Early fire detection and monitoring with an exportable algorithm

Many satellite-based methods for fire detection and monitoring have been developed to exploit data acquired by sensors. Their relatively low temporal resolution (hours) is, however, decidedly not adequate for detecting short-living events or fires characterised by a marked diurnal cycle and rapid evolution times.


Space technology against fires

In some North America states and in Europe, detecting fires as soon as possible is a priority to avoid their expansion and to mitigate the negative impacts they may have on citizens, infrastructures, and valuable vegetation areas.

Geostationary satellites have very high temporal resolution of 30 to 2.5 min and could, in principle, be more suitable for providing timely alerts and facilitating possible mitigation actions. However, such short revisit times are coupled with spatial resolutions of 3–5 km. This could represent a significant limitation for small fire detection and precise localization. However, unlike polar orbiting systems, the geostationary attitude assures very stable observational conditions at the pixel level.

This work describes in detail the Robust Satellite Techniques for FIRES detection and monitoring (RST-FIRES), a multi-temporal change-detection technique, and its application to the data of Spinning Enhanced Visible and InfraRed.

READ ALSO: Monitoring wildfires with spatial technology has limitations

Active fires began to be monitored at the global scale and on a regular basis by means of polar orbiting satellite sensors such as the Advanced Very High Resolution Radiometer (AVHRR) onboard the National Oceanic and Atmospheric Administration (NOAA) platforms within the World Fire Web (WFW) initiative of the JRC Space Application Institute; the Along Track Scanning Radiometer (ATSR) and the Advanced-ATSR (AATSR) onboard European Remote Sensing (ERS) satellites, which were used to produce the World ATSR Fire Atlas of the European Space Agency.

Presently, the most used and consolidated product, MOD14/MYD14, utilises Terra/Aqua-MODIS data based. This product forms the basis of several online tools such as the Fire Information for Resource Management System, Advanced Fire Information System, EFFIS, and DLRMODIS Fire Service.

MSG-SEVIRI image through which RST-FIRES can detect thermal anomalies. Source: Filizzola et al. (2017).

Timely detection, particularly of starting fires, could enable Forest Fire and Civil Protection Services to better manage their contra actions, minimizing the response time as well as the instrumental and human resources required to combat a fire when it is still at its first stage. In addition, in the presence of several contemporary events, prompt detection and quasi-continuous monitoring would help decision makers to optimise the management and deployment of exiguous ground and aerial resources.

Actually, the capability of detecting fires depends on different factors such as the fire’s burning temperature, its flaming area, and the sub-pixel fire position with respect to the SEVIRI instantaneous field of view

Everything is reduced to algorithms

Most of the algorithms developed for SEVIRI are derived from the methods applied to polar satellite data; only a few actually exploit the very high temporal resolution by using multi-temporal change detection analyses.

The exportation of fixed-threshold algorithms over SEVIRI is not immediate because the threshold values used for a specific polar satellite need to be modified accordingly.

In this paper, we describe the general multi-temporal RST-FIRES technique and its specific application to the MSG-SEVIRI data. The results of four years of experimentations, conducted by systematically checking ground hotspots detected by RST-FIRES, are also presented.

The Robust Satellite Techniques for FIRES detection and monitoring (RST-FIRES) algorithm is a specific application of the general Robust Satellite Technique approach to fire detection. The RST is based on a multi-temporal analysis of collocated satellite images and an automatic change-detection scheme. According to the RST approach, a signal may be considered ‘anomalous’ only when it statistically deviates significantly from its behaviour in the normal conditions of the specific place and time of observation.

Relative signal variations rather than absolute values are considered to detect anomalous values. As a consequence, a very intense signal does not generate anomalies if it represents a normal or frequently recurring condition in a specific site under specific observational conditions such as in the case of exposed rocks in summertime or highly reflecting fixed surfaces during particular hours of the day.

Moreover, the RST is based only on the analysis of satellite data with no need for additional information that may be difficult to find or that which is not available in every location. An example is the case of the Lasaponara et al. (2003) algorithm, which must be preliminarily trained on an independent catalogue of fires that previously occurred. In addition, no physical model and assumptions are behind the RST.

