Why Do Methane Emissions Occur? Towards a Predictive Framework for Risk-Targeted Mitigation in Oil and Gas Operations

Why Do Methane Emissions Occur? Towards a Predictive Framework for Risk-Targeted Mitigation in Oil and Gas Operations | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Why Do Methane Emissions Occur? Towards a Predictive Framework for Risk-Targeted Mitigation in Oil and Gas Operations Abdulmuiz Adekomi, Shuting Yang, Shannon Stokes, Arvind Ravikumar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7340966/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Addressing methane emissions from the oil and gas supply chain has emerged as a key near-term mitigation target. The past decade of research has improved our understanding of methane emissions, with a primary focus on quantifying emissions without describing their underlying causal mechanisms. In this work, we integrate source-specific methane emissions measurement from multiple large-area aerial surveys with source-tracked cause analyses to identify and analyze causal mechanisms that underlie observed emission patterns. Overall, 53% of all observed emissions can be attributed to specific causal categories, with the rest comprising normal operational emissions. While abnormal tank emissions were the most common cause, unloading events exhibited the highest average emission rate. Importantly, we find that large release events are not driven by fundamentally different causal mechanisms than those of small emitters, indicating that escalation due to specific operational conditions, rather than fundamentally distinct causes, drives high-magnitude emissions. In addition, we observe statistically significant quarterly and inter-operator variability in the prevalence of different causal categories, reinforcing the need for adaptive, operator-specific mitigation strategies. These findings support a shift in methane mitigation from generalized leak detection with one-size-fits-all solutions toward risk-targeted, process-informed mitigation. Earth and environmental sciences/Environmental sciences/Environmental impact Scientific community and society/Energy and society/Energy management Methane emissions Causal Analysis Predictive Analytics Super-emitters Risk: Mitigation Failure Mode and Effects Analysis (FMEA) Fault Tree Analysis (FTA) Appalachian Basin Escalation ratio Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Methane (CH₄) is a potent greenhouse gas with a global warming potential (GWP) 84–87 times greater than that of carbon dioxide (CO₂) over a 20-year period 1 , 2 . Methane alone accounts for nearly 30% of the global temperature rise since the pre-industrial era, making its mitigation crucial for addressing climate change 3 – 5 . Consequently, reducing methane emissions has become a key objective of global initiatives, such as the Global Methane Pledge, which targets a 30% reduction in methane emissions by 2030 6 . The oil and gas (O&G) industry represents one of the largest anthropogenic sources of methane, with emissions spanning the entire supply chain 7 . The Appalachian Basin, encompassing major production areas in Pennsylvania, Ohio, and West Virginia, is the largest natural gas-producing region in the United States, accounting for over 31% of the total U.S. dry natural gas production in 2024 8 . Given its significance, the basin has been the focus of several recent methane measurement campaigns. These studies have employed a range of top-down and bottom-up techniques, including aerial LiDAR, satellite retrievals, and on-site inspections, to quantify emission patterns across diverse facility types. While earlier research often focused on national-scale assessments, more recent measurement-informed analyses in Appalachia indicate that national inventories may not fully capture regional emissions, particularly reflecting an underrepresentation of episodic, high-emitting sources 9 – 11 . Although this basin generally exhibits lower methane intensity compared to oil-dominated regions like the Permian, its emission profile is nonetheless complex and often characterized by strong site-level heterogeneity 12 – 16 . However, these efforts have largely prioritized emission quantification, with limited emphasis on systematically uncovering the operational mechanisms underlying the observed emissions. This gap in causal understanding, particularly in a basin with such heterogeneous emission behavior, constrains the development of effective, cause-specific mitigation strategies. Addressing this knowledge gap requires a fundamental shift from quantifying how much methane is emitted to understanding why these emissions occur. Traditional quantification approaches, both bottom-up and top-down, are inadequate for diagnosing emission drivers. Top-down techniques, such as satellite and aerial observations, provide regional-scale insights but typically lack the resolution and operational information needed to attribute emissions to specific causes 17 – 22 . Conversely, bottom-up methods rely on emission factors and equipment counts that assume steady-state conditions, thereby failing to capture episodic, abnormal, or high-emission events 23 – 27 . Several studies have highlighted the significance of these overlooked sources. For instance, a study in the Fayetteville Shale attributed up to one-third of midday methane emissions detected via aircraft measurement to episodic activities like liquid unloading 28 . Similarly, routine unloading has been shown to produce highly skewed, heavy-tailed distributions owing to operational variability 29 . High-emission events have also been linked to specific operational failures, such as compromised thief hatches, malfunctioning controllers, and flare malfunctions 30 . These findings collectively highlight the limitations of conventional inventories in capturing the full spectrum and operational origins of emissions. Without a clear understanding of causation, most existing mitigation strategies remain reliant on generalized assumptions and tend to focus on reactive detection and response rather than proactive prevention guided by causal analysis. Bridging this gap requires methodologies capable of tracing emissions back to specific operational or equipment-related causes. For example, a large emission from a tank can be the result of an open thief hatch or point to upstream process issues unrelated to the tank, such as a stuck dump valve. Identifying these operational or failure pathways is important for accurate emissions accounting and designing targeted interventions. To address these limitations, this study introduces a causal analysis framework for methane emissions management. This approach provides a structured method for investigating and categorizing the operational mechanisms that drive methane emissions, facilitating the development of proactive mitigation strategies. Causal analysis techniques have been extensively utilized across diverse industries, such as industrial safety, engineering, aerospace, manufacturing, and healthcare. These methodologies provide a structured approach for diagnosing the underlying causal mechanisms and implementing corrective measures to prevent recurrence 31 . Specifically, this study integrates two established techniques from root cause analysis (RCA) and risk assessment: Failure Mode and Effects Analysis (FMEA), a semi-quantitative approach that evaluates and ranks potential failure modes by assessing their severity, likelihood, and detection 32 ; and Fault Tree Analysis (FTA), a logic-based hierarchical model that identifies critical failure events and the conditional pathways through which they lead to emission incidents 33 , 34 . In this study, we identify and analyze the causes of methane emission events detected through aerial measurement campaigns at oil and gas facilities in the Appalachian Basin. We combine measurement data with extensive operator engagement to identify and track underlying causes for observed emissions. We found that abnormal tank emissions is the most common cause of emissions. More importantly, super-emitter events are not caused by unique causal categories but rather represent escalations of common causes under specific operational conditions. The analysis also reveals significant quarterly and inter-operator variability in these causal categories and introduces a novel Escalation Ratio metric to quantify the likelihood of each category resulting in high-emission outcomes. This work represents one of the first basin-wide applications of causal analysis to methane emissions, offering a structured framework for shifting methane management beyond emission quantification toward causal understanding and prevention. These findings support the development of targeted, cause-specific mitigation strategies and contribute to the advancement of process-informed methane models. Ultimately, this approach provides practical guidance for designing proactive cause–driven methane mitigation strategies in the oil and gas sector, paving the way for predictive methane management. 2. Methods 2.1 Data Sources and Study Region Methane emissions data analyzed in this study were obtained as part of the 2024 Appalachian Methane Initiative (AMI), a basin-wide aerial measurement campaign focused on characterizing source-level emissions across the Appalachian Basin 35 . The study region spans approximately 54,000 square kilometers, covering the principal oil and gas production areas in Pennsylvania, Ohio, and West Virginia. This region includes a wide diversity of site configurations, equipment types, and operational practices, providing a representative setting for basin-wide emission analyses. Between April 2024 and December 2024, approximately 2,500 oil and gas production sites were selected for this analysis and surveyed across four quarterly aerial campaigns, comprising facilities operated by five major upstream companies participating in the AMI initiative. The detailed survey period is provided in Supplementary Table S1 . The survey timing was coordinated based on the weather conditions, flight logistics, and operational schedules. Each campaign generated an instantaneous snapshot of methane emissions at the time of overflight. The measurement dataset supports high-confidence attribution of emissions to specific equipment, forming the foundation for subsequent causal analysis. 2.2 Bridger Photonics Gas Mapping LIDAR Technology Bridger Photonics’ aircraft-mounted Gas Mapping LiDAR employs frequency-modulated continuous-wave (FMCW) laser technology, which sweeps perpendicular to the flight path of the aircraft. The system measures path-integrated methane concentrations in ppm-m. These concentrations were used to estimate methane emission rates based on wind speed and direction data obtained from nearby meteorological stations. The system overlays methane plumes on high-resolution aerial imagery to provide spatial context. Bridger's technology typically surveys dozens of sites each day, with the number of overflights per site varying based on the physical footprint. Larger sites often require multiple passes, resulting in some equipment being scanned more than once due to overlapping scan swaths. Consequently, a single emission source may produce multiple plume detections and corresponding emission rate estimates within the same visit. Bridger reports include both plume images and quality-controlled quantification estimates. The detection threshold is between 1–3 kg/h with a 90% probability of detection, depending on the atmospheric conditions. Uncertainty in emission rate estimates arises from several sources, including variability in wind speed and direction inputs, assumptions about plume dispersion geometry, and limitations in spatial resolution and aerial scan coverage. While quantitative uncertainty bounds are not calculated in this study, Bridger Photonics' system has undergone extensive validation through controlled release experiments 36 – 38 . 2.3 Cause Analysis Framework Causal analysis was employed to systematically identify and categorize the operational causes responsible for the observed methane emissions at AMI member companies. The process began with operator-led attribution based on internal site assessments. Each of the five participating operators investigated aerially detected emissions using their own protocols, drawing on site-level knowledge, operational data, and maintenance records to assign preliminary causal attributions. These initial attributions were subsequently reviewed by the research team, with additional information requested where necessary to ensure consistency and clarity across reporting. This process resulted in 1,669 emission events with operator-provided cause descriptions. The initial distribution of these events across broad operator-assigned categories, along with the data processing details, are presented in Supplementary Note S1. Operators used varied terminologies and levels of specificity in their descriptions (e.g., “dump valve stuck,” “valve left open,” “valve hung open”), resulting in 61 distinct cause labels. To address this heterogeneity, the research team developed a harmonized classification system by grouping similar descriptions into standardized causal categories. For instance, the examples above were consolidated under the category ‘incorrect valve position.’ The complete harmonized classification is presented in Table 1 . To focus the analysis on actionable emissions, two categories were excluded from further analysis. First, events attributed to ‘normal operations,’ defined as expected routine emissions from standard equipment functions (e.g., compressor rod packing vents operating as designed), were excluded. Episodic intermittent operations, including ‘well unloading,’ ‘maintenance,’ and ‘blowdown,’ are treated as distinct causal categories to analyze their specific emission patterns and are therefore not included under 'normal operations'. Additionally, events classified as ‘Unknown/third party,’ were also excluded (see Supplementary Fig. S2). The latter category included emissions for which no conclusive source could be identified on participating operator assets or emissions clearly originating from infrastructure owned by non-participating companies. After these exclusions, 1,011 emission events remained for detailed causal analysis. Their distribution across the final harmonized causal categories is shown in Supplementary Fig. S4. This dataset primarily comprises events for which operators conducted investigations and assigned initial causal attributions, which were subsequently categorized into standardized causal categories by the research team. These harmonized causal categories formed the analytical basis for subsequent quantitative assessments. In this study, 'frequency' refers to the observed event counts from the four quarterly aerial surveys, representing the relative prevalence of different causal categories within the measurement period, and not a statistically annualized recurrence rate. For the specific application of risk assessment tools, such as FMEA and FTA, these causal categories are treated as the primary failure modes for this analysis. Overall, this harmonized framework ensures consistency and supports a structured quantitative analysis of the emission causes. Table 1 Mapping of specific operator-reported cause descriptions into nine consolidated non-normal cause categories. This table consolidates 61 unique operator-provided cause descriptions from 1,669 emission events identified during the 2024 Appalachian Methane Initiative aerial surveys, into nine actionable causal categories, excluding normal operations and unknown/third-party sources, as detailed in Supplementary Note S1. Emission Type Detailed Subcategories Normal Operations Normal Operations, Venting Casings, Compressor Rod Packing, Combustion slip/Gas Engine Methane Slip, Pneumatic Controller Sensitivity, Flame Arrestor, Dehydrator Vents, Engine exhaust, Separator Vents, Pump, VDU, Pipeline, GPU, Generator. Drilling/Completions Completions, Drilling, Workover, Fracking, Horizontal Drilling, Well Ops & Completion. Mechanical Failures Packing Failures, O-ring Failures, Loose Fitting/Cracks, Venting from Internal Failure, Gasket, Corrosion, Blockage, Broken Part, Mechanical Leaks, Blockage, Regulator Failure, Pneumatic Controller Fault, GPU Box Malfunction, M&I leak. Incorrect Valve Position Dump Valve Hung Open, Dump Valve Cut, Incorrect Valve Position, Valve Left Open, Stuck Dump Valve. Abnormal Tank Emissions Thief Hatch / Tank Opening, Thief Hatch Failure, Tank Control Vents. Flare Unlit/Malfunction Unlit Flare, Burner, Flare Malfunction, Flare, Unlit Flame. Heater Unlit/Malfunction Heater Unlit, Heaters vents, Heater Malfunction. Maintenance Maintenance, Maintenance Vents, Flowline Repair. Blowdown Blowdowns. Well Unloading Well Unloading, Well Shutdown. Unknown/ Third Party Unknown, Unidentified, Third-Party Equipment 2.3.1 FMEA Analysis We conduct failure mode and effects analysis (FMEA) to systematically prioritize methane emission risk identified through the causal analysis. FMEA is a proactive method used to evaluate potential failure modes using three key parameters with individual scores: Severity (S), Occurrence (O), and Detection (D). All scores were normalized to a scale of 1 for the lowest risk to 10 for the highest risk to ensure comparability and interpretability across components. Severity (S) was derived from the average methane emission rate (kg/h) for each failure mode, representing its potential impact. Occurrence (O) was computed from the event count of each failure mode, reflecting its operational prevalence. For both Severity and Occurrence, the scores were normalized using a standard min-max scaling approach to achieve a 1–10 scale, as shown in Eq. 1: \(\:\:\:\:\:\:\:\:\:\:\:\:\:Score=1+\frac{\left(\alpha\:-{\alpha\:}_{min}\right)}{\left({\alpha\:}_{max}-{\alpha\:}_{min}\right)}(10-1)\) (Eq. 1) where \(\:Score\) is the normalized score between 1 and 10, \(\:\alpha\:\) is the observed value for the specific failure mode (i.e., the average emission rate for severity, or the event count for occurrence), \(\:{\alpha\:}_{min}\) is the minimum observed value for that parameter across all failure modes, \(\:{\alpha\:}_{max}\) is the maximum observed value for that parameter across all failure modes. Detection (D) represents the likelihood that an emission is captured or detected for control when it occurs. The scores were defined based on the average duration of the emission events for each failure mode. Given that shorter-duration events are less likely to be captured by snapshot-based aerial measurement campaigns, detection scores were inversely normalized. A reversed min-max scaling approach was used, assigning a score of 10 to the shortest average duration (most difficult to detect) and 1 to the longest average duration (easiest to detect), as shown in Eq. 2: \(\:\:\:\:\:\:\:\:\:\:\:\:\:{Score}^{*}=1+\frac{\left({\alpha\:}_{max}-\alpha\:\right)}{\left({\alpha\:}_{max}-{\alpha\:}_{min}\right)}(10-1)\) (Eq. 2) where Score* is the normalized detection score between 1 and 10, \(\:\alpha\:\) is the average duration for the specific failure mode, \(\:{\alpha\:}_{min}\) is the shortest average duration observed across all failure modes, and \(\:{\alpha\:}_{max}\) is the longest average duration observed. The composite Risk Priority Number (RPN) for each failure mode was then calculated by multiplying the three scores, as shown in Eq. 3. \(\:\:\:\:\:\:\:\:\:\:RPN=S\:\times\:O\:\times\:D\) (Eq. 3) The RPN provides a quantitative measure to rank and prioritize emission failure modes for targeted mitigation efforts. The detailed scoring inputs and results for this section are provided in Supplementary Table S2 and Supplementary Note S2. 