Project Canary is committed to continuously improving our hardware and software solutions through rigorous testing and refinement to create the most accurate facility-level methane emissions monitoring system.
Not all solutions and models are created equally. We know many on the market are touting “quantification,” but what goes into those models isn’t always clear. In the interest of full transparency, we want to share what makes our model stand out. In the process, we hope you learn more about what exactly goes into developing an emissions quantification model, how to differentiate between them, and what questions you can ask.
At a high level, Project Canary’s current version of quantification is designed to localize and quantify total site emissions. That includes small intermittent emissions from pneumatic devices and fugitive emissions that persist over long periods of time. Contact us if you want to dig deeper.
Critical Components of Quantification
Continuous monitoring solutions (CMS) transform raw sensor measurements (e.g., ambient ppm readings, wind speed, and wind direction) into composite data that are more informative and useful to operators, such as if, where, when, and at what rate emissions occurred on the facility. Additionally, it translates concentrations into quantification, a normalized value for true emissions isolated from atmospheric effects such as wind.
Accordingly, the quality and precision of a CMS system depends on the following:
1. Input data from sensors: this is impacted by sensor specs (such as detection limit), data frequency, and sensor placement.
2. Solution analytics (‘quantification model’): Model quality is critical as the translation of raw data to quantification measurement is based on inferences made by solution analytics. For models to accurately capture emissions, it must account for atmospheric conditions (wind), obstacles, and emission sources.
In short, high-quality sensor data + sophisticated analytics = more accurate quantification.
Project Canary Quantification Approach
Project Canary quantifies total site emissions for more accurate carbon accounting, comparison to bottom-up estimates, and tracking toward improved emissions performance. We combine state-of-the art hardware with industry-leading software solutions to give you the most precise measurements.
Our quantification model can be broken down into 3 primary components:
- Post-processing of sensor readings: We take the individual sensor readings and isolate the effect of wind direction on concentration utilizing a machine learning model.
- Forward direction model: Then, we take the leak locations we gained from our high-resolution drone imagery and simulate emissions from every potential leak source.
- Inverse solver: Once we have those two pieces, we run an inverse solving algorithm to find the best selection of emissions that match what the sensors are measuring.
Here are some of the attributes that make our CMS solution and quantification model unique:
Input Data from Sensors
Sensor Detection Limits
The model’s accuracy comes from the sensors’ precision and the data being used to generate the wind direction vs. concentration relationship. We can detect and quantify sub 0.125 ppm concentration uplifts above background, which is only possible using a high-precision sensor. Specifically: Canary X sensors use near-IR tunable diode laser absorption technology to detect ≤ 0.125 ppm with a 60s average reporting rate.
Sensor Placement
We optimize sensor density and placement to ensure higher fidelity readings with our Site Planning tool. We pull wind data from our extensive anemometer network and simulate emissions to ensure we have the highest probability of detection and the ability to triangulate/localize emission sources.
Data Duration
This approach can be done over a very long period of time (10 days) or a very short period of time (10min), but the accuracy of the output depends on the amount of data entered. That is why we use anywhere from 3-10 days of data to quantify the site emissions.
Solution Analytics: Quantification Model
High-Resolution Drone Imagery for 3D Site Modeling
To capture an accurate physical representation of a site, we use drones with sub-foot precision and at sub-inch per pixel resolution. When constructing site quantification, we use a multi-step approach for sub-meter precision to reconcile drone data with the ground control point (GCP). We can capture the distance and height of all equipment and obstacles to increase the accuracy of source attribution and overall observed mass quantification. We can identify the equipment group on a pad rather than viewing it as a single large emission source.
Typical alternatives are publicly available 2D imagery from often out-of-date satellite imagery, which cannot account for 3D obstacles.
Best Fit Algorithm
Our inverse solver algorithm will run simulations across large datasets to isolate the effect that wind direction has on elevated emissions concentrations. It considers forward direction, concentration, and wind direction and can account for complex scenarios such as pooling and multiple sources leaking simultaneously.
Our solver can also determine if an emission is coming from an off-pad source by examining an elevated concentration vs. wind distribution for all 360 degrees around the pad. All off-pad emissions are excluded by default.
Continued Improvements for Simulation Complexity
Our forward direction model is evolving rapidly. Our v1 model started with a simple Gaussian plume model. Like most current simulation models, the Gaussian model does not capture blockage or downwash effects caused by large buildings or obstructions – a known deficiency when estimating total emissions on site.
Our current quantification v2.5 model uses a 3D plume to account for the buoyancy of methane. Plume rise is a combination of the gases’ momentum and buoyancy that causes the gas to rise above the ambient air. Incorporating the vertical axis into the plume increased the accuracy of our quantification model, especially at higher release rates.
We continue to enhance our model to manage the complexity of atmospheric conditions and site configuration. In our upcoming v3 model, we have added a Navier Stokes physics model to account for vorticity and obstructions. See our high-fidelity dispersion simulator for a visual demonstration.
Testing Validation
We validate our CMS system with extensive testing, including third-party validation.
Third Party Testing
We have participated in testing done by the Methane Emissions Technology Evaluation Center (METEC) at Colorado State University since 2021. We have released our 2023 results proactively here. We also participate in ad-hoc testing through Stanford and NGIF.
First Party Testing
We have done a year of control released testing to develop and validate our model. We also partnered with METEC to conduct testing on their facility site with conditions simulating that of real operations at a well pad.