These reasons contribute to making the algorithms based on the RST approach more robust and more easily exportable to different geographic regions and climatic conditions as well as to different instrumental.

For some specific applications, several limitations were identified such as those related to the presence and spatial distribution of clouds across and also their excessive amount.

In fire-related applications a ‘thermal’ variable is used tipically as TMIR(x,y,t) which is the brightness temperature measured in the MIR channel of the sensor in correspondence to the ground resolution cell centred at (x,y) on the satellite image acquired at time t.

A fire is generally characterised by higher than normal MIR values so that the MIR(x,y,t) index is expected to be significantly greater than zero. High-intensity anomalies (i.e. high positive values of the MIR index) may be related to the presence of large/high temperature fires.

A time differential ALICE index may be used to identify the abrupt increase of theMIR signal at the start of a fire by considering the difference between two contiguous temporal slots. That index is particularly useful in case of geostationary satellite sensors.

Their very high temporal resolution enables detection of events at their very start, such as only 15 min for MSG-SEVIRI, even at very low extensions/absolute intensities owing to the very low value of σΔt(x,y).

A space differential ALICE index may be introduced mainly to increase the reliability, considering normal overheated conditions such as summer or particularly hot days, and the sensitivity to better highlight even small abnormal temperature increases.

Because it is based only on satellite data, it is quite straightforward to export RST-FIRES to different areas/conditions/fire regimes and different sensors onboard polar and geostationary platforms.

Tested in Italy

A) SEVIRI MIR channel of August 5, 2011 at 10:15 GMT. A square of four pixels (indicated by the arrow) has a TMIR greater than the surrounding pixels. B) Thermal anomalies are detected in that area by RST-FIRES and other independent fire detection products. Source: Filizzola et al. (2017).

RST-FIRES has been applied toMSG-SEVIRI since 2008 over Basilicata and Sicily, particularly in Palermo Province, which are among the most fire-affected areas in Italy.

Reference fields were computed on the basis of a long-term dataset of MSG-SEVIRI level 1.5 images acquired since 2004 during the 96 daily time slots from July to October. By this method, reference fields were computed on a monthly basis by series.

The use of recorded information enables a-posteriori validation in which the results depend strictly on the completeness and correctness of the catalogues, as well as on the minimum size of the recorded events that may vary in space and time according to the individual country’s policy.

Other differences could also exist between the fire databases compiled by national and local agencies. In Italy. On the whole, the result could be a significant difference in the number of recorded events.

The use of long time-series of satellite data is one of the key factors that make the RST-FIRES algorithm quite different from most of the other fire detection methods that use satellite data only in the spatial rather than the spatial-temporal dimension.

This enables characterization of ‘normal’ signals and detects ‘anomalies’ exclusively on the basis of satellite data.

One additional limitation of the RST-FIRES multi-temporal approach could be related to the short-term, changes in land cover such as recently burned areas, crop rotation, forest clearings, urban expansion, and solar panel installation, which are not immediately reflected into the reference fields that are updated on an annual basis.

READ ALSO: Modelling fire size of wildfires

 

 

A tool for take advantage

This study fully demonstrates the added value of the RST-FIRES technique for the early warning of fire events.

On the whole, RST-FIRES is shown to be 3 to 70 times more sensitive than all of the other considered SEVIRI-based products. This satisfactory result was not completely unexpected, considering the quite high false positive rate exhibited by RST-FIRES and the fact that other algorithms were designed to work in a wider area, from continental to the SEVIRI full disk.


Source: Filizzola, C., Corrado, R., Marchese, F., Mazzeo G., Paciello R., Pergola, N.,Tramutoli, V. RST-FIRES, an exportable algorithm for early-fire detection and monitoring: Description, implementation, and field validation in the case of the MSG-SEVIRI sensor. National Research Council (CNR), Institute of Methodologies for Environmental Analysis (IMAA), Potenza, Italy. University of Basilicata, School of Engineering (SI), Potenza, Italy.

Main photo: Example of a road event that is recorded in the SOR 2010 database. Although the active flame is so thin as to be virtually invisible, the presence of two small combustion spots is easily identifiable by smoke.

Leave a Reply