IEEE Geoscience and Remote Sensing Magazine - March 2021 - 271

the ancillary information was observed, providing an indication of the reasonableness of these estimates and the possibility of their integration.
More recently, Zhang et al. [138] proposed an alternate
method for deriving gas-flared volumes, based directly on
the nitrogen dioxide (NO2) and sulfur dioxide (SO2) emission rates provided by the OMI [146], a hyperspectral sensor
used for detecting and tracking urban-scale pollution sources. Using their equations requires knowledge of chemical
parameters [147]-[149], thus making them not easily applicable everywhere and adding a further source of uncertainty
to BCM computations. In this case, the method also resulted
in estimations that were in good agreement with both the
flared volumes reported by official regional databases and
the OLS- and VIIRS-based outcomes [29], [83].
In summary, although more sophisticated satellite products used for measuring GF activity could be developed/
used in the future, the key factor for improving the degree
of accuracy/consistency of satellite findings remains the
availability of a reliable ground truth.

measurements require an important preprocessing step
before being used. Coupling NO2 with SO2 OMI retrievals, Zhang et al. [138] found a better response of SO2 over
oceans, supporting the outcomes of previous studies [156],
[161]. Overall, satellite products are well correlated among
themselves [81], [159], with a tendency for underestimation
when compared with official
data sets (e.g., EDGAR) [138],
[161]. Especially when used to
investigate high-latitude arOBSERVATIONS ARE
eas, OMI retrievals may lead
LIMITED TO THE NIGHTTIME
to high levels of uncertainty
TO MINIMIZE THE EFFECTS
for the solar/satellite geomeOF SOLAR IRRADIATION,
tries, surface conditions (e.g.,
WHICH LEAD TO THE
snow, ice, and polar nights),
INCORRECT IDENTIFICATION
and the cloudiness affecting
OF FLARES CAUSED, FOR
those regions. Also, for modEXAMPLE, BY SUN GLINT
est production levels, OMIs
FROM HIGH-REFLECTANCE
can fail to recognize a signal
that is below the instrumenSURFACES.
tal detection limits [81], [159].
The OMI successor, TROPOMI, launched on the ESA's
Sentinel-5P satellite in October 2017, seems to overcome
these drawbacks, offering data with increasing levels of accuracy, spatial coverage, and sampling [81], [156], [162], as
recently demonstrated in [138] and [163]. Besides, the daily
global data from TROPOMI will be supplemented by even
higher ground-resolution hourly observations from new
sensors aboard recent [164] and future [165], [166] generations of geostationary satellites dedicated to tracking air
pollution [167], [168].
Among the " indirect " methods, the first ones developed
by the authors in [82] and [89] enabled the computing of

METHODS USED FOR ESTIMATING FLARING
EMISSIONS
The only current data sets that explicitly include GF emissions are the EDGAR [150] and ECLIPSE projects [13], [151],
[152], both of which use DMSP BCM estimates [83], [104]
for their assessments [152]-[154]. Nevertheless, significant
discrepancies between the two inventories as well as with
other global estimates have been noted, indicating that such
modeled data have to be used with care. As an example,
ECLIPSE concluded that the BC emissions resulting from
GF in 2010 were 228 [68] and 210 Gg [152]. At the other extreme, the authors in [142] reported a total global BC flaring
emission of 20 ± 6 Gg (one order of magnitude lower), which
is further reduced to 1.35 Gg if the EDGAR estimate is considered.
In this poor and confused framework,
TABLE 5. A SUMMARY OF THE REVIEWED METHODS USED FOR GF EMISSIONS
some authors developed satellite methods
ESTIMATION.
aimed at estimating flaring emissions [81],
[138], [139], [155]-[159]. They belong to two
KEY
SATELLITE
STUDY
YEAR OF
CATEGORY PARAMETER DATA
PERIOD
REFERENCES PUBLICATION
main categories (Table 5): 1) methods based
VIIRS
-
[82]
2013
I
CH4, CO2
on the direct measurements of oil- and gasI
CH4, CO2
VIIRS
January 2013- [89]
2015
related emissions from those satellite sensors
June 2014
able to infer the concentration of air pollutOMI, MODIS 2004-2015
[81]
2016
D
NO2, AOD
ants into the vertical profile of the atmosphere
I
BC
OLS
1994-2012
[69]
2016
[160] (labeled as " Category D " in Table 5) and
I
CO, CO2, NO, VIIRS
June/July
[130]
2017
2) the methods used to exploit flare pixel
NO2, SO2
2014, 2015
features (e.g., temperature, RH, and flared
OMI
2005-2015
[159]
2017
D
NO2
volumes) to retrieve several air pollutants as
I
CO2, CH4,
OLS
1995-2010
[139]
2019
NMVOCs,
well as GHGs (i.e., CO2 and CH4) emissions
CO, OC, BC,
(labeled as " Category I " in Table 5).
SO2, NOx
The OMI-NO2 observations showed great
D
NO2, SO2,
OMI
2005-2017
[138]
2019
potential to capture GF emissions for both
I
BC
SLSTR
2017
[115]
2019
offshore [81] and onshore [159] hot spots. BeCategories-I: indirect estimation method; D-direct estimation method. NMVOCs: non-CH 4 volatile
cause of the seasonality inherent in OMI-NO2
organic compounds.
data sets, as discussed in [81] and [159], such
MARCH 2021

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

271
IEEE Geoscience and Remote Sensing Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2021
