Fast horizontal radial plume mapping of N 2 O using open-path absorption spectroscopy with a quantum-cascade laser

: We demonstrate record fast horizontal radial plume mapping (HRPM) of nitrous


Introduction
One major challenge in the efforts to combat climate change is the development of sensitive and high-precision sensors for greenhouse gas (GHG) monitoring in various facilities and sites.
Analytical techniques such as fenceline gas monitoring (Naranjo and Baliga, 2012;Riddick et al., 2022;Hashmonay et al., 2008;Dobler et al., 2017) can provide a key tool for plant or facility operators to optimize their process lines, assess mitigation efforts, and ultimately reduce emissions.Locations where fenceline gas monitoring can be implemented include waste disposal sites, wastewater treatment plants, oil & gas facilities (on-shore and off-shore), gasbearing pipelines, and agricultural farms.Development of innovative remote GHG sensor systems for identification and quantification of GHG emissions is of great importance to society, providing governments and regulatory bodies with much needed tools to monitor compliance.
Measured over a 100-year period, the global warming potential of nitrous oxide (N 2 O) is 273 times that of carbon dioxide (CO 2 ) (IPCC, 2021), and it is a cause of ozone depletion (Ravishankara et al., 2009).N 2 O emissions are primarily produced by agricultural and industrial activities and there are ongoing research and policies aimed at reducing N 2 O emissions which therefore requires accurate and reliable sensors (Klimaplan, 2020;Faragó et al., 2022;Delre et al., 2017;Delre et al., 2019;Vergote et al., 2020).In wastewater treatment facilities, N 2 O is emitted during biological nitrogen removal through nitrification and J o u r n a l P r e -p r o o f denitrification.Methods for continuous monitoring of emission from this source would be of particular relevance, as the magnitude of emission is partially dependent on operational parameters (Daelman et al., 2015).This can be further complemented by measurements at a high resolution of the emission's spatial dynamics.
Over the past decade, the use of mid-infrared quantum cascade lasers (QCL) for detection of N 2 O have gained significant attention (Tao et al., 2012;Ma et al., 2013;Ren et al., 2014;Wei et al., 2014;Sohn et al., 2017;Zhang et al., 2021;Yang et al., 2021).This is because a number of strong N 2 O absorption lines (i.e.rotational-vibrational transitions) are present in the mid-infrared region around 4.5 and 7.8 µm with minimal interference from lines of other atmospheric gases such as CO 2 , methane (CH 4 ), carbon monoxide (CO) and water (H 2 O).
Researchers previously used 4.5 µm QCL for N 2 O sensing based on wavelength modulation spectroscopy (WMS) (Tao et al., 2012;Zhang et al., 2021).The sensor developed by Tao et al. was also demonstrated to measure CO, as the 4.5 µm region contains strong absorption lines of CO.For the case of CO 2 , its strongest absorption lines are found near the 2.0 µm, 2.7 µm, 4.3 µm and 4.8 µm wavelengths (Romaniello et al., 2020).The CO 2 lines around the 4.3 µm extend up to 4.5 µm albeit with reduced strengths.The N 2 O/CO sensor by cm long open multi-pass cell to achieve an effective sample path length of 16 m.However, this sensor is essentially a (quasi) point sensor like the commercial cavity ring down spectroscopy (CRDS) gas sensors that require transport (e.g. a van) in order to sample a larger area.
In this work, we demonstrate an open-path remote sensor based on tunable laser direct absorption spectroscopy (DAS) using a 4.5 µm distributed-feedback (DFB) QCL.The gas sensor has the ability to simultaneously measure path-integrated N 2 O and CO 2 concentrations.
Without using a multi-pass cell, tens of meters to 100 m of sampled path lengths are achieved in single-return paths by using reflectors placed at distances of 20 m to 50 m from the DAS system.The simplicity and high sensitivity of the N 2 O/CO 2 sensor allows for rapid scans of the QCL wavelength at a native rate of 2 kHz.The WMS method is a good alternative to the DAS method particularly when dealing with relatively weak absorbance levels.As the absorbance levels are enhanced by relying on strong mid-IR absorption lines and the use of tens of meters of total path lengths, we opted for a DAS implementation due to its signal processing simplicity and higher scan rates over WMS.As a result, the spectroscopically derived concentration measurements are more robust to sub-kilohertz noise, i.e. fluctuations and drifts at time scales longer than 1 ms.Previous WMS implementations with lower effective scan frequencies in the order of 10 Hz (Tao et al., 2012;Zhang et al., 2021) may suffer from these types of noise, resulting in spectral fit estimation errors.Either by averaging ten or taking one in ten successive absorption spectra, an effective data rate of 195 Hz is achieved by the sensor for simultaneous outdoor N 2 O and CO 2 concentration measurements.The high data rate and high-speed beamscanning of the sensor make it possible to steer the QCL beam to 16 reflectors in a 4 × 4 grid.
The sensor produces horizontal radial plume mapping (HRPM) image of N 2 O gas release from 16 path-integrated concentration measurements obtained in just 9 s of scan time -a large 60fold improvement over the 9-min scan time achieved in a recent work that used the HRPM method for CH 4 leak source localization (Soskind et al., 2023).The sensor may be adapted to other path-integrated gas concentration measurement techniques such as one-dimensional (fenceline) (Hashmonay et al., 2008) and vertical radial plume mapping (Hashmonay et al., 2008;Cossel et al., 2023) as well as other HRPM variants that employ mobile reflectors mounted on unmanned aerial vehicles (UAV) (Soskind et al., 2023;Cossel et al., 2023;Cossel et al., 2017) and unmanned ground vehicles (UGV) (Bennetts et al., 2013;Francis et al., 2022).

