Knowledge Resource Center for Ecological Environment in Arid Area
| 沙坡头人工固沙植被系统的跃变特征及早期信号分析 | |
| 其他题名 | Dynamics of regime shift and analyses of early warning signals of a sand-binding revegetation desert ecosystem in Shapotou |
| 陈宁 | |
| 出版年 | 2017 |
| 学位类型 | 博士 |
| 导师 | 王新平 |
| 学位授予单位 | 中国科学院大学 |
| 中文摘要 | 在全球气候变化和人类活动不断加剧的背景下，干旱区的荒漠化问题日益突出，当地居民的生活与福祉面临着前所未有的挑战，如何防治荒漠化对干旱区社会经济发展具有重大意义。荒漠化及其逆转过程可能是非线性的，在人为促进生态恢复的作用下，当环境条件发生变化时，生态系统可能会从一种稳定的状态骤然转换成另外一种状态，系统的结构和功能发生剧烈改变，这种现象被称为系统跃变（regime shift）。系统在发生跃变的过程中可能表现出一些特征，如临界放缓，可以此为基础构建早期信号（early warning signal）来预警系统跃变的发生。因此系统跃变及早期信号理论在系统水平为认识和解决生态学问题提供了新的视角，并为更加深入地认识与理解干旱生态系统的荒漠化及其逆转过程提供了重要理论依据。以往有关系统跃变及早期信号的研究在方法上以模型为主，长期观测研究相对较少，而且就研究对象而言，缺少跃变时间尺度比较大的系统，对干旱生态系统的研究尤为薄弱。腾格里沙漠东南缘地处草原化荒漠和荒漠草原的过渡带，自1956年建立沙坡头人工固沙植被生态系统以来，已从最初的流动沙丘演替为荒漠草原，作为一个大空间尺度的成功生态重建实例，积累了较长时间尺度上系统状态的连续观测资料，是研究干旱区脆弱生态系统生态重建的极佳案例。本文以腾格里沙漠东南缘沙坡头人工固沙植被生态系统为研究对象，基于灌木盖度、草本盖度和结皮盖度的长时间序列数据，结合不同植被恢复年限灌木的空间格局数据，系统分析了植被恢复进程中的系统跃变动态及变化类型，并通过检测时间尺度（1阶自相关系数、波谱密度比、标准差和偏度）和空间尺度早期信号（空间变异、空间偏度、空间自相关性和斑块大小分布）对系统跃变的预警能力，以期揭示干旱区生态恢复动态与机制，促进系统跃变及早期信号理论在干旱区恢复生态学领域的应用。主要研究结果如下：（1）人工植被演变过程中灌木盖度包括五个显著性的时间变化趋势，分别为恢复后3-8和11-14年的增加趋势，和恢复后16-19、28-30和37-40年的降低趋势（P<0.05）；而草本盖度则只经历了一个显著性的增加趋势,即恢复后的39-48年。在频度分布上灌木盖度和草本盖度呈双峰分布，分别为灌木盖度的15-20%和8-10%，和草本盖度的0-25%和30-50%。在恢复后的1-49年，灌木盖度和草本盖度分别经历了7和5个断点，其中恢复后17（19）、30（29-31）、38和46年在两者中都存在。因此系统变量的时间变化趋势是非线性的。（2）结皮厚度作为驱动力与灌木盖度的关系可以由不同截距的分段线性模型来拟合，其赤池信息量准则（Akaike Information Criterion，AIC）为216.98，低于线性模型的220.8；结皮厚度与草本盖度之间的关系可以由不同截距且不同斜率的分段线性模型来拟合，其AIC值为328.37，低于线性模型的333.33。灌木盖度和草本盖度的分段线性模型的分段点在结皮厚度上的位置分别为17 mm和20 mm处。（3）沙坡头人工固沙植被生态系统的系统变化类型为非连续型的系统跃变，恢复后37年到达系统跃变点，并于恢复后46年跃变至荒漠草原态，裸地态和荒漠草原态为该系统的两个稳定态。（4）以灌木盖度为系统变量时的时间尺度早期信号-1阶自相关系数、波谱密度比、标准差和偏度的趋势强度分别为0.410、0.410、0.652和-0.779（P≤0.05），相应草本盖度的时间尺度早期信号的趋势强度分别达到了0.794、0.802、0.808和0.500（P≤0.05），两个系统变量的四个时间尺度早期信号都能够预警系统跃变点的靠近。灌木盖度的四个早期信号的预警时间分别为13、13、16和17年，相应草本盖度的分别为16、16、20和20年。（5）相对于未去趋势处理，去趋势处理提高了时间尺度早期信号的趋势强度，灌木盖度的1阶自相关系数、波谱密度比、标准差和偏度分别提高了0.042、0.042、0.326和0.190，草本盖度的分别提高了0.383、0.374、0.234和0.103。去趋势处理将灌木盖度的标准差早期信号的预警时间提早了9年，而草本盖度的1阶自相关系数、波谱密度比和偏度分别提早了8、8和12年。（6）对于空间尺度早期信号，当空间格局数据包含三种主要灌木时（油蒿（Artemisia ordosica）、柠条锦鸡儿（Caragana korshinskii）和花棒（Hedysarum scoparium）），系统靠近跃变点的过程中（恢复后27-32年），空间变异从0.140增加至0.152，空间偏度从6.872降低至6.287，而空间自相关性则从0.638降低至0.595，斑块大小分布的拟合模型（幂定律、断尾幂定律和指数分布）无变化，均呈幂定律分布（P≤0.05）。当仅包含油蒿时，空间尺度早期信号表现出类似的趋势，空间变异从0.096增加至0.126，空间偏度从10.277降低至7.832，而空间自相关性则从0.478降低至0.462，斑块大小分布的拟合模型仍无变化，均呈断尾幂定律分布（P≤0.05）。使用两组不同空间格局数据，空间变异和偏度能够预警系统跃变的发生，而空间自相关性和斑块大小分布不能起到预警的作用。 |
| 英文摘要 | The question of drylands desertification becomes severer under the background of global climate changes and more intensified human activities. The lives and welfare of human beings in drylands encounter huge challenges, and thus preventing desertification in drylands is of great implications for the sociometric developments. The process and the reversed process of desertification may be nonlinear. When environmental conditions are altered under the effects of promoting restoration, an ecosystem may abruptly shifted from one alternative stable state to another one, accompanied with great changes in the structure, functions and services of the ecosystems. This phenomenon is called regime shift. Meanwhile, some indicators, which are named early warning signals (EWS), can be designed based on characteristics of ecosystems under regime shifts to predict the preceding regime