文献类型：期刊文献

英文题名：Methods for spatial and temporal detection of forest wildfire disturbance based on time series Eco-environment indicators

作者：Ji, Cuicui[1] Wu, Changbin[1] Li, Xiaosong[2,3] Sun, Fuyang[4] Sun, Bin[5]

第一作者：Ji, Cuicui

通信作者：Wu, CB[1]

机构：[1]Chongqing Jiaotong Univ, Sch Smart City, Chongqing 400074, Peoples R China;[2]Int Res Ctr Big Data Sustainable Dev Goals, Beijing 100094, Peoples R China;[3]Chinese Acad Sci, Aerosp Informat Res Inst, Beijing 100094, Peoples R China;[4]Minist Emergency Management, Forest Fire Warning & Monitoring Informat Ctr, Beijing 100054, Peoples R China;[5]Chinese Acad Forestry, Inst Forest Resource Informat Tech, Beijing 100091, Peoples R China

年份：2024

卷号：168

外文期刊名：ECOLOGICAL INDICATORS

收录：;EI(收录号：20244417272550);Scopus(收录号：2-s2.0-85207274975);WOS:【SCI-EXPANDED(收录号:WOS:001346205700001)】；

语种：英文

外文关键词：Forest wildfire disturbance information; Ecological environment; Time series analysis; PV; NPV; Change detection

摘要：Forest wildfire disturbance information extracting - extracting the changes in vegetation and the condition of the burned areas - is essential for post-fire management and effective forest recovery. This study derived ecological indicators from remote sensing time series data. Time series analysis methods and change detection algorithms were applied to assess these indicators, enabling the identification of spatiotemporal information of fire disturbances. We selected the Sen + Mann-Kendall model, Coefficient of variation, Hurst exponent and Slope trend analysis to analyze the long-term impacts of the indicators extracted from Landsat images, including photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), bare rocky (BR) and normalized burn ratio (NBR). We determined the spatial distribution and timing of wildfires by analyzing the variations and fluctuations in indicators. The variation patterns of the indicators following the fires are as follows: PV and NBR decreased, while NPV and BR initially increased and subsequently decreased. By analyzing the time series analysis results of PV, NPV, BR, and NBR, the spatio-temporal information of the fires could be determined. Additionally, we used the stacked convolution long short-term memory (Stacked ConvLSTM) neural network to extract the burned area. The area extraction accuracy of this algorithm is approximately 98.43 %. Finally, the ensemble empirical mode decomposition (EEMD) was utilized to unmix the monthly mean PV, thereby obtaining the periods of vegetation recovery over multiple years. The recovery period of vegetation post-fire ranges from 3 to 12 months. This study proposes a method for comprehensively extracting information on forest wildfire disturbances at a spatiotemporal scale and discusses the recovery period of vegetation following the wildfires, as well as future development trends. It's crucial for evaluating the impacts on the ecological environment and subsequent restoration.