Simulation Analysis of Unmanned Aerial Vehicle-Based Laser Remote Sensing for Methane Point Source Traceability and Leakage Quantification

Highlights

What are the main findings?

• A UAV-based laser remote sensing system combined with an enhanced Gaussian diffusion model and a GA-IPPF optimization algorithm can simultaneously and accurately estimate both the methane leakage rate and the spatial location of the source under various environmental conditions.
• The study quantitatively analyzes the impact of multiple error sources on the inversion results, revealing that wind direction and coordinate errors are the most critical factors affecting source localization accuracy.

What are the implications of the main findings?

• This research provides a practical and efficient methodology for the real-time, high-precision monitoring and quantification of methane point source leaks in industrial environments, overcoming the limitations of existing ground- and satellite-based methods.
• The GA-IPPF algorithm excels in handling wind uncertainties, providing an adaptable framework for mobile emission monitoring.

Abstract

Current methods for the high-precision real-time monitoring and parameter inversion of industrial methane point source leakage are insufficient. This research introduces a novel laser-based methane leakage monitoring approach for deployment on an unmanned aerial vehicle platform. An enhanced two-dimensional integral Gaussian diffusion model paired with a point sampling technique is employed to simultaneously determine the leakage rate and source location, integrating a genetic algorithm and an interior point penalty function algorithm for optimization. Simulations incorporating observational error sources are performed to quantitatively assess the accuracy of leakage parameter inversion under diverse errors, demonstrating the scheme’s viability. The accuracy of leakage parameter inversion achieved by the algorithm across various point sampling methods, gas plume characteristics, and wind speeds was examined, validating the assessment under multivariable influences in real observations. The proposed methodology was compared with two other leakage inversion optimization techniques, demonstrating its efficiency in addressing wind speed and directional effects. This study offers a practical method with significant implications for monitoring and quantifying industrial methane point source leakages.

1. Introduction

Methane (CH₄), the second most significant greenhouse gas, is critical to energy and climate change [1]. Research indicates that methane, despite having a shorter lifespan in the atmosphere, exerts a significantly stronger warming impact than carbon dioxide (CO₂) [2,3]. This implies that reducing methane emissions could mitigate global warming much faster than other greenhouse gases [4,5]. At present, methane emissions include those derived from natural and anthropogenic sources, such as agriculture, coal mining, combustion, waste disposal, and industrial processes [6,7]. Therefore, accurate methods for assessing methane emissions in various scenarios are essential. The oil and gas industry is the leading source of anthropogenic methane emissions, with leakages occurring during transportation, storage, and production [8]. As these leakages create significant uncertainties in emission estimates, the prompt monitoring of leakage sources and rates is essential for safety and quantitative monitoring within the industry.

Existing CH₄ point source leakage detection approaches for natural gas transmission pipelines and storage tanks employ various in situ methods such as the sound wave method, the distributed optical fiber sensor method, and the cross-correlation analysis method [9]. However, these in situ methods cannot be utilized for large-scale inspections, and do not provide real-time gas plume diffusion data or leakage parameters [10]. Although optical gas imaging (OGI) is also used for CH₄ leakage monitoring in industrial plants, the high cost of conventional cooled OGI limits its widespread industrial use [11,12]. Additionally, OGI struggles with complex background radiation, complicating CH₄ quantification despite advances in gas concentration inversion methods [13,14]. Effective CH₄ leakage monitoring equipment should have high spatial resolution and the capability for timely, quantitative detection. However, the low spatial resolution of satellites prevents timely CH₄ detection, and existing algorithms for quantifying CH₄ via satellite observations are not yet mature [15,16,17]. Therefore, a combination of leakage detection technologies may be required to achieve large-scale regional CH₄ point source emission measurements, individual site emission quantification, and major equipment emission monitoring [18,19].

The monitoring of CH₄ point source leakage parameters has led to the development of new technologies and experimental methods [20,21]. At present, CH₄ concentrations in pipelines are monitored by ground spotting, using in situ devices placed in critical areas such as valves. This method enables multi-point concentration measurements but struggles with quantifying leakage rates and identifying leak sources, and also has a limited observational range [22]. Moreover, CH₄ leakages in defined areas can be monitored using mobile platforms with in situ instruments [23,24]. While these systems detect CH₄ concentrations in plant areas, it is time-consuming for ground platforms to achieve large-scale observations and they are generally insufficient for quantifying point source leakage rates [25]. Ground-mobile CH₄ laser telemetry platforms face challenges in point source traceability and leakage rate calculations due to observational angle limitations and obstructions [26,27]. Airborne sensors, capable of low-altitude flights, can collect CH₄ concentration data and estimate leakage rates using cross-sectional mass balance or Gaussian diffusion techniques [28,29,30]. However, the high experimental costs and trajectory limitations of airborne platforms restrict their use for the continuous, real-time monitoring of point leakage sources in industrial settings.

In the context of CH₄ plant point source leakage monitoring, unmanned aerial vehicles (UAVs) demonstrate rapid maneuverability and automatic inspection capabilities [31,32,33,34]. Equipped with sensors, they can swiftly sample CH₄ concentrations at specific locations using two primary observation methods. The first method employs in situ CH₄ monitoring equipment, requiring UAVs to perform multi-point sampling on a diffusion cross-section of the leakage point source [35,36,37]. This method involves a complex path and samples only one cross-section at a time, with leakage rate calculations using the mass-balance methodology generally excluding the leakage source. The second method uses a CH₄ laser telemetry instrument for integral concentration detection of point source leakages [38,39,40]. This method allows for CH₄ detection downwind of the leakage without altering the UAV’s altitude, enabling the rapid acquisition of a two-dimensional integral distribution of the entire leakage concentration. Researchers have utilized a combination of ground-based laser telemetry equipment and UAVs to detect methane leak concentrations across extensive areas through UAV-mounted mirrors [41,42]. However, the available literature has reported limitations regarding quantitative and traceable detection methods for point sampling efficiency and leakage rates of CH₄ concentrations.

In summary, the existing techniques for monitoring CH₄ point source leakage often emphasize either the leakage rate or the source location, resulting in less effective outcomes. Enhancing leakage observations by integrating gas plume diffusion characteristics remains underdeveloped. To address these issues, this research introduces a novel detection scheme and optimized traceability method for CH₄ point source leakages based on the use of a UAV in order to determine leakage rates and sources. The methodology involves UAV-mounted CH₄ telemetry equipment and gas plume simulation models for multi-dimensional data sampling and rapid patrol observations. A genetic algorithm–interior point penalty function (GA-IPPF) computational optimization method quantifies leakage sources and rates. The study designs an experimental data collection scheme for continuous leakage from ground pipelines or CH₄ point sources. To evaluate the observation algorithm’s performance, various error sources are introduced, aiding in the assessment of the leakage parameter calculation method and offering a practical reference for future applications.

