We performed 7.5 weeks of path-integrated concentration
measurements of
Measurements of greenhouse gases, especially
Quantification of
More recently, several other approaches have also been applied to city-scale
emissions. Aircraft mass balance measurements
(White et al., 1976; Ryerson et al., 2001) have been used to determine city
emissions (Mays et al., 2009; Heimburger et al., 2017). However, the use of an aircraft is costly
and labor intensive and, therefore, not suited to long-term continuous
measurements. Column measurements from the Total Carbon Column Observation
Network (TCCON) were used to calculate total South Coast Air Basin (SoCAB)
CO and
As an alternative to these approaches, horizontal, kilometer-scale,
open-path instruments could in principle be used to determine
Here we present the quantification of city
Figure 1 shows the measurement layout for an initial campaign to quantify
Measurement layout. The two measurement paths are shown by red (reference) and black (over-city) lines. The two weather stations that provided wind speed and direction data are given by the green diamonds. The colored circles are turning movement count (TMC) locations, which are used as a proxy for the traffic source locations. Both color and size represent the number of traffic counts at each location. Dominant wind directions for the campaign overall (aqua) and the test case days (purple for 22 October and blue for 25 October) are given by colored arrows.
The DCS system was located on the top
floor of the National Institute of Standards and Technology (NIST) building
in Boulder, Colorado. This instrument has been described previously
(Truong et al., 2016; Waxman et al., 2017). The light from the combs is split to
generate two combined dual-comb outputs, one of which is transmitted over
the reference path and one of which is transmitted over the city path (see
Fig. 1). Here, we transmit 2–10 mW of light spanning 1.561 to 1.656
Typical 32 s spectrum measured over the 2 km reference path.
In previous work (Waxman
et al., 2017), we confirmed the high precision and accuracy possible with
open-path DCS. Two DCS instruments, constructed by different teams, measured
atmospheric air over adjacent paths over a 2-week period. The retrieved
path-averaged gas concentrations agreed to better than 0.6 ppm (0.14 %)
for
The reference and over-city paths had different path lengths and, therefore,
used slightly different telescopes and launch powers. For the reference
path, 2 mW of dual-comb light is launched from a 5 cm (2 in.) homebuilt off-axis
telescope (Cossel et al., 2017; Waxman et al., 2017). The light travels to a
6.35 cm (2.5 in.) retroreflector located on a hilltop 1 km to the southwest of NIST and then
is reflected back to a detector that is co-located with the launch telescope
for a
The measured spectra are analyzed as described in Rieker
et al. (2014) and Waxman et al. (2017) at 32 s intervals. Briefly, we fit a
seventh-order polynomial and HITRAN data to the measured spectrum in
100 GHz (0.333 cm
The variations in the retrieved concentrations are due to statistical
uncertainty, systematic uncertainty (discussed above), and the true
variations in the gas concentrations. Figure 8 of Waxman et al. (2017) quantified the statistical uncertainty in terms of the Allan
deviation over the 2 km reference path for both
Statistical uncertainty as quantified by the Allan deviations for
Meteorological data including pressure, wind direction, and wind speed are obtained from meteorological stations located at NCAR-Mesa
and NCAR-Foothills (
We measure a subset of Boulder traffic, so we use the city traffic data to
determine the fraction covered by our footprint (see Fig. 1). Traffic data
from the City of Boulder are freely available at:
ART measures traffic at 18 major intersections in Boulder for 5 days (1 work week, Monday through Friday) every year in 1 h bins to create a
diurnal cycle. The traffic counts for 2016 are shown in Fig. 4. We use these
data to scale our selected measurement time periods to a full day as
discussed in Sect. 3.3.4. Note that there is only a 10 %–20 % “peak” in
traffic counts at the standard commuter times with generally high traffic
levels from 07:00 to
City-wide traffic counts from the Boulder arterial count program (ART), normalized to a peak of unity.
TMC measures the number of vehicles at 140 intersections in Boulder for 1 work day per year during the hours of 07:45–08:45, 12:00–13:00, and 16:45–17:45. One-third of each of these sites is measured every year. We have scaled the 2014 and 2015 data to 2016 traffic levels by using total vehicle mile values available from the City of Boulder. We approximate city vehicle emissions by using the TMC locations as our source locations with a source strength scaled based on the location's fractional traffic count.