2.3.2 FTA Analysis Fault Tree Analysis (FTA) was employed as a structured, top-down technique to trace the logical pathways leading to methane emissions and to quantitatively estimate their likelihood. This approach uses a logic-based framework composed of AND/OR gates to model how individual or combined failures can escalate into emission events. The analysis began by defining "Methane Emissions (per year)" as the top-level event in the fault tree. The top event was then decomposed through an OR gate into nine Level 1 intermediate events, corresponding to the major failure modes such as mechanical failures. Selected Level 1 failure modes were further broken down into more specific sub-system or root causes using intermediate OR or AND logic gates, based on the specific sub-causes identified in the operator causal analysis data and established engineering principles. This hierarchical decomposition continued until the basic events were reached. These basic events represent the initiating failures for which probability data can be assigned, either from empirical observations or literature-based estimates. For instance, "Abnormal Tank Emissions" was decomposed into three intermediate contributors: vapor containment failure, control system failure, and liquid handling/overpressure events. These were further linked to basic events, such as thief hatches that were left open, which were assigned failure probabilities. The complete structure of the fault tree is presented in Supplementary Fig. S6. The failure probabilities were assigned to each basic event based on a combination of literature-based estimates for common component failures and empirical data extracted from the causal analysis dataset, where available, as summarized in Supplementary Table S3. The fault tree was then quantified to estimate the annual probability of the top-level emission event and assess the probabilistic importance of various failure pathways and basic events. This quantification enabled the identification of the most critical failure branches and basic events contributing to the overall emission risk, thereby informing targeted mitigation strategies. 2.4 Statistical Analysis and Uncertainty Statistical analyses were performed using Python (version 3.13.3), with the SciPy stats module for hypothesis testing. Chi-square tests of independence were applied to assess statistically significant differences in the distribution of emission causal categories across both temporal and spatial dimensions. A standard significance threshold of p < 0.05 was used to determine whether the observed differences were unlikely to have occurred by chance. Uncertainty in the dataset arises primarily from two sources: measurement limitations that may under-detect low-level, intermittent, or short-duration emissions due to detection thresholds, and inconsistencies in cause classification across operators. While these factors may affect resolution, their impact is mitigated by the large basin-wide dataset and consistent follow-up corrections with the participating operators to refine and standardize the categorizations. 3. Results 3.1 Basin-Wide Distribution of Methane Emission Causal Categories A basin-wide assessment of methane emissions revealed distinct patterns across both the frequency of emission causal categories and their contributions to the total emission rates. The emission events attributed to specific operational causes in this study represented approximately 61% of all distinct emission events detected during aerial surveys and accounted for 53% of the total methane emissions (see Supplementary Fig. S2 and S3 for further details on the emission distributions). Figure 1 shows the percentage of total emission events attributed to each causal category and their average emission rates (kg/h), overlaid with their proportional contributions to total basin-wide emissions. Abnormal tank emissions emerged as the most prevalent and impactful causal category, accounting for 32.1% of all events with documented cause and contributing 46.5% of total methane emissions from the analyzed causal categories, with an average rate of 52.6 kg/h. These findings are consistent with those of prior studies that identified tanks, particularly those with faulty hatches, vapor recovery issues, or uncontrolled venting, as critical contributors to upstream methane emissions 15 , 39 , 40 . The dual risk posed by high frequency and moderate-to-high intensity highlights the importance of targeted interventions, such as improved thief hatch seals and automated monitoring systems, in mitigating tank emissions. Mechanical failures were the second most common emission causal category, accounting for 20.8% of the emission events, with an average emission rate of 22.1 kg/h, contributing 12.4% to the total event-attributable emissions. These emissions are associated with degraded components, such as loose fittings or packing failures in our dataset, and result in lower intensity yet persistent emissions. In contrast, well unloading events, although accounting for only 9.4% of events, stand out as the most emission-intensive event, with an average emission rate of 71.3 kg/h, contributing 16.4% to the total event-attributable emissions. These episodic events, typically associated with the clearing of liquid from wells and the restoration of gas flow, are of short duration but show disproportionately large emission volumes. Prior studies have highlighted the significance of such episodic release during liquid unloading, noting that they may be insufficiently characterized in existing research 29 , 41 , 42 . Incorrect valve positions and flare unlit/malfunctions represent 13.6% and 10.7% of the total events, respectively, and exhibit average emission rates of 21.8 and 32.2 kg/h, respectively. Their total event-attributable emission contributions were 7.7% and 9.5%, respectively, indicating that although these emission causal categories occur relatively often, their individual impact is moderate compared to other event types. Notably, maintenance activities contributed a relatively small proportion of emission events (5.5%) and total event-attributable emissions (2.6%), with an average emission rate of 16.5 kg/h. These findings suggest that although emissions during maintenance are operationally expected, their impact is limited in the Appalachian Basin. Blowdowns, drilling/completions, and heater unlit/malfunctions each represented less than 5% of the total emission counts and cumulative event-attributable methane emissions in the dataset. Notably, emissions from heater unlit/malfunctions are underrepresented because many events fall below the minimum detection threshold of aerial technology. A recent study on heaters reported a median emission rate of 0.28 kg/h, which is significantly lower than the minimum detection threshold of aerial systems 43 . However, in this dataset, heater unlit/malfunction events that exceeded the detection threshold exhibited an average emission rate of 33.6 kg/h. This suggests that while such events are rare and often missed by snapshot detection methods, they can still pose a significant emission risk when they occur. These findings reveal a clear divergence between high-frequency, lower-intensity causes and low-frequency, high-intensity ones, suggesting the need for a dual-pronged mitigation strategy. On one hand, frequent, moderate-impact emissions, exemplified by mechanical failures and incorrect valve positions, may benefit from broad-based monitoring and operational improvements. On the other hand, infrequent yet high-impact events, such as well unloading, necessitate more targeted mitigation protocols that may not necessarily benefit from or require survey-type monitoring. 3.2 Super-Emitters and Skewed Emission Distributions Previous studies have consistently demonstrated that methane emissions in oil and gas operations are highly skewed, with a small number of high-emission events, referred to as super-emitters, accounting for the majority of the total emissions 17 , 44 , 45 . In this study, super-emitters are defined as large emission events with rates equal to or exceeding 100 kg/h. While this phenomenon has been well characterized in terms of emission magnitudes, it remains unclear how super-emitter behavior translates across different emission causes. Specifically, we want to examine if a small number of causes disproportionately contribute to super-emitters. Figure 2 a presents the cumulative distribution of methane emission rates across all analyzed emission events, highlighting the concentration of emissions among a small subset of large events. As shown, only a few events, comprising 7.2% of the emission events, exceeded the defined super-emitter threshold of 100 kg/h. Despite their low frequency, these super-emitter events contributed to approximately 80% of the total event-attributable methane emissions, confirming the heavy-tailed distribution of emissions. To investigate the cause of these super-emitter events, we disaggregated the super-emitter subset by causal category. Thus, Fig. 2 b shows the frequency distribution of super-emitter events across these causes. We found that abnormal tank emissions accounted for 36% of all super-emitter events. This aligns with prior findings that tanks, particularly those with compromised thief hatches, are leading sources of large methane releases in upstream operations 46 . Well unloading was the second most common cause of super-emitter events (21%), followed by mechanical failures (19%), illustrating that episodic, short-duration events and equipment degradation can contribute to extreme emissions. Other categories, such as incorrect valve position and flare unlit/malfunctions, appeared less frequently (each < 10%), yet still contributed to at least one super-emitter event. This indicates that even less frequent or typically lower-emitting causes can occasionally escalate into high-magnitude releases. These findings demonstrate that a small subset of causal categories is responsible for a disproportionate share of the super-emitter events. Additionally, they highlight the importance of linking emissions measurement data to operational causes, enabling proactive, causal-informed intervention before escalation occurs. Lastly, the data suggests that large emissions may not necessarily result from unique or anomalous causes, but rather from the escalation of common operational issues, a finding explored further in the subsequent discussion. 3.3 Comparing Causal Drivers of Large vs. Small Methane Emitters Building on the super-emitter analysis from the previous section, this analysis examines how emission patterns of emission causes differ between small (< 100 kg/h) and large (≥ 100 kg/h) emitters. To quantify the tendency of certain causal categories to be associated with high-magnitude emissions, we introduce an Escalation Ratio (ER). In this context, 'escalation' refers not to the real-time development of a single event but to the statistical overrepresentation of emission causes among large emitters compared to small emitters. The ER for a specific causal category is defined as $$\:{ER}_{i}=\:\frac{{f}_{i,large}/{N}_{large}}{{f}_{i,smal}/{N}_{small}}\:$$ where \(\:{f}_{i,large}\) and \(\:{f}_{i,small}\:\) represent the number of large and small emitter events, respectively, associated with causal category i, and \(\:{N}_{large}\) and \(\:{N}_{small}\) denote the total number of large and small emitter events across all categories. This ratio captures the relative overrepresentation of emission causes among high-magnitude emission events. The emission rate contributions for each category are provided in Supplementary Fig. S7. Figure 3 shows that abnormal tank emissions dominated both small and large emitter groups, appearing in over 30% of all events. An ER of 1.1 suggests a consistent frequency across emission scales, indicating that tanks pose a persistent, system-wide emission risk rather than being escalation-prone. Despite this consistency, tank emissions dominated the total methane contribution across both categories, accounting for nearly 50% of the total emissions among large emitters and 36% among small emitters. By contrast, well unloading events exhibited the highest escalation ratio of 2.8, showing a dramatic increase in frequency from small to large emitters. It is important to note that for episodic events like well unloading, the instantaneous emission rate can vary widely, meaning snapshot-based classifications may reflect observation timing rather than event differences. Nevertheless, the high ER demonstrates that unloading operations are disproportionately represented among the highest-rate emission events and highlights the importance of managing both the duration and intensity of their peak emission phase. Heater unlit/malfunctions with an ER of 2.6 and blowdowns with an ER of 1.4 also show an elevated presence among large emitters, despite being relatively infrequent overall. These categories represent emission releases that, although not as prevalent, can rapidly escalate under unfavorable containment conditions. For example, while heater malfunctions account for only a small share of total event-attributable emissions (< 1% overall) because they are mostly below the detection threshold of aerial surveys, their high ER suggests that when detected, they are more likely to be a super-emitter event. This likely reflects a binary emission pattern, where malfunctions, such as flameouts, cause sudden gas blow-by and overwhelm combustion systems, resulting in high-rate emissions. This step-change dynamic helps explain their elevated ER, though further diagnostics are needed to confirm all causal factors. Conversely, common operational causes, such as mechanical failures and incorrect valve positions, showed lower escalation tendencies, with ERs of 0.9 and 0.8, respectively. These causes are more prevalent among small emitters and tend to produce lower-magnitude emissions. This is expected as routine mechanical failures from wear and tear are not expected to result in super-emitters. Other causal categories displayed inverse escalation patterns with ER < 1. These categories appear more frequently among small emitters, reflecting consistent but lower-magnitude emissions. Overall, these results emphasize that super-emitter events are not separate anomalies but traceable to specific operational causes with quantifiable escalation potential. The ER introduced here serves as a valuable prioritization metric, highlighting which categories are disproportionately associated with large-scale releases. In addition, not all emission causes escalate equally. Therefore, mitigation planning should reflect this heterogeneity. Emission mitigation frameworks should apply differentiated strategies: one focused on routine prevention and detection for persistent low-escalation categories, and another aimed at escalation control for causal categories prone to producing super-emitter outcomes. 