Sentinel-3 active fire detection and FRP product performance - Impact of scan angle and SLSTR middle infrared channel selection
Graphical abstract
Introduction
The Sea and Land Surface Temperature Radiometer (SLSTR) builds on the heritage of the Along Track Scanning Radiometer (ATSR, Delderfield et al., 1986) and operates on two concurrently orbiting European satellites - Sentinel-3A and -3B (S3A and S3B) launched in 2016 and 2018 respectively. SLSTR provides dual-viewing angle, highly accurate imaging radiometry in multiple channels spanning the visible to longwave infrared spectral regions (Llewellyn-Jones et al., 2001; Smith et al., 2020). The primary role of SLSTR is to provide the dual view, highly radiometrically accurate data required to generate very high quality, daily mapping of Earth's sea and land surface temperature (SST and LST; Coppo et al., 2010, Coppo et al., 2015).
The widely used ESA World Fire Atlas (WFA; Arino and Rosaz, 1999) was generated from the ATSR and Advanced ATSR (AATSR) data records, forerunner instruments to SLSTR. SLSTR was equipped with certain new design features that would further improve the detection and characterisation of active fires beyond that provided by (A)ATSR. Specifically, the SLSTRs wider swath and two middle infrared (MIR) bands (the ‘standard’ S7 channel and the high dynamic range ‘F1’ fire channel; Coppo et al., 2010, Coppo et al., 2015; Wooster et al., 2012) make SLSTR able to provide greatly enhanced active fire (AF) information compared to (A)ATSR. This capability is more akin to that of the MODIS sensors deployed on the Terra and Aqua satellites, whose data are used to respectively generate the very widely used MODIS MOD14 and MYD14 AF products (Giglio et al., 2003, Giglio et al., 2016). Beyond AF detection, which was already included in the ESA WFA (Arino and Rosaz, 1999) but which is conducted using a far more comprehensive algorithm in the case of SLSTR (Wooster et al., 2012), the data from S7 and F1 allow routine calculation of the fire radiative power (FRP) of each detected AF pixel (Xu et al., 2020). FRP equates to the rate of the radiant heat output from a fire, and acts as a means of estimating rates of fuel consumption and smoke emission (Wooster et al., 2005). The pre-launch SLSTR AF product algorithm (Wooster et al., 2012) was recently significantly updated now that a number of years of real SLSTR data are available (Xu et al., 2020). Based on an operational version of this algorithm, Level 2 night-time AF products are now produced from SLSTR near nadir-view scan Level 1 data. Since the Sentinel-3 satellites (currently S3A and S3B) have a local equatorial crossing time (10:00 h and 22:00 h) that is very close to that of Terra (10:30 h and 22:30 h), the Sentinel-3 AF products are expected to ultimately replace those generated from the MODIS sensor onboard Terra as that mission reaches its end of life (expected early 2020s). Near real-time (NRT) versions of the night-time Sentinel-3 AF Detection and FRP Products have been publicly available from EUMETSAT since March 2020 (https://www.eumetsat.int/S3-NRT-FRP) and the non-time critical (NTC) product versions generated using the most up-to date ancillary information (e.g. the atmospheric profiles used for FRP atmospheric correction; Xu et al., 2020) are made operationally available by ESA with a short delay at the Copernicus Open Access Hub (https://scihub.copernicus.eu/dhus/#/home). Full daytime versions of both Level 2 NRT and NTC products are expected to be released in 2021. The purpose of the current work is to detail some key aspects of the SLSTR Level 1 data that influence generation of the Level 2 AF products, and specifically to evaluate how choices between how data collected in the two different dynamic range middle infrared bands (S7 and F1) are combined can affect the characteristics of the final AF product data record. In work akin to that conducted by Freeborn et al. (2011) for the MODIS AF products, we also evaluate how the SLSTR AF products characteristics are affected by the wide variations in view zenith angle (VZA; out to 55°) found around the instruments near-nadir scan, and how this varies depending on how the data from the S7 and F1 bands are combined. This work is designed to inform choices made to maximise the performance of the operational Sentinel-3 SLSTR AF products and their future ability to be merged with those from MODIS Terra to provide the type of long-term data record suitable for climate related investigations.
