Post-Fire assessment using Sentinel-2 images in French Mediterranean area - Office National des Forêts Yvon Duché, Jean-Luc Kicin, Benoît Reymond ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Post-Fire assessment using Sentinel-2 images in French Mediterranean area Office National des Forêts Yvon Duché, Jean-Luc Kicin, Benoît Reymond, Rémi Savazzi
ONF presentation ONF (National Forestry Board) is a State Public Body, under the joint supervision of the Ministries of Forestry and the Environment. ONF is responsible for forest management for both the state and local government, and has a public service remit to help the State and local authorities in protecting forests against fires. ONF has a specialized agency in the Midi-Mediterranean area to deal with wildfire issues that performs the following tasks: • Operational - monitoring, detection, first response, support to control activities • Development and maintenance work - roads, water points, areas cleared of undergrowth • Expertise support - hazard and fire mapping, fire risk management plans, equipment mapping … • Project management
Calculation method using Sentinel-2 data Within these missions, ONF establishes wildfire maps. Sentinel-2 images have been used from 2016 using differenced Normalized burnt Ratio (dNBR) calculation method. This method is based on the different spectral responses of Near InfraRed (NIR) and Shortwave Infrared bands (SWIR) for unburnt/burnt areas. Source : USDA Forest Service
Calculation method using Sentinel-2 data Example : Rognac fire, 10th of August 2016, 2 655 ha NBR = NBR pre fire NBR post fire Pre-fire image 3rd of August Post-fire image of 13th of August Sentinel 2A © ESA 2016 © CNES 2016 Sentinel 2A © ESA 2016 © CNES 2016 dNBR = NBR(pre-fire) - NBR(post-fire)
Fire outline mapping The dNBR calculation allows a quick fire outline mapping. Result of the dNBR calculation (raster format) Value > 0.1 No pixels shape generalizing Vector conversion
Fire outline mapping Accuracy of fire outline mapping has been validated on different kinds of situations. • Flat terrain • Steep terrain • Maximum outline difference < 20 m (i.e. pixel size) • 6 % surface difference between GPS/manual • 0.4 % surface difference between GPS record and cartography (unattainable areas) and dNBR map dNBR map Aix-en-Provence, 14th of June 2016 Châteauneuf-les-Martigues, 14th of July 2016
Fire outline mapping These tests have shown that : • dNBR mapping seems accurate for “big” (size/intensity) fires / in most summer conditions in French Mediterranean conditions equivalent to GPS survey • For big fires it allows a quick mapping of the fire outline • In steep terrain, dNBR is more accurate than traditional mapping • dNBR can also contribute to check and map small fire spots (20x20) more rapidly and more efficiently Red = false spot / Green = true spot
Fire outline mapping But depending on burn severity and vegetation type, some burnt areas can be missed by the dNBR calculation. This kind of situations often occur during the winter period when burn severity is lower than during summer time. Moustiers-Ste-Marie, 22nd of October 2017 Detected Undetected Moustiers-Ste-Marie, 22nd of October 2017 Roquefort des Corbières, 6th of September 2017
Fire outline mapping Therefore, fire outline mapping during winter period must complete dNBR calculation with GPS survey for instance. Detected Undetected Moustiers-Ste-Marie, 22nd of October 2017
Fire outline mapping In 2017, availability of Sentinel 2B images has reduced the fire mapping delay. In Mediterranean area, all fires of 50 ha and over have been mapped. The first one occurred on 24th of March, the last one on 30th of November. 44 fires from 50 to 2263 ha have been mapped using Sentinel images (16280 ha overall i.e. 83% of all burnt areas).
Assessing Post-fire vegetation damages using severity index The vegetation burn severity is defined as being the loss of aerial and subterranean organic material due to burning, by combustion or mortality. The classes of the severity index are defined from US fires, but can be used as first approximation to interpret the dNBR in Mediterranean conditions.
