Fig 1.
Flow chart of the steps used to create maps of leak indications and survey coverage.
The parenthetical numbers denote the section of the paper that describes the data processing step. Topics marked with an asterisk (*) denote parts of the algorithm that have been modified since v1.0. Icons are reprinted from The Noun Project under a CC BY license, with permission from icon authors. Individual author attributions are given in the references [7–19].
Table 1.
Summary of the data used by our algorithm and the data products it produces.
Fig 2.
An example of data processing from the CH4 survey.
The left panel shows the survey data after quality control. The right panel displays the elevated readings, observed peaks (blue markers), and leak indication (black pipe marker) information within the black rectangle marked in the left panel. The leak icon is reprinted from The Noun Project under a CC BY license, with permission from icon authors. Individual author attribution [19] is given in the bibliography. Maps were created using Leaflet for R [21].
Fig 3.
Histogram of the difference in location error.
The difference in location error is defined as unweighted error–weighted error. Positive values of this difference indicate that the weighted location estimate was closer to the actual leak. The mean difference was 0.38 m, and the weighted location was closer for 54% of the sites (26/48).
Fig 4.
Relationship between known emission rate and excess CH4 from controlled release experiments.
The axes are on the ln-ln scale.
Fig 5.
Histogram of the distance between the survey vehicle and NG leak expression points in Boston.
In this histogram, there were 468 passes of 19 known leaks. The typical distance between the car and leaks is 10–30 meters. Occasionally, the survey instruments detect leaks that are greater than 60 m from the car.
Table 2.
The number of road miles driven by survey effort in two cities.
Fig 6.
An example of a flute area (blue line).
Segments of pipe with multiple leaks can cause elevated CH4 levels to be detected over a large spatial extent, as illustrated by the blue polygon encircling multiple series of elevated readings. The scale is the log of the CH4 enhancement (log10 ppm over baseline). Map created using Leaflet for R [21].
Table 3.
A comparison of survey results between algorithm v1.0 and v2.0.
Fig 7.
Histograms of estimated sizes of 6125 leak indications developed from surveying in 15 cities.
The left figure shows a histogram of estimated emission rates truncated at 20 L/min. The right figure shows a histogram emission rates from all 6125 leak indications on the log10 scale.
Fig 8.
Cumulative emissions curve from the estimated sizes of 6125 leak indications.
The cumulative emissions curve indicates that largest 20% of leaks account for approximately 54% of total emissions.