Study on Local Power Plant Emissions Using Multi-Frequency Differential Absorption LIDAR and Real-Time Plume Tracking

Abstract

We present a new fiber-amplifier-based differential absorption light detection and ranging (DIAL) system for range-resolved detection of carbon dioxide (CO₂) and water vapor (H₂O) over a range of a few kilometers. The fiber amplifier chain is seeded with a 7.14 kHz fast-switching high-spectral purity wavelength source near 1572 nm to cover ten different frequencies across the CO₂/H₂O line pair in ≈1.4 ms. We demonstrate the system in a study of CO₂ emissions from a local power plant in Boulder, CO, USA. We use real-time wind information to predict the plume location for tracking and modeling of the CO₂ emission rate to compare with the reported data from the power plant over a 13 h period. There is overall agreement with the reported burn rate, but we see periods of bias towards underestimation of the CO₂ emission rate. We attribute the dropout periods to uncertainties between the measured and the plant’s local wind speed data that impact both the tracking location and the plume model predictions. Upcoming studies that will make use of real-time Doppler wind data are expected to significantly decrease these uncertainties.

1. Introduction

With the ever-increasing impact of climate change, new technology is needed to help monitor greenhouse gas (GHG) emissions, such as CO₂ and H₂O. Point sensors provide a weak link to remote sensing of these gases, and column-integrated measurements are only able to paint a partial picture when it comes to understanding CO₂ sources and sinks [1]. Differential absorption LIDAR (light detection and ranging, or DIAL) uses two or more LIDAR returns of the on- and off-resonance regions of an absorption line to obtain range-resolved GHG concentrations [2,3,4]. DIAL provides information not easily obtainable from other open-path techniques that provide column-integrated information, since range-resolved measurements enable direct construction of three-dimensional GHG concentration maps. These maps allow for a more in-depth understanding of atmospheric variations and a better quantification of GHG sources and sinks.

Emissions from power plants are ubiquitous examples of localized GHG sources. The process of burning natural gas to produce heat produces CO₂ and H₂O as waste products. The actual GHG concentrations located downwind from the source depend on local wind, temperature, atmospheric stability, and ambient levels. This variability makes it extremely difficult to track and quantify downstream CO₂ emission flux using static point sensors. Traditional flux estimates use large-scale networks of balloons, towers, and mobile point sensors but suffer from large spatial resolutions, making it difficult to accurately quantify local emissions [5,6]. Similarly, optical techniques, such as tunable absorption or Fourier transform infrared (FTIR) spectroscopy, require hard targets at multiple locations around the plume to measure flux [7,8]. A prior proof-of-concept measurement using a DIAL system in an industrial setting reported an increase in CO₂ at the plume but lacked sufficient local meteorological data to model flux [9]. Most recently, DIAL and Doppler LIDAR systems have been combined to provide simultaneous concentration and wind information to determine plume emission flux [10]. Using two-point DIAL and a high-efficiency superconducting nanowire detector, they showed increased CO₂ flux over the industrial park compared to a local park. Here, for the first time, we apply our novel ten-point, range-resolved DIAL measurements [11] with Gaussian plume modeling using local meteorological data to estimate the emission flux of CO₂ from a point source in an urban environment.

In this work, we present a fiber-amplifier-based DIAL system for performing standoff detection of CO₂ and H₂O concentrations over a 3 km range. The compact fiber-based components make the system ideal for deployment as a mobile platform for measurements at remote locations. In contrast to previous results demonstrated with a low-pulse-rate-frequency (PRF) free-space amplifier [12], the high-PRF (5 kHz to 10 kHz) DIAL system demonstrated here takes advantage of a low-dark-count photomultiplier tube (PMT) and a high-throughput detection system with near-continuous acquisition of returns. We demonstrate the DIAL system to observe the dry-air CO₂ concentrations from a local power plant’s stack emissions located at a distance of 1.13 km. Wind sensor data sampled above the lab are used in a Gaussian plume model to determine the pointing direction of the transceiver in real time to maximize overlap between the DIAL beam and the plume center. Finally, we use models to compare the observed increase in CO₂ concentrations to the reported natural gas burn rate from the University of Colorado (CU) East District Energy Plant (EDEP) power plant [13].

In Section 2.1, we give a description of the DIAL instrument including the fast frequency switching seed source and pulse generator. In Section 2.2, the method for plume modeling and tracking is discussed. In Section 2.3, we discuss the methods and approximations used to compare the DIAL-based emission rate with natural gas burn rate.