Optical arrangement
The experimental demonstrations described in this work were carried out using a DAS system illustrated in Fig. 1.As shown in Fig. 1(a), the system employs a 4.5 µm DFB-QCL (Thorlabs QD4500CM1) with a CW optical power of ~50 mW.The QCL beam is linearly polarized (p-J o u r n a l P r e -p r o o f polarized, >1000:1 extinction ratio).A collimating lens (not shown) collimates the QCL beam with 1/e 2 radius of 1 mm.75% of the QCL beam power is transmitted through a mid-infrared plate beamsplitter (BS, Thorlabs BSW511).To enhance its collimation (Rayleigh) length, the QCL beam is transmitted through a beam expander (lenses L1 and L2 with focal length of 20 mm and 150 mm, respectively).The expanded collimated beam has a 1/e 2 radius of 7.5 mm, corresponding to a Rayleigh length of 40 m.The expanded beam passes through a zero-order quarter-wave plate (QWP, Optogama), which turns the transmit beam polarization to a circularly polarized state.The transmit beam is angularly deflected or steered to a distant reflector (50 mm diameter) or an array of reflectors by a galvanometer scanner (GS, Cambridge Technology 6260H).The remotely mounted plane reflector(s) is (/are) aligned such that the return beam reverses the forward beam path and transmits through the QWP a second time.The return beam is consequently linear s-polarized.The available mid-infrared plate BS allows 50% of the s-polarized return signal to be reflected to the detection port.The s-polarized return signal efficiently transmits through a vertically aligned linear polarizer (LP) that also blocks ppolarized stray light.Thus, the combination of the QWP, plate BS and LP significantly minimizes unwanted fringing effects caused by stray light interacting with the detected return signal.A lens L3 focuses the s-polarized signal beam to an amplified mercury cadmium telluride photovoltaic detector (MCT-PD, Thorlabs PDAVJ5).The uncooled MCT-PD has an adjustable gain but we operate at its lowest conversion gain of 1500 V/W where maximal bandwidth (DC to 1 MHz) and low noise-equivalent power of 14 pW/Hz 1/2 are achieved.A typical mean power in the range of 10 µW to 100 µW is received by the MCT-PD.A flip mirror (FM) is also used to optionally deflect a green laser diode (LD) "guide beam" along the optical axis of the 4.5 µm transmit beam.The guide beam aids the optical alignment to the distant reflector(s).A top view photo of the breadboard with the optical components is shown in Fig. 1(b).Table 1 summarizes the optical parameters of the 4.5 µm QCL DAS sensor.For indoor measurements where air flow is sufficiently negligible, the pointing stability of the probe beam is primarily influenced by two factors: (1) the inherent pointing stability of the beam emitted from the QCL front facet, and (2) the temperature-dependent mechanical deflection angle of the closed-loop galvanometer scanner due to a finite "zero drift" of its angular position detector.Although we have not performed measurements of the beam pointing stability of our specific QCL, similar mid-IR QCLs with similar front facet collimation optics was previously studied in (Ryu et al., 2021), and were found to have post-collimation pointing stability of <0.125 mrad with varying drive current.Assuming the same level of stability for our QCL under sawtooth drive current, and a beam expander magnification factor of 7.5, we estimate the pointing stability of the expanded collimated beam incident on the GS module to be <17 µrad.The manufacturer-specified zero drift of the galvanometer scanner used is 15 µrad/C (maximum).If we consider a long-term drift of the indoor temperature of 1 C, this corresponds to an optical deflection angle deviation of 30 µrad.Under the above assumptions, the pointing stability indoors do not exceed 50 µrad, which corresponds to a negligible targeting error of <1 mm at 20 m distance (the distance separating the reflector and the sensor for indoor measurements).For the case of the HRPM field test, where outdoor temperature changed by ~5 C, the estimated pointing stability is <90 µrad, which corresponds to a targeting error of <5 mm at 53 m distance (the distance of the farthest 50 mm diameter reflector from the sensor in the HRPM field test).We also observed, from time to time, the influence of strong wind gusts, which increases the targeting error, but this could be mitigated by further improving the mechanical stability and ruggedness of the reflector mounts and the sensor housing and platform.Thus, for the HRPM field test, the pointing stability is satisfactory but not sufficient enough to allow us to average ten successive 2 kHz scans (i.e.improving SNR) before fitting an absorbance spectrum.This is because averaging ten successive 2 kHz scans that are amplitude modulated, albeit intermittently, due to lack of sufficient pointing stability results in fitting errors.We have found that these fitting errors are avoided by performing the fit to one in ten successive 2 kHz absorbance spectra instead -resulting in 195 Hz effective refresh rate.

