Automatic Fire Detection Success stories
&
unwanted "river of fire" /
"mangrove of fire" from Proba-V
VisioTerra is involved in Proba-V MEP (Mission Exploitation Platform). VisioTerra is developing enhanced processing and visualization tools with two services of automatic detection for fires and burnt areas.
Success stories
By computing the difference between the bands of the image of the day and "the rolling mean" of these bands, one define a change index. This index together with other criteria are analysed in a decision tree to detect fires or burnt areas. This automatic classification has been successfully applied for the detection of fires in Spain, Guinea and Sudan.
Fig.1: Fires observed on 7 August 2016 in Portugal. 2D_view3D_view
Fig.2: Fires observed on 5 February 2017 in Guinea. 2D_view3D_view
Scope of these preliminariy studies is to set up a multi-sensor alarm system for civil protection, farmers, NGOs, citizen... After a qualification phase, e-mail or SMS will be sent to the people for events occuring within their area of interest.
Rivers of fire
Not having a thermal band, Proba-V cannot unambiguously characterise fires. In particular a high backscattering coefficient may occur in the SWIR (Short-Wave Infrared) band (1564-1634 nm wavelength) that are not necessarely due to fires.
Figures below show for example the Senegal river seen in Guinea from Proba-V, Sentinel-2 and Landsat-7 instruments.
One may notice that "fire rivers" are only observed by Proba-V.
Fig.4: Senegal "fire river" seen from Proba-V (left), Sentinel-2 (middle) and Landsat-7 (right). 2D_animation
As shown in figure below, the high SWIR backscatter is due to specular reflection of Sun over the Senegal River by acting like a mirror.
The Solar irradiance vector (red) generates the specular reflection vector (green) as its symmetrical with respect to the surface normal (yellow).
This specular reflection defect could be avoided by acquiring the scene earlier in the morning (before 10 AM local time) or by reducing the FOV (Field Of View) and therefore on a smaller swath.
Fig.5: Reconstruction of the observation geometry of Proba-V on 22 April 2017 at 11:41:23 GMT. KML
Mangroves of fire
The same phenomenon is surprinsingly also observed in the border of mangroves in the image below observed by Proba-V on 11 April 2016.
As shown in figure below, the high SWIR backscatter is certainly due to the specular reflection of the Sun on moist surfaces acting as a mirror.
This rare phenomenon could be due to the small size of the shrubs constituting this mangrove swamp, itself due to high tidal range (5.66 m in Bissau) or a possible flood (?).
This interpretation is not absolutely certain and your contribution is welcome !
In this figure, the Solar irradiance vector (red) generates the specular reflection vector (green) as its symmetrical with respect to the surface normal (yellow).
Fig.7: Reconstruction of the observation geometry of Proba-V on 11 April 2016 at 12:01:20 GMT. KML