2. Materials and Methods

2.1. DIAL System

A schematic of the overall system is shown in Figure 1, and specifications of the components are given in

Table 1. Specifications [14] of the CO₂/H₂O fiber-amplifier-based DIAL system with a coaxial transceiver arrangement.

Frequency Converter:
Center wavelength	1571.9 nm	New Focus, ECDL
Arbitrary waveform generator	2 GS/s, 250 MHz bandwidth	AWG1, Tektronix, Model 3252
Electro-optic modulator	10 GHz bandwidth	EOM1, Thorlabs
Acousto-optic modulator	252 MHz	Brimrose
Arbitrary waveform generator	24 GS/s, 8 bits, 1 channel	AWG2, Tektronix, Model 7122C
Electro-optic modulator	20 GHz bandwidth	EOM2, EOSpace
Electro-optic switches	1 ns speed, 40 dB isolation	EOM3 + EOM4, EOSpace
Invar filter cavity	300 MHz free spectral range	Burleigh, CFP-500, l = 0.5 m
Number of frequencies	10	Δν = 48 GHz
Spectral purity	>99.9%	Filter cavity finesse ≈ 500
Booster optical amplifier	20 mW, <20 dB gain	Thorlabs, Model S9FC1082P
EDFA1	100 mW, <30 dB gain	Amonics, AMN-EDFA-L01/PM
EDFA2 + EDFA3	105 μJ/pulse, 1 W power	NP Photonics, Custom model
Pulse length	300 ns
Spectral linewidth	<1 MHz	Full width at half maximum
Pulse repetition rate	7.143 kHz	5 kHz to 10 kHz
Tx/Rx optics:
Tx beam diameter	41 mm	×10 beam expansion
Tx beam M2	<1.5	NP Photonics, Custom model
Tx beam divergence	<300 μrad
Elevation	1.6 degrees
Rx telescope diameter	279.4 mm	Schmidt-Cassegrain, Celestron
Bandpass filter	1.9 nm (FWHM)	>5 OD rejection, Alluxa
Neutral density filter	0.1, 0.3, 0.6, and 1 OD	Thorlabs
Fiber-coupled	300 μm aperture	1 m cable length
PMT detection	~2% QE	Hamamatsu H12397-75
Data acquisition:
Transimpedance amp gain	5 kV/A, τ3dB ≈ 1 ns	Femto HCA-400M-5K-C
Digitizer	12 bits, 1 GS/s, 2 channels	GaGe EON CSE123G2
Raw data storage	10 s accumulation	NetCDF4 file format
Range resolution	250 m to 500 m	Defined in post-processing
Temporal resolution	5 min	Defined in post-processing

. The frequency converter begins with a single, fiber-coupled, external-cavity diode laser (ECDL). The output is 90/10% split, where the 10% leg is frequency-offset in an acousto-optic modulator (AOM) fiber-coupled to an electro-optic phase modulator (EOM₁) and is then free-space-coupled to an Invar confocal filter cavity that is used in reflection to frequency-lock the laser (the PDH method) [12]. The 90% leg is sent to a second broadband modulator (EOM₂) that is driven by an arbitrary waveform generator (AWG). The frequencies and amplitudes of the AWG waveform are defined such that only one of the second-order comb teeth from among the generated sidebands and carrier is transmitted through the cavity [12] at any given time when modulated by a single-tone waveform. The output frequencies from the filter cavity are sequentially scanned over ten different tones every 140 μs to give a 7.14 kHz PRF. The spectral coverage is 48 GHz and probes the region containing the CO₂ line at 1572.335 nm (6359.97 cm⁻¹) and the H₂O line at 1572.257 nm (6360.28 cm⁻¹). As shown in Figure 1b, five points are selected on the CO₂ line (blue trace), one point on the H₂O line (red trace), and the remaining four points sample the background regions at both ends of the spectrum. By selecting five points across the top of the CO₂ line, we maintain a nearly equivalent signal-to-noise ratio (SNR) as in two-point DIAL and can use the line shape to fit the laser frequency [11,12]. For a range resolution of 250 m, determined by our bin size, the fractional absorption is near 4% for ambient levels (≈420 ppm) of CO₂.

