Quantifying uncertainties from mobile-laboratory-derived emissions of well pads using inverse Gaussian methods
<p>Mobile laboratory measurements provide information on the distribution of CH<sub>4</sub> emissions from point sources such as oil and gas wells, but uncertainties are poorly constrained or justified. Sources of uncertainty and bias in ground-based Gaussian-derived emissions e...
Main Authors: | , , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Copernicus Publications
2018-10-01
|
Series: | Atmospheric Chemistry and Physics |
Online Access: | https://www.atmos-chem-phys.net/18/15145/2018/acp-18-15145-2018.pdf |
Summary: | <p>Mobile laboratory measurements provide information on the distribution of
CH<sub>4</sub> emissions from point sources such as oil and gas wells, but
uncertainties are poorly constrained or justified. Sources of uncertainty and
bias in ground-based Gaussian-derived emissions estimates from a mobile
platform were analyzed in a combined field and modeling study. In a field
campaign where 1009 natural gas sites in Pennsylvania were sampled, a
hierarchical measurement strategy was implemented with increasing complexity.
Of these sites, ∼ 93 % were sampled with an average of 2 transects
in < 5 min (standard sampling), ∼ 5 % were sampled with an average
of 10 transects in < 15 min (replicate sampling) and ∼ 2 % were
sampled with an average of 20 transects in 15–60 min. For sites sampled
with 20 transects, a tower was simultaneously deployed to measure
high-frequency meteorological data (intensive sampling). Five of the
intensive sampling sites were modeled using large eddy simulation (LES) to
reproduce CH<sub>4</sub> concentrations in a turbulent environment. The LES
output and LES-derived emission estimates were used to compare with the results
of a standard Gaussian approach. The LES and Gaussian-derived emission rates
agreed within a factor of 2 in all except one case; the average difference
was 25 %. A controlled release was also used to investigate sources of
bias in either technique. The Gaussian method agreed with the release rate more
closely than the LES, underlining the importance of inputs as sources of
uncertainty for the LES. The LES was also used as a virtual experiment to
determine an optimum number of repeat transects and spacing needed to produce
representative statistics. Approximately 10 repeat transects spaced at least
1 min apart are required to produce statistics similar to the observed
variability over the entire LES simulation period of 30 min. Sources of
uncertainty from source location, wind speed, background concentration and
atmospheric stability were also analyzed. The largest contribution to the
total uncertainty was from atmospheric variability; this is caused by
insufficient averaging of turbulent variables in the atmosphere (also known
as random errors). Atmospheric variability was quantified by repeat
measurements at individual sites under relatively constant conditions.
Accurate quantification of atmospheric variability provides a reasonable
estimate of the lower bound for emission uncertainty. The uncertainty bounds
calculated for this work for sites with > 50 ppb enhancements were
0.05–6.5<i>q</i> (where <i>q</i> is the emission rate) for single-transect sites and
0.5–2.7<i>q</i> for sites with 10+ transects. More transects allow a mean
emission rate to be calculated with better precision. It is recommended that
future mobile monitoring schemes quantify atmospheric variability, and
attempt to minimize it, under representative conditions to accurately
estimate emission uncertainty. These recommendations are general to
mobile-laboratory-derived emissions from other sources that can be treated as point
sources.</p> |
---|---|
ISSN: | 1680-7316 1680-7324 |