Control electronics and data acquisition
The DFB-QCL is temperature-stabilized at 25 C by a thermoelectric cooler (TEC) controller and driven by a current controller as shown in Fig. 1(a).These TEC and current controllers,

Sensor sensitivity -Allan deviation analysis
To assess the sensitivity of the 4.5 µm DAS system, simultaneous measurements of indoor N 2 O and CO 2 concentrations were performed.The DAS system and retroreflector were 20 m apart, hence sampling a total path length of 40 m of indoor air.The system was operated remotely to avoid the effects of human breath on the measurement.Figure 3    but is interpolated and smoothed to one with higher pixel density for visual purposes (Soskind et al., 2023).We have followed the guidelines and procedure described by Soskind et al. to derive the HRPM 2D map, including the use of background cells, i.e. 2 × 4 white cells in Fig. J o u r n a l P r e -p r o o f Supplementary data), the improved temporal resolution of our sensor allows us to detect the start and end (estimated by our sensor at 14:09:35 and 14:43:27, respectively) of the gas release in matters of seconds instead of several minutes.Note that "clean" HRPM image frames, each corresponding to stable symmetric "staircase" pattern of the path-integrated concentration, are obtained before 14:09:35 (start of release) and after 14:43:27 (end of release) -indicating homogeneous near-ambient N 2 O distribution over the probed area.In the video, a few instances of missed measurements (data loss) in the path-integrated concentration are observed when the fitting algorithm is unable to produce a reasonable fit, for example, during periods when the probing beam is coincidentally blocked by a passing person that regulates the gas leak rate within the grid.