The output of the fast-switching wavelength source is amplitude-modulated (Δt ≈ 10 μs) in a booster optical amplifier (BOA). The pulsed output is further amplified in an erbium-doped fiber amplifier (EDFA) and coupled through two EOM-based polarization switches to generate 300 ns pulses with >40 dB rejection. The fast-switching wavelength source is sent sequentially through two additional EDFA stages (EDFA₁ + EDFA₂), where the final amplifier is doped to reduce stimulated Brillouin scattering. The final amplifier beam is specified with a 1/e² waist of 4 mm, an M² value < 1.5, and an output power of 900 mW. The output beam is sent through a polarization-independent isolator and two anti-reflection-coated telescopes to give an effective x10 beam expansion and a uniform 41 mm output beam in near-field. The transmitted beam was collimated using a shear plate and has a measured divergence at 1 km of ≈300 μrad. The average transmitted power is near 750 mW to maintain class 1 M operation. The transmit beam is directed to the center of the 279 mm Schmidt–Cassegrain receiving telescope that focuses the DIAL retrievals into a 300 µm optical fiber for detection using a fiber-coupled photo-multiplier tube.

The amplified PMT output is sampled by a 1 GS/s analog-to-digital converter. The final amplifier, optics, and telescope are fixed to an alt-azm mount for positional control of the transmitter/receiver (Tx/Rx) system. Daytime operation is enabled by the insertion of a 1.2 nm narrowband optical filter before the PMT having a >5 OD (optical density) out-of-band rejection and a filter wheel with up to 1 OD of broadband optical attenuation.

The recorded data stream was processed in real time to generate two stored records every 10 sec: one of them accumulated photon counts; the second of them accumulated voltage (proportion to the PMT current). The two channels enhance the linear detection range of the PMT to give a dynamic range of >10⁵. In post-processing, the ten colors are separated and averaged over predetermined range bins for each record. The background points are averaged together to serve as the common reference for the signal ratios. The natural log of the DIAL derivatives over the range bins (not necessarily consecutive bins; see below) are evaluated to give the absorption line shapes of CO₂ and H₂O. The specific record used in the fit of the range-resolved concentrations depends on whether pulse pile-up error is significant in the photon count channel, in which case the current channel is used. The spectral fits were performed using a nonlinear least-squares algorithm to determine the CO₂ and H₂O concentrations at 5 min intervals. We combined these two fit results to calculate the dry-air CO₂ concentration maps. The SNR is primarily determined by the square root of the number of equivalent photons detected for the on-line signal. The size of each range bin is increased from 250 m to 500 m as the range increases from 400 m to 3 km to compensate for the inverse-square-law decrease in returned signal response.

Unlike the most common DIAL response, where the off-line backscatter returns decrease with range, R, as ≈1/R², the backscatter returns from a plume can increase abruptly when passing through the condensate, especially for temperatures at or below freezing. The large transient responses (spikes) observed for the on-line and off-line regions at the 1.13 km range (stack position) are shown in Figure 2. This spike can cause significant discontinuities in the derivatives used to extract the power-independent concentrations which, in turn, result in rapidly changing biases for at least two of the range bins that include this condensate return. To work around this, we use a discontinued derivative approach that uses instead the two range bins on either side of the spike, thereby obtaining the CO₂ concentration increase from the plume that is devoid of distortions from this plume response. Figure 2 shows how the bins, shaded in green, are chosen to avoid the plume spike. It is also important to note that while the additional scatter from the spike causes a decrease in the downrange signal that is proportional to power, the relative decrease is independent of frequency over this narrow spectral region; therefore, the DIAL derivatives provide a true measure of the CO₂ concentration in the plume. In other words, we can recover the integrated path concentrations between any two points along the DIAL path regardless of the signal distortions from plume saturation effects.

2.2. Gaussian Plume Model and Tracking

The goal of this work is to compare the estimated plume emissions determined through DIAL measurements with the bottom-up CO₂ emissions derived from the natural gas burn rate of a local power plant. To achieve this, we need to both measure the concentration of CO₂ in the plume and the velocity of the plume above the stack. The latter requires knowledge of local wind velocity as well as environmental factors, such as time of day, cloud cover, ambient temperature, and power-plant-dependent variables that include exit velocity, stack height, and exit temperature. The wind data are from a local weather station (Vaisala, Model WXT536 [14]) located about two meters above the transmitter. While building interference can alter local wind conditions, the measurements near the lab are assumed to be in good agreement with the local wind at the smokestacks and are used in a Gaussian plume model to estimate the increase in the CO₂ concentration from the power plant.

