Study On The Forest Fire Risk Early Warning In Guangdong Province Based On MODIS

Forest fire is a natural disaster under the combined action of human factors and natural factors.It will damage a large number of vegetation resources,cause huge social and economic losses,and destroy the ecological balance.Using remote sensing technology for real-time early warning of forest fire risk can provide scientific support for effective prevention of forest fire.After the completion of large-scale afforestation of barren hills in Guangdong Province,the forest area is greatly increased,the proportion of young and middle-aged forests is increased,the tree species structure is single,and the natural villages in the suburbs are scattered,roads extend in all directions,and human activities are frequent,which makes forest fires easily occurBased on the comprehensive analysis of the research results at home and abroad,this paper constructs the primary index set of forest fire risk early warning factors.Based on the historical data of forest fire satellite hotspots,the principal component analysis method is used to screen the forest fire risk early warning factors,and the index system of forest fire risk early warning factors is constructed.The forest fire risk early warning model is constructed by comprehensively using the forest fire dynamic risk index and natural disaster risk index.Using MODIS data to calculate the forest fire risk value of each pixel,draw the thematic map of forest fire risk real-time distribution,and realize the real-time warning of forest fire risk.The results are as follows:(1)The spatial-temporal analysis of satellite hot spot data of forest fires in Guangdong Province from 2009 to 2019 shows that the spatial-temporal distribution of historical forest fires in Guangdong Province is quite different.In terms of spatial distribution,the high density areas of forest fire hot spots cover Qingyuan City,Shaoguan City,Heyuan City,Meizhou City,Chaozhou City,Jieyang City,Shanwei City,Huizhou City and Zhaoqing City,and the density of forest fire hot spots in Northeast Guangdong Province is the highest;In terms of time distribution,there are 12210 forest fires in Guangdong Province,with an average of 1110 per year.Among them,2011 is the year with the largest number of forest fires,followed by 2013 and 2015,and 2019 is the year with the least number of forest fires;In terms of seasonal distribution,the number of forest fires is the most in winter,9867 in spring,7905 in spring,less in autumn and almost no in summer.The results of spatiotemporal analysis can be used to calculate the spatiotemporal distribution density of historical forest fires,which is one of the important parameters for building forest fire risk early warning model.(2)On the basis of comprehensive analysis of domestic and foreign research results,the primary index set of forest fire risk early warning factors was constructed,including normalized vegetation index(NDVI),Temperature Vegetation Drought Index(TVDI),normalized infrared index(ndii7),elevation(GC),slope(PD),aspect(PX),spatial distribution density of historical forest fire(MD),tree species combustion type(RSLX),and forest fire risk early warning factors(DBLX)and other 9 factors.Using 5980 historical forest fire spots as samples,the primary index values were extracted and principal component analysis was conducted.The contribution values of each factor were 20.456%,17.63%,17.545%,11.362%,10.641%,9.217%,8.525%,3.253%and 1.371%respectively.The results show that:the cumulative contribution value of the first seven factors is 95.376%,so we choose NDVI,ndii7,TVDI,GC,PD,PX,MD and other seven factors to build the forest fire risk early warning index system.(3)Taking NDVI,TVDI and NDII7 as independent variables,an improved forest fire dynamic risk index model FDDI was constructed;Taking FDDI as the dependent variable and GC,PD,PX and MD as the independent variables,correlation analysis was carried out to obtain the weight values of GC,PD,PX and MD,which were 0.15501,0.093185,0.040017 and 0.074882 respectively.Four factors with weight,such as GC,PD,PX and MD,were added to the model to construct the forest fire risk early warning model NFDDI.Using 500 historical fire points and 500 random non fire points as samples,the ROC test of the model is carried out.The test results show that the AUC value of the forest fire risk early warning model is 0.714(>0.7),and the fitting effect of the model is good,which can effectively carry out forest fire risk early warning.(4)Based on MODIS data,DEM data and historical data of forest fire from February 14 to 17,2021,the early warning factors were extracted,and the forest fire risk values of each pixel were calculated by using the forest fire risk early warning model.The forest fire risk early warning map of Guangdong Province from February 15 to 18,2021 was drawn.From the map,it can be seen that there are high forest fire risk areas in all cities of Guangdong Province for four consecutive days,Among them,Guangzhou,Foshan,Zhongshan,Dongguan,Shantou and Shenzhen have larger areas of high forest fire risk areas.Over time,the degree of forest fire risk gradually decreases,and the area of high forest fire risk areas gradually decreases.In this paper,forest fire risk early warning in Guangdong Province is studied and tested based on MODIS data.The research results have application value.It will be the future research focus to integrate multi-source remote sensing data to carry out forest fire risk early warning.This paper solves the limitations of the dynamic forest fire risk index model due to ignoring the static factors.Future research will also consider more forest fire risk warning factors(such as lightning strike,wind speed,wind direction,etc.)to establish a more perfect forest fire risk warning model.