The remainder of this article is structured as follows: Section 2 discusses the proposed UAV CH₄ observation system’s structure and composition, the Gaussian integral diffusion model, and the UAV observation program, along with the composition of simulation data and the algorithmic flow. Section 3 verifies the feasibility of the CH₄ point source leakage parameter estimation method using the Gaussian diffusion model and simulation data, assessing the quantitative impacts on leakage rate and source location under different error conditions. It also compares the proposed algorithm’s advantages with two other traceability methods. Section 4 summarizes the study’s contributions, offering an observational solution for monitoring and preventing CH₄ leakage from point sources in plant areas and providing technical support for further development of CH₄ leakage quantification and traceability using mobile platforms.

2. Materials and Methods

2.1. Methane Detection System

A UAV-based CH₄ leakage monitoring system is presented, comprising a UAV, CH₄ detection equipment, and an environmental parameter measuring instrument. As illustrated in Figure 1a, the UAV monitors a CH₄ point source leakage at an altitude of 30–50 m and detects the CH₄ concentration downwind. The UAV, equipped with GPS and altitude measurement tools, synchronously records its coordinates and altitude. To complement the ground-based weather station in collecting environmental data, the wind field for this study is characterized using a ground-based weather station measuring at a height of 10 m. This approach is justified for two primary reasons: First, the core objective of this work is to validate the inversion algorithm under a controlled and simplified scenario, focusing on near-surface methane leaks where the wind field can be reasonably approximated using a single representative value. Second, this method reflects a common and practical operational setup in many industrial monitoring contexts, where relying on existing ground station data is more feasible than equipping every drone with sophisticated anemometry devices. The CH₄ monitoring equipment is mounted on the UAV, as shown in Figure 1b. The CH₄ detection system consists of a 1653.7 nm tunable laser, a collimated transmitter, a Risley scanning module, a receiver lens, a near-infrared detector, and a data reception and control module. The control module regulates the rotational speed and direction of the wedge mirrors. The reception module captures concentrations and sample coordinates, while the data processing module stores data in real-time. The CH₄ instrument utilizes high-speed laser sweeping and Risley scanning mirrors for efficient point sampling. The UAV pauses to observe downwind CH₄ leakage, avoiding the extensive flight patterns required by in situ CH₄ monitoring equipment. This methodology enables rapid and efficient concentration sampling and, in conjunction with the CH₄ leak parameter calculations detailed in Section 2.3, enhances CH₄ leakage monitoring efficiency. The simulated analysis in this study was performed using software of MATLAB (v2024).

2.2. Gas Diffusion Model

The gas diffusion model was employed to simulate CH₄ point source leakage, examining gas concentration variations under different wind speeds, leakage rates, and atmospheric stability. For typical pipeline CH₄ point source diffusion, this study uses the adaptable and widely used Gaussian diffusion model for theoretical simulation. The Gaussian plume model—also known as the normal distribution model—is applicable for modeling the concentration of a neutral gas leakage in the air, and is suitable for emission sources with a flat, continuous, and stable downgradient surface, without considering the simulation of chemical deposition. In the general diffusion equation, it is assumed that the leakage source of air mass diffusion is at the origin of the whole coordinate system—i.e., the point (0, 0) in the coordinate axis—however, the location of the leakage point is not considered to be at the origin in the next diffusion, but is instead at an arbitrary location within the observation plant area, as shown in Figure 2. At this time, the coordinates of the source of the leakage of the air mass are (x0, y0). In addition, considering the effect of wind speed on the leakage process, the direction of diffusion of the air mass is related to the change in wind speed and, as such, it is necessary to convert the coordinates according to the direction of wind speed and the location of the leakage source. Thus, the concentration point within the air mass is converted to the direction of wind speed for modeling. In this way, the angle between any point in the air mass and the origin of the established coordinate system and the angle between the wind speed and the origin of the coordinate system are derived. The process of coordinate transformation and derivation of the diffusion equation is as follows:

Firstly, the leak source coordinates need to be normalized to the coordinate origin. When any point on the air mass with coordinates (x_i, y_i) is obtained, the origin coordinates are converted using the following equation:

$$\begin{array}{l}{x}_1=x_i-x0\\ y_1=y_i-y0\end{array}$$

(1)

After that, based on the direction of diffusion of the air mass, the relative angles can be obtained with respect to the coordinates, which can be either the observed coordinate system or an arbitrarily determined coordinate system. When only the diffusion of the air mass is taken into consideration, the coordinate system is established due east and due north. In simulated observations, the coordinate system can be converted according to the observation azimuth angle and wind direction angle to determine the angle of the air mass diffusion direction relative to the simulated coordinate system, which will be discussed later. The present equations are based on the coordinate systems in the east and north directions. The wind speed is decomposed in terms of the coordinate components, allowing the downwind and crosswind distances of the concentration points within the air mass to be determined. Subsequently, the wind speed components are calculated as follows:

$$\begin{array}{l}\beta =\alpha -180\\ u_x=u\cdot \sin \beta \cdot \pi /180\\ u_y=u\cdot \cos \beta \cdot \pi /180\end{array}$$

(2)

Here, the meteorological wind direction starts from due north and turns clockwise in the direction from which the wind is coming, the angle is calculated using the wind angle, and u is the wind speed vector. Next, to calculate the air mass downwind and crosswind distance, it is necessary to calculate the angle between the wind direction and the line connecting the origin and any plume point, which is calculated using vector multiplication and then divided by the modulus length to obtain the cos value of the angle. Then, the inverse of cos is taken to obtain the angle θ, which is calculated as follows:

$$\theta =\arccos \left({\displaystyle \frac{u_x\cdot x_1+u_y\cdot y_1}{u\cdot \mathrm{sqrt}\left(x_1^2+y_1^2\right)}}\right)$$

(3)

where x and y correspond to the coordinate values of any concentration point within the air mass. Next, the downwind distance R_D and crosswind distance R_C are calculated as follows:

$$\begin{array}{l}{R}_C=\cos \theta \cdot \mathrm{sqrt}\left(x_1^2+y_1^2\right)\\ R_D=\sin \theta \cdot \mathrm{sqrt}\left(x_1^2+y_1^2\right)\end{array}$$