To measure the emission flux from the power plant reliably, we need to consistently sample the center line of the plume that is changing with wind speed, direction, and air temperature. To solve this, we need to actively track the plume to determine the ideal position to point the DIAL system to measure plume CO₂ flux in a variety of conditions. Here, we use a Gaussian plume model [15,16,17] that is dependent on multiple atmospheric conditions, including temperature, solar irradiance, and wind conditions. The wind speed and direction data acquired in real time by the weather station are used to actively predict the plume location based on a running vector of the most recent 150 s of data. These predictions are then used to reposition the transceiver system every 45 s to 60 s. The tracking speed is ultimately limited by the weather station measurement time, which varies between 15 s and 45 s per record. The Gaussian plume model is used to calculate the center-line locations in 3D space, such that the beam passes through the plume at a predetermined distance away from the stack (depending on temperature; see below). The weighted average of the position vector is calculated to determine the final pointing location. Once the best plume sampling line is calculated, the coordinates are transferred to the lab frame and offset for the transceiver base constraints (pointing 10.7° off north and platform tilt of 9°). Video of the tracking system and model is given in the Supplementary Materials.

The optimal performance of the tracking system is found in moderate winds (2 m/s to 4 m/s). Low winds lead to larger deviations between the data from the weather station and the actual conditions at the stacks located 1.13 km away, while high winds dilute the CO₂ concentration increase, making detection above ambient levels more difficult. Generally, there will be high to moderate winds during the day and early evening, and low-speed, calm conditions in the late evening and morning (see Figure A2). Additionally, while some additional scattering from the water condensate in the plume is acceptable (see Figure 2) for temperatures below freezing, the plume can become non-transparent. To compensate for changing temperatures, the distance away from the stack is varied between 30 m and 150 m depending on the density of the condensate.

2.3. Comparison with EPED Emission Data

The data reported by the EDEP plant are for consumption of natural gas (NG) fuel in pounds per hour. Methane (CH₄) makes up about 75% to 95% of NG, the balance being largely ethane (C₂H₆, < 15%) and propane (C₃H₈, < 10%). To estimate the CO₂ flow from the plant, the following mass balanced flow rate can be approximated as:

$${\dot{m}}_{\mathrm{C}{\mathrm{O}}_2}=\left[{\eta }_{\mathrm{C}\mathrm{H}4}\frac{M_{\mathrm{C}{\mathrm{O}}_2}}{M_{\mathrm{C}{\mathrm{H}}_4}}+{\eta }_{C2H6}\frac{2M_{\mathrm{C}{\mathrm{O}}_2}}{M_{{\mathrm{C}}_2{\mathrm{H}}_6}}+{\eta }_{C3H8}\frac{{3M}_{\mathrm{C}{\mathrm{O}}_2}}{M_{{\mathrm{C}}_3{\mathrm{H}}_8}}\right]{\dot{m}}_{NG}\left[\frac{\mathrm{g}}{\mathrm{s}}\right]$$

(1)

where η_CH4, η_C2H6, and η_C3H8 are the percentages by weight of methane, ethane, and propane, with methane estimated to be 77.2% of the natural gas, and the subscripted Ms designate the molar masses of CO₂, methane, ethane, and propane.

Equation (1) gives the flow rate of CO₂ from the power plant in g/s. To translate our DIAL data into a flow rate, we use the mass flux and plume area to define the flow rate [10]. The estimated flow rate of the plume can be described as:

$${\widehat{\dot{m}}}_{\mathrm{C}{\mathrm{O}}_2}=j_p\left[\frac{\mathrm{g}}{{\mathrm{m}}^2\mathrm{s}}\right]A_p\left[{\mathrm{m}}^2\right]$$

(2)

where j_p is the mass flux of the CO₂ plume and A_p is the plume area. The plume area is calculated from the Gaussian plume model (see Appendix A). Other methods attempted, such as scanning the DIAL beam across the plume, reduced the observed increase in CO₂ below detectable levels. The sporadic nature of the plume dynamics makes this sort of classification difficult. Even with a dedicated LIDAR system for plume profiling, modeling would still be required to track the average plume center. The mass flux is defined as the product between the increase in CO₂ by the plume and the plume speed:

$$j_p={\mathsf{\Delta }}_p\mathrm{C}{\mathrm{O}}_2\left[\frac{\mathrm{g}}{{\mathrm{m}}^3}\right]u_p\left[\frac{\mathrm{m}}{\mathrm{s}}\right]$$

