Quantitative Classification of Forest Fire in Far East Forest Area in Russia
俄罗斯远东林区林火过程数值分类研究

MODIS is broadly used to detect forest fire because of its spatial resolution and its large coverage,and it has more bands than AVHRR and frequency.In order to support the fire rescue and quick response in forest fire,the analysis of forest fire process and identification of the fire status become more important.This paper focuses on a quantitative method to classify forest fire process.We take Far East Russia as an example to describe how the method could be applied for identifying the forest fire status and real time response.Firstly,five types of forest fire in its burning process were identified as follows: 1) The active fire was identified as the fire in visible flames;2) Hidden fire was recognized as the fire with high temperature but invisible flames.In general,hidden fire is considered to be a status before the active fire in forest burning process;3) The over fire followed the active fire,normally it keeps high temperature without visible flames,but it is different from the hidden fire in vegetation;4) The burnt surface was identified as the area which does not have high temperature and has little vegetation;5) The fifth type is unfired land.Secondly,two indicators were selected in order to identify types of forest fire process,which are brightness temperature and vegetation.After pre-processing of MODIS including bowtie deduction,the brightness temperature were calculated in each pixel based on the brightness temperature algorithm using MODIS CH21 and CH31,and the vegetation index were calculated in each pixel with the algorithm of NDVI using MODIS CH1 and CH2.Thirdly,brightness temperature and vegetation index which satisfied with certain conditions were selected to identify each type of fire process.Finally,the supervised classification and decision-tree classification methods were adopted.We classified the forest fire process in Far East Russia.In order to conduct the accurate assessment of classification system,this paper adopted time series analysis on April(22 2003),May 8 2003,and June 7 2003.This method is innovative since this is the first study to attempt to classify the fire status,which can be used for near-real time identification of the fire. Also,the method has an advantage in visualization of the forest fire process status in an effective way.From the case study of Far East Russia,brightness temperature and vegetation index from MODIS data were revealed to be suitable indicators for classification of forest fire process.Moreover,the supervised classification and a decision-tree classification method can provide the easy and simple tools to identify each of forest fire process types.The time series analysis could be a great help to understand the forest fire process.