Funding Opportunity Flux Inversion Algorithms for Continuous Monitoring
The Collaboratory to Advance Methane Science (CAMS) is seeking responses to a funding opportunity to create a transparent, freely available, vendor agnostic model to quantify and localize methane emissions at oil and gas production sites. The work will help to improve our understanding of trends in emissions and our ability to predict emissions in the future. Improving the accuracy of inverse dispersion modeling to quantify emission rates and localize emission sources gives industry regulators, as well as operators, a powerful tool for evaluating bottom-up emissions. Importantly, this will allow policymakers and regulators to have higher confidence in emissions estimates using modeling techniques that can be continuously field tested and improved.
Responding entities should submit full proposals for projects by December 09, 2022. It is anticipated that one award will be made with funding ranging from $450,000-$500,000.
Primary Study Objectives
The primary objective of the funded work will be to develop a non-proprietary, open-source algorithm that utilizes methane concentration data detected by continuous monitoring point sensors. Training and validation of the model can be performed using a combination of simulated “data” and/or field data collected at controlled release testing sites.
This research will allow CAMS to build on past successes and continue to advance the science around methane emissions.
This will be accomplished by sponsoring and engaging in research targeted at identifying, understanding, and supporting the most viable technologies to ultimately reduce overall methane emissions.
Specific study objectives include to build, train and apply an inverse dispersion model to:
- Localize an emissions source at the equipment category level (e.g. storage tank, separator, compressor) with 90%+ confidence;
- Determine whether and under what conditions low-cost sensors can achieve reasonable quantification.
- Improve emission rate quantification with uncertainty comparable to the best available technology today, including specified boundary conditions in specified test environments and operating parameters (i.e. ±30% or better1);
- Determine emissions category (e.g. intermittent, continuous, routine, fugitive, etc.).
The study should be repeatable, publishable and allow opportunities for future work, preferably through publication of the model as open source. The model should perform with a high degree of computational efficiency / usability (should be able to report back on near real-time basis using minimal computational resources).
value chain. Point concentration sensors at fixed ground locations will provide continuous monitoring capabilities of emissions for the purposes of training, validation, and application of the model. Actual sensors used in the study will be provided by CAMS subject to availability and cost.
The successful project should build on previous analyses of emission estimates based on dispersion modeling downwind. AI / ML approaches have been shown to achieve higher accuracy and efficiency than other inverse modeling approaches in studies at larger scales and at small spatial scales for efficient source localization of indoor airborne contaminants.2 This should create a context to understand high quantification uncertainties and improve previously reported error bounds.3
Project proposals should include a discussion of the intended approach, criteria for sensor selection, and a strategy for data acquisition. The results of this effort will benefit:
- Regulators and policymakers concerned with methane emissions;
- Measurement technology vendors seeking to improve proprietary algorithms; and
- Operators and environmental organizations seeking to understand the accuracy and limitations of their data, take advantage of rapidly emerging sensor technologies, and democratize access to technology options outside of proprietary algorithms.
Inverse dispersion modeling is an approach that allows operator to estimate total methane emissions from point and area sources and temporal trends when measurements are made continuously. Inverse dispersion models look for answers to the question: What would source emission rates have to be to cause the observed increase in downwind concentrations?
The approach involves measurement of downwind methane concentrations with estimated or measured meteorological parameters (e.g. wind speed, wind direction and air turbulence) to back-calculate emissions rates using atmospheric dispersion models. An emissions rate is estimated by using meteorological parameters and atmospheric models to calculate how a plume would disperse downwind to achieve the measured concentration.
Methods combining wind information with an inverse plume model have been demonstrated on aircraft (Hirst et al., 2013), vehicles (Yacovitch et al., 2015, Rella et al., 2015), and unmanned aerial systems (Nathan et al., 2015), with data often collected at some distance downwind.
As sensor technologies have evolved, some continuous monitoring technologies perform well in detection (as demonstrated through METEC, EDF Methane Detectors Challenge, and Astra). Nevertheless, these technologies have fallen short of the goals for measurement and attribution, as demonstrated by the results seen in flux rates compared with a controlled release and, to a lesser extent, in attribution.
While ARPA-E MONITOR has focused more on the development of cheap, high-accuracy sensor technologies, there has been less of a focus on the development and improvement of inverse flux algorithms. As a result, emission estimates obtained using inverse dispersion modeling demonstrate uncertainty. Previous analysis of emission estimates based on dispersion modeling downwind reports error bounds of +117/−46 percent (Robertson et al., 2017). Another recent study by CIRES that implemented a HYSPLIT inverse dispersion model to estimate emissions found that uncertainties for inverse dispersion modeling in the range of 30-40% (Angevine 2020).
One explanation for the relatively high quantification uncertainties (Golston et al., 2018) associated with continuous methane monitoring during field campaigns is the complexity of the inverse flux algorithms employed by technology providers. Technology vendors employ proprietary plume inversion models or inverse flux algorithms based on site-specific data collected from operators and estimate the size of the leak.
- An open and transparent model to detect, localize and quantify emissions for short-range dispersion effects.
- Model(s) should include localization and emission rate quantification algorithms that use a machine learning approach.
- Source code should be made available for public use.