(4)

According to Pasquill’s classification method, atmospheric stability can be divided into six grades (A–F). The meteorological conditions of A, B, and C are unstable, E and F are stable, and D is neutral [43,44]. The downwind distance R_D is introduced into the equations in Table 1 to calculate the diffusion coefficient under different atmospheric stabilities, allowing the diffusion coefficient of ${\sigma }_y$ in the crosswind direction to be obtained. Given that UAV observations are used in this study, the integral concentration of CH₄ is detected and the relevant two-dimensional concentration integral equation for the z-direction of the Gaussian diffusion model is as follows:

$$\begin{array}{l}{C}_y(x_i,y_i)={\displaystyle \frac{Q}{\sqrt{2\pi }{\sigma }_y\left(R_D(u,\theta ,x_i,y_i,x0,y0)\right)\cdot u}}\\ \cdot \exp \left(-{\displaystyle \frac{1}{2}}\cdot \left({\displaystyle \frac{R_C\left(u,\theta ,x_i,y_i,x0,y0\right)}{{\sigma }_y\left(R_D(u,\theta ,x_i,y_i,x0,y0)\right)}}\right)\right)\cdot {\displaystyle \frac{R\cdot T}{\mathrm{M}\cdot P}}\end{array}$$

(5)

where $C_y\left(x_i,y_i\right)$ indicates 2D diffusion-integrated concentrations at the corresponding coordinates, mg/m², as shown in Figure 2; Q represents the leakage rate, mg/s; u is average wind speed, m/s; ${\sigma }_y$ is the diffusion coefficient of the corresponding air mass in the cross-section direction of wind direction, m; $x_s$ and $y_s$ represent the spatial coordinates of the leakage source; R is the gas constant, $R=8.314\mathrm{P}\mathrm{a}·{\mathrm{m}}^3/\left(\mathrm{m}\mathrm{o}\mathrm{l}·\mathrm{K}\right)$; T is the ambient temperature; P is the ambient pressure; and M is the molar mass of CH₄, 16 g/mol. The Gaussian model is built based on the following assumptions: (1) the wind field is stable, the wind speed is constant and greater than 1 m/s, and there is no violent change in the wind direction during the observation time; (2) diffuse gas plumes move in space with normal distribution; (3) gas plumes follow the law of conservation of mass in diffusion, and the reflection from the ground is 1; and (4) the leakage rate is uniform and continuous. The calculation of atmospheric diffusion coefficients at different atmospheric stabilization (AS) levels is shown in Table 1.

Figure 2. Atmospheric dispersion model showing the wind direction coordinate transformation relationship: (a) before conversion and (b) after conversion. The green dashed line represents the wind direction component.

Figure 2. Atmospheric dispersion model showing the wind direction coordinate transformation relationship: (a) before conversion and (b) after conversion. The green dashed line represents the wind direction component.

Table 1. Calculation of atmospheric diffusion coefficients at different AS levels.

AS	${\mathit{\sigma }}_{\mathit{y}}$
A	0.22 RD/(1 + 0.0001 RD)0.5
B	0.16 RD/(1 + 0.0001 RD)0.5
C	0.11 RD/(1 + 0.0001 RD)0.5
D	0.08 RD/(1 + 0.0001 RD)0.5
E	0.06 RD/(1 + 0.0001 RD)0.5
F	0.04 RD/(1 + 0.0001 RD)0.5

Table 1. Calculation of atmospheric diffusion coefficients at different AS levels.

AS	${\mathit{\sigma }}_{\mathit{y}}$
A	0.22 RD/(1 + 0.0001 RD)0.5
B	0.16 RD/(1 + 0.0001 RD)0.5
C	0.11 RD/(1 + 0.0001 RD)0.5
D	0.08 RD/(1 + 0.0001 RD)0.5
E	0.06 RD/(1 + 0.0001 RD)0.5
F	0.04 RD/(1 + 0.0001 RD)0.5

2.3. Leakage Data Monitoring

The utilization of UAVs equipped with CH₄ telemetry instruments for observing CH₄ concentrations at leakage points represents an effective methodology for rapid surveys, demonstrating significant potential for monitoring leaks in the oil and gas industry. The UAV is equipped with a device to determine its flight position. Initially, an observation coordinate plane is established and the UAV conducts a survey. The UAV observes the near-surface leakage plume from the air and obtains the integral concentration in the Z direction, ultimately obtaining the two-dimensional concentration distributions in the coordinate axes X and Y. The model connecting the raw instrument signal to the path-integrated methane concentration is based on the Beer–Lambert law for laser absorption spectroscopy. The measured light attenuation at the specific methane absorption line is directly proportional to the number of methane molecules along the optical path. This path-integrated concentration (in units of ppm·m) is the primary data product for each line-of-sight measurement. This process effectively translates the one-dimensional path measurements into a two-dimensional field, representing the integrated concentration along the path direction at each point in the plane. When the CH₄ concentration exceeds a predetermined threshold, a point source leakage is identified. In Figure 3a, an observation scheme for UAV-mounted CH₄ telemetry is proposed based on ground CH₄ point source leakage characteristics. The UAV operates at a specified altitude and, upon detecting a leakage, the telemetry instrument samples concentration points within its field of view using a specialized Risley scanning device, obtaining corresponding concentration coordinates. Steeper-angled scans result in a longer atmospheric path length, leading to greater laser beam dispersion and energy attenuation. This increases the measurement error and can cause the signal-to-noise ratio to drop below the instrument’s detection threshold, resulting in data gaps (‘missing spots’). Our analysis explicitly accounts for this by defining a valid detection threshold. Only measurements with a signal strength above this threshold are used in the 2D reconstruction and subsequent inversion. The discrete angular steps of the scanner inherently limit the spatial resolution of the reconstructed plume. While this may prevent the perfect recreation of an ideal continuous scan pattern, our scanning protocol and angular resolution were designed to be sufficient to capture the core structure of typical methane plumes. In Figure 3b, (x, y) represents the UAV’s position relative to the ground, while scanning point coordinates (x’, y’) are computed using the offset (∆x, ∆y). Leakage parameters are subsequently determined by combining the proposed leakage rate and source localization methods. This sampling scheme facilitates stationary concentration sampling using a UAV, eliminating complex flight path planning and enabling efficient leakage parameter inversion with fewer observations. In comparison to in situ CH₄ measurement equipment, the UAV’s observation path is more straightforward and practical, significantly simplifying the sampling process and enhancing its application value.

