国家公园是中国自然保护地中最重要的类型，建立国家公园体制是我国生态文明体制建设的重要内容。稳定性的维持对国家公园生态系统保持或恢复自身结构和功能、防止物种灭绝、维持生态系统服务具有关键作用。然而，在气候和土地利用/覆盖变化的双重驱动下，生态系统的稳定性可能发生改变。因此，需要通过模型预测的方法预知国家公园及其周边区域未来的环境变化以及这种变化所引起的生态系统稳定性的改变，进而制定合理的措施以应对未来气候变化对国家公园生态系统的影响。基于以上背景，考虑“保险假说”，本研究构建了“未来环境变化模拟”——“物种栖息地分布变化模拟”——“基于栖息地破碎化和食物网结构的稳定性评价”——“稳定性维持方案构建”的方法体系，以期对国家公园稳定性变化预测及维持提供方法体系参考。
本研究选择的历史时期为2000-2017年的近20年时间。这一时期的数据主要用于建模，并与未来情况做对比。未来气候变化情景为SSP3-7.0和SSP5-8.5情景，时间分别为2030年和2060年。通过Delta统计降尺度方法对历史时期以及CMIP6气候模式输出的SSP3-7.0情景和SSP5-8.5情景的气候数据进行降尺度。通过FLUS模型对湖泊、永久积雪、耕地、城乡工矿居民用地进行模拟。通过BP神经网络方法，模拟归一化植被指数NDVI的变化情况。
本研究的稳定性以食物网结构和物种栖息地的空间破碎化/连通性来表征。首先，结合环境变化模拟结果，通过获取的物种分布的点位数据，考虑气候、地形和资源等因素，运用最大熵模型模拟大型食肉动物、小型食肉动物、大型食草动物、小型食草动物和鸟类等28种三江源物种在历史时期以及SSP3-7.0和SSP5-8.5情景下2030年和2060年的空间分布情况。将研究区分为20个单元，基于物种之间的捕食关系分规划单元、分情景、分时间构建食物网，并统计食物网节点数、连接数、连接密度等参数。运用Fragstats软件，计算物种栖息地在不同情景和不同时间气候与土地利用/覆盖变化下被分割和聚合的程度。选择景观格局指数中的斑块占景观面积的比例、斑块数量、边界密度、最大斑块指数、形状指数、周长-面积分维数、聚合度指数、结合度指数、景观分离度等。
基于“保险假说”，即生物多样性有助于在面对环境波动时维持稳定性，因此本研究稳定性维持的目标是增强食物网的多样性（复杂程度）以提升/维持生态系统的稳定性。考虑到建设用地、耕地和道路等对栖息地的破碎化以及对物种交流的阻隔作用，本研究的稳定性维持方案部分旨在通过运用电路理论和Linkage Mapper软件构建生态廊道加强捕食者与被捕食者在被分隔的栖息地之间的流动，增强捕食者与被捕食者之间相互作用。本研究取得的主要结论如下：
未来环境变化方面，根据BCC-CSM2-MR降尺度结果，SSP3-7.0情景和SSP5-8.5情景温度和降水量都呈现增加趋势，SSP5-8.5情景增加的更多；土地利用/覆盖变化方面，SSP3-7.0情景和SSP5-8.5情景下积雪冰川面积均呈现减少趋势，湖泊面积、建设用地面积、耕地面积和NDVI也呈现增加趋势。
从物种栖息地连通性和破碎化程度来看，在未来需要得到进一步关注和保护的主要是濒危的大型食肉或食草动物。对于食肉动物来说，栖息地面积随时间减少，破碎化增加。对于食草动物来说，聚集度和连接度也都降低栖息地面积下降，并且破碎化程度加重，最大斑块面积降低，聚集度、连接度也都呈现降低趋势，分离度增加。对于食物网结构的变化，总体上食物网结构有变简单的趋势，但是变化并不剧烈。SSP5-8.5情景中，节点数呈减少趋势的斑块比SSP3-7.0的斑块更多。SSP3-7.0和SSP5-8.5情景，连接数的变化则是比较接近的。
生态系统稳定性维持构建方案中，本研究所构建的历史时期大型捕食者与被捕食者相互作用关系维持的生态廊道共46条；SSP3-7.0情景下2030年43条，2060年57条；SSP5-8.5情景下2030年48条、2060年51条。所构建的历史时期小型捕食者与被捕食者相互作用关系维持的生态廊道共62条；SSP3-7.0情景下2030年73条、2060年70条；SSP5-8.5情景下2030年90条、2060年70条。不同时间和情景的捕食者与被捕食者相互作用关系维持的生态夹点区域主要分布在研究区的黄河源园区、研究区南部和研究区西南部。

National parks are the most important type of protected area in China, and the establishment of national park system is an important part of the construction of ecological civilization system in China. The maintenance of stability plays a key role in maintaining or restoring the structure and function of the national park ecosystem, preventing species extinction, and maintaining ecosystem services. However, under the dual driving forces of climate and land use/cover change, the stability of the ecosystem may change. Therefore, it is necessary to predict the future environmental changes in the national park and its surrounding areas and the changes in ecosystem stability caused by such changes through model prediction. Then, reasonable measures should be formulated to cope with the impact of future climate change on the national park ecosystem. Based on the above background and taking into account the “insurance hypothesis”, this study has established a methodology of “simulation of future environmental change” - “simulation of species habitat distribution change” - “stability assessment based on habitat fragmentation and food web structure” - “construction of stability maintenance plan”, with a view to providing a methodological reference for the prediction and maintenance of stability change in national parks..