- A peer-reviewed manuscript describing the study design, modeling approach, methods and results.
- A report that summarizes findings and provides a gap assessment and next steps for improving accuracy quantification (e.g. data limitations, hardware / sensing limitations, physical limitations).<sup?4
Funding awarded under this announcement may be used by the project team to meet cost share requirements, where eligible, for Area of Interest 4 (AOI 4) in DOE-FOA-0002616.
Project proposals should indicate if the primary applicant intends to apply for funding under AOI 4 of the DOE FOA.
For applicants who plan to apply for funding under AOI 4 of the DOE FOA and are seeking to allocate funding awarded under this announcement toward cost share commitments required under AOI 4 of the DOE FOA, proposals must be accompanied by as much information about the cost sharing budget and resources as is necessary for internal evaluation of CAMS’ ability to fulfill the cost sharing obligation.
The solicitation is open to public and private entities. The study objectives can be accomplished with an assembled team that can include academics, national laboratories, environmental organizations, technology vendors, etc. A multi-disciplinary research team comprising two or more entities is strongly preferred. The ideal team will combine skills, experience and demonstrated expertise in:
- Working with a range of instrumentation including methane sensors and 3-D sonic anemometers
- AI/ML model development and application
- Atmospheric modeling, dispersion processes and inverse methods
- Field measurement campaigns
RFP Release Date: 09/27/2022
Deadline to submit proposals: 12/06/2022
Anticipated Award Date: 01/30/2023
Anticipated Project Start Date: 03/20/2023
Anticipated Period of Performance: 12 months
Submit proposals to: firstname.lastname@example.org
Available Funding: $450,000 to $500,000, no matching funds are required
Questions on the funding opportunity should be directed to:
CAMS Program Administrator
The solicitation is open to public and private entities. The successful contracting group should consist of a team of experts with data analysis and statistical experience and field design and execution experience at oil and gas operations.
Proposal should not exceed 20 pages in length and must include the following:
- Point of Contact (name, title, business address, phone, email)
- Executive Summary (project description, team members)
- Scope of Work (goals, objectives, technical approach and methods to be used, task level descriptions, deliverables and milestones)
- Budget (breakdown by task, labor, M&S, subcontractors, consultants). Contract will be time and materials based, labor rates should be included in proposal
- Team Qualifications (include relevant work and publications, similar projects, resource experience, resources and capabilities)
- One page public project summary (will be used if awarded)
The project will be awarded, contracted and overseen by the CAMS program administrator, GTI Energy. The project will engage with a technical committee and steering committee which will include representatives of CAMS sponsors, the PI’s study team and the Project Manager. The contractor will also have access to an independent review by the CAMS External Scientific Advisory Board.
Proposals will be evaluated based on the following criteria:
Scientific and Technical Merit (30%)
- Applicant’s approach to achieving the goals and objectives of the RFP.
- Overall clarity and completeness of the proposal, including the appropriateness, clarity, rationale and completeness of the technical approach and work scope.
- Understanding of technical/scientific problem, challenges, limitations of the current state of knowledge or technology relative to addressing the problem based on RFP response.
- The degree to which the proposed work is based on sound scientific and engineering principles.
Technical Approach and Understanding (30%)
- Thoroughness of the description of the proposed development approach and degree to which the proposed approach or methodology meets the stated objectives of the RFP.
- The degree to which the Applicant provides detail that clearly outlines the technical benefits, technical challenges, and feasibility of the proposed approach.
- The reasonableness of the project schedule to integrate all tasks/subtasks and achieve key project objectives as reflected by well-defined, quantifiable, and verifiable critical path milestones and key project decision points.
- Adequacy and completeness of the work scope and task level descriptions, including identification of project risks and strategies for mitigation of those risks.
Qualifications, Experience and Capabilities (20%)
- Demonstrated experience of the applicant and partnering organizations in the technology areas addressed in the application and in managing projects of similar size, scope, and complexity.
- Clarity of design and likely effectiveness of the project team, including subcontractors or partners, for successful completion of the proposed research.
- Capability and availability of proposed personnel, facilities, and equipment for the performance of defined project tasks and sub-tasks.
- Demonstrated experience managing in program management appropriate for the project.
Budget and Cost Effectiveness (20%)
- Reasonableness of proposed budget relative to the project goals, objectives, and tasks.
- Detailed breakdown of budget by labor, materials, equipment and travel.
1 Johnson et al, Blinded evaluation of airborne methane source detection using Bridger Photonics LiDAR, Remote Sensing of Environment, Vol 259, 2021, 112418, ISSN 0034-4257, https://doi.org/10.1016/j.rse.2021.112418.
2 Bryan Travis, Manvendra Dubey, Jeremy Sauer, Neural networks to locate and quantify fugitive natural gas leaks for a MIR detection system, Atmospheric Environment: X, Volume 8, Dec 2020, 100092, ISSN 2590-1621, https://doi.org/10.1016/j.aeaoa.2020.100092.
3 Previous reports have estimated at +117/−46 percent (Robertson et al., 2017).
4 Ideally the study should break down the components that affect quantification accuracy, including what can be improved through better sensors or better models, and what are the absolute limitations on accuracy that are determined by physics.