2.4. Data Composition and Processes

Figure 4 introduces the simulation data composition and process. Initially, the distribution of ground CH₄ point source leakage plume is simulated using a Gaussian diffusion model. Following this, the UAV and CH₄ telemetry instrument gather data including flight altitude, Global Positioning System (GPS) coordinates, CH₄ concentration, and Risley scanning angle. The flight altitude, Risley mirror scanning angle, and GPS data are used to calculate calibration coordinates for concentration points within the plume. These coordinates, along with CH₄ concentration and meteorological parameters (wind speed, wind direction, pressure, humidity, temperature), are used to form a calibration data matrix. This matrix is then fed into a leakage retrieval model, which combines a Gaussian diffusion model with an optimization algorithm to determine the leakage rate and the exact source location. To ensure the reliability and accuracy of methane leak detection, this study employs a multi-stage adaptive detection strategy. The detection threshold is set at a path-integrated concentration of >100 ppm·m—a value significantly higher than the atmospheric background level of approximately 100 ppm·m at a 50 m observation height, thereby effectively minimizing false positives. To mitigate instrument noise and concentration jitter near the instrument’s lower detection limit of ~5 ppm·m, the system utilizes a dual-averaging approach for noise suppression: spatial averaging at each waypoint during the coarse survey cruise, and temporal averaging through time-series accumulation at each scan point during the hovering fine-scanning phase to enhance the signal-to-noise ratio. Furthermore, the system features an adaptive scanning capability, increasing the scan speed for strong plumes and automatically extending the integration time for weak signals to ensure data quality. This approach achieves an optimal balance between survey coverage and measurement precision, providing a solid foundation for subsequent inversion calculations.

2.5. Location and Quantization Algorithm

Figure 5 demonstrates the enhanced GA-IPPF algorithm for calculating leakage parameters using two-dimensional integral concentrations of gas plumes. This approach integrates theoretical Gaussian diffuse plumes with simulated sampling points, incorporating CH₄ detection coordinates into Equation (1) to determine theoretical concentrations. Concurrently, the UAV CH₄ system measures the simulated concentrations at these coordinates. Optimization is achieved by comparing the theoretical and measured concentrations. The leakage parameters are selected as the optimization values, which are Q: x(1), x0: x(2), y0: x(3). The minimization function is defined as follows:

$$\begin{array}{l}f(x_i,y_i)=\left\{{\displaystyle \frac{x(1)}{\sqrt{2\pi }u{\sigma }_y(R_D(u,\theta ,x_i,y_i,x(2),x(3)))}}\cdot \exp (-{\displaystyle \frac{R_C^2(u,\theta ,x_i,y_i,x(2),x(3))}{2{\sigma }_y^2(R_D(u,\theta ,x_i,y_i,x(2),x(3)))}}\right\}\\ \cdot {\displaystyle \frac{R\cdot T}{\mathrm{M}\cdot P}}-C_y^{\prime }(x_i,y_i)\end{array}$$

(6)

where $C_y^{\prime }\left(x_i,y_i\right)$ are the measured concentrations. Further, we construct the optimization function F(x) as:

$$F(x)={{\displaystyle \sum _{i=1}^m\left[C_y(x_i,y_i)-C_y^{\prime }(x_i,y_i)\right]}}^2$$

(7)

Firstly, an initial matrix, $x_0$, containing the location of the leakage source to be solved and the leakage rate is determined; then, we set the initial value of penalty factor as $r_0$, the rate of change in the penalty factor as q, the given error as $\epsilon >0$, and k = 1. The penalty function equation is constructed as follows:

$$G\left(x,r_k\right)=F(x)-r_k{\displaystyle \sum _{j=1}^p\log (g_j(x))}$$

(8)

where $g_j\left(x\right)$ represents the inequality constraints, $j=1,\dots p$, where p represents the number of constraints. The boundaries are set as $lb\le x_j\le ub,j=1,2,3$. Next, the unconstrained optimization method is applied to find the penalty function, $G\left(x,r_k\right)$, and the extreme point is denoted as $x^{*}\left(r_k\right)$.

$$G(x^{\ast },r_k)=\min (G(x,r_k))$$

(9)

Afterwards, the extreme points of the penalty function are checked to determine whether they satisfy convergence to the optimal solution of the problem, with the judgment criterion defined as follows:

$$abs\left\{rk{\displaystyle \sum _{j=1}^p\log (g_j(x))}\right\}\le \epsilon$$

(10)

If the extreme point satisfies the condition, $x^{*}\left(r_k\right)$ is the optimal solution to the constraint and we stop iterating; otherwise, we continue to find extreme points and reduce the penalty factor r, setting $r_{k-1}=q{r}_k$, and take $x^{*}\left(r_k\right)$ as the starting point of the next iteration. At the final iteration, the convergence condition $x^k$ is output, which comprises the optimal leakage source location and leakage rate parameters. Figure 5 illustrates the algorithm’s process flow, factoring in various error sources during the inversion of simulated leakage parameters: CH₄ concentration (C) error, wind speed (WS) error, wind direction (WD) error, and coordinate location (CL) error. Multiple error sets were applied to the simulated measurement data to quantitatively assess different error sources, offering more insights for practical observations.

Figure 5. Genetic algorithm process for leakage traceability and leakage rate quantification. The colored rings represent the algorithm’s search range, gradually narrowing in on the leak location.

Figure 5. Genetic algorithm process for leakage traceability and leakage rate quantification. The colored rings represent the algorithm’s search range, gradually narrowing in on the leak location.

The simulation parameters employed in this study are presented in

Table 2. Simulation parameter settings.

Parameters	Value	Unit
Equipment Observation altitude	50	m
Risley angle	10	°
Leakage source coordinates	(x0 = 0, y0 = 1)	m
Leakage rate	100, 300, 500	mg/s
Atmospheric stability	A, D, F	-
Wind speed	2, 5, 7	m/s
Wind direction	270	°
Atmospheric Temperature	300	K
Atmospheric pressure	101,325	Pa

. The UAV, equipped with a CH₄ telemetry device, operates at 50 m above ground level with a Risley mirror angle of 10°. The flight altitude of 50 m was selected for this initial validation study as it represents a challenging, yet realistic operational scenario. At this relatively high altitude for ground-source quantification, the path-integrated methane signal is weaker and the spatial resolution of the plume scan is lower, when compared to closer-range flights. By demonstrating the efficacy of our GA-IPPF algorithm under these ‘worse-case’ conditions, we aimed to conduct a robust and conservative validation of its core capabilities. Success under these conditions strongly suggests that the algorithm would perform equally well or better at lower altitudes, where signal strength and data density are inherently higher. The simulated leak source coordinates are at (x0 = 0, y0 = 1), which avoids initializing the source at the origin (0, 0) of the coordinate system. Based on typical leakage rate detection limits, the leakage rates for three simulated gas plumes were chosen as 100/300/500 mg/s. AS was categorized as A/D/F, and wind speeds were set at 2/5/7 m/s. This study examines the quantification ability of the proposed algorithm regarding leakage rates and sources under different leakage and atmospheric conditions (e.g., wind speeds). The ambient wind direction was fixed at 270°, with temperature and pressure at 300 K and 101,325 Pa, respectively.

