Research On Building Fire Potential Index Model Using GF1 And Landsat8 Images

Fire has bring about significant danger and severe damage for forest and human.When satellite technology can monitor forest fire the nearly real-time,the images were extracted to get ranges of fire area.It is based on remote sensing to research mostly in northeast and northwest China region,seldom in south China region using high resolution image.Thus,it is limited to national fire danger rating system researching on remote sensing.The study has developed intensive research on forest fire using high resolution image.The present research chooses Landsat8 images,GF1 images and average relative humidity as the data source to build FPI model.Taking a case study of Dongjiang forest management area in Zijin county of Heyuan,it is developed to interpolate Temperature on Landsat8 images with Universal Kr iging,the resolution of which is the same as that of GF1 adding average relative humidity.Then,it causes FPI model graphs integrating many kinds of vegetation index from 2013 to 2015.The main research content as follows: 1 Temperature retrieval based on Radioactive Transfer Equation(RTE)Land Surface Temperature(LST)is a very important parameter on the FPI model.It is a very effective method to retrieval LST or Land Surface Emissivity(LSE).It is calculated by RTE on Landsat8 image and analyzed precisely.The retrieval Temperature of range is from 24.85? to 34.35? in 2013,from 24.44? to 28.05? in 2014,from 26.79? to 28.98? in 2015.2 Universal Kriging interpolation on Lansat8 temperature imageThe temperature retrieval image on Landsat8 were scaling up to image of which resolution is equal to the resolution of GF1.To validate the reasonability of the temperature image scaling,It is verified by Moran 'I coefficient and normal distribution.The Moran 'I coefficient of images are 0.96,0.96 and 0.97 respectively.Furthermore,the p-value of Shapiro-Wilk test of temperature image were 0.1857,0.09968 and 0.08534,which meeting the requirement of normal distribution method.The average error coefficient is decreased by 5.73% and 5.57% compared to that of MODIS products,which is decreased by 5.63% by that of field test.The results show that it is reasonable that scaling up the temperature image having used universal kriging interpolation.3)Constructing FPI model based on vegetation indexThe vegetation index der ived from GF1 image is building FPI model affiliating relative humidity like meteorological factors.The difference of average deviation on every fire rank of GNDVI and OSAVI is 2.37% and 2.55%.Moreover,the average coefficient of correlation is 0.9423 and 0.9376,which is high correlation to NDVI.Therefore,they are similar to NDVI so that they can replace NDVI to build FPI model.And the average coefficient of correlation of ARVI and GARI are 0.7368 and 0.6689,which are signif icant to NDVI.However,the average deviation of EVI and EVI2 are 41.98% and 41.97%,the average coefficient of correlation of that are 0.4624 and 0.4340.It showed a lack of correlation to NDVI.In addition,the statistical results of MNLI and VARI are totally incorrect to NDVI.The GNDVI,OSAVI,ARVI and GARI that significant correlation indexes can replace NDVI build FPI model in high resolution image.The percent of areas catching fire easily of which FPI model based on NDVI is less than 1% every year,because there are high relative humidity and long time lag.The range catching fire easily in study area is far fewer than that of other local regions,such as northeastern forest areas,which accord to reality.