Outdoor field test -DAS sensor vs. CRDS sensor
J o u r n a l P r e -p r o o f Another possible reason for the intermittent appearance of hot spots of high N 2 O levels in the HRPM frames is the fact that the scanned laser beam is measuring the gas concentration in discrete lines of sight across a very thin plane.It is likely that the gas plume is not always sufficiently intersecting the measurement plane or the beam as the release point was slightly higher than the laser paths (at 0.8 m height of reflectors).By averaging over longer durations (e.g. in the order of minutes), a more statistically meaningful gas mapping data is obtained, as discussed in (Fritz, et al., 2005).In Fig. 7, we show the combination of the HRPM data with the wind statistics (wind rose) during three different periods of the field campaign.
J o u r n a l P r e -p r o o f An excessively low duty-cycle degrades the sensitivity of the gas sensor.However, the dutycycle can be regained by using a beam-scanner with shorter switching times, e.g.galvanometer scanners with smaller or lighter mirrors.The persistent pattern between beam paths shown in the plots of Fig. 8 is a result of the fitting algorithm being continuously active even during the transition from one beam path to the next and is considered an artifact.During the transition, the return signal is highly attenuated so the fitting algorithm is essentially attempting a fit to a J o u r n a l P r e -p r o o f noisy data.We have opted to keep the fitting algorithm continuously active (instead of deactivating it during the transition) as a means of empirically determining the transition time (or dead time) between adjacent beam positions.Hence, the artifact data during the transition are sufficiently filtered out in the HRPM data as shown, for example, in Fig. 6.
By selecting telescope lenses L1 and L2 with appropriate focal lengths and aperture sizes, beam collimation can be extended and path lengths of several hundreds of meters can be achieved, extending the utility of the sensor to more applications such as fenceline monitoring of larger facilities.For longer path lengths, it may be beneficial to select weaker gas absorption lines in order to maintain a high dynamic range of the sensor.During all our measurements, the temperature of the sampled air varied from 20 C to 25 C.For all our curve fitting, we have used HITRAN reference absorbance spectra derived at a fixed 20 C setting.Although the change in peak absorbance of reference spectra from 20 C to 25 C is relatively small (less than 2%), future implementation of our DAS sensor could include atmospheric temperature and pressure sensors and utilize HITRAN-derived reference spectra (lookup table) calculated at various combinations of temperature and pressure.
The presented DAS sensor may also be combined with wind information and dispersion models to infer emissions flux (Soskind et al., 2023), configured for vertical radial plume mapping that is also used for emissions flux estimation (Hashmonay et al., 2008;Cossel et al., 2023) or gas sensing systems that are integrated with UAVs and UAGs for enhanced flexibility and mapping coverage (Soskind et al., 2023;Cossel et al., 2023;Cossel et al., 2017;Bennetts et al., 2013;Francis et al., 2022).Moreover, the commercial availability of QCLs (as well as interband cascade lasers) of various wavelengths across the LWIR, MWIR and SWIR regimes may be exploited to configure the sensor and its spatial gas mapping capability to other target gases of interest.