3. Results

3.1. Simulation Data Sampling

Figure 6 depicts the distribution of selected simulation data points using Risley mirrors and a gas plume at various rotational speed ratios, with the mirrors rotating in opposite directions. One mirror’s speed was fixed at 0.9 r/s, while the other mirror’s speed ratios were −1.75, −2.25, and −2.7, with the negative sign indicating opposite rotation. The observation height was 50 m, and the UAV-based observation model had an 8 m scanning radius. Increasing the Risley mirror speed can reduce total scan time; however, for CH₄ telemetry, a lower speed is maintained to meet practical data sampling needs, balancing CH₄ sampling frequency and signal-to-noise ratio. The point distribution density varies with different speed ratios, affecting the concentration sampling density. The distribution characteristics of concentration points within the gas plume at three different speeds were analyzed in terms of their impact on quantitative accuracy and error in leakage parameter estimation, considering variations in sampling point density.

The distribution and concentration of the gas plume under various conditions were simulated, as depicted in Figure 7, illustrating variations in the gas plume for different leakage rates, AS, and wind speeds. The coordinates of the leakage source were established as (x0 = 0, y0 = 1), and the wind direction was designated as 270° to standardize the calculations. Figure 7a–c demonstrate the gas plume distributions at a leakage rate of 300 mg/s and wind speed of 2 m/s, corresponding to three AS (A, D, and F). When the AS increases, the range of concentration distribution in the wind’s tangential direction decreases, resulting in a higher CH₄ concentration distribution. Figure 7d–f present the gas plume distributions under wind speeds of 2, 5, and 7 m/s at a leakage rate of 300 mg/s and AS F. With increasing wind speed, CH₄ is transported downwind more rapidly and the concentration distribution becomes more uniform. Figure 7g–i illustrate the gas plume distributions at leakage rates of 100, 300, and 500 mg/s, with a wind speed of 2 m/s and AS D. Higher leakage rates result in increased CH₄ concentrations at the same downwind distance, with a consistent distribution shape. These higher-concentration data points facilitate simulated observation data sampling, enabling the adjustment of UAV observations downwind from the leakage source.

Simulations and analyses under varying Risley mirror scanning mode and gas plume distribution are presented in Figure 8. In particular, the UAV observes ground-level gas plumes with varying distributions, utilizing different rotational speed ratios. Increased AS elevates the downwind gas plume concentration, while a rotational speed ratio of −2.25 facilitates high-concentration point sampling but reduces data points along wind shear. For low AS (A), longitudinal sampling points increase while the number of high-value concentration points decrease, necessitating higher measurement accuracy. As the rotational speed ratio shifts from −2.25 to −2.7, the point sampling density increases, enhancing data recovery of the gas cluster leakage, affecting the inversion of leakage parameters, and introducing potential error sources.

3.2. Inversion Error Effect Analysis

Subsequently, this study analyzed the impact of simulating multiple error sources on the inversion results derived from the UAV-observed CH₄ point source leakage rate and source location based on the GA-IPPF algorithm. The error sources include relative error in the measured concentration and absolute errors in the wind speed, wind direction, and coordinate position. In this study, we assumed that all major random error sources (e.g., sensor measurement noise, wind field fluctuations) follow a Gaussian (normal) distribution. This assumption is based on the central limit theorem and is a common model for describing random measurement noise. In our analysis, these error sources were treated as statistically independent. This is fundamental to our use of a controlled variable approach to isolate the impact of each error type. Specifically, when investigating the influence of a single error source (e.g., wind speed error) on the inversion results, we introduced random perturbations conforming to a Gaussian distribution only to that input, while holding all other inputs (e.g., wind direction, concentration) at their ideal, error-free values. This methodology ensures that any observed change in the output is unequivocally attributable to the specific error source under investigation, thereby guaranteeing the reliability and clarity of the causal analysis. Additionally, the study evaluated the algorithm’s ability to quantify leakage parameters under different rotational speed ratio, ambient wind speed, leakage rate, and AS conditions, considering these as four error sources, serving as a reference for practical observations. Among them, the typical inversion concentration error of the methane telemetry instrument is 5–10% and the error range was set to a maximum of 30%, considering the stability of the UAV flight and the different reflectivity changes on the ground. The typical anemometer error is ±5%. Considering the set wind speed value, the wind speed error was considered to be in the range of 0.1–0.5 m/s. The wind direction setting was based on a selected range from 10° to 50°, which effectively covers the range of natural fluctuations in wind direction under a wide range of environmental conditions, from stable to extremely unstable. In addition, this range also incorporates the additional instrumental errors that can occur with low-cost wind speed and direction sensors in complex environments, and this broad sensitivity analysis allowed us to comprehensively assess the response of the model outputs to wind direction uncertainties, ensuring the robustness of the study’s conclusions. Taking into account the scanning angle error as well as the UAV flight steady-state, the error assessment can be maximized by setting a larger range of positional coordinate errors for the ground observation. The coordinate absolute error range (0.1–2.5 m) models combined georeferencing uncertainties from GPS inaccuracy, platform pointing/jitter, and sensor temporal misalignment. The deviation settings for the four error sources are shown in

Table 3. Error parameter settings.

Parameters	Value	Unit
Relative error of concentration	5, 10, 30	%
Absolute error of wind speed	0.1, 0.3, 0.5	m/s
Absolute error of wind direction	10, 30, 50	°
Absolute error of sample coordinate location	0.1, 1, 2.5	m

.