shifts. As a new aspect of studying ecological questions in system level, the theory of regime shift and associated EWS could provide important theoretical supports in comprehensively understanding the problem of desertification and also restoration in drylands. Furthermore, previous studies on regime shifts and EWS mainly relied on models, and were in lack of long-term observational data. As to study objects, the ecosystems with long timescale of regime shifts like drylands were still scarce in the studies of regime shift and EWS. The southeastern fringe of the Tengger Desert is the ecotone of steppified desert and desertified steppe. After re-vegetation since 1956, Shapotou sand-binding re-vegetation ecosystem is set up and then shifts from the original barren moving sand dune to the desertified steppe. As a successful large-scale restoration case with long-term continuous observations of system states, this ecosystem is a perfect case to study the question of desertification and restoration in drylands. With the ecosystem as an example case, this combined with long-term temporal time series of shrub cover, grass cover and biocrust thickness, and spatial patterns of shrubs in different years to investigate the dynamics and type of the ecosystem. And then we explored whether the preceding regime shift could be forecasted by the EWS in temporal (including autocorrelation coefficient at lag 1, spectral density ratio, standard deviation and skewness) and spatial scale (spatial variance, spatial skewness and spatial autocorrelation, and patch size distribution). This study on one hand can unfold the dynamics and mechanisms of dryland desertification and restoration, on the other hand advances the developments and applications of regime shift and associated EWS in drylands. The main results are summarized as following:(1) There were five significant temporal trends in time series of shrub cover, located in year 3-8 and 11-14 after re-vegetation as increasing trends, in year 16-19, 28-30 and 37-40 after re-vegetation as decreasing tends (P < 0.05). There were only one significant increasing trend in time series of grass cover, year 39-48 after re-vegetation (P < 0.05). Shrub cover and grass cover tended to be bimodal in frequency distributions. The two modes lay in 15-20% and 8-10% for shrub cover, and 0-25% and 30-50% for grass cover. Shrub cover and grass cover respectively experienced seven and five breakpoints after re-vegetation, and they shared four of the breakpoints - year 17(19), 30(29-31), 38 and 46 after re-vegetation. Therefore, the temporal patterns of system variables were nonlinear.(2) Among the fitted models of linear and separated linear, separated linear model with different intercepts (AIC = 216.98) was better than the linear one (AIC = 220.8) in fitting the biocrust thickness-shrub cover relationship. For the relationship of biocrust thickness and grass cover, separated linear model with different intercepts and different slopes (AIC = 328.37) performed better than the linear one (AIC = 333.33). The breakpoints for the separated linear models of shrub cover and grass cover located in 17 and 22 mm in biocrust thickness.