Spectral characterisation and discrimination of burnt areas

Abstract

Abstract. Spectral properties of recent burns are characterised, in the visible, near infrared, mid-infrared, thermal infrared, and microwave spectral domains. Fireinduced reflectance changes are also compared for various ecosystems and biomes, and discussed in terms of the ecological effects of phytomass combustion. The spectral signatures of combustion products and of burnt areas are compared with those of various plant material and land cover types, in order to graphically represent relevant aspects of burnt area spectral discrimination. A series of colour composite images, based on Landsat Thematic Mapper data is used to illustrate the appearance of burnt surfaces in various tri-spectral spaces, and in contrast with healthy forests, agricultural fields, and urban areas. The temporal evolution of the spectral properties of burns is also demonstrated, with a five-year time series of Thematic Mapper images of two conifer forest burns in central Portugal. Finally, a series of conclusions is proposed, concerning the distinctive spectral properties of burnt surfaces, and implications for discrimination and mapping of such areas. IntroductionVegetation fires are common in tropical, temperate, and boreal biomes. In the tropics fire is often used as a land management tool, employed in shifting agriculture, hunting practices, to prevent the invasion of grasslands by shrubs, and in the conversion of primary forest to other uses. In temperate regions fire is also extensively used for slash burning, but is most often considered a hazard when it burns in ecosystems that are prized for their economic and ecological value. In both tropical and temperate biomes, a great majority of fires, whether desired or not, occur as a result of human activities, but in boreal regions natural causes (i.e. lightning) still account for a significant proportion of the total area burnt, and fire is a very important ecological factor influencing plant community dynamics.These various kinds of fire activity generate a range of economic, ecological, atmospheric and climatic impacts, with magnitudes that are strongly dependent on the areal extent of the burns. Detailed and current information concerning the location and extent of the burnt areas is important for assessing economic losses and ecological effects, monitoring land use and land cover changes, and modelling E. Chuvieco (ed.), Remote Sensing of Large Wildfires

Burned Area Mapping in the Brazilian Savanna Using a One-Class Support Vector Machine Trained by Active Fires

“…Several studies have shown that the spectral space of short wave infra-red and middle infra-red channels in savanna areas is advantageous, especially at the 2.1 and 3.9 µm wavelengths, respectively [69][70][71][72][73]. Thus, the implementation of the approach presented herein with spectral indexes presenting greater separability between burned and unburned surfaces may improve the results.…”

“…Several studies relied on the use of remote sensing to map burned areas at a global/regional scale [5][6][7][8][9][10][11]. However, the variable persistence of burn scars within different vegetation types, and the spectral confusion with other phenomena (e.g., cloud shadowing) are some of the problems that still hamper accurate burned area mapping [12]. Accordingly, users of burned area maps have stressed the need to improve product accuracy, namely in order to refine current estimates of burned areas, thus providing input to global analysis of ecological impacts of fires to better understand the relations between fire occurrence and biodiversity, and to improve the assessment of atmospheric emissions derived from vegetation fires [13,14].…”

Use of Unmanned Aerial Vehicle for Multi-temporal Monitoring of Soybean Vegetation Fraction

“…An M value between the pixel groups lower than 1 means that the two histograms overlap each other and have a poor separability, whereas an M value higher than 1 means that the histograms are well discriminated. (Pereira et al, 1999). Figure 9 shows sample histograms used in the M-statistics.…”

Mapping the daily progression of large wildland fires using MODIS active fire data

“…The standard MODIS burnt area product Roy et al 2005) makes use of post-fire reflectance changes in the near infrared (NIR) and short-wave infrared (SWIR) spectral regions. These reflectance changes are caused by the removal of vegetation and deposition of charcoal and ash by fire (Pereira et al 1999;Trigg and Flasse 2001). In addition to spatial burn extent information, the algorithm also provides the approximate day of burning, with a nominal uncertainty of up to eight days (Roy et al 2005).…”