To evaluate the algorithm and the impacts of error sources, random errors were introduced to the simulation data and the rotational speed ratio of the double wedge scanning mirror was varied to modify the sampling strategy. Twenty calculations were conducted, and the Q value and the location of the leakage source (x0, y0) were statistically analyzed, with the results presented in Figure 9; specifically, Figure 9a,e,i illustrate the effects of deviations in Q and x0, y0 for different speed ratios and relative concentration errors. When the relative concentration error was below 10%, the mean values of the three leakage parameters aligned closely with the theoretical baseline values. At a 30% relative concentration error, the mean values of the inversion parameters for the three speed ratios remained near the baseline values, although the number of error points and the distribution range of deviation values increased, with minimal difference between the inversion results at the three rotational speeds. For wind speeds below 0.1 m/s, the inversion values deviated from the theoretical values by approximately 3%. When the wind speed deviation ranged from 0.3 to 0.5 m/s, the leakage rate deviation varied from 16 to 27%, with a median deviation from the theoretical value ranging from 3 to 10%. Increasing the rotational speed ratio brought the median closer to the baseline value, indicating that adjusting the rotational speed ratio to increase the number of point samples positively influences the inverted leakage rate in the presence of wind speed errors. The leakage source location inversion results indicate that wind speed variations from 0.1 to 0.5 m/s have minimal impact, with results close to baseline values. Figure 9c,g,k show that the absolute wind direction error affects the leakage parameter. When the wind direction changes were below 30°, the inversion results for the leakage rate Q were near the baseline. At a 50° change, the absolute deviation in leakage rate ranged from 100 to 200 mg/s, increasing with higher rotational speed ratios. For x0, the inversion results matched the baseline for wind errors of 10–30°. At 50°, the deviation in x0 ranged from −10 to 20 m. For y_s, the median inversion results remained close to the baseline for wind errors of 10–50°, with absolute deviation increasing from 5 m to 40 m, while rpm ratio sampling point changes had minimal effect. Figure 9d,h,i illustrate that the absolute error in sample position affects leakage parameters. With coordinate errors of 0.1–1 m, the inversion results were near the baseline. At a 2.5 m error, the inversion error and the median deviation from the baseline increased, with the median leakage rate reaching 320 mg/s and a relative deviation of 6% from 300 mg/s, an x0 coordinate error of 8 m, and a deviation in y0 of 1.1 m. The y0 absolute error showed less variation, indicating that Q and x0 are more sensitive to coordinate errors than y0. The results shown in Figure 9 demonstrate that, when the error-influencing factors are constant, the change in the rotational speed ratio does not have a significant effect on the inversion results.

The ability to quantify gas plume leakage rates for different wind speeds was examined along with the impacts of four error sources, as depicted in Figure 10. Figure 10a,e,i show the variations in leakage rate and source location relative to errors in measured concentration and wind speed. As the relative error in measuring CH₄ concentration increased, the range of Q inversion errors widened. At a 10% measurement error, the Q deviation was 10 mg/s. With a 30% measurement error, the deviation was smallest at a wind speed of 2 m/s and increased at 5 m/s and 7 m/s, with a maximum deviation between 260 and 320 mg/s and relative deviation of 6–20%. Higher wind speeds significantly affected Q when relative concentration errors were 30% or higher. The inversion values of x₀ and y0 also increased with increasing measurement errors. While a 10% measurement error had minimal inversion effects, the deviation range increased with wind speed at 30%, causing larger inversion deviations. Figure 10b,f,j show the inversion results with varying absolute wind speed. As the absolute wind speed error increased from 0.1 to 0.5 m/s, the range of inversion deviation values for Q increased. With a 0.5 m/s error and 2 m/s ambient wind speed, inversion errors ranged from 230 to 360 mg/s, peaking at a 20% deviation relative to 300 mg/s. Higher ambient wind speeds reduced Q inversion errors, indicating that low wind speeds amplify the impact of wind speed measurement errors on Q. For x0 and y0, the inversion results remained close to the baseline and were insensitive to wind speed errors of 0.1–0.5 m/s. Figure 10c,g,k reveal that inversion biases for the three leakage parameters generally increased with absolute wind direction errors. For Q and x0, the inversion biases were smaller at a 10° wind direction error, while y0 showed a larger bias. At a 30° wind direction error, Q and x0 deviated more with a diffuse gas plume and 7 m/s ambient wind speed. With a 50° wind direction error, the Q and x0 inversion biases decreased with rising ambient wind speeds, while y0 showed consistent inversion error biases across 10–50° wind errors. Figure 10d,h,l display the inversion results under absolute sample coordinate errors. As the position errors increased from 0.1 to 2.5 m, the inversion errors for Q, x0, and y0 increased. At a 0.1 m positional deviation, the inversion results were close to the baseline with minimal impact. With a 2.5 m positional error, the Q and x0 inversion results approached the baseline as wind speed increased, although the deviation intervals widened. For y0, both the median and deviation range increased at higher wind speeds but the relative baseline deviation was smaller than that for Q and x0, indicating less susceptibility to positional coordinate errors. Figure 10 reveals that, among the various parameters, alterations in the acquired sample coordinate error have the most pronounced effect on both the inverted leakage rate and the leakage source location when the wind speed of the simulated plume is modified.

The accuracy of leakage parameter inversion for the gas plume was evaluated across different leakage rates and AS conditions using a sampling strategy with a rotational speed ratio of −2.7. Figure 11 and Figure 12 present the inversion results for two leakage rates—100 mg/s and 500 mg/s—under the influence of four error sources. Generally, the absolute deviation of the inverted leakage rate Q for both rates was comparable, increasing proportionally with the four error sources. The maximum deviation due to absolute errors in measured concentration and wind speed was around 20% for the three AS levels. For x0 and y0, the consistency of the four error sources increased, indicating that the inversion method provides more reliable results as the gas plume leakage rate rises from 100 to 500 mg/s. Regarding AS, the relative errors show that, for the 5% and 10% cases, Q, x0, and y0 were roughly equivalent to the baseline. In contrast, the 30% concentration deviation revealed greater deviations under higher AS (A). As the wind speed error increased, the deviation of Q also increased, with a maximum relative deviation of about 20% compared to 100 mg/s. Meanwhile, x0 and y0 were less affected by wind speed errors. For wind direction errors, a 10° error significantly impacted Q and x0, and the inversion result bias increased with stability from A to F for 30° and 50° errors. For y0, the inversion bias remained relatively consistent across the three stability levels. Figure 11 and Figure 12 demonstrate that the inversion results exhibited good agreement for gas plumes with leakage rates ranging from 100 to 500 mg/s. Furthermore, the inversion results for the F stability gas plume were more susceptible to wind direction variations compared to other stability levels.

