IEEE Geoscience and Remote Sensing Magazine - March 2021 - 274

◗◗ Low-intensity and/or more intermittent flares will likely

be undetected if the sensor does not have suitable spectral bands and frequent revisit time.
◗◗ For passive optical sensors, clouds may hamper target
detection, and the integration of multiple sensors may
improve sampling opportunities. Furthermore, pixels
containing gas flares are frequently marked as clouds
(e.g., the VIIRS cloud mask [29]); in such cases, an adequate cloud detection procedure becomes a key factor
in effectively reducing false cloudiness.
Other drawbacks may depend on the inherent characteristics of the sensor/data used:
◗◗ Gas flares are point sources, and some are omitted due
to the sensor's too-coarse spatial resolution.
◗◗ The spectral resolution of satellite sensors is a determining factor that affects the system's capability to recognize gas flares: as widely explained in this article, the
SWIR band is more suitable for the detection of hightemperature flares, while the MWIR channel may furnish acceptable results for lower temperatures.
◗◗ When only MWIR bands are available, the flare-identification process can be limited by the level of saturation
of these channels [113], which usually affects AVHRR,
MODIS, and VIIRS (imagery band) sensors but not, for
example, SLSTR sensors, which are equipped with two
detection channels that feature saturation levels capable
of exceeding 500 K.
◗◗ Off-nadir acquisitions may introduce inaccuracy in the
measured signal. This is an issue that significantly impacts sensors like MODIS, which suffer from the " bow tie
effect " at the edge of the scan. In addition, the authors in
[29] demonstrated that higher
radiances are measured for
flares viewed at high VIIRS
THE AVAILABILITY OF
satellite zenith angles; this efGLOBAL OBSERVATIONS
fect is not as pronounced for
FROM MULTIPLE SENSORS
SLSTR, which have a narrower
GUARANTEED BY CURRENT
range of viewing angles [86].
AND FUTURE SATELLITE
Obviously, some of these
MISSIONS MAKES IT
lat te r d rawbac k s may be
POSSIBLE TO EXTEND AND
further highlighted or limited depending on the accuFURTHER IMPROVE THE
racy/sensitivity of the used apCONSISTENCY AND
proach/algorithm.
RELIABILITY OF GAS FLARE
Concer ning f lared volMONITORING.
umes and flaring emissions,
in addition to the aforementioned aspects, it should be
mentioned that the discrepancies noted among satellite
methods and current inventories depend on many factors
and are mainly linked to the differences in methodologies, EFs, and temporal and spatial scales. The limited access to official flaring-volume records and the lack of field
measurements of representative EFs do not allow one to
correctly calibrate the satellite estimates, generating differences among them. Nonetheless, even if the uncertainties
274

remain abundant, the developed techniques that combine
satellite data with ancillary data provide a further source for
improving the performance of air quality or climate models
in regions where flaring is important.
Finally, we believe that, to increase the availability and
reliability of information on GF and its emissions, future
works should be conducted with the following goals:
◗◗ Improve the detection opportunity: The use of a multisource
system that maximizes the chance to collect information on a global scale with the highest level of spatial
and temporal resolution will increase the amount of information gathered in determining whether a given gas
flare is active or not.
◗◗ Enhance the reliability of flared-volume quantification: Given
the highly uncertain nature of the reported flaring volumes and the limited amount of reliable information
(often at the single-flaring station level), a methodology
that is totally independent of the reported GF volumes
would be advantageous.
◗◗ Improve emission estimations: The use of information on
the completeness of combustion, together with thermodynamic considerations, would help in developing a
more realistic, dynamic system of EF determination; inverse modeling could be useful in these two last points.
◗◗ Develop a fully automated system: The development of an
operator-independent algorithm would allow for the inclusion of GF emissions in models that provide air quality information in near-real time
◗◗ Integrate the data acquired by different sensors for a full characterization of GF: This process would start from activity
detection and move toward the physical characterization of the flare (i.e., the size, temperature, and thermal
intensity) up to the evaluation of the amount and kind
of gas emission.
◗◗ Make further substantial efforts for using daytime images: Implementing this process would allow for a more comprehensive overview of the GF phenomena.
Forthcoming studies should also encourage the combination of multiscale, multisource systems for a better
understanding of the proliferation and status of OG platforms. This information would be very useful for decision
and policy makers at national and international levels in
an effort to accurately quantify and monitor reduction
and control strategies, which are also recommended by the
Sendai Framework for Disaster Risk Reduction 2015-2030
[172]. For this to happen, we need more awareness of the
impact of GF at the local and global levels. We see it as a
fundamental step toward the implementation of a common policy for GF management.
CONCLUSIONS
This article examined the literature pertaining to the analysis of GF with satellite-based observations from the 1970s
through July 2019. We classified the methodologies into
three main categories: those devoted to identifying gas
flares, those aimed at their characterization (temperature,
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