Conclusion
In this work we demonstrated simultaneous measurement of N 2 O and CO 2 concentrations in the atmosphere (at 195 Hz outdoor and at 2 kHz indoor) rate using a retroreflected mid-infrared beam in an open-path direct absorption spectroscopy system.The system employs a tunable DFB-QCL operating around 4.5 µm.Using Allan deviation analysis, our 4.5 µm DAS system J o u r n a l P r e -p r o o f
combined in one unit (Thorlabs ITC4002QCL), are controlled by a module that comprises analog-to-digital converters (ADC), digital-to-analog converters (DAC) and a fieldprogrammable gate array (FPGA) board.The module also controls the GS (for beam-scanning) and samples the MCT-PD voltage signal at 250 kSamples/s.The FPGA enables the synchronization between the beam steering, current ramp generation and detector signal acquisition which is required for reliable measurements.A program in the real-time FPGA controller and in a PC -both written in LabVIEW (National Instruments) -enables control of the process, ramping of the laser current, readout of the data (i.e.MCT-PD voltage), evaluation of the concentration and finally storage on the PC.The FPGA and the PC are connected via an J o u r n a l P r e -p r o o f Ethernet link.Similar to most DAS implementations, the DFB-QCL wavelength is scanned around the absorption lines of the target gases.Fast wavelength scanning (or chirp) from 4.467 µm to 4.471 µm is accomplished by applying a sawtooth drive current with a mean current of 348.5 mA, a peak-to-peak current of 200 mA and a scan frequency of 1953 Hz to the DFB-QCL.For each period of the sawtooth current, a corresponding sampled MCT-PD voltage signal is obtained as shown in Fig.2(a), which is a nonlinearly modulated form of the gas transmission spectrum.The nature of nonlinearities is two-fold.First, an x-axis or wavelength axis nonlinearity exists due to the nonlinear relation between drive current and wavelength as the ability to modulate the DFB-QCL wavelength with current at high scan frequency is limited by the thermal time constant of the laser(Tombez et al., 2013).We implement a 2 nd -order polynomial correction of the x-axis nonlinearity by using a reference N 2 O gas cell (Wavelength References Inc.) and comparing the locations of minima and maxima of its measured spectrum to those found in a HITRAN database/simulator.Second, a y-axis modulation occurs due to laser-current-dependent beam power of the DFB-QCL (causing an apparent tilt in the transmission spectrum), residual interference or fringing effects due to spurious reflections, and the nonlinear mapping of incident power to MCT-PD voltage.In order to derive accurate and precise concentration readings from the DAS sensor, it is important to correct or account for these signal amplitude modulations.In the DAS optical design presented in this work, fringing effects have been greatly diminished by the fact that dominant spurious reflections from antireflection-coated surfaces of optical components are p-polarized and thus get blocked by the vertically aligned linear polarizer (LP) in front of the MCT-PD.The same LP allows the spolarized return signal to reach the MCT-PD efficiently.Hence, the y-axis modulation is dominated by the laser-current-dependent return power profile and the nonlinear relation between incident optical power on to the MCT-PD and its output voltage, which are eliminated using a one-time calibration procedure.The correction of the nonlinear relation between incident power and the MCT-PD output voltage is determined by comparing the scan measured at high mean incident power of 100 µW to its attenuated counterpart (i.e. at 20 dB optical attenuation using a properly rotated additional LP in front of the MCT-PD, which reduces the mean incident power to ~1 µW at which the MCT-PD response is sufficiently linear) and performing a 3rd-order polynomial fit.The DC offset of the MCT-PD is also accounted for by measuring the detector voltage reading when the laser is turned off.