A statistical analysis was conducted on the relative error ranges for Q and the absolute error distributions for x0 and y0, examining leakage parameter errors from various sources at leakage rates of 100 and 500 mg/s, as shown in Figure 13. The results show a consistent deviation distribution across both leakage rates under different gas plume conditions and error sources. Wind direction errors significantly impacted Q, x0, and y0, with Q showing a relative error range of −60% to 40%. Absolute wind speed error was the second most significant error source for Q. These findings highlight the necessity of high-resolution measurements in actual tests to reduce the effects of wind speed and direction, validating the UAV-based high-speed sampling used in this research. For x0, wind direction errors caused substantial inversion bias, while coordinate position errors were the second most influential factor affecting x0 inversion accuracy. For y0, an error range of 10–50° resulted in an absolute position deviation of −10 to 40 m, indicating the persistent significance of a 50° deviation. Thus, wind field stability and rapid data acquisition during the simulation process have more significant impacts on y0 positioning than on x0. According to Figure 13, the inversion of the leakage rate is most significantly affected by errors in wind speed and direction; for the x0 coordinate of the leak source, the primary influencing factors are wind direction and sample coordinate errors; and the y0 coordinate of the leakage source is predominantly impacted by wind direction errors.

3.3. Multi-Algorithm Performance Evaluation

Following its development, this study analyzed the impact of simulating multiple error sources on the inversion results of UAV-observed CH₄ point source leakage rates and source locations based on the GA-IPPF algorithm. The error sources included relative error in the measured concentration and absolute errors in the wind speed, wind direction, and coordinate position. Additionally, the study evaluated the algorithm’s ability to quantify leakage parameters under different conditions of rotational speed ratio, ambient wind speed, leakage rate, and AS, considered as error sources, which provides a reference for practical applications.

This study then proceeded to select two alternative leakage parameter inversion algorithms for a comparative evaluation with GA-IPPF. This analysis was conducted with the aim of examining the computational errors of leakage parameters across four error sources, evaluating the performance and merits of each algorithm, and ultimately provide a reference for algorithmic applications in practical observational settings. As shown in Figure 14, the GA-IPPF, interior point penalty function (IPPF), and genetic algorithm–particle swarm optimization (GA-PSO) algorithms were utilized to statistically calculate the leakage parameter inversion results under the four error sources. The GA-PSO algorithm combines the global search capability of the Genetic Algorithm and the local convergence speed of the Particle Swarm Optimization (PSO) algorithm. In this hybrid framework, the genetic algorithm’s population evolves through selection, crossover, and mutation operations, while the speed–position update mechanism of PSO is introduced to optimize the individuals in the offspring population, thus effectively avoiding premature convergence and accelerating the optimization search process. To ensure that all the compared algorithms (GA-IPPF, IPPF, GA-PSO) were evaluated under fair conditions, we set a uniform benchmark for comparison: all algorithms used the same random seeds to generate the initial population and, for GA-IPPF and GA-PSO, the control parameters (e.g., crossover rate, mutation rate, inertia weight) were set with reference to typical values in the relevant literature. After setup, they were preliminarily tested to ensure their basic performance on this research problem. Each algorithm was run independently the same number of times on each test case, and its statistical results were recorded to eliminate the effect of randomness. Figure 14a,e,i display the Q, x0, and y0 parameters calculated using the different algorithms under varying relative measured concentration errors. The inversion deviation increased progressively as the relative measured concentration error rose from 5% to 30%. The inversion results of the three algorithms are fundamentally similar, with minimal variation in their optimization effect on the measured concentration error. Figure 14b,f,j illustrate the impacts of absolute wind speed on the inversion results. For Q, the GA-IPPF algorithm showed superior optimization compared to the other algorithms for equivalent errors, with a median near the baseline of 300 mg/s. The x0 and y0 parameters were minimally affected by wind speed errors between 0.1 and 0.5 m/s. Figure 14c,g,k show the optimization results of the three algorithms under varying wind speed errors. Under a low error of 10° wind direction, the GA-IPPF and GA-PSO algorithms provided more favorable optimization results—closer to the baseline when compared to the IPPF algorithm, which showed a significant deviation. The IPPF algorithm showed limitations for wind errors of 30° and 50°, whereas the GA-IPPF algorithm optimized the leakage parameters more effectively under such wind errors.

Errors and data distributions were subsequently calculated for inversions with various error sources and algorithms. Figure 15a–c show the relative error in Q and the absolute coordinate errors in x0 and y0, indicating that GA-IPPF and IPPF outperformed GA-PSO in wind speed optimization, with smaller deviation ranges in relative errors. IPPF was less effective in wind speed error inversion, whereas GA-IPPF optimized both wind speed and direction. The x0 and y0 results reveal substantial wind direction errors in IPPF, while GA-IPPF and GA-PSO yielded better results. Figure 15d–f illustrate the mean and error distributions of Q, x0, and y0. The relative error of the measured concentration had a minimal impact on the inverted leakage rate, which was closest to the theoretical value of 300 mg/s across all algorithms. IPPF showed a more significant deviation in mean wind speed, and all algorithms exhibited larger errors compared to C. For wind direction errors, GA-IPPF’s calculated leakage rate was closest to the theoretical 300 mg/s, outperforming IPPF and GA-PSO. Wind direction errors introduced more discrepancies in y0, with GA-IPPF showing the smallest error. The GA-IPPF algorithm proposed in this study demonstrated optimal performance in handling wind direction errors, allowing for effective optimization under the influence of wind direction and speed, thereby controlling the leakage rate calculations and showing excellent comprehensive performance.

4. Discussion

The proposed UAV-based laser methane leakage monitoring system demonstrates significant potential for accurate and efficient quantification of methane point source emissions. By integrating an enhanced Gaussian diffusion model with genetic algorithm–interior point penalty function (GA-IPPF)-based optimization, this study addresses key challenges in simultaneous leakage rate estimation and source localization. Below, we discuss the implications of our findings and highlight the methodological strengths, limitations, and future directions.