(3)

where u_p is the speed of the plume at the measurement location and Δ_pCO₂ is the increase in CO₂ due to the plume. This can be related to the DIAL measurements using:

$${\mathsf{\Delta }}_p\mathrm{C}{\mathrm{O}}_2=\frac{L_{DIAL}}{L_p}{\mathsf{\Delta }}_{DIAL}\mathrm{C}{\mathrm{O}}_2\left[\frac{\mathrm{g}}{{\mathrm{m}}^3}\right]$$

(4)

where L_DIAL is the ≈400 m range bin, L_p is the calculated plume width parallel to the beam, and Δ_DIALCO₂ is the increase in CO₂ over the DIAL range bin. The speed of the plume is calculated using the weather station data and plume rise equations. With the DIAL system operating close to the plume source, the total plume speed will depend on both stack exit velocity and plume rise, as well as the average wind speed. The total wind speed is defined as:

$$u_p=\sqrt{u_z^2+u_x^2}$$

(5)

where u_z is the vertical plume velocity and u_x is the downwind velocity. The downwind component is z-dependent and related to the point sensor reading using the power law:

$$u_x\left(z\right)=u_0{\left(\frac{z}{z_0}\right)}^{\rho }$$

(6)

where u₀ is the wind speed at the reading height, z₀, and the exponent, ρ, depends on atmospheric stability. To estimate the vertical plume velocity, we calculate the final plume rise height, where u_z = 0, and assume a linear decrease in velocity from the plume source exit velocity:

$$u_z\left(z\right)=\left(1-\frac{z}{\mathsf{\Delta }h}\right)u_{exit}$$

(7)

where u_exit is a constant. Details on how the plume rise is calculated are given in Appendix A. Finally, from the DIAL data and local metrological data, the calculated CO₂ flow rate, ${\widehat{\dot{m}}}_{C{O}_2}$, is determined from:

$${\widehat{\dot{m}}}_{\mathrm{C}{\mathrm{O}}_2}=\frac{L_{DIAL}}{L_p}{\mathsf{\Delta }}_{\mathrm{D}\mathrm{I}\mathrm{A}\mathrm{L}}\mathrm{C}{\mathrm{O}}_2\sqrt{{\left(1-\frac{z}{\mathsf{\Delta }h}\right)}^2{u}_{exit}^2+u_0^2{\left(\frac{z}{z_0}\right)}^{2\rho }}{A}_p$$

(8)

The results from Equation (8) are compared to the reported EDEP emissions data below. The local wind data, u₀, are used in multiple instances, and, as such, contribute the largest source of uncertainty to both the plume tracking and plume flow-rate calculations. The full list of constants and assumptions is detailed in Appendix A.

3. Results

The DIAL measurements presented are from 10 February 2023, when an increase in CO₂ levels due to the emissions from a local power plant was well above ambient levels. The backscatter profile from 400 m to 3.3 km over the 13 h measurement period is shown in Figure 3a, where the increase in backscatter at the plume location is easily seen. To illustrate how well we track the plume, we look at the backscatter signal that shows an increase in signal return at the plume depth. After subtracting the background LIDAR returns, we can recover the total plume backscatter. Figure 3b shows the log backscatter returns at the plume, where the signal above the grey region is considered plume backscatter, showing near-continuous observation of the plume throughout the dataset. However, this also shows we miss the plume at irregular intervals, meaning that flow-rate estimates will be biased low during these less-than-ideal atmospheric conditions. The conventional CO₂ concentration map for this dataset is shown in Figure 3c, where derivatives were taken over consecutive range bins through the plume. The increase in concentration (shown in red) is clearly evident at a range of 1.13 km. However, there are multiple spots where the plume backscatter badly biases the fit concentrations, showing up as the darker red spots around the relative times of 2 h, 8 h, and 10 h, and in Figure 3b when the plume backscatter goes into the highlighted red zone.