3.4 FMEA and FTA Analysis This study presents the first application of FMEA to quantitative risks associated with different causal categories of measured methane emission events at a basin scale. The Risk Priority Number (RPN) was computed for each failure mode, as described in the Methods section, providing a composite score to guide mitigation prioritization. Figure 4 presents the resulting RPN values, alongside the individual scores for severity, occurrence, and detection (see Supplementary Table S2). Abnormal tank emissions emerged as the highest-priority failure mode with an RPN of 487, driven by a high severity score of 7.3, maximum occurrence score of 10, and detection score of 6.6, indicating moderate difficulty in detection. This finding reinforces the importance of tank emissions to basin-wide emissions and highlights their status as a high-risk, high-frequency, and difficult-to-detect emission source. Well unloading ranked as the next highest priority with an RPN of 295. Although it accounted for only 9.4% of emission events, it exhibited the highest average emission rate and was close to the maximum detection score of 10, indicating its high-impact, low-detectability profile. Flare unlit/malfunction ranked third with an RPN of 142, primarily driven by its high detection difficulty score of 8.5, alongside moderate severity and occurrence values. Mechanical failures and incorrect valve positions yielded mid-range RPNs of 90 and 79, respectively, driven primarily by lower severity scores of 2.78 and 2.73, respectively, despite relatively higher occurrence and detection difficulty scores. Although not the most individually consequential failure modes, their risk profiles are elevated by the combined influence of human and operational errors, coupled with limited detectability. Blowdowns, with an RPN of 85, was less frequent but exhibited maximum detection difficulty (10.0), emphasizing the challenges of capturing short-duration events. Lower-priority failure modes with RPN values ≤ 50 include heater unlit/malfunctions, maintenance, and drilling/completions. These categories were either infrequent, associated with lower emission rates, or characterized by longer durations that improve detection, suggesting that standard operational protocols may be sufficient for their management. Together, these FMEA results provide a quantitative, data-informed, and structured framework for prioritizing mitigation strategies beyond frequency or emission rates alone. By integrating three risk dimensions, FMEA provides a more robust risk profile and reorders priorities. For example, mechanical failures were the second most frequent event; they were de-prioritized while flare unlit/malfunction was ranked higher due to detection scores. This illustrates how FMEA highlights intermittent failure modes that might otherwise be overlooked. High-RPN failure modes, such as tank emissions and well unloading, warrant focused attention through enhanced automation, improved detection strategies, and predictive interventions. In contrast, moderate and low-risk failure modes may be incorporated into broader operational workflows. While this represents a simplified FMEA for interpretability, the approach demonstrates the value of integrating emissions data into structured causal analysis tools. This framework also lays the groundwork for advanced FMEA models. To quantitatively assess the likelihood of methane emissions arising from the identified operational and equipment failures, we conducted FTA. The FTA modeled the logical escalation pathways from the three most frequent failure categories: abnormal tank emissions, mechanical failures, and incorrect valve positions, to the top event, annual methane emission. The complete fault tree structure is provided in Supplementary Fig. S6, with the associated basic event probabilities listed in Supplementary Table S3. The FTA propagation yielded an estimated annual probability of 7.2% for methane emissions arising from these three failure categories. When disaggregated, abnormal tank emissions exhibited the highest individual probability at 2.7%, followed by mechanical failures at 2.5%, and incorrect valve position at 2.0%. Within the tank emissions branch, vapor containment failures contributed most significantly, with an annual probability of 1.4%, primarily driven by thief hatch or tank opening events (1.0%). Control system failures and liquid handling/overpressure incidents contributed 0.85% and 0.41%, respectively. For incorrect valve position, human error categorized as "valve left open", emerged as the dominant pathway, with an annual probability of 2.0%. This figure is over 100 times greater than the combined contribution of other valve faults within this failure category, highlighting the disproportionate role of operational issues. In the mechanical failure branch with a total annual probability of 2.5%, loose fittings and structural cracks represented the primary contributors. To assess the validity of the bottom-up FTA model, we compared model-derived annual probabilities with empirical event frequencies from the survey data. For tank-related events, the FTA-predicted annual probability of 2.7% was in close agreement with the empirically observed annual event frequency of 2.0%. In stark contrast, a similar analysis for flare malfunctions (detailed in Supplementary Note S3) yielded a theoretical annual probability of ~ 0.2% based on literature reliability data, nearly 50 times lower than the empirically observed annual frequency of 7.6% in our dataset. This strong alignment for tanks validates the model structure, whereas the discrepancy for flares highlights the critical importance of using basin-specific, measurement-informed data over generic failure rates to accurately assess real-world emission risks. Overall, these FTA findings illustrate how emissions result from a convergence of equipment degradation, procedural lapses, and human error, highlighting critical nodes that drive emission risk and informing targeted mitigation strategies. 3.5 Quarterly Variation in the Distribution of Emission Causes Understanding the quarterly distribution of emission causes is critical for identifying operational or seasonal patterns that can inform predictive maintenance and targeted mitigation strategies. Figure 5 presents the quarterly distribution of methane emission events, expressed as a percentage of total emission events. The associated p-values from a chi-square test are included to indicate whether the observed differences are statistically significant across quarters. The raw unnormalized event counts are provided in Supplementary Fig. S4. Several emission causes, including blowdown, flare unit/malfunction, incorrect valve position, and tank emissions, did not exhibit statistically significant temporal differences (p > 0.05). These patterns suggest a relatively constant rate of occurrence throughout the year, possibly reflecting equipment types or causal mechanisms that are insensitive to seasonal influences. For instance, incorrect valve positions remained relatively consistent across all quarters, while tank emissions showed minor fluctuations; highest in Q2, lowest in Q3, but without statistical significance (p = 0.33). This shows the persistent nature of these causal categories, indicating that they are likely tied to regular operational workflows rather than external drivers. By contrast, three causal categories: maintenance events, mechanical failures, and well unloading exhibited statistically significant temporal variation. Mechanical failures peaked in Q2 with approximately 32% of the category’s total, showing significant variation across quarters (p = 0.04). Notably, maintenance events showed a sharp spike in Q3, accounting for 30% of all events in this category (p = 0.00008). This concentration is notable, given the shorter 14-day survey duration in Q3 (Supplementary Table S1 ), making an increase in planned maintenance activities unlikely. The statistically significant peak suggests a genuine, yet unexplained, cluster of maintenance-related emissions during this period. Well unloading showed a sharp increase in Q4, with moderate but overall significant variability across quarters (p = 0.0075). Although unloading events are often assumed to typically occur at a relatively uniform rate, this Q4 increase may relate to year-end production optimization or target-based management efforts, a pattern further examined in the next section. These patterns suggest that while many emissions causal categories are temporally invariant, others may be influenced by seasonal operational cycles or maintenance scheduling. Overall, the statistically significant trends emphasize the non-steady state nature of some emission causes and support a differentiated approach to mitigation. For instance, time-sensitive patterns such as seasonal maintenance peaks and Q4 unloading surges suggest a need for time-stratified intervention strategies, whereas persistent emission causes may be best addressed through routine inspection protocols and systematic detection efforts. 3.6 Inter-Operator Variability of Emission Causes Understanding differences in the distribution of emission causes among operators is essential for identifying operator-specific patterns that may reflect variations in facility design, maintenance procedures, or operational practices. Figure 6 presents the distribution of emission causes across four anonymized operators as companies A through D in the Appalachian basin. The data are expressed as the percentage of total emission events that are attributable to each operator, for each causal category. Statistical analysis confirms that the distribution of events for all major causal categories exhibits statistically significant differences across operators (all p < 0.05). This finding aligns with previous research suggesting that comprehensive sampling across operators may be more effective for reducing basin-wide uncertainty than repeated temporal measurements 47 . The distribution of emission causes reveals substantial heterogeneity in emissions behavior among the companies. Flare unlit/malfunction emissions were heavily concentrated in Company A, which accounted for nearly 49% of all such events, a proportion significantly different from other operators (p < 0.00001). Abnormal tank emissions showed a notable skew, with Companies B and C collectively responsible for over 80% of all tank emissions. Mechanical failures were also unevenly distributed, with Company B contributing nearly 37% of these failures, potentially reflecting differences in equipment condition or operational stress. Well unloading events were most prevalent in Company D, which reported approximately 37% of all such events (p < 0.00001). This aligns with the spike in Q4 discussed earlier and may indicate company-specific production strategies. While incorrect valve position was more evenly distributed, it still exhibited a statistically significant difference (p = 0.005), with Company D reporting elevated levels (about 25%). Finally, maintenance-related emissions were more frequently reported by Company C, and although blowdown events were less frequent overall, they were notably higher in Company A (11%) with significant variation across companies (p = 0.0012). These findings highlight the operator-specific nature of methane emission causes. The non-uniform distribution of emission causes across operators corroborates findings that emissions are influenced by differences in factors like infrastructure age, automation levels, and maintenance practices 15 , 30 , 48 , 49 . This reinforces the need for tailored mitigation strategies. For instance, an operator predominantly experiencing flare-related events might prioritize investments in flare reliability and maintenance, whereas another operator with a high incidence of tank emissions would focus on tank integrity programs and vapor recovery systems. Incorporating causal analysis-based causal insights into inventory models can improve both granularity and accuracy, enabling more targeted and process-informed methane mitigation efforts at both company and regional levels. 4. Discussion This study offers one of the first basin-wide applications of causal analysis techniques to methane emissions in upstream oil and gas operations. By linking aerial detection data to classified emission causes, we introduce a process-based approach that shifts the analytical focus from quantifying emission magnitudes to diagnosing emission causation. This method demonstrates that methane emissions are largely traceable to identifiable, recurring causes, many of which exhibit clear temporal, operator-specific, and escalation-related trends, highlighting the need for process-informed mitigation. Several key insights emerge from this work, each carrying significant implications for understanding emission drivers, their variability, and process-based risk mitigation planning. First, across both the full dataset and the super-emitter subset, tank emissions consistently emerged as the most frequent and highest-contributing causal category. Importantly, our data reveal that super-emitter events are not driven by fundamentally different causes. Instead, common causes, especially those related to abnormal tank emissions, well unloading, and mechanical degradation, can escalate into high-magnitude emissions under specific operational conditions. This consistency suggests that mitigation efforts should not only focus on the type of emissions, but also on understanding and controlling the operational conditions that cause a routine emission to escalate into high-emission rate events. Consequently, this understanding supports a shift in mitigation strategies away from purely anomaly detection or qualitative segregation towards risk-based monitoring and prevention focused on known, high-escalation emission causes. Second, statistically significant quarterly and inter-operator variation was observed in the distribution of emission causes. For example, seasonal peaks in maintenance-related events highlight the influence of time-dependent operational practices. Similarly, strong divergence in the prevalence of these causes across operators demonstrates that emissions are not uniformly distributed. These patterns contradict a “one-size-fits-all” approach and emphasize the importance of proactive adjustments and tailored strategies aligned with seasonal operational variations and operator-specific differences. Third, a comprehensive mitigation strategy must also prioritize sources based on cumulative emissions, considering rate, duration, and frequency 50 . Our analytical framework supports this balanced approach. The structured application of FMEA enabled prioritization of emission causes beyond just occurrence. By integrating frequency, severity, and detection difficulty, FMEA highlights that high-risk, low-frequency failure modes may warrant rapid intervention strategies, such as remote shutoff systems and predictive alerts, while persistent but lower-intensity ones may require routine monitoring and maintenance protocols. To complement this FTA integrates annual emissions probabilities, offering a time-based risk perspective. Together, FMEA and FTA provide a practical, process-informed pathway for identifying how simple malfunctions may compound into significant emission events and mitigation levers, thereby facilitating more cost-effective and targeted mitigation. A novel contribution of this study is the introduction of the Escalation Ratio (ER), a metric quantifying how much more frequently an emission cause occurs among super-emitters compared to small emitters. Emissions categories such as well unloading exhibited high ER values, highlighting their disproportionate presence in high-emission cases. The ER metric offers a new layer of insight into causal analysis by identifying causal categories with higher escalation risks and lays the groundwork for future development of predictive scoring systems that can inform the design of proactive interventions. The collective findings of this research support a paradigm shift in methane mitigation: from a predominantly reactive reliance on leak detection to an integrated approach emphasizing proactive cause prevention. By embedding causal insights into inventory development, LDAR scheduling, and broader operational risk management frameworks, this work lays the groundwork for more effective, operationally grounded, and risk-informed emissions management strategies. Our results strongly suggest that many super-emitter events arise from the escalation of common operational issues, suggesting a path toward more proactive management by identifying the specific causal categories most prone to escalation. Furthermore, the FMEA and FTA models presented in this work are intentionally simplified, primarily serving to illustrate methodological integration and provide a foundational framework for integrating causal analysis with emissions monitoring. Future research can build on these findings. Key priorities include better standardization of causal classification protocols across operators to improve cross-comparability and the development of more sophisticated probabilistic FTA models that can dynamically simulate failure propagation pathways. Furthermore, predictive analytics, such as Poisson regression for failure frequency and machine learning for causal classification, offer promising pathways for real-time risk scoring and automated emissions attribution. When integrated with maintenance systems, these tools can transform methane mitigation from a reactive discipline into a proactive, process-embedded component of methane management in the oil and gas sector. Declarations Author contributions Abdulmuiz A. Adekomi: Methodology, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. Shuting Yang: Methodology, Validation, Conceptual framing, Writing – review & editing. Shannon Stokes: Methodology, Formal analysis, Conceptual framing, Writing – review. Arvind P. Ravikumar: Conceptualization, Writing – review & editing, Supervision, Project administration, Funding acquisition. Notes A.P.R. is currently a member of the Gas Pipeline Advisory Committee of the US Department of Transportation; in this role, he is a Special Government Employee. A.P.R. has current research support from the US Department of Energy, Environmental Defense Fund, and sponsors of the Energy Emissions Modeling and Data Lab (EEMDL). Acknowledgements This work was funded in part by the U.S. Department of Energy under Grant No. DE-FE0032311 and the Appalachian Methane Initiative. Data Availability The datasets and codes generated and/or analyzed in this study are publicly available. Additional data supporting the findings, including emission event records and causal classification outputs, are provided in the Supplementary Information. References Masson-Delmotte V, Zhai P, Pirani A et al (2021) EdsV. Masson-Delmotte, P. Zhai, A. Pirani, Eds. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change . https://www.ipcc.ch/report/ar6/wg1/ Sobanaa M, Prathiviraj R, Selvin J, Prathaban M (2023) A comprehensive review on methane’s dual role: effects in climate change and potential as a carbon–neutral energy source. Environ Sci Pollut Res 31:10379–10394 Tibrewal K et al (2024) Assessment of methane emissions from oil, gas and coal sectors across inventories and atmospheric inversions. Commun Earth Environ 5:26 Control methane to (2021) slow global warming — fast. Nat Editor 2021 August 25 Adekomi AA, Carrasco-Jaim OA, Weerasooriya UP, Okuno R (2024) Enhanced CO 2 Capture-Desorption by Surface-Active Amine. Energy Fuels 38:14435–14448 Shindell D, Kuylenstierna JCI, Michalopoulou E, Höglund-Isaksson L, Zhang Y, Seltzer K et al (2021) United Nations Environment Programme (UNEP) and Climate and Clean Air Coalition (CCAC). Global Methane Assessment. 10.5281/zenodo.4777096 Alvarez RA et al (2018) Assessment of methane emissions from the U.S. oil and gas supply chain. Science 361:186–188 U.S natural gas production remained flat in 2024 - U.S. Energy Information Administration (EIA). Goetz JD et al (2015) Atmospheric Emission Characterization of Marcellus Shale Natural Gas Development Sites. Environ Sci Technol 49:7012–7020 Allen DT et al (2013) Measurements of methane emissions at natural gas production sites in the United States. Proc. Natl. Acad. Sci. 110, 17768–17773 Peischl J et al (2015) Quantifying atmospheric methane emissions from the Haynesville, Fayetteville, and northeastern Marcellus shale gas production regions. J Geophys Res Atmos 120:2119–2139 Omara M et al (2022) Methane emissions from US low production oil and natural gas well sites. Nat Commun 13:2085 Omara M et al (2024) Constructing a measurement-based spatially explicit inventory of US oil and gas methane emissions (2021). Earth Syst Sci Data 16:3973–3991 Williams JP et al (2025) Small emission sources in aggregate disproportionately account for a large majority of total methane emissions from the US oil and gas sector. Atmospheric Chem Phys 25:1513–1532 Omara M et al (2016) Methane Emissions from Conventional and Unconventional Natural Gas Production Sites in the Marcellus Shale Basin. Environ Sci Technol 50:2099–2107 Sherwin ED et al (2024) US oil and gas system emissions from nearly one million aerial site measurements. Nature 627:328–334 Brandt AR et al (2014) Methane Leaks from North American Natural Gas Systems. Science 343:733–735 Thorpe MJ et al (2024) Deployment-invariant probability of detection characterization for aerial LiDAR methane detection. Remote Sens Environ 315:114435 Wang JL et al (2022) Multiscale Methane Measurements at Oil and Gas Facilities Reveal Necessary Frameworks for Improved Emissions Accounting. Environ Sci Technol 56:14743–14752 Daniels W et al (2023) Towards multi-scale measurement-informed methane inventories: reconciling bottom-up site-level inventories with top-down measurements using continuous monitoring systems. Preprint at. https://doi.org/10.26434/chemrxiv-2023-jp5nt-v2 Zavala-Araiza D et al (2015) Reconciling divergent estimates of oil and gas methane emissions. Proc. Natl. Acad. Sci. 112, 15597–15602 Varon DJ et al (2019) Satellite Discovery of Anomalously Large Methane Point Sources From Oil/Gas Production. Geophys Res Lett 46:13507–13516 Allen DT (2014) Methane emissions from natural gas production and use: reconciling bottom-up and top-down measurements. Curr Opin Chem Eng 5:78–83 Schwietzke S et al (2016) Upward revision of global fossil fuel methane emissions based on isotope database. Nature 538:88–91 Rutherford JS et al (2021) Closing the methane gap in US oil and natural gas production emissions inventories. Nat Commun 12:4715 Eleweke I et al (2025) The Convergence of Machine Learning and Power Systems: Enhancing Efficiency, Optimization, and Sustainability in Energy Management. 8 Varon DJ et al (2023) Continuous weekly monitoring of methane emissions from the Permian Basin by inversion of TROPOMI satellite observations. Atmospheric Chem Phys 23:7503–7520 Schwietzke S et al (2017) Improved Mechanistic Understanding of Natural Gas Methane Emissions from Spatially Resolved Aircraft Measurements. Environ Sci Technol 51:7286–7294 Zaimes GG et al (2019) Characterizing Regional Methane Emissions from Natural Gas Liquid Unloading. Environ Sci Technol 53:4619–4629 Zavala-Araiza D et al (2017) Super-emitters in natural gas infrastructure are caused by abnormal process conditions. Nat Commun 8:14012 Latino MA, Latino RJ, Latino KC (2020) Root Cause Analysis: Improving Performance for Bottom-Line Results. CRC Press / Taylor & Francis Group, Boca Raton, FL Mascia A et al (2020) A failure mode and effect analysis (FMEA)-based approach for risk assessment of scientific processes in non-regulated research laboratories. Accredit Qual Assur 25:311–321 Ruijters E, Stoelinga M (2015) Fault tree analysis: A survey of the state-of-the-art in modeling, analysis and tools. Comput Sci Rev 15–16:29–62 Li X, Liu W, Zhou N, Yuan X (2025) Probability Analysis of Hazardous Chemicals Storage Tank Leakage Accident Based on Neural Network and Fuzzy Dynamic Fault Tree. Appl Sci 15:3504 Sharafutdinov E (2024) Spatiotemporal Variation in Anthropogenic Methane Emissions in the Appalachian Basin. UT Electron Theses Diss Johnson MR, Tyner DR, Szekeres AJ (2021) Blinded evaluation of airborne methane source detection using Bridger Photonics LiDAR. Remote Sens Environ 259:112418 Bell C et al (2022) Single-blind determination of methane detection limits and quantification accuracy using aircraft-based LiDAR. Elem Sci Anthr 10:00080 Conrad BM, Tyner DR, Johnson MR (2023) Robust probabilities of detection and quantification uncertainty for aerial methane detection: Examples for three airborne technologies. Remote Sens Environ 288:113499 Johnson MR, Tyner DR, Conrad BM (2023) Origins of Oil and Gas Sector Methane Emissions: On-Site Investigations of Aerial Measured Sources. Environ Sci Technol 57:2484–2494 Caulton DR et al (2023) Abnormal tank emissions in the Permian Basin identified using ethane to methane ratios. Elem Sci Anthr 11 Allen DT et al (2015) Methane Emissions from Process Equipment at Natural Gas Production Sites in the United States: Pneumatic Controllers. Environ Sci Technol 49:633–640 Vaughn TL et al (2018) Temporal variability largely explains top-down/bottom-up difference in methane emission estimates from a natural gas production region. Proc. Natl. Acad. Sci. 115, 11712–11717 Festa-Bianchet SA, Mohammadikharkeshi M, Tyner DR, Johnson MR (2024) Catalytic Heaters at Oil and Gas Sites May be a Significant yet Overlooked Seasonal Source of Methane Emissions. Environ Sci Technol Lett 11:948–953 Lyon DR et al (2016) Aerial Surveys of Elevated Hydrocarbon Emissions from Oil and Gas Production Sites. Environ Sci Technol 50:4877–4886 Zavala-Araiza D et al (2015) Toward a Functional Definition of Methane Super-Emitters: Application to Natural Gas Production Sites. Environ Sci Technol 49:8167–8174 Festa-Bianchet SA, Milani ZR, Johnson MR (2023) Methane venting from uncontrolled production storage tanks at conventional oil wells—Temporal variability, root causes, and implications for measurement. Elem Sci Anth 11:00053 Chen Y et al (2024) Reconciling ultra-emitter detections from two aerial hyperspectral imaging surveys in the Permian Basin. Preprint at https://doi.org/10.31223/X5G68V Roscioli JR et al (2015) Measurements of methane emissions from natural gas gathering facilities and processing plants: measurement methods. Atmospheric Meas Tech 8:2017–2035 Fox TA, Barchyn TE, Risk D, Ravikumar AP, Hugenholtz CH (2019) A review of close-range and screening technologies for mitigating fugitive methane emissions in upstream oil and gas. Environ Res Lett 14:053002 Cardoso-Saldaña FJ (2023) Tiered Leak Detection and Repair Programs at Simulated Oil and Gas Production Facilities: Increasing Emission Reduction by Targeting High-Emitting Sources. Environ Sci Technol 57:7382–7390 Additional Declarations Yes there is potential Competing Interest. A.P.R. is currently a member of the Gas Pipeline Advisory Committee of the US Department of Transportation; in this role, he is a Special Government Employee. A.P.R. has current research support from the US Department of Energy, Environmental Defense Fund, and sponsors of the Energy Emissions Modeling and Data Lab (EEMDL). Supplementary Files AdekomiMethaneCausalAnalysisSI20250810.docx Adekomi_Methane_Causal_Analysis_SI_20250810 Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7340966","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":509298453,"identity":"8b1948ee-7242-4cd5-a280-0390eb020647","order_by":0,"name":"Abdulmuiz Adekomi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA70lEQVRIiWNgGAWjYFACxgYgwcwgASQPfICLshGp5eAMBgNitIABRAszDzFadGckN39gqLGWk2zvTjxs2/ZHzuD48QcMH8oO49RidiOxwYDhWLqxNM/ZDYdz2wyMJXtyDBhnnMOvJYGB7XDiPIlcsJbEfoYcBmbeNvxaDjD8O1wP1mLZZlDfxv/8AfNf/FoaGxjbDidIg7Qwthkk8EskGDAz4tNy5mEzQ2JfuuHMnrMbDvacMzacOeONAZCRjlvL8fTHHz58s5aXON67+cOPMjl5g/PpDx/8KLPGqQUMEtAFDuBXPwpGwSgYBaOAEAAAknNarbs/O0wAAAAASUVORK5CYII=","orcid":"","institution":"University of Texas at Austin","correspondingAuthor":true,"prefix":"","firstName":"Abdulmuiz","middleName":"","lastName":"Adekomi","suffix":""},{"id":509298454,"identity":"b685a553-11e0-4b91-a6ad-5245f37b7356","order_by":1,"name":"Shuting Yang","email":"","orcid":"","institution":"Center for Energy and Environmental Systems Analysis (CEESA), The University of Texas at Austin.","correspondingAuthor":false,"prefix":"","firstName":"Shuting","middleName":"","lastName":"Yang","suffix":""},{"id":509298455,"identity":"5a3c1ee5-379e-4574-b6e8-4672c2b514fb","order_by":2,"name":"Shannon Stokes","email":"","orcid":"","institution":"Center for Energy and Environmental Systems Analysis (CEESA), The University of Texas at Austin.","correspondingAuthor":false,"prefix":"","firstName":"Shannon","middleName":"","lastName":"Stokes","suffix":""},{"id":509298456,"identity":"7190d7f6-940f-4f27-8b6a-28e1aa386f17","order_by":3,"name":"Arvind Ravikumar","email":"","orcid":"https://orcid.org/0000-0001-8385-6573","institution":"The University of Texas at Austin","correspondingAuthor":false,"prefix":"","firstName":"Arvind","middleName":"","lastName":"Ravikumar","suffix":""}],"badges":[],"createdAt":"2025-08-10 21:50:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7340966/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7340966/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90552508,"identity":"1fcde6aa-ccef-4c1c-af10-07f458e22d49","added_by":"auto","created_at":"2025-09-04 03:29:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":146134,"visible":true,"origin":"","legend":"\u003cp\u003eBasin-wide assessment of methane emission causal categories. \u003cstrong\u003e(a)\u003c/strong\u003e Distribution of event frequency, showing the percentage of total events attributed to each category. \u003cstrong\u003e(b)\u003c/strong\u003e Dual-axis plot showing the average emission rate in kg/h (purple bars, left y-axis) for each category, alongside the percentage contribution of each category to the total emissions (orange line with markers, right y-axis).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/552f86605c0d35b9ad549abd.png"},{"id":90552663,"identity":"5831931b-ee5a-4f85-a109-0a0ff599a419","added_by":"auto","created_at":"2025-09-04 03:37:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":105999,"visible":true,"origin":"","legend":"\u003cp\u003eSkewed distribution of methane causes and super-emitter contributions \u003cstrong\u003e(a)\u003c/strong\u003e Cumulative frequency distribution of methane emissions by emission rate across all detected events in the basin-wide dataset. Each point represents an individual emission event colored by the causal category. The vertical red line marks the super-emitter threshold at 100 kg/h, while the horizontal blue line indicates the point at which 80% of the total emissions are accounted for. Only 7.2% of events exceeded the super-emitter threshold, yet they contributed to the majority of event-attributable emissions, highlighting the heavy-tailed nature of events and the importance of targeting high-emission causes. \u003cstrong\u003e(b)\u003c/strong\u003e Count/frequency distribution of super-emitter events (≥100 kg/h) by causal category, showing the disproportionate contribution of a few causal categories to extreme emission events.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/7636566b4c71ff6729a82239.png"},{"id":90552665,"identity":"b7bc54dc-c800-4248-930b-6f3d631aa90d","added_by":"auto","created_at":"2025-09-04 03:37:29","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":101727,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of methane emission causes by emitter size. Bar charts comparing the distribution of emission cause frequencies between small (\u0026lt;100 kg/h) and large (≥100 kg/h) methane emitters. The escalation ratio (ER) annotated above each category indicates how much more frequently that category occurs among large emitters compared to small emitters. The corresponding percentage contribution to total emissions mass by causal category for the small and large emitter events is presented in Supplementary Fig. S7.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/ea57e434f4e2b37d4eeeeda9.png"},{"id":90553208,"identity":"61234041-3025-462c-9667-34ac69525db2","added_by":"auto","created_at":"2025-09-04 03:45:29","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":96959,"visible":true,"origin":"","legend":"\u003cp\u003eBasin-wide FMEA ranking of methane emission causal categories. The main bar plot shows the Risk Priority Number (RPN), sorted from highest to lowest. Inset plots show the corresponding individual scores for Severity (S), Occurrence (O), and Detection (D), scaled from 1 to 10. RPN = S × O × D. Higher RPN values indicate greater overall risk prioritization. A detailed breakdown of the scoring inputs and normalization factors for S, O, and D is provided in Supplementary Note S2.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/8d1534f927750d2a3407739a.png"},{"id":90552510,"identity":"29c79c53-520d-42c5-b253-2888350cf80b","added_by":"auto","created_at":"2025-09-04 03:29:29","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":105916,"visible":true,"origin":"","legend":"\u003cp\u003eA quarter-by-quarter comparison of emission causes (with at least one event in each quarter), with p-values indicating whether the distribution for each category significantly differs statistically. Bars on the left show p-values above the threshold, suggesting no significant variation across quarters, whereas bars on the right display p-values below the threshold, indicating meaningful differences. The 'heater unlit/malfunction' and 'drilling/completions' categories were excluded due to insufficient emission events for meaningful statistical analysis.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/a071b88b95e020437bb278f8.png"},{"id":90552516,"identity":"def0c33a-df7a-4075-9bdf-d9a9bce30209","added_by":"auto","created_at":"2025-09-04 03:29:29","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":96417,"visible":true,"origin":"","legend":"\u003cp\u003eComparative breakdown of emission causes across four anonymized companies (A, B, C, and D). Each bar reflects the percentage of total emission events attributed to a specific cause for each company. The variation in bar heights indicates how certain causes dominate in some companies more than others. P-values from Chi-squared tests are shown above each category. A fifth participating operator was excluded from this inter-operator comparison due to an insufficient number of attributable events for a meaningful statistical analysis.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/3c029fa31e53865624804544.png"},{"id":90553997,"identity":"6ed6f811-281b-467e-9299-e08ad708098f","added_by":"auto","created_at":"2025-09-04 03:53:33","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1448897,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/044425fc-53e5-45f2-9a96-2111c8413b72.pdf"},{"id":90552514,"identity":"5dfe9202-ff60-436e-bc1e-9f108398730c","added_by":"auto","created_at":"2025-09-04 03:29:29","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":751382,"visible":true,"origin":"","legend":"Adekomi_Methane_Causal_Analysis_SI_20250810","description":"","filename":"AdekomiMethaneCausalAnalysisSI20250810.docx","url":"https://assets-eu.researchsquare.com/files/rs-7340966/v1/ac5e112a6739d8809c889b64.