基于小波变换的卧龙国家级自然保护区植被时空变化分析

Analysis of temporal and spatial changes in vegetation cover using wavelet transform method in Wolong Natural Reserve

提出一种基于小波变换从长时间序列、大范围遥感数据中快速、自动化检测植被动态变化的方法。以MODIS 500m空间分辨率,16d合成的NDVI数据为数据源,对受到2008年5月12日汶川地震严重影响的卧龙国家级自然保护区内2003年至2012年的植被动态变化进行时空分析,为保护生态多样性及生态系统的稳定性提供依据。研究表明:1)地震后保护区内植被指数减少的面积大范围增加,且波动较震前更为明显,统计分析结果能够更为直观地反映地震及其次生灾害等极端现象对该地区植被的破坏程度;2)保护区内植被指数极值变化多发生在夏季或秋季,较低海拔地区极值变化多发生在夏季,而在高海拔地区则多发生在秋季;3)在大熊猫最适宜栖息的区域(2600—2800m)植被指数极值减少量大于0.4的范围大于增加量大于0.4的范围,反映出植被在震后的恢复状况并没达到理想的水平。同时发现在该海拔区域范围内植被指数减少的面积在春夏两季较大,表明在该时间段卧龙地区大熊猫最适宜生存区域的植被情况较为不稳定,需更为关注其动态,采取适当的保护措施。

On May 12, 200.8, the Wenchuan earthquake struck the Sichuan province, triggering strong continuous aftershocks in its surrounding regions, which led to secondary geological disasters, such as landslides, collapses, and mudslides, in more than 120 regions in Wolong National Natural Reserve. This bioreserve consists of broad-leaved deciduous forests, deciduous coniferous mixed forest, and coniferous forests, and is considered an important habitat of giant pandas in China. The population of pandas in this area accounts for approximately 10% of the total number of pandas in the entire country. However, vegetation in this bioreserve was adversely affected by the 2008 earthquake. With the development of earth observation techniques over the last decade, long-term observations of natural vegetation have detected large-scale land-cover changes vegetation changes ~ Among the various long-term satellite data, MODIS data have significant potential in detecting from regional to global scale because of their high temporal resolution. In this research, we proposed a method based on wavelet transform to extract information regarding vegetation changes using normalized difference vegetation index （NDVI） time series data in Wolong National Natural Reserve. In particular, MODIS 16-daily products recorded from 2003 to 2012 with 500-m resolution were selected for investigation, and the effects of the earthquake and secondary geological disasters on vegetation in Wolong National Natural Reserve were quantified~ We obtained numerous relevant findings. （ 1 ） The proportion of the areas where the NDVI greatly decreased after the earthquake is significantly larger than before~ Moreover, by using the trend-seasonal model and breakpoint detection method from the time series data, we discovered that the NDVI curves fluctuated more obviously after the earthquake. （2） By performing a seasonal analysis on the extreme values of NDVI time series, we determined that the increased extreme values mostly occurred in summer because the vegetation cover during this season was the highest and also the most unstable one in a year. The second most important season was autumn. Compared with the increased values, the decreased NDVI extreme values often emerged in summer and spring. The results specified that the area with decreased NDVI data was the third highest when the Wenchuan earthquake transpired. Nonetheless, the statistical results obtained from comparing the inter-annual difference were biased because of the identified moderate cloud effect on NDVI data in 2007 and 2012. （3） The Wenchuan earthquake that occurred on May 12, 2008 was an intra-continental shallow-focus earthquake, which extensively affected the environment and areas surrounding Sichuan province. In this research, the data were also analyzed at land elevations where giant pandas are most comfortable in （2600--2800 m） to quantify the influence of the earthquake and secondary geological disasters on the habitat of the giant pandas. Accordingly, we realized that the extent of decreased NDVI extreme values remarkably exceeded the instances of increased NDVI. Vegetation recovery in Wolong National Natural Reserve did not achieve an expected level Based on the above findings, the method introduced in this paper can be used to detect significant changes in long-time NDVI data; however, it cannot identify whether the causative factor is a disastrous natural phenomenon （e.g., earthquake, fire, etc. ） or a human activity （e.g., deforestation, urbanization, etc.）. In future, we intend to apply our method in a wider scope when data-intensive calculation technique is required.