Fig. 2 .
Fig. 2. (a) Measured detector signal during one period of the 1953 Hz sawtooth drive current that repeatedly scans the QCL wavelength from 4.467 µm to 4.471 µm.(b) Corresponding DAS absorbance spectrum data derived from the detector signal, the fitted curve, and the HITRANderived reference spectra of N 2 O, CO 2 and H 2 O used by the fitting algorithm.The reference spectra use the nominal global average concentrations (at sea level) of N 2 O, CO 2 , and H 2 O, atmospheric temperature of 20 C, pressure of 1013 hPa, and a total path of 40 m.
(a) and Fig. 3(b) show the time series of N 2 O in ppb and CO 2 in ppm, respectively, at 1953 Hz update rate.The corresponding histograms are shown in Fig. 3(c) and Fig. 3(d), which indicate that both time series generally have a normal (Gaussian) distribution.The Allan deviation plots of the N 2 O and CO 2 measurements are also shown in Fig. 3(e) and Fig. 3(f), respectively.The measured N 2 O precision at 1953 Hz is 0.6 ppb and the minimum N 2 O Allan deviation occurs at an averaging time  = 1 s; demonstrating an excellent precision of 25 ppt at a mean concentration of 296ppb.This is an improvement over the achieved sensitivity of ~50 ppt (at mean of 326 ppb) at 5 s averaging time by a previous WMS based N 2 O sensor(Tao et al., 2012), which relied on an open multi-pass cell with a 16 m total sampled path.In terms of open-path sensitivity, our DAS based N 2 O sensor has a 1 ppb-m/Hz 1/2 sensitivity while the WMS based N 2 O sensor in(Tao et al., 2012) achieved a sensitivity of 2 ppb-m/ Hz 1/2 .For the case of CO 2 , the precision at 1953 Hz is 0.6 ppm and the minimum Allan deviation is also around 1 s where a precision of 30 ppb (or 1.2 ppm-m/Hz 1/2 open-path sensitivity) is achieved at a mean concentration of 463 ppm -a performance at par with benchmark tunable laser spectroscopy-based CO 2 sensors previously developed(Xiang et al., 2013).For  > 1 s, the CO 2 Allan deviation quickly increases, which is due to drifts in the actual indoor CO 2 concentration as seen in Fig.3(b).A commercially available N 2 O/CO 2 sensor using a mid-infrared QCL and an enclosed multi-pass cell (13 m sampled path) has a sensitivity of 160 ppt for N 2 O and 200 ppb for CO 2 at 1 s integration time or an equivalent open-path sensitivity of 2.08 ppb-m/Hz 1/2 for N 2 O and 3.2 ppm-m/Hz 1/2 for CO 2 (MIRA Ultra N 2 O/CO 2 High Accuracy Analyzer, Aeris Technologies, Inc.).The turnaround in the Allan deviation for N 2 O (and also for CO 2 ) is potentially caused by a combination of QCL relative intensity noise (1/f noise), the inherent variation in the path-J o u r n a l P r e -p r o o f integrated concentration of the gas molecules along the beam, as well as the low-frequency (≤1 Hz) beam pointing instabilities of the galvanometer scanner induced by mechanical vibration and temperature drift in the room.

Fig. 3 .
Fig. 3. Time series of (a) N 2 O and (b) CO 2 concentration simultaneously measured by the 4.5 µm QCL based DAS system.Histogram of (c) N 2 O and (d) CO 2 data.Allan deviation of (e) N 2 O and (f) CO 2 concentration versus averaging time .Total path length: 40 m.

Figure 4
Figure 4(a) and Fig. 4(b) show photographs of the experimental configuration in the field where the open-path DAS sensor was deployed near the aeration tanks of a wastewater treatment plant near Copenhagen, Denmark (BIOFOS Spildevandscenter Avedøre) on the 11 th of August 2022.The comparison of the data obtained by our DAS sensor and that of a reference point sensor -

Fig. 4 .
Fig. 4. (a) Photograph of the field test site and setup.(b) Aerial view of the site taken from Google Maps showing a distance of 20 m from the DAS sensor to the retroreflector.(c) N 2 O and (d) CO 2 path-averaged concentration measured by the 4.5 µm DAS system and the corresponding measurements of the CRDS (Picarro) reference point sensor.With regards to the general intensity differences between the measurements of our DAS sensor and the CRDS point sensor in Fig.4, we note that the DAS sensor appears to underestimate the N 2 O concentration during two aeration events: first, during 0 to 0.3 hr, and second, during 1.25 hr to 1.75 hr.Interestingly, these are two periods in which increased mean levels in measured CO 2 are also observed.Hence, one potential reason for the underestimation of N 2 O emissions by the DAS sensor is the degree of crosstalk between the N 2 O and the CO 2 concentration estimation as the DAS sensor relies on multispecies spectral fitting involving broadened gas absorption lines that have some degree of overlap.Note that during another aeration event from 0.3 hr to 1.0 hr the CO 2 level (and hence, any associated crosstalk) is relatively lower and the average N 2 O reading of the DAS sensor matches that of the CRDS reference sensor.

