20世纪初以来沂蒙山区森林植被动态及其对气候变化的响应——基于遥感和树轮的研究

Combining remote sensing and tree ring techniques to investigate forest vegetation dynamics and its response to climate change in Yimeng mountainous area from the early 20th century

关键词:  树轮  遥感  NDVI  沂蒙山区  森林植被动态

Key words:  tree ring  remote sensing  NDVI  Yimeng mountainous area  forest vegetation dynamics

摘要:

遥感能够直接反映植被生长信息，但目前观测时间较短；树轮资料可以弥补遥感手段的不足，但往往缺乏与植被生长状况（如植被生产力）的直接关联。已有研究往往计算不同植被指数指标和树轮指数的相关性，并基于此来重建长时间尺度的植被动态信息。但基于相关性所选取的指标在不同地区存在较大差异，不利于空间对比研究。以沂蒙山区为研究区，选取能够有效表征植被年初级生产力的生长季NDVI累积值指标，利用Bootstrap法建立了其与树轮宽度的关系，重建了20世纪初期以来植被动态时间序列，并分析了变化特征及其与气候变化的关系。结果表明：沂山地区生长季NDVI累积值的多年平均值为7.36，低于蒙山和塔山地区；蒙山地区植被状况呈现好转趋势，特别是20世纪90年代之后植被好转趋势更加明显，而塔山和沂山地区植被无显著变化。小波分析结果表明研究区植被动态存在较显著的2 a、4 a或8 a尺度的周期变化，与生长季帕尔默干旱指数及平均温度的周期变化相一致，但干旱指数与植被动态具有更高的相关性。本研究综合运用树轮与遥感技术发展了长时间尺度植被动态时间序列重建方法，有助于更好地理解植被动态及其对气候变化的响应。

Abstract:

Background, aim, and scope Remote sensing is an effective approach of vegetation monitoring, and remote sensing-based vegetation indices can well capture vegetation growth curve. However, the longest remote sensing observations are only available for about 40 years, which limits their use in research on long-term vegetation dynamics. Tree ring data usually have longer records, which overcomes the limitations of remote sensing observations, but there is no direct link between tree ring and vegetation properties (e.g., gross primary productivity, biomass). Previous studies combine tree ring width and satellite-based vegetation index to reconstruct vegetation dynamics for a long period. However, the selection of indicators depends much on correlation, and thus different indicators might be chosen for different regions, which is not suitable for comparison spatially. Here, we took Yimeng mountainous area as the study area, reconstructed long-term accumulative NDVI values of the growing season, an effective indicator of annual gross primary productivity, from the earlier 20th century, and characterized their temporal changes. Materials and methods (1) 225 tree ring cores were collected from three sites within our study area: Mengshan Mountain, Tashan Mountain and Yishan Mountain. The ring widths of all the cores were measured using MeasureJ2X professional measurement software under the LINTAB tree ring width meter (measuring accuracy 0.01 mm), and then COFECHA program was used to test the cross-dating quality. Finally, the standardized chronology (STD), difference chronology (RES) and autoregressive standardized chronology (ARS) were established by the ARSTAN program. (2) To derive the time series of accumulative NDVI of the growing season, we extracted NDVI time series at 15-day interval for each sampling site, used the TIMESAT software to determine the start, end, and length of the growing season, and finally got the sum of NDVI values within the range of the growing season. (3) The Bootstrap method was used to establish the empirical relationships between tree ring width and accumulative NDVI, and reconstruct the time series of accumulative NDVI from early 20th century. (4) Wavelet analysis was utilized to identify the underlying periods within the long-term time series of climate variables and accumulative NDVI. Results The mean accumulative NDVI of growing season for Yishan Mountain area is 7.36, which was lower than that in Mengshan and Tashan Mountain areas. Vegetation productivity in Mengshan mountain area showed an increasing trend, and the magnificence of increase was particularly large after 1980s, but there was no significant trend in Tashan and Yishan Mountain areas. There were 2 years, 4 years and 8 years of period underlying the time series of accumulative NDVI for three sites. Variations in mean wavelet power between 2 and 8 years for accumulative NDVI was more consistent with those for PDSI than temperature. Discussion Our research found that tree ring width significantly correlated with accumulative NDVI of growing season, an effective indicator of vegetation productivity. Meanwhile, our used nonparametric method, Bootstrap regression, was more robust than traditional statistical method, which could address the situations when the sample size was small or the distribution of samples was not normal. Therefore, our research provides a framework which accurately reconstructs vegetation dynamics for a long time period. We also found that vegetation dynamics within our study area were determined by combined water and temperature, as indicated by the highly consistence between variations in accumulative NDVI and PDSI. Conclusions We concluded that integrating remote sensing and tree ring techniques could effectively reconstruct long-term vegetation dynamics, and accumulative NDVI of growing season was useful indicator to be chosen for reconstruction given its high correlation with tree ring width and its close link with vegetation productivity. Vegetation dynamics in the Yimeng Mountainous areas were determined by water and temperature factors. Recommendations and perspectives We develop a framework to accurately reconstruct vegetation dynamics by combined remotely sensed data and tree ring materials, which could be extended to other research areas. Our findings that both water and temperatures are important to determine vegetation productivity are useful for explaining and predicting vegetation dynamics under climate change especially in the warm temperate Yimeng mountainous regions.