IR3.9, Night Microphysics RGB, Convection RGB, Natural Colour RGB, Fire Temperature RGB, LSA SAF FRP Pixel
Drought conditions across Spain supported wildfire development in several regions.
Last Updated
10 July 2023
Published on
13 April 2023
By Ivan Smiljanic, Djordje Gencic, Carla Barroso and Jochen Kerkmann
A forest fire started on 23 March, and quickly developed into a very intense episode that raged through thousands of hectares of land, less than 100km north of Valencia, eastern Spain. Initially, the fire was not detected by Meteosat's SEVIRI instrument at 11:45 UTC that day, however, only 15 minutes later a strong signal was apparent — suggesting a fast intensification of initial burning (Figure 1).
SEVIRI infrared comparison
23 March 11:45 UTC
23 March 12:00 UTC
Figure 1: Comparing fire signals between two consecutive SEVIRI scans of Meteosat-10, 23 March 11:45 and 12:00 UTC
An hour after the first SEVIRI signal, the fire intensified enough to saturate the IR3.9 'fire channel' (saturation at 336K), seen in the animated imagery as flickering around the core of the most intense fires.
The high intensity of the fire was also manifested through a blinding effect, where artificial black lines appeared in the scanning direction of the SEVIRI instrument (Figure 2).
Figure 2: Blinding of SEVIRI instrument by high-intensity fires, seen as linear black feature in IR3.9 channel on 23 March 14:00 UTC
The animation in Figures 3-5 show an active wildfire in the Castellón region over several days, through a single IR3.9 channel or RGB products that utilise this channel.
Figure 3: Fire dynamics captured through IR3.9 channel (time step 15 min), 23 March 12:00 UTC–28 March 08:00 UTC
Figure 4: Fire dynamics captured through Night Microphysics RGB (time step 15 min.), 23 March 12:00 UTC–28 March 08:00 UTC.
Figure 5: Fire dynamics captured through Severe Convection RGB (time step 15 min.), 23 March 12:00 UTC–28 March 08:00 UTC
Due to the small particle size in respect to wavelength, smoke is almost invisible in the infra-red region of electromagnetic waves. For that reason none of the above animations show the smoke signal (or only slightly show it). Whereas, the Natural Colour RGB, and, to a degree, the Cloud Phase RGB, utilising solar channels, start to show smoke signals. Smoke is seen mostly in blue shades, since blue is the RGB component with the shortest wavelength in both products.
Figure 6: Natural Colour RGB showing smoke cloud (blue) and fire hot spot (red) during 23 March, 11:00–17:30 UTC
As well as the IR3.9 channel, the NIR2.25 and NIR1.6 channels are also sensitive to fires. Wien's displacement law (peak wavelength inversely proportional to radiating body temperature) suggests that these two channels can better detect fires with higher/highest temperatures, respectively, hence, also adding the fire temperature information. As well as showing area affected by fire, both the Cloud Phase RGB and (to a higher degree) the Fire Temperature RGB, qualitatively provide more data about the intensity of fires. The Cloud Phase RGB provides a more comprehensive view on the fire scene, revealing both surrounding cloud classification and, to a degree, smoke and fire temperature information, but, without the NIR3.9 channel in the RGB combination, it misses all the weaker/smaller fire sources. Fires are picked up by green to yellow, and smoke in vague blue shades. Burnt areas are also detectable through slightly different shades of cloud-free ground.
Cloud Phase RGB comparison
25 March 23:59 UTC
27 March 23:59 UTC
Figure 7: Cloud Phase RGB view on the Castellon wildfire on 25 and 27 March
The Fire Temperature RGB (Figure 8) can also provide important information. Its main purpose is to assess fire strength, by estimating its temperature based off the strength of the signal in three NIR channels — red: reflected component of IR3.9, green: NIR2.25, and blue: NIR1.6. Since the IR3.9 is the most sensitive to the smallest temperature changes, even from the sub-pixel fires, it is expected that even the small ones will be picked up and show as red in the RGB. Plus, as the temperature of the fire rises, it shows more in the green component, NIR2.25, displaying in RGB as yellow or orange. For the strongest/hottest fires, a significant signal also shows in the blue component, the NIR1.6, and the fire displays almost white in the RGB.
In the Figure 8, a comparison between SEVIRI IR3.9 (left) and the SLSTR Fire Temperature RGB (right) has been made. Note the difference in resolution where SEVIRI has 3km at Nadir, while the SLSTR Fire Temperature RGB consists of a mix of 500m and 1km channels. Thus, the fire only shows as four darker pixels in the SEVIRI IR3.9 channel, while the Fire Temperature RGB gives a much better idea of the shape of the fire and the most intensive spots — orange, or even yellow, meaning that fire has a significant contribution in both red and green components.
The SLSTR Fire Temperature RGB can be considered as a good proxy for future MTG FCI capabilities.
Figure 8: Comparison between MSG SEVIRI IR 3.9 um (left) and Sentinel-3A SLSTR Fire Temperature RGB (right), 25 March 10:43UTC
To quantitatively assess the fire intensity, we consulted some of the geophysical products that are based on the satellite data. Figure 9 shows the LSA SAF Fire Radiative Power (FRP) product over the Iberian Peninsula, in which fire intensity is expressed in MW units. The product reveals several pixels with high values of FRP estimates from 23–27 March. The most intense fires reached radiative power values of around 800MW on 23 March and 900MW on 27 March. Several other smaller fires in the wider drought-affected region could also be seen.
Figure 9: LSA SAF Fire Radiative Power (FRP) product over Iberian peninsula, 23–27 March