The historical period selected for this study is nearly 20 years from 2000 to 2017. The data of this period is mainly used for modeling and comparison with the future situation. The future climate change scenarios are SSP3-7.0 and SSP5-8.5, and the period are 2030 and 2060. The climate data of the SSP3-7.0 scenario and SSP5-8.5 scenario output by the CMIP6 climate model in the historical period are scaled down by the Delta statistical scaling method. The Future Land Use Simulation (FLUS) model is used to simulate changes of lakes, permanent snow cover, arable land, urban and rural industrial and mining residential land. The change of NDVI was simulated by back propagation (BP) neural network.
The stability of this study is characterized by the structure of food web and the spatial fragmentation/connectivity of species habitat. First, combined with the simulation results of environmental change, the maximum entropy model was used to simulate the spatial distribution of 28 species from the Three-River-Source Region, including large carnivores, small carnivores, large herbivores, small herbivores and birds, in the historical period and in 2030 and 2060 under the scenarios of SSP3-7.0 and SSP5-8.5 by taking into account the factors of climate, terrain and resources. The study area was divided into 20 units. Based on the predator-prey relationship between species, the food network is constructed by planning units, scenarios, and time, and the parameters such as the number of nodes, connections, and connection density of the food network are counted. The Fragstats software was used to calculate the degree of fragmentation and aggregation of species habitat under different scenarios and different times of climate and land use/cover changes. The selected landscape metrics include Percentage of Landscape, the number of patches, the edge density, the largest patch index, the shape index, the perimeter area fractal dimension, the aggregation index, the patch cohesion index, the landscape sdivision index, etc.
Based on the “insurance hypothesis”, that is, biodiversity helps to maintain stability in the face of environmental fluctuations, therefore the goal of stability maintenance in this study is to enhance the complexity of the food web to improve/maintain the stability of the ecosystem. Taking into account the fragmentation of habitat and the blocking effect of construction land, arable land and roads on species exchange, the optimization part of this study aims to strengthen the flow of predators and prey among separated habitats and enhance the interaction between predators and prey by using circuit theory and Linkage Mapper software to build ecological corridors. The main conclusions of this study are as follows:
In terms of future environmental changes, according to the BCC-CSM2-MR downscaling results, the temperature and precipitation of SSP3-7.0 and SSP5-8.5 scenarios show an increasing trend, and SSP5-8.5 scenarios increase more; In terms of land use/cover change, the area of snow cover under SSP3-7.0 and SSP5-8.5 scenarios shows a decreasing trend, and the area of lakes, construction land, arable land and NDVI also shows an increasing trend.
From the perspective of the connectivity and fragmentation of the species habitat, the endangered large carnivores or herbivores that need further attention and protection in the future. For carnivores, habitat area decreases and fragmentation increases over time. For herbivores, the degree of aggregation and connectivity also decreased the habitat area, and the degree of fragmentation increased, the largest patch area decreased, the degree of aggregation and connectivity also showed a downward trend, and the degree of separation increased. As for the changes in the structure of the food web, the structure of the food web tends to be simpler, but not dramatic. In the SSP5-8.5 scenario, the number of patches with a decreasing trend is more than that of SSP3-7.0. In the SSP3-7.0 and SSP5-8.5 scenarios, the change of the number of connections is relatively close.
In the construction scheme of ecosystem stability maintenance, 46 ecological corridors were constructed to maintain the interaction between large predators and their prey in the historical period; 43 and 57 ecological corridors for 2030 and 2060 under SSP3-7.0 scenario; 48 and 51 ecological corridors for 2030 and 2060 under SSP5-8.5 scnario. 62 ecological corridors for the maintenance of the interaction between small predators and prey in historical period; 73 and 70 ecological corridors for 2030 and 2060 under SSP3-7.0 scenario; 90 and 70 ecological corridors for 2030 and 2060 under SSP5-8.5 scenario. The ecological pinch areas are mainly distributed in the Yellow River Source Park, the south and the southwest of the study area.