基于时序突变检测的植被空间变化特征识别方法——以海河北部山区为例

长时序植被变化趋势分析是分析植被生长与退化状况的主要手段之一，但难以精细揭示分析时段内植被变化动态的时空特征。将归一化植被指数（NDVI）时序突变检测与趋势分析结合起来，建立了精细识别植被空间变化特征的方法，并以海河北部山区为案例，开展了实证研究。案例分析结果表明：2000-2018年间，海河北部山区有78.1%的区域NDVI发生了突变，其中1.7%的面积发生退化。应用基于时序突变检测的植被变化识别方法进一步发现：植被变化在2011年出现突变拐点，突变前（2000-2011年）在西北草原区和东南林农区存在1.6%的显著退化，突变后（2011-2018年）在东部林农区存在1.2%的显著退化。分析表明，海河北部山区植被整体改善与气象因子的关系不大，局部恶化则是气象因子与人类活动共同作用的结果。与常规长时序植被变化趋势分析相比，基于时序突变检测的植被空间变化特征识别方法揭示了评估时段内植被空间变化的动态趋势，可为明确区域植被的变化以及差异化的植被恢复策略提供决策信息。

Recognition of vegetation spatial variation based on time-series mutation detection: A case study of the mountainous area of Northern Haihe River Basin

Long-term trend analysis (LTA) is a typical approach to understand the growth and degradation of vegetation in the area. However, the dynamic of the spatio-temporal characteristics of the vegetation is often difficult to be captured by the LTA. This study combines LTA with a time-series mutation detection of normalized vegetation index (NDVI) for identifying the spatial variation of vegetation, where northern mountainous region of Haihe River Basin is used as a case. The results show that 78.1% of the NDVI in the area changed abruptly from 2000 to 2018, among which 1.7% was degraded. An inflection point of NDVI, which occurred in 2011, was found by time-series mutation detection. A significant degradation of 1.6% in the northwest steppe and southeast forest-farmland region, and 1.2% of degradation in the east forest-farmland region are also observed before the inflection point (i.e. 2000-2011) and after (i.e. 2011-2018). The results also show that the overall improvement of vegetation in the mountainous area of the Northern Haihe River has minor correlation with the meteorological factors, meanwhile local deterioration results from meteorological factors and human activities. Compared with the conventional LTA, the proposed method gains insight of the dynamic trend of vegetation spatial changes in the study area and therefore providing useful information for vegetation recovery strategies.