The GA-IPPF algorithm outperforms traditional methods (e.g., IPPF and GA-PSO) in handling wind speed and directional uncertainties, achieving median leakage rate deviations of <10% under moderate error conditions (Figure 14 and Figure 15). This robustness stems from the hybrid optimization framework, which combines global search capabilities (genetic algorithm) with local refinement (interior point penalty function). The two-dimensional integral Gaussian model further improves accuracy by accounting for plume dispersion dynamics, particularly under variable atmospheric stability (AS) conditions (Figure 7, Figure 11 and Figure 12). Our error analysis revealed that wind direction errors (>30°) and coordinate inaccuracies (>1 m) dominate inversion uncertainties (Figure 9 and Figure 10). For instance, a 50° wind direction error introduces a 40 m deviation in source localization (y0), while concentration measurement errors (>20%) disproportionately affect low wind speed scenarios (2 m/s). These findings underscore the need for high-frequency wind field measurements to mitigate temporal variability and precision georeferencing (e.g., RTK-GPS) to minimize coordinate errors. Adaptive sampling strategies were employed, where higher Risley mirror rotational speeds (−2.7 ratio) improved point density in high-concentration regions (Figure 6). This study primarily evaluated the proposed methodology under a set of controlled simulation scenarios, including steady wind conditions and continuous leakage sources. While these idealized parameters enabled a systematic and quantifiable analysis of the algorithm’s core performance and its sensitivity to key factors, such as atmospheric stability, they inevitably simplify the full complexity of real-world environments. Operational conditions may involve intermittent leaks, complex terrain, and highly turbulent wind fields that were not explicitly tested here. Consequently, the direct applicability of our current findings to all possible scenarios—particularly those with extreme turbulence or calm conditions—requires further investigation. Nonetheless, the demonstrated robustness across a spectrum of common atmospheric stabilities (Classes A–F) suggests that the algorithm can be considered as a solid foundation for addressing a wide range of common emission scenarios. The framework established in this work provides a critical baseline and a validated starting point for future research aimed explicitly at incorporating these additional layers of real-world complexity.

Compared to satellite-based or ground mobile platforms, UAV systems offer (1) higher spatial resolution (8 m scanning radius at 50 m altitude) for pinpointing small-scale leaks; (2) rapid deployment, enabling real-time monitoring without complex path planning (Section 2.3); and (3) cost-effectiveness, as it avoids the need for extensive ground sensors or airborne campaigns. However, limitations persist, including susceptibility to extreme weather (e.g., winds > 7 m/s) and reliance on plume dispersion assumptions (e.g., steady wind fields). To transition from simulation to field applications, future work should (1) validate the proposed algorithm using real-world leaks, incorporating turbulence and complex terrain effects; (2) integrate multi-sensor data (e.g., LiDAR, hyperspectral imaging) to enhance plume characterization performance; and (3) optimize UAV flight patterns dynamically based on real-time plume detection.

While this study demonstrated the potential of the proposed method through comprehensive simulations under a range of scenarios, it is important to acknowledge its primary limitation: the validation is currently simulation-based. As with any computational model, while our simulations incorporate realistic parameters and uncertainty analyses (e.g., for coordinate errors, wind field variability, and concentration retrieval), they cannot fully capture the full spectrum of complexities inherent in real-world operational environments. Factors such as sensor drift, UAV platform instability under gusty conditions, and small-scale intermittent turbulence represent critical challenges that must be addressed for robust field deployment. Future work must therefore prioritize rigorous field validation campaigns. The insights gained from this simulation study—particularly regarding the sensitivity of the algorithm to specific error sources—provide an invaluable framework for designing those future experiments. They allow for the targeted collection of real-world data and the focused refinement of the algorithm under actual operational conditions, which is the essential next step towards transitioning this promising approach from simulation to practical application.

We acknowledge that higher flight altitudes are currently constrained by the signal-to-noise ratio of our sensor and the need for a sufficient number of valid data points to reconstruct the plume. Lower-altitude flights would alter the background concentration perception and improve data quality, theoretically enhancing quantification accuracy. Exploring this performance trade-off across a range of altitudes is a primary objective of our ongoing research. We intend to conduct systematic multi-altitude flight tests in subsequent experiments. This work will be essential for optimizing flight strategies and further refining the algorithm’s performance for a wider range of scenarios. It is important to note that the current simulation represents an idealized scenario. Several real-world factors that would introduce additional noise and uncertainty into the measurements were not simulated in this study. Variations in ground surface reflectivity at the laser wavelength would cause fluctuations in the returned signal strength, which could be misinterpreted as concentration changes. The impact of angled scans, which lead to a larger footprint and laser energy dispersion, is only partially accounted for by our path-length correction. More sophisticated modeling that includes beam shape and incidence angle effects on energy distribution is needed for higher-fidelity simulations. Our current work focuses on validating the core inversion algorithm against a well-defined benchmark. The aforementioned effects represent critical sources of error that should be characterized and mitigated in future work aimed at transitioning the technology to field deployment.

5. Conclusions

To achieve more effective surface CH₄ point source leakage monitoring, a crucial task in the oil and gas industry, a UAV-based methodology was proposed to detect CH₄ leakage concentrations and spatial locations. For this purpose, Gaussian diffusion models and optimization algorithms were integrated for ground-based CH₄ leakage simulation. The validity of the proposed methodology was confirmed through modeling and an optimized observation scheme, allowing for an analysis of the impacts of various errors on the leakage rate and source location inversion results during actual ground-based CH₄ leakage scenarios. The results indicated that, as error sources increase, the inverted leakage parameters present greater deviations. For wind speed errors ranging from 0.1 to 0.5 m/s, the coordinates of the leakage source closely align with the baseline, exhibiting minimal errors. The density of Risley scanning mirror sampling coordinates, ambient wind speed, leakage rate, and AS significantly impact leakage rate inversion. Adequate sampling ensures consistent inversion results, highlighting the importance of strategic sampling point density design. Ambient wind speed and AS influence the gas plume concentration point and wind tangential data point distributions. Lower wind speeds result in more concentrated gas plume concentration points, affecting the sampling accuracy. Additionally, the GA-IPPF algorithm’s efficiency was analyzed under varying leakage rates, demonstrating reliable inversion results within the range of 100–500 mg/s, thereby ensuring the reliability of leakage parameter inversion. Compared to another two algorithms, the GA-IPPF algorithm was shown to excel in terms of the inversion of wind speed and direction, effectively reducing errors due to wind field variations, thus showing promise to benefit practical applications.

This observation scheme mitigates the need for additional sensors along the pipeline, streamlines ground CH₄ leakage monitoring, and enables the more efficient simultaneous positioning of leakage rate and source location, outperforming single-parameter inversion methods. The developed UAV monitoring technology and traceability method offer innovative solutions for CH₄ point source monitoring. Future efforts will focus on constructing the UAV monitoring system, conducting field experiments, performing comparative studies, refining the algorithms and equipment parameters, and providing further references for future CH₄ point source monitoring.