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nA.P.R. is currently a member of the Gas Pipeline Advisory Committee of the US Department of Transportation; in this role, he is a Special Government Employee. A.P.R. has current research support from the US Department of Energy, Environmental Defense Fund, and sponsors of the Energy Emissions Modeling and Data Lab (EEMDL).","formattedTitle":"Why Do Methane Emissions Occur? Towards a Predictive Framework for Risk-Targeted Mitigation in Oil and Gas Operations","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eMethane (CH₄) is a potent greenhouse gas with a global warming potential (GWP) 84\u0026ndash;87 times greater than that of carbon dioxide (CO₂) over a 20-year period\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Methane alone accounts for nearly 30% of the global temperature rise since the pre-industrial era, making its mitigation crucial for addressing climate change\u003csup\u003e\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Consequently, reducing methane emissions has become a key objective of global initiatives, such as the Global Methane Pledge, which targets a 30% reduction in methane emissions by 2030\u003csup\u003e6\u003c/sup\u003e. The oil and gas (O\u0026amp;G) industry represents one of the largest anthropogenic sources of methane, with emissions spanning the entire supply chain\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe Appalachian Basin, encompassing major production areas in Pennsylvania, Ohio, and West Virginia, is the largest natural gas-producing region in the United States, accounting for over 31% of the total U.S. dry natural gas production in 2024\u003csup\u003e8\u003c/sup\u003e. Given its significance, the basin has been the focus of several recent methane measurement campaigns. These studies have employed a range of top-down and bottom-up techniques, including aerial LiDAR, satellite retrievals, and on-site inspections, to quantify emission patterns across diverse facility types. While earlier research often focused on national-scale assessments, more recent measurement-informed analyses in Appalachia indicate that national inventories may not fully capture regional emissions, particularly reflecting an underrepresentation of episodic, high-emitting sources\u003csup\u003e\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Although this basin generally exhibits lower methane intensity compared to oil-dominated regions like the Permian, its emission profile is nonetheless complex and often characterized by strong site-level heterogeneity\u003csup\u003e\u003cspan additionalcitationids=\"CR13 CR14 CR15\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. However, these efforts have largely prioritized emission quantification, with limited emphasis on systematically uncovering the operational mechanisms underlying the observed emissions. This gap in causal understanding, particularly in a basin with such heterogeneous emission behavior, constrains the development of effective, cause-specific mitigation strategies.\u003c/p\u003e\u003cp\u003eAddressing this knowledge gap requires a fundamental shift from quantifying how much methane is emitted to understanding why these emissions occur. Traditional quantification approaches, both bottom-up and top-down, are inadequate for diagnosing emission drivers. Top-down techniques, such as satellite and aerial observations, provide regional-scale insights but typically lack the resolution and operational information needed to attribute emissions to specific causes\u003csup\u003e\u003cspan additionalcitationids=\"CR18 CR19 CR20 CR21\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Conversely, bottom-up methods rely on emission factors and equipment counts that assume steady-state conditions, thereby failing to capture episodic, abnormal, or high-emission events\u003csup\u003e\u003cspan additionalcitationids=\"CR24 CR25 CR26\" citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. Several studies have highlighted the significance of these overlooked sources. For instance, a study in the Fayetteville Shale attributed up to one-third of midday methane emissions detected via aircraft measurement to episodic activities like liquid unloading\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Similarly, routine unloading has been shown to produce highly skewed, heavy-tailed distributions owing to operational variability\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. High-emission events have also been linked to specific operational failures, such as compromised thief hatches, malfunctioning controllers, and flare malfunctions\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. These findings collectively highlight the limitations of conventional inventories in capturing the full spectrum and operational origins of emissions. Without a clear understanding of causation, most existing mitigation strategies remain reliant on generalized assumptions and tend to focus on reactive detection and response rather than proactive prevention guided by causal analysis. Bridging this gap requires methodologies capable of tracing emissions back to specific operational or equipment-related causes. For example, a large emission from a tank can be the result of an open thief hatch or point to upstream process issues unrelated to the tank, such as a stuck dump valve. Identifying these operational or failure pathways is important for accurate emissions accounting and designing targeted interventions.\u003c/p\u003e\u003cp\u003eTo address these limitations, this study introduces a causal analysis framework for methane emissions management. This approach provides a structured method for investigating and categorizing the operational mechanisms that drive methane emissions, facilitating the development of proactive mitigation strategies. Causal analysis techniques have been extensively utilized across diverse industries, such as industrial safety, engineering, aerospace, manufacturing, and healthcare. These methodologies provide a structured approach for diagnosing the underlying causal mechanisms and implementing corrective measures to prevent recurrence\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Specifically, this study integrates two established techniques from root cause analysis (RCA) and risk assessment: Failure Mode and Effects Analysis (FMEA), a semi-quantitative approach that evaluates and ranks potential failure modes by assessing their severity, likelihood, and detection\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e; and Fault Tree Analysis (FTA), a logic-based hierarchical model that identifies critical failure events and the conditional pathways through which they lead to emission incidents\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIn this study, we identify and analyze the causes of methane emission events detected through aerial measurement campaigns at oil and gas facilities in the Appalachian Basin. We combine measurement data with extensive operator engagement to identify and track underlying causes for observed emissions. We found that abnormal tank emissions is the most common cause of emissions. More importantly, super-emitter events are not caused by unique causal categories but rather represent escalations of common causes under specific operational conditions. The analysis also reveals significant quarterly and inter-operator variability in these causal categories and introduces a novel Escalation Ratio metric to quantify the likelihood of each category resulting in high-emission outcomes. This work represents one of the first basin-wide applications of causal analysis to methane emissions, offering a structured framework for shifting methane management beyond emission quantification toward causal understanding and prevention. These findings support the development of targeted, cause-specific mitigation strategies and contribute to the advancement of process-informed methane models. Ultimately, this approach provides practical guidance for designing proactive cause\u0026ndash;driven methane mitigation strategies in the oil and gas sector, paving the way for predictive methane management.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Data Sources and Study Region\u003c/h2\u003e\u003cp\u003eMethane emissions data analyzed in this study were obtained as part of the 2024 Appalachian Methane Initiative (AMI), a basin-wide aerial measurement campaign focused on characterizing source-level emissions across the Appalachian Basin\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. The study region spans approximately 54,000 square kilometers, covering the principal oil and gas production areas in Pennsylvania, Ohio, and West Virginia. This region includes a wide diversity of site configurations, equipment types, and operational practices, providing a representative setting for basin-wide emission analyses. Between April 2024 and December 2024, approximately 2,500 oil and gas production sites were selected for this analysis and surveyed across four quarterly aerial campaigns, comprising facilities operated by five major upstream companies participating in the AMI initiative. The detailed survey period is provided in Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. The survey timing was coordinated based on the weather conditions, flight logistics, and operational schedules. Each campaign generated an instantaneous snapshot of methane emissions at the time of overflight. The measurement dataset supports high-confidence attribution of emissions to specific equipment, forming the foundation for subsequent causal analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Bridger Photonics Gas Mapping LIDAR Technology\u003c/h2\u003e\u003cp\u003eBridger Photonics\u0026rsquo; aircraft-mounted Gas Mapping LiDAR employs frequency-modulated continuous-wave (FMCW) laser technology, which sweeps perpendicular to the flight path of the aircraft. The system measures path-integrated methane concentrations in ppm-m. These concentrations were used to estimate methane emission rates based on wind speed and direction data obtained from nearby meteorological stations. The system overlays methane plumes on high-resolution aerial imagery to provide spatial context. Bridger's technology typically surveys dozens of sites each day, with the number of overflights per site varying based on the physical footprint. Larger sites often require multiple passes, resulting in some equipment being scanned more than once due to overlapping scan swaths. Consequently, a single emission source may produce multiple plume detections and corresponding emission rate estimates within the same visit. Bridger reports include both plume images and quality-controlled quantification estimates. The detection threshold is between 1\u0026ndash;3 kg/h with a 90% probability of detection, depending on the atmospheric conditions. Uncertainty in emission rate estimates arises from several sources, including variability in wind speed and direction inputs, assumptions about plume dispersion geometry, and limitations in spatial resolution and aerial scan coverage. While quantitative uncertainty bounds are not calculated in this study, Bridger Photonics' system has undergone extensive validation through controlled release experiments\u003csup\u003e\u003cspan additionalcitationids=\"CR37\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Cause Analysis Framework\u003c/h2\u003e\u003cp\u003eCausal analysis was employed to systematically identify and categorize the operational causes responsible for the observed methane emissions at AMI member companies. The process began with operator-led attribution based on internal site assessments. Each of the five participating operators investigated aerially detected emissions using their own protocols, drawing on site-level knowledge, operational data, and maintenance records to assign preliminary causal attributions. These initial attributions were subsequently reviewed by the research team, with additional information requested where necessary to ensure consistency and clarity across reporting. This process resulted in 1,669 emission events with operator-provided cause descriptions. The initial distribution of these events across broad operator-assigned categories, along with the data processing details, are presented in Supplementary Note S1. Operators used varied terminologies and levels of specificity in their descriptions (e.g., \u0026ldquo;dump valve stuck,\u0026rdquo; \u0026ldquo;valve left open,\u0026rdquo; \u0026ldquo;valve hung open\u0026rdquo;), resulting in 61 distinct cause labels. To address this heterogeneity, the research team developed a harmonized classification system by grouping similar descriptions into standardized causal categories. For instance, the examples above were consolidated under the category \u0026lsquo;incorrect valve position.\u0026rsquo; The complete harmonized classification is presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. To focus the analysis on actionable emissions, two categories were excluded from further analysis. First, events attributed to \u0026lsquo;normal operations,\u0026rsquo; defined as expected routine emissions from standard equipment functions (e.g., compressor rod packing vents operating as designed), were excluded. Episodic intermittent operations, including \u0026lsquo;well unloading,\u0026rsquo; \u0026lsquo;maintenance,\u0026rsquo; and \u0026lsquo;blowdown,\u0026rsquo; are treated as distinct causal categories to analyze their specific emission patterns and are therefore not included under 'normal operations'. Additionally, events classified as \u0026lsquo;Unknown/third party,\u0026rsquo; were also excluded (see Supplementary Fig. S2). The latter category included emissions for which no conclusive source could be identified on participating operator assets or emissions clearly originating from infrastructure owned by non-participating companies.\u003c/p\u003e\u003cp\u003eAfter these exclusions, 1,011 emission events remained for detailed causal analysis. Their distribution across the final harmonized causal categories is shown in Supplementary Fig. S4. This dataset primarily comprises events for which operators conducted investigations and assigned initial causal attributions, which were subsequently categorized into standardized causal categories by the research team. These harmonized causal categories formed the analytical basis for subsequent quantitative assessments. In this study, 'frequency' refers to the observed event counts from the four quarterly aerial surveys, representing the relative prevalence of different causal categories within the measurement period, and not a statistically annualized recurrence rate. For the specific application of risk assessment tools, such as FMEA and FTA, these causal categories are treated as the primary failure modes for this analysis. Overall, this harmonized framework ensures consistency and supports a structured quantitative analysis of the emission causes.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eMapping of specific operator-reported cause descriptions into nine consolidated non-normal cause categories. This table consolidates 61 unique operator-provided cause descriptions from 1,669 emission events identified during the 2024 Appalachian Methane Initiative aerial surveys, into nine actionable causal categories, excluding normal operations and unknown/third-party sources, as detailed in Supplementary Note S1.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEmission Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDetailed Subcategories\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNormal Operations\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNormal Operations, Venting Casings, Compressor Rod Packing, Combustion slip/Gas Engine Methane Slip, Pneumatic Controller Sensitivity, Flame Arrestor, Dehydrator Vents, Engine exhaust, Separator Vents, Pump, VDU, Pipeline, GPU, Generator.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDrilling/Completions\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCompletions, Drilling, Workover, Fracking, Horizontal Drilling, Well Ops \u0026amp; Completion.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMechanical Failures\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePacking Failures, O-ring Failures, Loose Fitting/Cracks, Venting from Internal Failure, Gasket, Corrosion, Blockage, Broken Part, Mechanical Leaks, Blockage, Regulator Failure, Pneumatic Controller Fault, GPU Box Malfunction, M\u0026amp;I leak.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eIncorrect Valve Position\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDump Valve Hung Open, Dump Valve Cut, Incorrect Valve Position, Valve Left Open, Stuck Dump Valve.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAbnormal Tank Emissions\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThief Hatch / Tank Opening, Thief Hatch Failure, Tank Control Vents.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFlare Unlit/Malfunction\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUnlit Flare, Burner, Flare Malfunction, Flare, Unlit Flame.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eHeater Unlit/Malfunction\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHeater Unlit, Heaters vents, Heater Malfunction.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMaintenance\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMaintenance, Maintenance Vents, Flowline Repair.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eBlowdown\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBlowdowns.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eWell Unloading\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWell Unloading, Well Shutdown.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eUnknown/ Third Party\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUnknown, Unidentified, Third-Party Equipment\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.3.1 FMEA Analysis\u003c/h2\u003e\u003cp\u003eWe conduct failure mode and effects analysis (FMEA) to systematically prioritize methane emission risk identified through the causal analysis. FMEA is a proactive method used to evaluate potential failure modes using three key parameters with individual scores: Severity (S), Occurrence (O), and Detection (D). All scores were normalized to a scale of 1 for the lowest risk to 10 for the highest risk to ensure comparability and interpretability across components. Severity (S) was derived from the average methane emission rate (kg/h) for each failure mode, representing its potential impact. Occurrence (O) was computed from the event count of each failure mode, reflecting its operational prevalence. For both Severity and Occurrence, the scores were normalized using a standard min-max scaling approach to achieve a 1\u0026ndash;10 scale, as shown in Eq.\u0026nbsp;1:\u003c/p\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:Score=1+\\frac{\\left(\\alpha\\:-{\\alpha\\:}_{min}\\right)}{\\left({\\alpha\\:}_{max}-{\\alpha\\:}_{min}\\right)}(10-1)\\)\u003c/span\u003e\u003c/span\u003e(Eq.