Study On Typical Forest Disturbance Identification Using Satellite Remote Sensing And Fire Severity Driving Force

The development of forest disturbance information monitoring research based on domestic satellite data is one of the important measures to promote the application of domestic satellite data in forest resource supervision.In this paper,multi-source data has been used and two typical forest disturbance elements,forest fires and buildings in the forest,have been selected as the research objects.The satellite remote sensing identification of typical forest disturbance and fire severity driving force have been carried out.This paper included four aspects: first,a building recognition method that combines pixel-based and object-based methods in the forest district has been proposed baesd on the GF-2 PMS images;second,eleven indexes have been selected based on GF-6 WFV images to systematically evaluate the ability of these indexes and index differences to identify burned areas and explore the appropriate spectral bands and indexes;third,the differenced Normalized Burn Ratio(d NBR)of the large fires were calculated based on Landsat TM/ETM+/OLI data in Northeast China from 1998 to 2018 and the fire severity was classified,then the spectra of vegetation restoration of different disaster levels on GF-6 WFV images have been analysed by using the idea of space instead of time;last,the Geodetector model has been used to analyze the driving force of forest fire severity in Northeast China,and the Random Forest model has been used for comparison.The main conclusions are as follows:(1)The building detection method by using GF-2 images that combined the pixel-level and object-level not only retained the advantages of simplicity and ease,but also avoided the salt and pepper phenomenon.Compared with the pixel-level or object-level method,the correctness,completeness and quality of the result were all significantly improved.In terms of the quality evaluation indicator,this method was 0.20 and 0.13 higher than pixel-level and object-level methods,respectively.(2)In the comparative analysis of the indexes using GF-6 WFV data to identify burned areas,Burned Area Index(BAI)and Global Environment Monitoring Index(GEMI)had the best identification results,followed by Normalized Difference Vegetation Index(NDVI),Enhanced Vegetation Index(EVI),Soil-Adjusted Vegetation Index(SAVI),Modified Soil-Adjusted Vegetation Index(MSAVI),Normalized Difference Water Index(NDWI)and Modified Chlorophyll Absorption Ratio Index 2(MCARI2)with medium performance,while Modified Normalized Difference Soil Index(MNDSI),Normalized Difference Red Edge Index(NDRE1)and MERIS Terrestrial Chlorophyll Index(MTCI)performing poorly.(3)It took 15 years for the moderately and severely burned areas in Northeast China to recover to normal vegetation status,while the lightly burned areas recovered earlier;Between the 14 th and 16 th years of recovery,the lightly burned area might be affected by other disturbances.Its spectral information changes were different from that of normal vegetation,and there was a significant change in each band;It was consistent for the restoration of vegetation in the moderately and severely burned areas between the 14 th and 16 th years,while the spectral value changes in the lightly burned areas were different from those in the moderately and severely burned areas and normal vegetation..(4)Elevation,vegetation coverage,longitude and latitude were the four important factors affecting the distribution of forest fire severity in Northeast China from 1998 to 2018.Where the more dense vegetation coverage,higher altitude and lower slope was,the more severity the burned degree by forest fires would be;There were significant differences in the distribution of fire severity between different longitudes and altitudes,and elevation was more explanatory than other terrain driving factors such as slope,aspect and Topographic Position Index(TPI).