All 7.5 weeks of DCS measurements of
A total of 7.5 weeks of dual-comb spectroscopy data for the reference path
(red) and the over-city path (black) smoothed to 5 min time intervals.
Enhancements in the over-city path relative to the reference path are
observed in
The diurnal cycle of
Diurnal cycle analysis. Data are the median of the full 7.5 weeks.
The diurnal cycle of the reference path
The diurnal cycle of the reference path
To select test case days to estimate the city emissions, we filter the
Footprint calculations and time series data for the 2 case study
days. Left column – Saturday, 22 October 2016 data; right column – Tuesday,
25 October 2016 data. Upper panels
In order to confirm that the reference path measured clean background air and the over-city path measured city emissions, we calculated footprints for the two test case time periods using the Stochastic Time-Inverted Lagrangian Transport (STILT-R) model (Fasoli et al., 2018). The input meteorology file consisted of a uniform wind field with wind data from the NCAR-Foothills lab, boundary layer height from the North American Regional Reanalysis (NARR), and uniform turbulent velocity variance calculated from the Pasquill stability class (determined from wind speed and solar insolation) from the ground up to the boundary layer. We also used hyper near-field scaling described in Fasoli et al. (2018). Average footprints for the two time periods are shown in Fig. 7. The footprint for the reference path covers undeveloped areas extending from the near foothills into the mountains. The footprint for the over-city path also has contributions from the same general mountain region. In addition, this path has sensitivity to an extended area within the city and, therefore, to a large fraction of the traffic emissions. Note that the open-path geometry leads to a much larger extended footprint for this path than would be the case for a single point sensor located at the same height within the city.
The variability in the reference
To convert from the measured enhancement to an emissions rate, we require a model that connects the source strength to the plume concentration. Since we do not have a high-resolution, spatially resolved inventory for Boulder similar to the Hestia model for Salt Lake City (Mitchell et al., 2018), we use the existing Boulder traffic inventory (see Sect. 2.3) in conjunction with a Gaussian plume model.
The standard Gaussian plume model that includes total reflection at the
Earth's surface is as follows (Seinfeld and Pandis, 2006):
We modify this equation in several ways: (1) since we measure the
column-integrated concentration over a finite beam path at an angle to the
wind direction, we integrate the plume concentration along this beam path
and then normalize to the length of the beam path; (2) we sum over the
emissions locations in the city that contribute emissions to our
measurements. Thus, our overall measurement equation is as follows:
To calculate Eq. (
The rotated Eq. (
Seven measured parameters factor into the emissions calculation of
Parameters used to calculate the emission rate from Eq. (
In addition, there are assumptions and possible uncertainties inherent to
the Gaussian plume model. First, the model does not include the effects of
buildings, trees, or other objects that could break up the plume between the
emissions location and the beam path. Second, we assume that all
Further, we ran plume calculations in STILT-R using both wind fields derived
from the local meteorological stations shown in Fig. 1 and using the North
American Mesoscale Forecast System (NAM,
There are a number of non-traffic sources of
We first consider power plants. There are two power generation facilities on
the Department of Commerce (DOC) campus located near the NIST building that
houses the dual-comb spectrometer: the site's Central Utilities Plant (CUP),
and the National Oceanic and Atmospheric Administration (NOAA) building's
boilers. To calculate their average
The University of Colorado also has a power plant that falls within the main
footprint associated with the over-city beam path, shown in Fig. 7, and
whose emissions are expected to intersect our over-city beam path. The EPA
Greenhouse Gas Reporting Program (GHGRP,
The large Valmont power station lies just outside the city limits to the east of Boulder; however, given its location and the dominant selected westerly wind, emissions from this source do not reach our beam paths. There are no other power generation facilities within the city that report to the GHGRP, so we make no further corrections based on power plants.