\u0026nbsp;1)\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Score\\)\u003c/span\u003e\u003c/span\u003e is the normalized score between 1 and 10, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e is the observed value for the specific failure mode (i.e., the average emission rate for severity, or the event count for occurrence), \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{min}\\)\u003c/span\u003e\u003c/span\u003e is the minimum observed value for that parameter across all failure modes, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{max}\\)\u003c/span\u003e\u003c/span\u003e is the maximum observed value for that parameter across all failure modes.\u003c/p\u003e\u003cp\u003eDetection (D) represents the likelihood that an emission is captured or detected for control when it occurs. The scores were defined based on the average duration of the emission events for each failure mode. Given that shorter-duration events are less likely to be captured by snapshot-based aerial measurement campaigns, detection scores were inversely normalized. A reversed min-max scaling approach was used, assigning a score of 10 to the shortest average duration (most difficult to detect) and 1 to the longest average duration (easiest to detect), as shown in Eq.\u0026nbsp;2:\u003c/p\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:{Score}^{*}=1+\\frac{\\left({\\alpha\\:}_{max}-\\alpha\\:\\right)}{\\left({\\alpha\\:}_{max}-{\\alpha\\:}_{min}\\right)}(10-1)\\)\u003c/span\u003e\u003c/span\u003e(Eq.\u0026nbsp;2)\u003c/p\u003e\u003cp\u003ewhere Score* is the normalized detection score between 1 and 10, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\alpha\\:\\)\u003c/span\u003e\u003c/span\u003e is the average duration for the specific failure mode, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{min}\\)\u003c/span\u003e\u003c/span\u003e is the shortest average duration observed across all failure modes, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{max}\\)\u003c/span\u003e\u003c/span\u003e is the longest average duration observed.\u003c/p\u003e\u003cp\u003eThe composite Risk Priority Number (RPN) for each failure mode was then calculated by multiplying the three scores, as shown in Eq.\u0026nbsp;3.\u003c/p\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:\\:\\:\\:\\:\\:\\:\\:\\:RPN=S\\:\\times\\:O\\:\\times\\:D\\)\u003c/span\u003e\u003c/span\u003e(Eq.\u0026nbsp;3)\u003c/p\u003e\u003cp\u003eThe RPN provides a quantitative measure to rank and prioritize emission failure modes for targeted mitigation efforts. The detailed scoring inputs and results for this section are provided in Supplementary Table S2 and Supplementary Note S2.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.3.2 FTA Analysis\u003c/h2\u003e\u003cp\u003eFault Tree Analysis (FTA) was employed as a structured, top-down technique to trace the logical pathways leading to methane emissions and to quantitatively estimate their likelihood. This approach uses a logic-based framework composed of AND/OR gates to model how individual or combined failures can escalate into emission events. The analysis began by defining \"Methane Emissions (per year)\" as the top-level event in the fault tree. The top event was then decomposed through an OR gate into nine Level 1 intermediate events, corresponding to the major failure modes such as mechanical failures. Selected Level 1 failure modes were further broken down into more specific sub-system or root causes using intermediate OR or AND logic gates, based on the specific sub-causes identified in the operator causal analysis data and established engineering principles. This hierarchical decomposition continued until the basic events were reached. These basic events represent the initiating failures for which probability data can be assigned, either from empirical observations or literature-based estimates. For instance, \"Abnormal Tank Emissions\" was decomposed into three intermediate contributors: vapor containment failure, control system failure, and liquid handling/overpressure events. These were further linked to basic events, such as thief hatches that were left open, which were assigned failure probabilities. The complete structure of the fault tree is presented in Supplementary Fig. S6. The failure probabilities were assigned to each basic event based on a combination of literature-based estimates for common component failures and empirical data extracted from the causal analysis dataset, where available, as summarized in Supplementary Table S3. The fault tree was then quantified to estimate the annual probability of the top-level emission event and assess the probabilistic importance of various failure pathways and basic events. This quantification enabled the identification of the most critical failure branches and basic events contributing to the overall emission risk, thereby informing targeted mitigation strategies.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Statistical Analysis and Uncertainty\u003c/h2\u003e\u003cp\u003eStatistical analyses were performed using Python (version 3.13.3), with the SciPy stats module for hypothesis testing. Chi-square tests of independence were applied to assess statistically significant differences in the distribution of emission causal categories across both temporal and spatial dimensions. A standard significance threshold of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was used to determine whether the observed differences were unlikely to have occurred by chance. Uncertainty in the dataset arises primarily from two sources: measurement limitations that may under-detect low-level, intermittent, or short-duration emissions due to detection thresholds, and inconsistencies in cause classification across operators. While these factors may affect resolution, their impact is mitigated by the large basin-wide dataset and consistent follow-up corrections with the participating operators to refine and standardize the categorizations.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Basin-Wide Distribution of Methane Emission Causal Categories\u003c/h2\u003e\u003cp\u003eA basin-wide assessment of methane emissions revealed distinct patterns across both the frequency of emission causal categories and their contributions to the total emission rates. The emission events attributed to specific operational causes in this study represented approximately 61% of all distinct emission events detected during aerial surveys and accounted for 53% of the total methane emissions (see Supplementary Fig. S2 and S3 for further details on the emission distributions). Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the percentage of total emission events attributed to each causal category and their average emission rates (kg/h), overlaid with their proportional contributions to total basin-wide emissions.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAbnormal tank emissions emerged as the most prevalent and impactful causal category, accounting for 32.1% of all events with documented cause and contributing 46.5% of total methane emissions from the analyzed causal categories, with an average rate of 52.6 kg/h. These findings are consistent with those of prior studies that identified tanks, particularly those with faulty hatches, vapor recovery issues, or uncontrolled venting, as critical contributors to upstream methane emissions\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. The dual risk posed by high frequency and moderate-to-high intensity highlights the importance of targeted interventions, such as improved thief hatch seals and automated monitoring systems, in mitigating tank emissions.\u003c/p\u003e\u003cp\u003eMechanical failures were the second most common emission causal category, accounting for 20.8% of the emission events, with an average emission rate of 22.1 kg/h, contributing 12.4% to the total event-attributable emissions. These emissions are associated with degraded components, such as loose fittings or packing failures in our dataset, and result in lower intensity yet persistent emissions. In contrast, well unloading events, although accounting for only 9.4% of events, stand out as the most emission-intensive event, with an average emission rate of 71.3 kg/h, contributing 16.4% to the total event-attributable emissions. These episodic events, typically associated with the clearing of liquid from wells and the restoration of gas flow, are of short duration but show disproportionately large emission volumes. Prior studies have highlighted the significance of such episodic release during liquid unloading, noting that they may be insufficiently characterized in existing research\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eIncorrect valve positions and flare unlit/malfunctions represent 13.6% and 10.7% of the total events, respectively, and exhibit average emission rates of 21.8 and 32.2 kg/h, respectively. Their total event-attributable emission contributions were 7.7% and 9.5%, respectively, indicating that although these emission causal categories occur relatively often, their individual impact is moderate compared to other event types. Notably, maintenance activities contributed a relatively small proportion of emission events (5.5%) and total event-attributable emissions (2.6%), with an average emission rate of 16.5 kg/h. These findings suggest that although emissions during maintenance are operationally expected, their impact is limited in the Appalachian Basin.\u003c/p\u003e\u003cp\u003eBlowdowns, drilling/completions, and heater unlit/malfunctions each represented less than 5% of the total emission counts and cumulative event-attributable methane emissions in the dataset. Notably, emissions from heater unlit/malfunctions are underrepresented because many events fall below the minimum detection threshold of aerial technology. A recent study on heaters reported a median emission rate of 0.28 kg/h, which is significantly lower than the minimum detection threshold of aerial systems\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. However, in this dataset, heater unlit/malfunction events that exceeded the detection threshold exhibited an average emission rate of 33.6 kg/h. This suggests that while such events are rare and often missed by snapshot detection methods, they can still pose a significant emission risk when they occur.\u003c/p\u003e\u003cp\u003eThese findings reveal a clear divergence between high-frequency, lower-intensity causes and low-frequency, high-intensity ones, suggesting the need for a dual-pronged mitigation strategy. On one hand, frequent, moderate-impact emissions, exemplified by mechanical failures and incorrect valve positions, may benefit from broad-based monitoring and operational improvements. On the other hand, infrequent yet high-impact events, such as well unloading, necessitate more targeted mitigation protocols that may not necessarily benefit from or require survey-type monitoring.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Super-Emitters and Skewed Emission Distributions\u003c/h2\u003e\u003cp\u003ePrevious studies have consistently demonstrated that methane emissions in oil and gas operations are highly skewed, with a small number of high-emission events, referred to as super-emitters, accounting for the majority of the total emissions\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. In this study, super-emitters are defined as large emission events with rates equal to or exceeding 100 kg/h. While this phenomenon has been well characterized in terms of emission magnitudes, it remains unclear how super-emitter behavior translates across different emission causes. Specifically, we want to examine if a small number of causes disproportionately contribute to super-emitters.\u003c/p\u003e\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea presents the cumulative distribution of methane emission rates across all analyzed emission events, highlighting the concentration of emissions among a small subset of large events. As shown, only a few events, comprising 7.2% of the emission events, exceeded the defined super-emitter threshold of 100 kg/h. Despite their low frequency, these super-emitter events contributed to approximately 80% of the total event-attributable methane emissions, confirming the heavy-tailed distribution of emissions. To investigate the cause of these super-emitter events, we disaggregated the super-emitter subset by causal category. Thus, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb shows the frequency distribution of super-emitter events across these causes. We found that abnormal tank emissions accounted for 36% of all super-emitter events. This aligns with prior findings that tanks, particularly those with compromised thief hatches, are leading sources of large methane releases in upstream operations\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. Well unloading was the second most common cause of super-emitter events (21%), followed by mechanical failures (19%), illustrating that episodic, short-duration events and equipment degradation can contribute to extreme emissions. Other categories, such as incorrect valve position and flare unlit/malfunctions, appeared less frequently (each \u0026lt;\u0026thinsp;10%), yet still contributed to at least one super-emitter event. This indicates that even less frequent or typically lower-emitting causes can occasionally escalate into high-magnitude releases.\u003c/p\u003e\u003cp\u003eThese findings demonstrate that a small subset of causal categories is responsible for a disproportionate share of the super-emitter events. Additionally, they highlight the importance of linking emissions measurement data to operational causes, enabling proactive, causal-informed intervention before escalation occurs. Lastly, the data suggests that large emissions may not necessarily result from unique or anomalous causes, but rather from the escalation of common operational issues, a finding explored further in the subsequent discussion.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Comparing Causal Drivers of Large vs. Small Methane Emitters\u003c/h2\u003e\u003cp\u003eBuilding on the super-emitter analysis from the previous section, this analysis examines how emission patterns of emission causes differ between small (\u0026lt;\u0026thinsp;100 kg/h) and large (\u0026ge;\u0026thinsp;100 kg/h) emitters. To quantify the tendency of certain causal categories to be associated with high-magnitude emissions, we introduce an Escalation Ratio (ER). In this context, 'escalation' refers not to the real-time development of a single event but to the statistical overrepresentation of emission causes among large emitters compared to small emitters. The ER for a specific causal category is defined as\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:{ER}_{i}=\\:\\frac{{f}_{i,large}/{N}_{large}}{{f}_{i,smal}/{N}_{small}}\\:$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{i,large}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{i,small}\\:\\)\u003c/span\u003e\u003c/span\u003erepresent the number of large and small emitter events, respectively, associated with causal category i, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{N}_{large}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{N}_{small}\\)\u003c/span\u003e\u003c/span\u003e denote the total number of large and small emitter events across all categories. This ratio captures the relative overrepresentation of emission causes among high-magnitude emission events. The emission rate contributions for each category are provided in Supplementary Fig. S7.\u003c/p\u003e\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows that abnormal tank emissions dominated both small and large emitter groups, appearing in over 30% of all events. An ER of 1.1 suggests a consistent frequency across emission scales, indicating that tanks pose a persistent, system-wide emission risk rather than being escalation-prone. Despite this consistency, tank emissions dominated the total methane contribution across both categories, accounting for nearly 50% of the total emissions among large emitters and 36% among small emitters. By contrast, well unloading events exhibited the highest escalation ratio of 2.8, showing a dramatic increase in frequency from small to large emitters. It is important to note that for episodic events like well unloading, the instantaneous emission rate can vary widely, meaning snapshot-based classifications may reflect observation timing rather than event differences. Nevertheless, the high ER demonstrates that unloading operations are disproportionately represented among the highest-rate emission events and highlights the importance of managing both the duration and intensity of their peak emission phase.\u003c/p\u003e\u003cp\u003eHeater unlit/malfunctions with an ER of 2.6 and blowdowns with an ER of 1.4 also show an elevated presence among large emitters, despite being relatively infrequent overall. These categories represent emission releases that, although not as prevalent, can rapidly escalate under unfavorable containment conditions. For example, while heater malfunctions account for only a small share of total event-attributable emissions (\u0026lt;\u0026thinsp;1% overall) because they are mostly below the detection threshold of aerial surveys, their high ER suggests that when detected, they are more likely to be a super-emitter event. This likely reflects a binary emission pattern, where malfunctions, such as flameouts, cause sudden gas blow-by and overwhelm combustion systems, resulting in high-rate emissions. This step-change dynamic helps explain their elevated ER, though further diagnostics are needed to confirm all causal factors. Conversely, common operational causes, such as mechanical failures and incorrect valve positions, showed lower escalation tendencies, with ERs of 0.9 and 0.8, respectively. These causes are more prevalent among small emitters and tend to produce lower-magnitude emissions. This is expected as routine mechanical failures from wear and tear are not expected to result in super-emitters. Other causal categories displayed inverse escalation patterns with ER\u0026thinsp;\u0026lt;\u0026thinsp;1. These categories appear more frequently among small emitters, reflecting consistent but lower-magnitude emissions.