Forest Disturbance Detection By Remote Sensing

Forest ecosystems, which are major parts of the terrestrial biosphere, play an important role in terrestrial carbon cycle and storage. However, the accuracy of regional forest carbon-flux estimation is greatly influenced by the lack of forest disturbance data. Thanks to large coverage, comprehensive and dynamic monitoring, remote sensing is becoming a useful tool for disturbance monitoring. How to use a long time-series data to extract disturbance information is one of the significant problems for current study.Using MOD09A1surface reflectance data and MCD12Q1surface coverage data from2001to2011, forest disturbances in jiangxi are monitored by DI index time-series algorithm in recent decade respectively. The algorithm was evaluated by comparing with the existing burned area products and the TM data. The forest disturbance area was extracted from2001to2011in Jiangxi Province, and the spatial, temporal patterns and driving factors are analyzed. The conclusions are as follows:(1) Seven spectra bands of TM data and remote sensing indexs are analysised between before and after forest combustion. The results show that TM4, TM7and TM6band have higher distinguishing ability, which reflecting the characteristics of the burned forest more precisely.NBR、NSTV1、NSTV2and NSEV1among18different remote sensing indexes have higher separability value, and the extraction accuracy for burned forest is higher than85%,in addition,extraction capacity of EVI、 BAI、MIRBI、SAVIT and NSEV2is relatively weak. The improvement burned index NSTV2is highest as TM6band is introduced, indicating that properly introduction of thermal infrared band can improve the extraction capacity of burned forest.(2) A normalized DI index is constructed by MODIS tasseled cap transformation coefficients and MCD12Q1data,then the mean state and the natural variation of each pixel are got through time-series DI data sets. If the change of a pixel is beyond natural variability, the pixel is regarded as forest disturbance.forest disturbance monitoring algorithm is proposed based on dynamic threshold DI index of long sequences.(3) Compared by the MODIS burned area product, the proposed method can better detect forest fire disturbance area of Jiangxi Province, through TM data of2007and2008,burned area in northeast of Ganzhou is extracted by using NSTV2index, the result is used to verify DI algorithm by18selected plaques area, with average accuracy of85%. The result DI monitor is consistented with MODIS and TM color composite images in space.(4) The total area of Forest disturbance is6129.5KM2from2001to2011in Jiangxi, its major forest disturbances years are2003,2004,2008and2009, the average annual disturbed area is802.31KM2, double to non-major disturbance years.Forest disturbance in Jiangxi is mainly distributed in southern and central region, with relatively fewer area in Northern of Jiangxi, and mainly located in low-altitude, low slope and near river area.