As discussed above, we avoid this problem by performing DIAL calculations over two discontinuous range bins on either side of the plume spike, as shown in Figure 2. The measured CO₂ across the plume is shown in Figure 4a (blue) for a 13 h overnight period. For comparison, the ambient CO₂ concentrations sampled from above the lab by a cavity ringdown point sensor (CRDS) are shown in red in Figure 4a. The increases in CO₂ concentration across the plume are shown in Figure 4b for this same period after subtraction of the ambient levels determined from the average over bins downrange of the stack. The error bars from the fit residuals are shown superimposed in Figure 4 and illustrate that the CO₂ levels are consistently higher than the ambient levels. We only see up to about a 100 ppm increase in CO₂ or equivalently a maximum of 6% total absorption over the range bin. These low values minimize the impact of saturated absorption effects [18]. The error bars shown in Figure 4 range from 5 ppm to 20 ppm and are nearly independent of the overlap with the plume center.

4. Discussion

This measured increase in CO₂ is translated into an estimated flow rate using the process described above. The calculated values of the final plume rise, Δh, plume sampling distance, z, plume speed, u_p and plume area, A_p, used here are shown in Figure 5. The atmospheric conditions and stack parameters determine the final plume rise shown in Figure 5a. This plume rise is combined with the location of the DIAL beam relative to the stack shown in Figure 5b and the local wind speed to calculate the total plume speed shown in Figure 5c. The area of the plume used to translate the mass flux into a flow rate in Figure 5d shows large fluctuations during the first two daytime hours. These values are calculated at a 1 min resolution and then averaged to match the 5 min DIAL resolution to calculate flow rates.

The estimated flow rate of CO₂ is shown in Figure 6a, with values ranging from 0 g/s to 1500 g/s of CO₂ being emitted from the CU power plant with a 5 min temporal resolution. The reported EDEP data from this year (2023) were used to generate average values of the emission flux over the month of February and are shown in Figure 6b for the two side-by-side stacks at 1 h resolution. Figure 6b shows the average reported natural gas burn rate in pounds per hour for each of the two stacks, with the error bars representing the variation at each hour over the month. From these data, a small relative increase in the CO₂ flux is evident at dawn and dusk. The DIAL-based estimated CO₂ flux with 1 h resolution is shown in Figure 6c, in orange, and compared with the values from the burn rate for that day, in purple. Over the 13 h dataset, we find relatively good agreement, with nearly half of the DIAL flux predictions well within the combined uncertainties of the measured and reported data. We note that (i) the total backscatter signal decreased significantly for the latter part of the night, resulting in somewhat reduced SNR, and (ii) the periods of low wind speed can easily lead to the observed dropouts in the measured flux rate because of increased uncertainties in the predicted plume location. Nevertheless, the DIAL flux rates are seen to never exceed the reported values over the full measurement period, as expected. Gaussian plume model equations show that the CO₂ increase corresponds to an average tracking error of less than two times the z-dependent dispersion parameter, σ_z (see details in Appendix A).

Current efforts are directed towards incorporating real-time Doppler LIDAR data, which are expected to significantly reduce the plume model uncertainties. Unlike the current weather station location near the lab, a Doppler LIDAR system allows for range-resolved monitoring of the wind field at the plume location. As the wind speed is the largest source of uncertainty in our tracking and flow-rate estimates, reducing this error will lead to significant improvements in the flux estimates. There have been recent advances reported for systems that obtain the wind field and gas concentration simultaneously for CO₂ and CH₄ [19,20]. We note that to recover the 3D wind field, these techniques must scan over a sufficiently large area [10]. With the small concentration increases reported in this application, such scanning would quickly lower the detected plume emission below the background environment fluctuations. The robustness of measuring both the wind field and the range-resolved CO₂ concentrations with standoff techniques will be effective for many GHG source types and surrounding environments.

5. Conclusions

In this paper, we have demonstrated a compact fiber-amplifier-based DIAL system to monitor the dry-air CO₂ concentration enhancements across an emission plume from a natural gas power plant in the Boulder, Colorado, area. The DIAL system samples the spectral region at 10 different frequencies and near the highest possible repetition rate for a direct detection method. The DIAL data together with wind-field predictions from a Gaussian plume model were used to estimate the CO₂ emission flux for a 13 h nighttime period on 10 February 2023. A detailed discussion has been given of the procedure used to actively track the plume using meteorological data sampled near the transceiver and the models used to estimate the CO₂ emission flux from this point source in an urban environment. The DIAL-derived flux was in relatively good agreement over the 13 h monitoring period with the actual emission flux determined from the reported natural gas burn rate. Underestimation bias and challenges of operating in sub-optimal atmospheric conditions have been discussed. Future work will include data from a Doppler LIDAR system to measure the 3D wind field near the stack to minimize plume tracking and model uncertainties.