IEEE Geoscience and Remote Sensing Magazine - March 2021 - 270

when using a polynomial regression with a cubic function of
temperature and a linear function of source area to better predict BCMs than linear regression. The main drawback of using
regression approaches is that the model calibration is possible
only when official data on BCMs are available.
FLARED VOLUMES AND FLARING EMISSIONS
With reference to the satellite parameter, it has been
RS techniques revealed their capability for estimating the
defined using both night light [77], [83], [104] and/or IR
flaring amount directly and/or indirectly via multispectral
signals [29], [82], [87], [95], [96], [110], [115], [145] meadata [82], [130], providing a twofold advantage. On the
sured as radiances or brightness temperatures in pixels conone hand, they can supplement/complement the analyses
taining the gas flare(s), whose location has to be known a
based on in situ data and, on the other hand, fill gaps in the
priori. Several metrics (e.g., FRP and RH, described in the
absence of routine monitor " Methods for Gas Flares Characterization " section), aging efforts [131]-[135]. Moregregated on both an annual and monthly scale, have been
over, as high‐density satellite
FOR PASSIVE OPTICAL
used to quantify GF activity.
observations become more
SENSORS, CLOUDS MAY
The monthly index was developed using the most recent
widely available, multipollutHAMPER TARGET
satellite sensors, such as VIIRS [29], [82] and SLSTR [96],
ant satellite data analysis will
[115], which allow for characterizing the pixel behavior
be an increasingly useful apDETECTION, AND THE
with more detail thanks to their better spectral resolution
proach
to
independently
track
INTEGRATION OF MULTIPLE
(one or two bands in the SWIR region) and spatial resoemission
activity
and
evaluate
SENSORS MAY IMPROVE
lution (reduced to 500 m when compared to the 1 km of
related policy interventions
SAMPLING OPPORTUNITIES.
MODIS). The coefficients from the resulting linear regression
[136]-[138]. Nevertheless, the
provided by [87], [95], and [110] are site-specific, namely, the
main challenge in using satellite products for estimating GF
Niger Delta region, and not applicable to other locations. On
activity is their limited access to official activity records
one side, this may explain the high accuracy of their correas well as their lack (or the poor availability) of field measponding BCM estimates when compared with independent
surements of representative emission factors (EFs), thus
sources; however, on the other side, it requires a recalibration
making their calibration/validation difficult [132], [139],
of the prediction model when applied to other zones. For
[140], [142]-[145].
global-scale models, like the ones developed by the authors
in [29] and [96], high-biased BCM values may be derived
METHODS USED FOR ESTIMATING FLARED VOLUMES
when applied at the country level, as verified by [29].
To the best of our knowledge, satellite-based methods (sumUsing the linear equation calibrated on a monthly sum
marized in Table 4) are the only data source to globally asof RH estimates generated unrealistically high ﬂared gas
sess the amount of gas-flared volumes. In most of the cases, a
volumes because of a logarithmic growth of RH against
linear regression model coupling a satellite parameter, chosen
known BCMs. This nonlinear relationship was introduced
as representative of flaring sites' emissive power (e.g., FRP and
by applying an exponent to the source area during the calRH) and the known data on gas-flared volumes, is applied.
culation of a modified RH estimate (RH') for each flare. Its
In addition, Franklin et al. [145] explored the performances
value was tuned to achieve the best possible R 2 between the
reported flare volumes and RH'.
In general, the potential causes of inconsistent
estimates are due to both satellite products
TABLE 4. A SUMMARY OF THE REVIEWED METHODS USED FOR GAS-FLARED
and
official
BCM values. There are surely errors
VOLUME COMPUTATION.
in the reported GF data used in the calibration.
SATELLITE STUDY
YEAR OF
Such information, when indicated by different
KEY PARAMETER
REFERENCES PUBLICATION
DATA
PERIOD
sources for the same country, often differs, with
Sum of lights index
OLS
1995-2008
[77], [104]
[2007],
[2009]
no clear way to determine the best reference
Monthly sum of RH
VIIRS
2012-2017
[82], [29]
[2013], [2016]
value. In general, it is difficult to ascertain this
MODIS
2009
[110]
[2014]
Annual FRPMIR
uncertainty, as the errors introduced by the ofSquare of the sum of radi- MODIS
2000-2014
[95]
[2015]
ficial data are unavailable [77], [83], [96], [104].
ances by their standard
Satellite data also introduce noise in the results
deviation
for the possible misidentification of flares and
Annual BTE
MODIS
2000-2016
[87]
[2018]
noncontinuous sampling (observations are reMonthly sum of RH
SLSTR
2017
[115]
[2019]
stricted to nighttime conditions and screened
ATSR,
1993-2012, [96]
[2019]
Monthly sums of FRP SWIR
SLSTR
2017
for factors such as sunlight, moonlight, clouds,
Temperatures and areas
VIIRS
2012-2016
[145]
[2019]
and view angle). Nevertheless, a generally good
OMI
2005-2017
[138]
[2019]
NO2-SO2 emission rates
fit between the annual flared gas volumes reported by the methods collected in Table 4 and
is possible to characterize gas flares consistently using estimates of the radiant power provided by both categories of
the methods described [82], [111].

270

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2021
IEEE Geoscience and Remote Sensing Magazine - March 2021

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