In addition, there are also likely diffuse emissions from residential and
commercial furnaces and water heaters that use natural gas. The City of
Boulder Community Greenhouse Gas Emissions Inventory reports 20 %
of the city emissions, or
Once leaf senescence has completed, neither plants nor soil respiration
contribute to
In order to compare with the city inventory, we scale our results to an
annual total. To do this, we use the hourly traffic data of Fig. 4 to scale
The scaling relies heavily on the traffic count data supplied by the city of
Boulder, which do not have an associated uncertainty value. A comparison
of these data over several years shows a typical 7 % statistical variation
at a given TMC location after removing a linear trend. We assume this
reflects day-to-day fluctuations in traffic. In addition, there will be
seasonal variations, which are not captured in the extrapolation from our 2 test case days to the annual emissions. Due to the lack of seasonal data for
Boulder traffic, we use the detailed Hestia traffic inventory for Salt Lake
City, UT, given in Fig. 2 of Mitchell et
al. (2018). These data show a variation of
Including the additional uncertainty on the scaling to annual emissions, we
estimate an annual emission rate of
The city vehicle emissions estimate comes from total vehicle miles traveled
based on data from the transportation department, miles per gallon inputs
from the EPA state inventory tool, and vehicle type distribution from the
Colorado Department of Public Health and the Environment (Kimberlee Rankin,
City of Boulder, personal communication). The City of Boulder estimates total
vehicle emissions of
In comparison, we estimate
Future improvements should include additional and different beam paths, selected based on prevailing wind directions. Our initial assumption that the mountain path would generally act as a reference path was incorrect since the prevailing daytime winds during this time of year are not out of the west but rather the southeast. An east–west running beam north of the city and one south of the city would allow us to utilize a larger fraction of the data as the predominant midday wind direction during the fall is out of the north to the northeast (see Fig. 1). Even longer beam paths would also interrogate a larger fraction of the city and measure a correspondingly larger fraction of the vehicle emissions. Vertically resolved data from, e.g., a series of stacked retroreflectors would better test the assumption of vertically dispersing Gaussian plumes.
Additionally, more extensive modeling to cover variable wind directions and speeds would allow the incorporation of a much larger fraction of the data than the 2 days selected here. An inversion-based model similar to Lauvaux et al. (2013) could potentially be applied to a small city like Boulder; however, this would depend heavily on the quality of the bottom-up emissions inventory used to generate the priors. Indeed, one of the major future improvements would be to generate a detailed Hestia inventory of Boulder, CO, similar to that generated for Salt Lake City, UT (Mitchell et al., 2018).
We demonstrate the use of an open-path dual frequency comb spectroscopy system for quantifying city emissions of carbon dioxide. We send light over two paths: a reference path that samples the concentration of gases entering the city from the west, and an over-city path that measures the concentrations of gases after the air mass has crossed approximately two-thirds of the city, including two major commuter arteries. The measured diurnal cycle shows a significant traffic-related enhancement in the carbon dioxide signal during weekdays in the over-city path compared to the reference path. We select 2 case study days with appropriate wind conditions and apply Gaussian plume modeling to estimate the total vehicular carbon emission. We then scale these results up to annual city-wide emissions using traffic data from the City of Boulder. We find overall traffic-related carbon emissions that are approximately 1.4 times greater than the city's bottom-up traffic emissions inventory but with an uncertainty that encompasses the city inventory estimate. Further improvements to this method should include improved design of reference and over-city paths and a more detailed inventory model for Boulder CO, which together should further reduce the overall uncertainty in the estimate.
As per NIST regulations, all data are archived at NIST and available upon request.
Equation (1) is the standard Gaussian plume equation as discussed in Sect. 3.3.2 (Seinfeld and Pandis, 2006). It is reproduced
here:
The DCS returns the average concentration along a line path. We denote
distance along this path by the variable
Rather than a single source at (
We assume that the point source emissions locations are 1 m above ground
(
EMW, KCC, IC, and NRN designed the experiment. WCS helped build the hardware for the open-path measurements. EMW, KCC, and GWT ran the experiment. FG wrote the processing code for the data analysis. EMW processed the data and did the Gaussian plume modeling. KCC did the STILT-R modeling. EMW, KCC, IC, and NRN cowrote the manuscript.
The authors declare that they have no conflict of interest.
We thank Kimberlee Rankin, Randall Rutsch, Bill Cowern, and Chris Hagelin from the City of Boulder for city inventory and traffic information; Anna Karion for assistance with STILT-R modeling; and Dave Plusquellic and Caroline Alden for assistance with the manuscript. This work was funded by Defense Advanced Research Program Agency DSO SCOUT program and James Whetstone and the NIST Special Programs Office. Eleanor M. Waxman and Kevin C. Cossel are partially supported by National Research Council postdoctoral fellowships.
This paper was edited by Ronald Cohen and reviewed by three anonymous referees.