| Forest fire disturbance is an important natural driving factor for dynamic changes of forest structure and function,and it is also a worldwide forestry disaster.It is of great significance to understand the occurrence law of forest fire and to plan forest fire management and prevention and control strategy.At present,there is a lack of spatial and temporal information of speckle scale for the forest fire data mainly collected in various provinces and cities(regions)in China.However,the products based on remote sensing can provide time-continuous and spatially clear speckle information,but the accuracy of the products is often limited by the low spatial resolution.Based on Great Xing’ an Mountains forest sites in 2001-2017 as the research object,based on historical burns were extracted from multi-source remote sensing data set information,can quantitative characterization of forest fire situation relevant characteristic parameters(the occurrence time,area,fire intensity,etc.),analyzed the intcrannual between dynamic changes of forest history in the study area,such as spatial distribution,seasonal characteristics.The main work and conclusions are as follows:(1)A new method for extracting the information of forest burned area is proposed.Based on the assimilation and improvement of the 250-meter global fire trail product(Fire_CCI)using the 30-meter global forest change product(GFC)to obtain the forest fire spot product(GFC_CCI),the use of manually interpreted fire spot data to verify the accuracy The results show that:GFC_CCI products’ fire spot recognition rate(RR),variation interpretation amount(R2),root mean square error(RMSE),mean square absolute percentage error(MAPE)and other precision indicators(80%,0.85,0.41,respectively)11.13%)are better than the original Fire_CCl products(73%,0.77,0.60,18.13%)and MODIS fire trace products(MCD64A1)(60%,0.77,0.65,19.11%).The improvement effect is especially obvious in the recognition of small and medium-sized fire spots.(2)Based on time series Landsat data set,an quantitative evaluation of fire disturbance intensity on the original cloud shadow interference such as problems in remote sensing image,using CFMask cloud detection algorithm for removing,information for the missing part is using similar periods of remote sensing image to fill values using threshold segmentation method of band operation for fire index(dNBR)has carried on the quantitative evaluation,divided into high,medium and low three levels of intensity.(3)Mann-Kendall trend test spatial information statistics and other methods were used to systematically analyze the overall dynamic changes of forest fires in the study area.From 2001 to 2017,a total of 1.4686 million hectares of forest fires occurred in Great Xing’an Mountains forest area,and the interannual forest fire area showed a significant decreasing trend(Z=-3.91,P<0.01),the distribution law of east more west less north more south is presented as a whole;The study area was seriously disturbed by forest fire from 2003 to 2006,and showed an obvious decreasing trend year by year after 2006,and maintained at a low level after 2012.Forest fires mainly occurred in spring and fall,mainly in April and October,accounting for 14%,24%and 44%of the whole year respectively.(4)The spatial aggregation technology was used to realize grid processing,and the information of forest fire area ratio,month,intensity,and so on was extracted by the grid to clarify the spatial pattern of forest fire situation.The analysis results show that the forest fire situation in different regions shows obvious differences:in the southeast region,the burned area is dominated by low fire frequency,while the middle and high fire frequency areas are mainly distributed in the northeast and a small number of central areas;63%The burned area is spring forest fires,mainly concentrated in the southeast,while autumn and summer forest fires arc mainly distributed in the northeast and the middle;the forest fire intensity generally shows the distribution pattern of"northwest high,southeast low",northeast and southeast Moderate and low intensity is dominant,while the western and northwestern regions are dominated by medium and high intensity.. |
PDF Full Text Request