Section snippets
Sentinel-3 SLSTR instrument scans
As introduced in Section 1, a key aspect of SLSTR is the collection of data in two scans of the Earth within a few minutes of one another but made under very different view zenith angle (VZA) conditions. These two near contemporaneous scans deliver the data required to very accurately account for atmospheric effects on the measured infrared brightness temperatures, facilitating the SST retrieval process in particular (Coppo et al., 2015). The first scan made of an Earth location by SLSTR is the
SLSTR observation modelling methodology
The SLSTR S7 and F1 pixel footprints shown in Fig. 3, Fig. 4 were generated using an observation simulator from STFC Rutherford Appleton Laboratory which is akin to that used to simulate ATSR-2 data by Godsalve (1995). The ground sampling distance of the footprint maps used within the observation simulator is 40 m, and an input surface-leaving radiance map representing the ground target whose SLSTR view is to be simulated is generated at the same 40 m resolution. Two different ground targets
Modelled impact of scan position on retrieved FRP
Building on the simulations made at the near-nadir view centre and edge of scan positions shown in Fig. 6, Fig. 8, results from all scan positions across the full SLSTR near nadir swath for both the 800 K sub-pixel (37.2 MW) artificial fire and the real (184 MW) boreal forest fire are shown in Fig. 9. At each scan position, the graphs show the ratio of the retrieved FRP compared to that retrieved at the scan centre, based on statistics derived for 28 neighbouring pixel sampling locations around
Summary and conclusions
The Sea and Land Surface Temperature Radiometer (SLSTR) operating onboard the Sentinel-3A and -3B satellites provides dual-view, global observations of the Earth every day in multiple VIS, MIR and -LWIR spectral channels. Observations from the SLSTR near nadir view scan are being used to generate a set of Level 2 Active Fire Detection and FRP Products based on an operational version of the algorithm presented in Xu et al. (2020). The S3 satellites equatorial crossing time of ~10:00 h and
Credit author statement
W.Xu: Data curation, Investigation, Methodology, Formal Analysis, Writing - original draft, Visualisation. M.J. Wooster: Conceptulisation, Funding Aquistion, Methodology, Project Administration, Resources, Supervision, Validation, Writing - review & Editing; Visualisation; E. Polehampton: Methodology, Software. R. Yemelyanova: Methodology, Software, Investigation. T. Zhang: Methodology, Software, Investigation.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported by NERC National Capability funding though the UK's National Centre for Earth Observation (NE/R016518/1) and by the European Space Agency FIDEX project (ESA Contract 4000122813/17/I-BG). The NERC Airborne Research Facility (NARF) and British Antarctic Survey are thanked for support in use of the twin otter aircraft. ACRI ST are thanked for processing of the Level 2 SLSTR data products used herein.
References (26)
- et al.
Multi-decade global gas flaring change inventoried using the ATSR-1, ATSR-2, AATSR and SLSTR data records
Remote Sens. Environ.
(2019) - et al.
Addressing the spatiotemporal sampling design of MODIS to provide estimates of the fire radiative energy emitted from Africa
Remote Sens. Environ.
(2011) - et al.
An enhanced contextual fire detection algorithm for MODIS
Remote Sens. Environ.
(2003) - et al.
The collection 6 MODIS active fire detection algorithm and fire products
Remote Sens. Environ.
(2016) - et al.
Global characterization of biomass-burning patterns using satellite measurements of fire radiative energy
Remote Sens. Environ.
(2008) - et al.
The new VIIRS 375 m active fire detection data product: algorithm description and initial assessment
Remote Sens. Environ.
(2014) - et al.
Fire radiative energy for quantitative study of biomass burning: derivation from the BIRD experimental satellite and comparison to MODIS fire products
Remote Sens. Environ.
(2003) - et al.
Sentinel-3 SLSTR active fire detection and FRP product: pre-launch algorithm development and performance evaluation using MODIS and ASTER datasets
Remote Sens. Environ.
(2012) - et al.
First study of Sentinel-3 SLSTR active fire detection and FRP retrieval: night-time algorithm enhancements and global intercomparison to MODIS and VIIRS AF products
Remote Sens. Environ.
(2020) - et al.
Approaches for synergistically exploiting VIIRS I-and M-band data in regional active fire detection and FRP assessment: a demonstration with respect to agricultural residue burning in eastern China
Remote Sens. Environ.
(2017)
1997 and 1998 world ATSR fire atlas using ERS-2 ATSR-2 data
SLSTR: a high accuracy dual scan temperature radiometer for sea and land surface monitoring from space
J. Mod. Opt.
Sea and land surface temperature radiometer on Sentinel-3
Cited by (11)
Global burned area mapping from Sentinel-3 Synergy and VIIRS active fires
2022, Remote Sensing of EnvironmentCitation Excerpt :A full daytime version was released in 2021, although for now only the night-time version is available as standard global, monthly summarised files (https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-fire-radiative-power?tab=overview, last accessed July 2022). The SLSTR sensor can complement MODIS or VIIRS active fire information (Xu et al., 2021). This paper presents the development of a global burned area algorithm based on SYN reflectance and VIIRS 375 m active fire products.
Hourly biomass burning emissions product from blended geostationary and polar-orbiting satellites for air quality forecasting applications
2022, Remote Sensing of EnvironmentSatellite remote sensing of active fires: History and current status, applications and future requirements
2021, Remote Sensing of EnvironmentCitation Excerpt :Contextual algorithms still remain the most commonly used AF detection method. Since their initial development (Flasse and Ceccato, 1996; Giglio et al., 1999; Kaufman et al., 1998), they have been applied to data from numerous LEO sensors, including VIIRS (Schroeder et al., 2014; Csiszar et al., 2014; Zhang et al., 2017), MODIS (Kaufman et al.,1998; Giglio et al., 2016), the BIRD Hot Spot Recognition System (HSRS, Zhukov et al., 2006), the TRMM Visible and Infrared Scanner (VIRS, Giglio et al., 2000), SLSTR (Wooster et al., 2012b; Xu et al., 2020), and Landsat (Schroeder et al., 2016; Kumar and Roy, 2018), and also GEO sensors such as those carried by Meteosat (e.g. Wooster et al., 2015; Amraoui et al., 2010; Di Biase and Laneve, 2018), GOES (Prins et al., 1998; Xu et al., 2010; Schmit et al., 2017; Xu et al., 2021a), Himawari (Xu et al., 2017; Wickramasinghe et al., 2016), and FengYun (Xu et al., 2010). The majority of AF detection algorithms are applied to single date imagery, with some adding basic temporal constraints to remove possible false alarms (e.g., Prins et al., 1998; Xu et al., 2010; Kumar and Roy, 2018).