5
(b).Each background cell has an approximate dimension of 5.6 m × 9.7 m.Similar to the work by Soskind et al., we also made the assumption that the first measurement cell (lower right corner of HRPM image) has the same concentration as the background cells.This was satisfied for most part of the experiment duration as the mean wind direction mostly prevented significant increases in the concentration of the 1 st measurement cell and the 8 background cells from ambient levels (~0.3 ppm).To test the unprecedented fast HRPM capability of the DAS N 2 O sensor, two controlled gas release experiments from gas cylinders were performed.The first was a single-source N 2 O release at 1 m height at the location (0, 40 m) as shown by the red triangle in Fig. 5(b).The emission rate was measured to be 1.26 kg/h.The second release was a double-source with each source having an emission rate of 0.92 kg/h also at 1 m height at two locations, (−6 m, 48 m) and (6 m, 40 m), denoted by the orange triangles in Fig. 5(b).

Fig. 5 .
Fig. 5. (a) Aerial view of the controlled N 2 O release test where the HRPM method was used for source localization.(b) The 4 × 4 grid geometry showing the 16 reflector positions (black squares), the 16 measurement cell centers (white crosshairs), and the 16 beam paths (red lines).The 4.5 µm QCL based N 2 O sensor is placed at the origin (0, 0).The approximate locations of the gas release sources (red triangle for single-source and orange triangles for double-source) and the wind sensor (green circle) are also marked.Results of the single-source release test are shown in Fig. 6.The upper panel shows a graph of the path-integrated N 2 O concentration measured by our sensor at 195 Hz as the beam is directed to the 16 reflectors in the 4 × 4 grid.With 16 path-integrated concentration measurements and a matrix or HRPM kernel (Soskind et al., 2023) with 16 × 16 elements constructed from path lengths in the measurement and background cells, a 4 × 4 -pixel HRPM image can be calculated every 9 s.An interpolated and Gaussian-filtered version of the HRPM image with increased pixel count is shown in Fig. 6 lower panel.The scan cycle of the systempresented in this work is much faster than that of the benchmark(Soskind et al., 2023) due to two reasons: (1) the system relies on the strongest N 2 O absorption lines around 4.5 µm, which results in high SNR, while the benchmark system utilized the near-infrared absorption overtone of CH 4 around 1.654 µm, and (2) the beam-scanning performance of the GS employed in our system allows for faster switching and stabilization of the beam to adjacent plane reflectors without relying on any feedback control loop that was required in the implementation used by Soskind et al.As can be observed from the video file associated to Fig.6(see Appendix A.

Fig. 6 .
Fig. 6.A video frame of the dynamic evolution of the path-integrated concentration (in ppb-m) in each of the 16 beam positions and the corresponding HRPM image (in ppm).Each video frame corresponds to a 9 s scan duration with only 0.1 s transition time between adjacent beam directions.See Appendix A. Supplementary data for the video file.The detection of N 2 O plume between 14:09:35 (start of release) and 14:43:27 (end of release) appears intermittently -perhaps indicative of the complex and dynamic nature of gas transport in an outdoor setting where the wind velocity varies in magnitude and direction.Wind turbulence was also possible due to presence of trees and structures near the measurement area.

Fig. 7 .
Fig. 7. Nine-minute averaged HRPM image (left) for the case (a) before the N 2 O release, (b) during a single-source N 2 O release, and (c) during a double-source N 2 O release.The wind rose during each period is also shown (right).

Fig. 8 .
Fig. 8. Indoor N 2 O level in ppb measured by the DAS sensor as the beam is alternately directed towards two reflectors about 5 apart (each reflector is 20 m from the sensor).Three cases are shown: (a) 1 s, (b) 0.33 s, and (b) 0.15 s per beam position.The time per beam position includes the transition time of the beam between reflectors, which is 0.09 s for all cases.