(3) The type of system dynamics of Shapotou sand-binding re-vegetation ecosystem belonged to discontinuous regime shift, rather than type of smooth or threshold. The tipping point of the regime shift located in year 37 after re-vegetation. The ecosystem shifted to the desertified steppe state completely after 46 years, and also the bare state and the desertified steppe were two alternative stable states for the ecosystem.(4) The trend strengths of temporal EWS - autocorrelation coefficient at lag 1, spectral density ratio, standard deviation and skewness - were 0.410, 0.410, 0.652 and -0.770 (P < 0.05), respectively, for shrub cover, and 0.794, 0.802, 0.808 and 0.500 (P < 0.05), respectively, for grass cover. Autocorrelation coefficient at lag 1, spectral density ratio, standard deviation and skewness could signal the preceding regime shift 13, 13, 17 and 16 years before the tipping points for shrub cover, and 16, 16, 20 and 20 years before the tipping points for grass cover.(5) Relative to non-detrending treatments, detrending could improve the trend strength of temporal EWS for both shrub cover and grass cover. Autocorrelation coefficient at lag 1, spectral density ratio, standard deviation and skewness increased 0.042, 0.042, 0.326 and 0.190, respectively, for shrub cover, and increased 0.383, 0.374, 0.234 and 0.103, respectively, for grass cover. Meanwhile, detrending advanced the warning time of standard deviation of shrub cover another 9 years earlier from 8 years before the tipping point when shrub cover was not detrended. For grass cover, detrending played a role in advancing the warning time of autocorrelation coefficient at lag 1, spectral density ratio and skewness, for which the advancements were 8, 8 and 12 years, respectively.(6) For spatial early warning signals, as approaching the tipping point from year 27 to 32 after re-vegetation, spatial variance increased from 0.140 to 0.152, spatial skewness decreased from 6.872 to 6.287, spatial autocorrelation decreased from 0.638 to 0.595, and patch size were always power law distributed with power law, power law with exponential cutoff and exponential models as candidates (P < 0.05), when spatial patterns included all three dominant shrubs (i.e., Artemisia ordosica, Caragana korshinskii and Hedysarum scoparium). When only A.ordosica was used, there were similar trends of spatial EWS as approaching the tipping point. Spatial variance increased from 0.096 to 0.126. Spatial skewness decreased from 10.277 to 7.832. Spatial autorcorrelation decreased from 0.478 to 0.462. There were also no changes in fitting model of patch size distribution between year 27 and 32 after re-vegetation. Whereas the best fitting model of patch size distribution was power law with exponential cutoff (P < 0.05). For both data of spatial patterns, spatial variance and skewness could signal the approaching of tipping point, while spatial autocorrelation and patch size distribution could not. |
| 中文关键词 | 干旱区 ; 荒漠化 ; 系统跃变 ; 早期信号 ; 跃变点 |
| 英文关键词 | drylands desertification regime shift early warning signal tipping point |
| 语种 | 中文 |
| 国家 | 中国 |
| 来源学科分类 | 生态学 |
| 来源机构 | 中国科学院西北生态环境资源研究院 |
| 资源类型 | 学位论文 |
| 条目标识符 | http://119.78.100.177/qdio/handle/2XILL650/287841 |
| 推荐引用方式 GB/T 7714 | 陈宁. 沙坡头人工固沙植被系统的跃变特征及早期信号分析[D]. 中国科学院大学,2017. |
| 条目包含的文件 | 条目无相关文件。 | |||||
| 个性服务 |
| 推荐该条目 |
| 保存到收藏夹 |
| 导出为Endnote文件 |
| 谷歌学术 |
| 谷歌学术中相似的文章 |
| [陈宁]的文章 |
| 百度学术 |
| 百度学术中相似的文章 |
| [陈宁]的文章 |
| 必应学术 |
| 必应学术中相似的文章 |
| [陈宁]的文章 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
