in order to collect a sufficient number of time traces and provide an average value for this time period. To improve the measurement sensitivity, the beam path location has been changed to a longer interaction path and closer to a busy single lane street. The absorption path length used in this second set is 500 m (light blue background). For the last set of experiments (dark blue background), the path was significantly lengthened to 900 m by adding two additional reflections to increase sensitivity while remaining close to the street. See Fig. 3(b) for a simplified campus map with measurement paths and locations.

For all data sets shown, the repetition rate detuning was set to 4 Hz. This amount of detuning compromises between a sufficient spectral resolution (50 GHz) and short apodization windows of a few milliseconds making the relatively slow fluctuations in air (∼ 0.1 s) negligible [37]. After averaging up to 240 subsequent single-trace interferograms, a temporal resolution of up to one minute is achieved.

All experimental results are compared with an official NO₂ monitoring station, operated by the Styrian Government, located about 1 km southeast of our location (indicated by the brown arrow in Fig. 3(b)). For the last four measurement days presented, an identical NO₂ monitoring station at our location was used to gather half-hourly averaged concentration values, shown as red circles in Fig. 3. Our results agree well with both reference measurements for the hourly and half-hourly averaged data over the last year. As expected, the deviation to the official monitoring station is larger than to an equivalent reference at our site. This indicates the influence of the location on the measured value. However, the universal consistency of the NO₂ concentrations within less than 10 ppb at the different locations also indicates a uniform behavior in the relative change during the day across different city districts.

In addition to the daily concentration peaks, the changes in nitrogen dioxide concentration over the course of the day were examined. Figure 4 depicts the half-hourly average concentration values measured at different times during a single day. The background color indicates sunrise and sunset. Solar radiation strongly influences the NO₂ concentrations due to photo-dissociation of NO₂ below 400 nm [38,39]. The results of the different measurement techniques agree very well and show a strong peak in concentration during the morning hours, while the concentration remains almost constant afterwards. The NO₂ concentration typically peaks in the morning hours due to a lower boundary layer and a smaller mixing volume when compared to the late afternoon, in combination with heavy traffic in the morning [40].

 

Fig. 4. NO₂ concentration measured during the 21st of March 2023 with sunrise and sunset visualized in the background. Dual-comb spectroscopy data (green bars) is compared to official data from the Environment Agency Austria (EAA) [35], measured 1 km away from our location, (red stars) and to chemiluminescence measurement results (Chem.) at our location (Petersgasse 16, red circles).

Download Full Size | PDF

In order to examine the dependence of the concentration on local NOx emitters, a portable chemiluminescence measurement device has been borrowed from the department for air pollution control that meets the “Air Pollution Control Act” and placed it close to the DCS setup measuring the NO₂ concentration outside of the building. In the chemiluminescence device, a steady stream of approximately 1.1 l/min is directed through a tube into a measurement chamber. NUV light photolytically converts NO₂ to NO [41], which is detected by first generating an excited nitrogen dioxide molecule by $\textrm{NO} + {\textrm{O}_3} \to {\; \textrm{NO}}_2^\mathrm{\ast } + \; {\textrm{O}_2}$ and then by counting the photons of wavelengths higher than 590 nm resulting from the relaxation process $\textrm{NO}_2^\mathrm{\ast } \to \; \textrm{N}{\textrm{O}_2} + h\nu $ [42]. A calibration test showed uncertainties of up to 5% of the measurement. As shown in Fig. 4, our DCS data is in excellent agreement with the chemiluminescence measurements. Comparing our chemiluminescence data with the DCS results in an average difference of 16% with a correlation value of 0.96 while the EAA data differ on average by 38% and has a correlation of 0.71 with the DCS measurements.

For higher temporal resolution, we conduct continuous data acquisition with a dual-comb signal every 250 ms and perform windowed averaging over different time intervals. Figure 5(a) shows the concentration for 30-minute windows corresponding to 7200 averaged traces in each window. The uncertainty is calculated from the statistical standard deviation of the one-minute intervals. The upward trend, the peak between 6:30 am and 7:00 am, and the subsequent downward trend are all consistent with the chemiluminescence reference installed at our site.

 

Fig. 5. (a) NO₂ concentrations of 30-minute windows calculated from a continuous measurement on 4^th of May 2023 by averaging over one-minute intervals. Dual-comb spectroscopy data points (green bars) are compared to a chemiluminescence monitoring device at the same location (red circles). The total absorption path length is 900 m for the dual-comb trace and the chemiluminescence device only samples one point in space. The sunrise is indicated by the color shading in the background. (b) One-minute time-resolved NO₂ concentration from a continuous measurement in the morning (green bars), compared to the standard three-minute interval monitoring resulting from a chemiluminescence measurement device. The white bars in the background represent alignment work of our set-up where no data was recorded.