\u003c/p\u003e\u003cp\u003eOverall, these results emphasize that super-emitter events are not separate anomalies but traceable to specific operational causes with quantifiable escalation potential. The ER introduced here serves as a valuable prioritization metric, highlighting which categories are disproportionately associated with large-scale releases. In addition, not all emission causes escalate equally. Therefore, mitigation planning should reflect this heterogeneity. Emission mitigation frameworks should apply differentiated strategies: one focused on routine prevention and detection for persistent low-escalation categories, and another aimed at escalation control for causal categories prone to producing super-emitter outcomes.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.4 FMEA and FTA Analysis\u003c/h2\u003e\u003cp\u003eThis study presents the first application of FMEA to quantitative risks associated with different causal categories of measured methane emission events at a basin scale. The Risk Priority Number (RPN) was computed for each failure mode, as described in the Methods section, providing a composite score to guide mitigation prioritization.\u003c/p\u003e\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the resulting RPN values, alongside the individual scores for severity, occurrence, and detection (see Supplementary Table S2). Abnormal tank emissions emerged as the highest-priority failure mode with an RPN of 487, driven by a high severity score of 7.3, maximum occurrence score of 10, and detection score of 6.6, indicating moderate difficulty in detection. This finding reinforces the importance of tank emissions to basin-wide emissions and highlights their status as a high-risk, high-frequency, and difficult-to-detect emission source. Well unloading ranked as the next highest priority with an RPN of 295. Although it accounted for only 9.4% of emission events, it exhibited the highest average emission rate and was close to the maximum detection score of 10, indicating its high-impact, low-detectability profile. Flare unlit/malfunction ranked third with an RPN of 142, primarily driven by its high detection difficulty score of 8.5, alongside moderate severity and occurrence values.\u003c/p\u003e\u003cp\u003eMechanical failures and incorrect valve positions yielded mid-range RPNs of 90 and 79, respectively, driven primarily by lower severity scores of 2.78 and 2.73, respectively, despite relatively higher occurrence and detection difficulty scores. Although not the most individually consequential failure modes, their risk profiles are elevated by the combined influence of human and operational errors, coupled with limited detectability. Blowdowns, with an RPN of 85, was less frequent but exhibited maximum detection difficulty (10.0), emphasizing the challenges of capturing short-duration events. Lower-priority failure modes with RPN values\u0026thinsp;\u0026le;\u0026thinsp;50 include heater unlit/malfunctions, maintenance, and drilling/completions. These categories were either infrequent, associated with lower emission rates, or characterized by longer durations that improve detection, suggesting that standard operational protocols may be sufficient for their management.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTogether, these FMEA results provide a quantitative, data-informed, and structured framework for prioritizing mitigation strategies beyond frequency or emission rates alone. By integrating three risk dimensions, FMEA provides a more robust risk profile and reorders priorities. For example, mechanical failures were the second most frequent event; they were de-prioritized while flare unlit/malfunction was ranked higher due to detection scores. This illustrates how FMEA highlights intermittent failure modes that might otherwise be overlooked. High-RPN failure modes, such as tank emissions and well unloading, warrant focused attention through enhanced automation, improved detection strategies, and predictive interventions. In contrast, moderate and low-risk failure modes may be incorporated into broader operational workflows. While this represents a simplified FMEA for interpretability, the approach demonstrates the value of integrating emissions data into structured causal analysis tools. This framework also lays the groundwork for advanced FMEA models.\u003c/p\u003e\u003cp\u003eTo quantitatively assess the likelihood of methane emissions arising from the identified operational and equipment failures, we conducted FTA. The FTA modeled the logical escalation pathways from the three most frequent failure categories: abnormal tank emissions, mechanical failures, and incorrect valve positions, to the top event, annual methane emission. The complete fault tree structure is provided in Supplementary Fig. S6, with the associated basic event probabilities listed in Supplementary Table S3.\u003c/p\u003e\u003cp\u003eThe FTA propagation yielded an estimated annual probability of 7.2% for methane emissions arising from these three failure categories. When disaggregated, abnormal tank emissions exhibited the highest individual probability at 2.7%, followed by mechanical failures at 2.5%, and incorrect valve position at 2.0%. Within the tank emissions branch, vapor containment failures contributed most significantly, with an annual probability of 1.4%, primarily driven by thief hatch or tank opening events (1.0%). Control system failures and liquid handling/overpressure incidents contributed 0.85% and 0.41%, respectively. For incorrect valve position, human error categorized as \"valve left open\", emerged as the dominant pathway, with an annual probability of 2.0%. This figure is over 100 times greater than the combined contribution of other valve faults within this failure category, highlighting the disproportionate role of operational issues. In the mechanical failure branch with a total annual probability of 2.5%, loose fittings and structural cracks represented the primary contributors.\u003c/p\u003e\u003cp\u003eTo assess the validity of the bottom-up FTA model, we compared model-derived annual probabilities with empirical event frequencies from the survey data. For tank-related events, the FTA-predicted annual probability of 2.7% was in close agreement with the empirically observed annual event frequency of 2.0%. In stark contrast, a similar analysis for flare malfunctions (detailed in Supplementary Note S3) yielded a theoretical annual probability of ~\u0026thinsp;0.2% based on literature reliability data, nearly 50 times lower than the empirically observed annual frequency of 7.6% in our dataset. This strong alignment for tanks validates the model structure, whereas the discrepancy for flares highlights the critical importance of using basin-specific, measurement-informed data over generic failure rates to accurately assess real-world emission risks. Overall, these FTA findings illustrate how emissions result from a convergence of equipment degradation, procedural lapses, and human error, highlighting critical nodes that drive emission risk and informing targeted mitigation strategies.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Quarterly Variation in the Distribution of Emission Causes\u003c/h2\u003e\u003cp\u003eUnderstanding the quarterly distribution of emission causes is critical for identifying operational or seasonal patterns that can inform predictive maintenance and targeted mitigation strategies. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the quarterly distribution of methane emission events, expressed as a percentage of total emission events. The associated p-values from a chi-square test are included to indicate whether the observed differences are statistically significant across quarters. The raw unnormalized event counts are provided in Supplementary Fig. S4.\u003c/p\u003e\u003cp\u003eSeveral emission causes, including blowdown, flare unit/malfunction, incorrect valve position, and tank emissions, did not exhibit statistically significant temporal differences (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). These patterns suggest a relatively constant rate of occurrence throughout the year, possibly reflecting equipment types or causal mechanisms that are insensitive to seasonal influences. For instance, incorrect valve positions remained relatively consistent across all quarters, while tank emissions showed minor fluctuations; highest in Q2, lowest in Q3, but without statistical significance (p\u0026thinsp;=\u0026thinsp;0.33). This shows the persistent nature of these causal categories, indicating that they are likely tied to regular operational workflows rather than external drivers.\u003c/p\u003e\u003cp\u003eBy contrast, three causal categories: maintenance events, mechanical failures, and well unloading exhibited statistically significant temporal variation. Mechanical failures peaked in Q2 with approximately 32% of the category\u0026rsquo;s total, showing significant variation across quarters (p\u0026thinsp;=\u0026thinsp;0.04). Notably, maintenance events showed a sharp spike in Q3, accounting for 30% of all events in this category (p\u0026thinsp;=\u0026thinsp;0.00008). This concentration is notable, given the shorter 14-day survey duration in Q3 (Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), making an increase in planned maintenance activities unlikely. The statistically significant peak suggests a genuine, yet unexplained, cluster of maintenance-related emissions during this period. Well unloading showed a sharp increase in Q4, with moderate but overall significant variability across quarters (p\u0026thinsp;=\u0026thinsp;0.0075). Although unloading events are often assumed to typically occur at a relatively uniform rate, this Q4 increase may relate to year-end production optimization or target-based management efforts, a pattern further examined in the next section. These patterns suggest that while many emissions causal categories are temporally invariant, others may be influenced by seasonal operational cycles or maintenance scheduling.\u003c/p\u003e\u003cp\u003eOverall, the statistically significant trends emphasize the non-steady state nature of some emission causes and support a differentiated approach to mitigation. For instance, time-sensitive patterns such as seasonal maintenance peaks and Q4 unloading surges suggest a need for time-stratified intervention strategies, whereas persistent emission causes may be best addressed through routine inspection protocols and systematic detection efforts.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Inter-Operator Variability of Emission Causes\u003c/h2\u003e\u003cp\u003eUnderstanding differences in the distribution of emission causes among operators is essential for identifying operator-specific patterns that may reflect variations in facility design, maintenance procedures, or operational practices. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents the distribution of emission causes across four anonymized operators as companies A through D in the Appalachian basin. The data are expressed as the percentage of total emission events that are attributable to each operator, for each causal category. Statistical analysis confirms that the distribution of events for all major causal categories exhibits statistically significant differences across operators (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). This finding aligns with previous research suggesting that comprehensive sampling across operators may be more effective for reducing basin-wide uncertainty than repeated temporal measurements\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe distribution of emission causes reveals substantial heterogeneity in emissions behavior among the companies. Flare unlit/malfunction emissions were heavily concentrated in Company A, which accounted for nearly 49% of all such events, a proportion significantly different from other operators (p\u0026thinsp;\u0026lt;\u0026thinsp;0.00001). Abnormal tank emissions showed a notable skew, with Companies B and C collectively responsible for over 80% of all tank emissions. Mechanical failures were also unevenly distributed, with Company B contributing nearly 37% of these failures, potentially reflecting differences in equipment condition or operational stress. Well unloading events were most prevalent in Company D, which reported approximately 37% of all such events (p\u0026thinsp;\u0026lt;\u0026thinsp;0.00001). This aligns with the spike in Q4 discussed earlier and may indicate company-specific production strategies. While incorrect valve position was more evenly distributed, it still exhibited a statistically significant difference (p\u0026thinsp;=\u0026thinsp;0.005), with Company D reporting elevated levels (about 25%). Finally, maintenance-related emissions were more frequently reported by Company C, and although blowdown events were less frequent overall, they were notably higher in Company A (11%) with significant variation across companies (p\u0026thinsp;=\u0026thinsp;0.0012).\u003c/p\u003e\u003cp\u003eThese findings highlight the operator-specific nature of methane emission causes. The non-uniform distribution of emission causes across operators corroborates findings that emissions are influenced by differences in factors like infrastructure age, automation levels, and maintenance practices\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. This reinforces the need for tailored mitigation strategies. For instance, an operator predominantly experiencing flare-related events might prioritize investments in flare reliability and maintenance, whereas another operator with a high incidence of tank emissions would focus on tank integrity programs and vapor recovery systems. Incorporating causal analysis-based causal insights into inventory models can improve both granularity and accuracy, enabling more targeted and process-informed methane mitigation efforts at both company and regional levels.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study offers one of the first basin-wide applications of causal analysis techniques to methane emissions in upstream oil and gas operations. By linking aerial detection data to classified emission causes, we introduce a process-based approach that shifts the analytical focus from quantifying emission magnitudes to diagnosing emission causation. This method demonstrates that methane emissions are largely traceable to identifiable, recurring causes, many of which exhibit clear temporal, operator-specific, and escalation-related trends, highlighting the need for process-informed mitigation. Several key insights emerge from this work, each carrying significant implications for understanding emission drivers, their variability, and process-based risk mitigation planning.\u003c/p\u003e\u003cp\u003eFirst, across both the full dataset and the super-emitter subset, tank emissions consistently emerged as the most frequent and highest-contributing causal category. Importantly, our data reveal that super-emitter events are not driven by fundamentally different causes. Instead, common causes, especially those related to abnormal tank emissions, well unloading, and mechanical degradation, can escalate into high-magnitude emissions under specific operational conditions. This consistency suggests that mitigation efforts should not only focus on the type of emissions, but also on understanding and controlling the operational conditions that cause a routine emission to escalate into high-emission rate events. Consequently, this understanding supports a shift in mitigation strategies away from purely anomaly detection or qualitative segregation towards risk-based monitoring and prevention focused on known, high-escalation emission causes.\u003c/p\u003e\u003cp\u003eSecond, statistically significant quarterly and inter-operator variation was observed in the distribution of emission causes. For example, seasonal peaks in maintenance-related events highlight the influence of time-dependent operational practices. Similarly, strong divergence in the prevalence of these causes across operators demonstrates that emissions are not uniformly distributed. These patterns contradict a \u0026ldquo;one-size-fits-all\u0026rdquo; approach and emphasize the importance of proactive adjustments and tailored strategies aligned with seasonal operational variations and operator-specific differences.\u003c/p\u003e\u003cp\u003eThird, a comprehensive mitigation strategy must also prioritize sources based on cumulative emissions, considering rate, duration, and frequency\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. Our analytical framework supports this balanced approach. The structured application of FMEA enabled prioritization of emission causes beyond just occurrence. By integrating frequency, severity, and detection difficulty, FMEA highlights that high-risk, low-frequency failure modes may warrant rapid intervention strategies, such as remote shutoff systems and predictive alerts, while persistent but lower-intensity ones may require routine monitoring and maintenance protocols. To complement this FTA integrates annual emissions probabilities, offering a time-based risk perspective. Together, FMEA and FTA provide a practical, process-informed pathway for identifying how simple malfunctions may compound into significant emission events and mitigation levers, thereby facilitating more cost-effective and targeted mitigation.\u003c/p\u003e\u003cp\u003eA novel contribution of this study is the introduction of the Escalation Ratio (ER), a metric quantifying how much more frequently an emission cause occurs among super-emitters compared to small emitters. Emissions categories such as well unloading exhibited high ER values, highlighting their disproportionate presence in high-emission cases. The ER metric offers a new layer of insight into causal analysis by identifying causal categories with higher escalation risks and lays the groundwork for future development of predictive scoring systems that can inform the design of proactive interventions.