Assessing Post-fire vegetation damages using severity index Fields measures on the Rognac fire indicate that : • For a same severity index, impacts can vary with the type of vegetation in place before fire : • Wooded type (forest) vegetation height > 3m • Brush type (moor) vegetation height < 3m • Grass type • For a same severity index impacts on the vegetation vary with the vegetal cover density • Test of a composite index mixing both the type of vegetation and of severity index Densité / couvert de Sévérité du feu sur la végétation (dNBR) la végétation Faible Moyen Fort Houppier vert ou Etage arborée légèrement roussi sur la Houppier totalement roussi totalement ou Forêt partie inférieure (présence de quelques partiellement brulé (Sous étage et litière sujets encore vert possible) (feu de cime) brulée) Faible Végétation totalement brulé Arbustif Moyenne Strate arbustive roussie Végétation totalement brulée Strate arbustive verte et Strate arbustive brulée et Végétation Dense roussie (en mélange) roussi (en mélange) totalement brulé Végétation rase Végétation haute Herbacée totalement brulée totalement brulée
Assessing Post-fire vegetation damages using severity index Sentinel-2 images can also be used to qualify the type of damaged vegetation, analyzing pre-fire situation. 1 : Non combustible (mineral eau) 1 : Incombustible (Minéral, Eau) 2 : Herbacée 2 : Herbacee 3 : Arbustif dense 3 : Arbustif Dense 4 : Résineux hors IFN 4 : Résineux 5 : Feuillus 5 : Feuillus hors IFN 6 : Vigne Pre-fire image Image classification 7 : Verger 8 : Haie Pre-fire land cover 41 : Résineux indifférenciés 42 : Pin Alep 43 : Pin Maritime 44 : Pin Laricio ou Noir 45 : Pin Pignon 51 : Feuillus indifférenciés External GIS 52 : Chêne liège databases 53 : Chêne vert
Assessing Post-fire vegetation damages using severity index Densité / couvert de Sévérité du feu sur la végétation (dNBR) la végétation Faible Moyen Fort Houppier vert ou Etage arborée légèrement roussi sur la Houppier totalement roussi totalement ou Forêt partie inférieure (présence de quelques partiellement brulé (Sous étage et litière sujets encore vert possible) (feu de cime) brulée) Faible Végétation totalement brulé Finally, the use of Sentinel-2 images Arbustif Moyenne Dense Strate arbustive roussie Strate arbustive verte et Végétation totalement brulée Strate arbustive brulée et Végétation allows a vegetation post-fire Herbacée roussie (en mélange) Végétation rase totalement brulée roussi (en mélange) Végétation haute totalement brulée totalement brulé Burn severity damages assessment. They Pre-fire land cover represent a quick and accurate tool to evaluate the risk of potential Post-fire secondary effects of wildfires. damages Non Combustible Vignes parcourue Verger parcourue assessment Haie brulée herbacée brulée Arbustif dense vert en mélange avec roussi Arbustif dense roussi Arbustif dense totalement brulé Chêne liège parcouru encore verts Chêne liège roussis Chêne liège totalement brulés Feuillus Indifférenciés parcouru encore verts Pin Maritime totalement brulés Feuillus Indifférenciés roussis Pin Pigon parcouru encore vert Feuillus Indifférenciés totalement brulés Pin Pignon roussis Feuillus hors IFN parcouru encore verts Pin Pignon totalement brulés Feuillus hors IFN roussis Résineux Indifférenciés parcouru encore vert Arbustif dense vert en mélange avec roussi Feuillus hors IFN totalement brulés Résineux Indifférenciés roussis Pin Alep parcouru encore vert Résineux Indifférenciés totalement brulés Pin Alep roussis Résineux hors ifn parcouru encore vert Pin Alep totalement brulés Résineux horsIFN roussis Pin Maritime parcouru encore vert Résineux hors IFN totalement brulés Pin Maritime roussis
Conclusion Sentinel-2 has brought a great change in extended wildfire mapping thanks to good : • revisit delay • resolution • available bands It doesn’t replace draft operational cartography but allows to draw the final map accurately with limited field checking. Some post-fire damages can also be come up to. This assessment would need complementary data bases (soil) to evaluate the erosion and floods risk. Further work on images has to be made from now on to evaluate vegetation regrowth within 2017 burnt areas.
You can also read