Download Full Size | PDF

We record and average 240 time traces every minute, which yields sufficient sensitivity to detect NO₂ and determine its concentration values in the local atmosphere. Figure 5(b) shows a typical DCS measurement sequence with one-minute temporal resolution outperforming an available standard chemiluminescence monitoring device (Horiba APNA-370), which yielded a temporal resolution of three minutes (e.g., [43], red circles). The large peak at 8:15 a.m. coincides with a motorcycle coming to a stop just beneath the measurement path, with its engine running for several minutes. This event could not be reproduced. The combination of one-minute temporal resolution and 90-meter spatial resolution allows for NO₂ field monitoring, potentially enabling the identification of short-term NO₂ sources, such as those produced during industrial processes, and generally the detection of transient concentration changes in real-time.

Both the half-hour and one-minute time-resolved concentrations are in excellent agreement with the corresponding reference values from the chemiluminescence instrument at our location. Since the DCS results are path-integrated values, unlike the chemiluminescence measurement, which samples only a single point in space, we expect some deviation between the two fundamentally different detection systems. The DCS absorption measurement along the full interaction path length results in a precise determination of the NO₂ concentration along the street as opposed to the single-point measurement of the chemiluminescence instrument.

The white bars in the background of Fig. 5(b) indicate downtime due to manually performed realignment procedures of the DCS set-up on both, the sender/receiver and the reflector side. In particular, alignment was required due to temperature changes caused by the rising sun. This downtime will be eliminated in the future by a motorized realignment procedure, constantly compensating any drift. It should be noted that the chemiluminescence monitoring device did not store the three-minute averages during the last 20 minutes of the measurement shown in Fig. 5(b), only the half-hourly aggregated values could be retrieved for this time period.

For all data sets presented, we estimate the current detection limit of the mobile dual-comb spectrometer to be approximately (5 ± 2) ppb for a one-minute averaging window. This detection limit is calculated from the 1σ standard deviation of the noise in transmission. A signal-to-noise ratio of two was chosen as the limit, below which no concentration could be reliably determined. The uncertainty corresponds to the statistical uncertainty evaluated for the different days of measurement. This is sufficient for reliable NO₂ detection as annual mean concentrations at traffic stations in European cities typically do not fall below 10 ppb [44]. Specifically, at the NO₂ monitoring station closest to the university campus, half-hourly mean NO₂ levels dropped below 5 ppb only 15 times on a total of 3 days during the observation period between March 2022 and May 2023.

4. Conclusion

In this work, a mobile device (120 cm x 75 cm) based on dual-comb spectroscopy operating in the visible spectral region was introduced as a reliable field monitoring system of the atmospheric NO₂ concentration in the city of Graz. The new device is capable of determining the NO₂ concentration daytime-independent with a detection limit of (5 ± 2) ppb, a spatial resolution of 90 m and a temporal resolution of up to one minute. This work employs a folded path geometry, which enables future experiments with a spatial resolution of less than 100 m. This advanced spatial resolution can be advantageous for identifying the location of short-term NO₂ sources, e.g., in industrial areas. Similar to a light house, a centralized dual comb spectrometer with rotating beam direction combined with an array of retroreflectors can cover a significant detection area in contrast to a point detector measuring the concentration at a single point in space only. The variable, path-integrated method allows for an efficient location of potential gas sources or leaks [45]. Our data is in excellent agreement with the monitoring systems operated by the Styrian Government. The presented spectrometer currently outperforms the commercial standard device provided by the Air Pollution Prevention Unit of the Styrian Government by a factor of three and will be further improved. This improvement will be implemented by optimizing the beam geometry over the long path length to reduce beam divergence and by switching to reflective optics at the receiver port. Further miniaturization will improve portability. A promising, smaller alternative to fiber lasers are quantum-cascade-laser [46] and silicon-chip-based laser sources [47] for DCS. Full stabilization, i.e., including carrier envelope offset frequencies of both oscillators, could increase the detection sensitivity by at least one order of magnitude. However, advanced stabilization would require an extension of the optical and RF setup. Single cavity dual-comb lasers could be an alternative as they share the resonator and will exhibit similar noise and drift behaviors [48]. Finally, spectral broadening will enable the detection of several molecular species with a single fieldable dual-comb spectrometer. Especially the simultaneous measurement of atmospheric gasses like NO₂, O₃ and SO₂ with a single portable instrument will pave the way for tracking complex correlations in atmospheric reaction cycles.