\u003c/p\u003e\u003cp\u003eThe collective findings of this research support a paradigm shift in methane mitigation: from a predominantly reactive reliance on leak detection to an integrated approach emphasizing proactive cause prevention. By embedding causal insights into inventory development, LDAR scheduling, and broader operational risk management frameworks, this work lays the groundwork for more effective, operationally grounded, and risk-informed emissions management strategies. Our results strongly suggest that many super-emitter events arise from the escalation of common operational issues, suggesting a path toward more proactive management by identifying the specific causal categories most prone to escalation.\u003c/p\u003e\u003cp\u003eFurthermore, the FMEA and FTA models presented in this work are intentionally simplified, primarily serving to illustrate methodological integration and provide a foundational framework for integrating causal analysis with emissions monitoring. Future research can build on these findings. Key priorities include better standardization of causal classification protocols across operators to improve cross-comparability and the development of more sophisticated probabilistic FTA models that can dynamically simulate failure propagation pathways. Furthermore, predictive analytics, such as Poisson regression for failure frequency and machine learning for causal classification, offer promising pathways for real-time risk scoring and automated emissions attribution. When integrated with maintenance systems, these tools can transform methane mitigation from a reactive discipline into a proactive, process-embedded component of methane management in the oil and gas sector.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor contributions\u003c/h2\u003e\n\u003cp\u003eAbdulmuiz A. Adekomi: Methodology, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eShuting Yang: Methodology, Validation, Conceptual framing, Writing – review \u0026amp; editing.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eShannon Stokes: Methodology, Formal analysis, Conceptual framing, Writing – review.\u003c/p\u003e\n\u003cp\u003eArvind P. Ravikumar: Conceptualization, Writing – review \u0026amp; editing, Supervision, Project administration, Funding acquisition.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNotes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA.P.R. is currently a member of the Gas Pipeline Advisory Committee of the US Department of Transportation; in this role, he is a Special Government Employee. A.P.R. has current research support from the US Department of Energy, Environmental Defense Fund, and sponsors of the Energy Emissions Modeling and Data Lab (EEMDL).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was funded in part by the U.S. Department of Energy under Grant No. DE-FE0032311 and the Appalachian Methane Initiative.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The datasets and codes generated and/or analyzed in this study are publicly available. Additional data supporting the findings, including emission event records and causal classification outputs, are provided in the Supplementary Information.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMasson-Delmotte V, Zhai P, Pirani A et al (2021) EdsV. Masson-Delmotte, P. Zhai, A. Pirani, Eds. \u003cem\u003eClimate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ipcc.ch/report/ar6/wg1/\u003c/span\u003e\u003cspan address=\"https://www.ipcc.ch/report/ar6/wg1/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSobanaa M, Prathiviraj R, Selvin J, Prathaban M (2023) A comprehensive review on methane\u0026rsquo;s dual role: effects in climate change and potential as a carbon\u0026ndash;neutral energy source. Environ Sci Pollut Res 31:10379\u0026ndash;10394\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTibrewal K et al (2024) Assessment of methane emissions from oil, gas and coal sectors across inventories and atmospheric inversions. Commun Earth Environ 5:26\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eControl methane to (2021) slow global warming \u0026mdash; fast. Nat Editor 2021 August \u003cem\u003e25\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAdekomi AA, Carrasco-Jaim OA, Weerasooriya UP, Okuno R (2024) Enhanced CO\u003csub\u003e2\u003c/sub\u003e Capture-Desorption by Surface-Active Amine. Energy Fuels 38:14435\u0026ndash;14448\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShindell D, Kuylenstierna JCI, Michalopoulou E, H\u0026ouml;glund-Isaksson L, Zhang Y, Seltzer K et al (2021) United Nations Environment Programme (UNEP) and Climate and Clean Air Coalition (CCAC). Global Methane Assessment. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.5281/zenodo.4777096\u003c/span\u003e\u003cspan address=\"10.5281/zenodo.4777096\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlvarez RA et al (2018) Assessment of methane emissions from the U.S. oil and gas supply chain. Science 361:186\u0026ndash;188\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eU.S natural gas production remained flat in 2024 - U.S. Energy Information Administration (EIA). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003c/span\u003e\u003cspan address=\"http://www.eia.gov/todayinenergy/detail.php?id=65025\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGoetz JD et al (2015) Atmospheric Emission Characterization of Marcellus Shale Natural Gas Development Sites. Environ Sci Technol 49:7012\u0026ndash;7020\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen DT et al (2013) Measurements of methane emissions at natural gas production sites in the United States. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 110, 17768\u0026ndash;17773\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeischl J et al (2015) Quantifying atmospheric methane emissions from the Haynesville, Fayetteville, and northeastern Marcellus shale gas production regions. J Geophys Res Atmos 120:2119\u0026ndash;2139\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOmara M et al (2022) Methane emissions from US low production oil and natural gas well sites. Nat Commun 13:2085\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOmara M et al (2024) Constructing a measurement-based spatially explicit inventory of US oil and gas methane emissions (2021). Earth Syst Sci Data 16:3973\u0026ndash;3991\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWilliams JP et al (2025) Small emission sources in aggregate disproportionately account for a large majority of total methane emissions from the US oil and gas sector. Atmospheric Chem Phys 25:1513\u0026ndash;1532\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOmara M et al (2016) Methane Emissions from Conventional and Unconventional Natural Gas Production Sites in the Marcellus Shale Basin. Environ Sci Technol 50:2099\u0026ndash;2107\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSherwin ED et al (2024) US oil and gas system emissions from nearly one million aerial site measurements. Nature 627:328\u0026ndash;334\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBrandt AR et al (2014) Methane Leaks from North American Natural Gas Systems. Science 343:733\u0026ndash;735\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThorpe MJ et al (2024) Deployment-invariant probability of detection characterization for aerial LiDAR methane detection. Remote Sens Environ 315:114435\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang JL et al (2022) Multiscale Methane Measurements at Oil and Gas Facilities Reveal Necessary Frameworks for Improved Emissions Accounting. Environ Sci Technol 56:14743\u0026ndash;14752\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDaniels W et al (2023) Towards multi-scale measurement-informed methane inventories: reconciling bottom-up site-level inventories with top-down measurements using continuous monitoring systems. Preprint at. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.26434/chemrxiv-2023-jp5nt-v2\u003c/span\u003e\u003cspan address=\"10.26434/chemrxiv-2023-jp5nt-v2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZavala-Araiza D et al (2015) Reconciling divergent estimates of oil and gas methane emissions. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 112, 15597\u0026ndash;15602\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVaron DJ et al (2019) Satellite Discovery of Anomalously Large Methane Point Sources From Oil/Gas Production. Geophys Res Lett 46:13507\u0026ndash;13516\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen DT (2014) Methane emissions from natural gas production and use: reconciling bottom-up and top-down measurements. Curr Opin Chem Eng 5:78\u0026ndash;83\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchwietzke S et al (2016) Upward revision of global fossil fuel methane emissions based on isotope database. Nature 538:88\u0026ndash;91\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRutherford JS et al (2021) Closing the methane gap in US oil and natural gas production emissions inventories. Nat Commun 12:4715\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEleweke I et al (2025) The Convergence of Machine Learning and Power Systems: Enhancing Efficiency, Optimization, and Sustainability in Energy Management. 8\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVaron DJ et al (2023) Continuous weekly monitoring of methane emissions from the Permian Basin by inversion of TROPOMI satellite observations. Atmospheric Chem Phys 23:7503\u0026ndash;7520\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchwietzke S et al (2017) Improved Mechanistic Understanding of Natural Gas Methane Emissions from Spatially Resolved Aircraft Measurements. Environ Sci Technol 51:7286\u0026ndash;7294\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZaimes GG et al (2019) Characterizing Regional Methane Emissions from Natural Gas Liquid Unloading. Environ Sci Technol 53:4619\u0026ndash;4629\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZavala-Araiza D et al (2017) Super-emitters in natural gas infrastructure are caused by abnormal process conditions. Nat Commun 8:14012\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLatino MA, Latino RJ, Latino KC (2020) Root Cause Analysis: Improving Performance for Bottom-Line Results. CRC Press / Taylor \u0026amp; Francis Group, Boca Raton, FL\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMascia A et al (2020) A failure mode and effect analysis (FMEA)-based approach for risk assessment of scientific processes in non-regulated research laboratories. Accredit Qual Assur 25:311\u0026ndash;321\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRuijters E, Stoelinga M (2015) Fault tree analysis: A survey of the state-of-the-art in modeling, analysis and tools. Comput Sci Rev 15\u0026ndash;16:29\u0026ndash;62\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi X, Liu W, Zhou N, Yuan X (2025) Probability Analysis of Hazardous Chemicals Storage Tank Leakage Accident Based on Neural Network and Fuzzy Dynamic Fault Tree. Appl Sci 15:3504\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSharafutdinov E (2024) Spatiotemporal Variation in Anthropogenic Methane Emissions in the Appalachian Basin. UT Electron Theses Diss\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJohnson MR, Tyner DR, Szekeres AJ (2021) Blinded evaluation of airborne methane source detection using Bridger Photonics LiDAR. Remote Sens Environ 259:112418\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBell C et al (2022) Single-blind determination of methane detection limits and quantification accuracy using aircraft-based LiDAR. Elem Sci Anthr 10:00080\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eConrad BM, Tyner DR, Johnson MR (2023) Robust probabilities of detection and quantification uncertainty for aerial methane detection: Examples for three airborne technologies. Remote Sens Environ 288:113499\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJohnson MR, Tyner DR, Conrad BM (2023) Origins of Oil and Gas Sector Methane Emissions: On-Site Investigations of Aerial Measured Sources. Environ Sci Technol 57:2484\u0026ndash;2494\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCaulton DR et al (2023) Abnormal tank emissions in the Permian Basin identified using ethane to methane ratios. Elem Sci Anthr 11\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen DT et al (2015) Methane Emissions from Process Equipment at Natural Gas Production Sites in the United States: Pneumatic Controllers. Environ Sci Technol 49:633\u0026ndash;640\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVaughn TL et al (2018) Temporal variability largely explains top-down/bottom-up difference in methane emission estimates from a natural gas production region. \u003cem\u003eProc. Natl. Acad. Sci.\u003c/em\u003e 115, 11712\u0026ndash;11717\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFesta-Bianchet SA, Mohammadikharkeshi M, Tyner DR, Johnson MR (2024) Catalytic Heaters at Oil and Gas Sites May be a Significant yet Overlooked Seasonal Source of Methane Emissions. Environ Sci Technol Lett 11:948\u0026ndash;953\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLyon DR et al (2016) Aerial Surveys of Elevated Hydrocarbon Emissions from Oil and Gas Production Sites. Environ Sci Technol 50:4877\u0026ndash;4886\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZavala-Araiza D et al (2015) Toward a Functional Definition of Methane Super-Emitters: Application to Natural Gas Production Sites. Environ Sci Technol 49:8167\u0026ndash;8174\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFesta-Bianchet SA, Milani ZR, Johnson MR (2023) Methane venting from uncontrolled production storage tanks at conventional oil wells\u0026mdash;Temporal variability, root causes, and implications for measurement. Elem Sci Anth 11:00053\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Y et al (2024) Reconciling ultra-emitter detections from two aerial hyperspectral imaging surveys in the Permian Basin. Preprint at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.31223/X5G68V\u003c/span\u003e\u003cspan address=\"10.31223/X5G68V\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoscioli JR et al (2015) Measurements of methane emissions from natural gas gathering facilities and processing plants: measurement methods. Atmospheric Meas Tech 8:2017\u0026ndash;2035\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFox TA, Barchyn TE, Risk D, Ravikumar AP, Hugenholtz CH (2019) A review of close-range and screening technologies for mitigating fugitive methane emissions in upstream oil and gas. Environ Res Lett 14:053002\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCardoso-Salda\u0026ntilde;a FJ (2023) Tiered Leak Detection and Repair Programs at Simulated Oil and Gas Production Facilities: Increasing Emission Reduction by Targeting High-Emitting Sources. Environ Sci Technol 57:7382\u0026ndash;7390\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Methane emissions, Causal Analysis, Predictive Analytics, Super-emitters, Risk: Mitigation, Failure Mode and Effects Analysis (FMEA), Fault Tree Analysis (FTA), Appalachian Basin, Escalation ratio","lastPublishedDoi":"10.21203/rs.3.rs-7340966/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7340966/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAddressing methane emissions from the oil and gas supply chain has emerged as a key near-term mitigation target. The past decade of research has improved our understanding of methane emissions, with a primary focus on quantifying emissions without describing their underlying causal mechanisms. In this work, we integrate source-specific methane emissions measurement from multiple large-area aerial surveys with source-tracked cause analyses to identify and analyze causal mechanisms that underlie observed emission patterns. Overall, 53% of all observed emissions can be attributed to specific causal categories, with the rest comprising normal operational emissions. While abnormal tank emissions were the most common cause, unloading events exhibited the highest average emission rate. Importantly, we find that large release events are not driven by fundamentally different causal mechanisms than those of small emitters, indicating that escalation due to specific operational conditions, rather than fundamentally distinct causes, drives high-magnitude emissions. In addition, we observe statistically significant quarterly and inter-operator variability in the prevalence of different causal categories, reinforcing the need for adaptive, operator-specific mitigation strategies. These findings support a shift in methane mitigation from generalized leak detection with one-size-fits-all solutions toward risk-targeted, process-informed mitigation.\u003c/p\u003e","manuscriptTitle":"Why Do Methane Emissions Occur? Towards a Predictive Framework for Risk-Targeted Mitigation in Oil and Gas Operations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-04 03:29:24","doi":"10.21203/rs.3.rs-7340966/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"98be7fb6-b912-4a88-8767-f4550b7de315","owner":[],"postedDate":"September 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":54092940,"name":"Earth and environmental sciences/Environmental sciences/Environmental impact"},{"id":54092941,"name":"Scientific community and society/Energy and society/Energy management"}],"tags":[],"updatedAt":"2026-04-20T19:55:52+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-04 03:29:24","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7340966","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7340966","identity":"rs-7340966","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}