旱区生态系统受水资源匮乏、低植被覆盖率等环境条件的限制，对气候变化和人类活动的干扰十分敏感，生态环境非常脆弱。全球变化背景下，中国旱区生态系统面临着前所未有的压力，生态系统处在急剧变化中，生态系统结构和功能的非线性变化易导致生态系统发生稳态转换，影响区域生态安全与可持续发展。稳态转换会改变生态系统的反馈调节机制，影响生态系统风险响应能力，转换后的系统因具有较高的稳定性而难以逆转。土壤水蚀是中国旱区典型的退化风险类型，也是生态系统治理的重点内容。因此，分析生态系统多稳态及稳态转换特征，界定生态系统阈值并对生态系统突变进行早期预警，探讨稳态转换对水蚀的影响，对于旱区生态系统的突变应对和管理至关重要。本研究在多稳态的理论框架下，系统评估3种植被指数、6种遥感地表反照率与旱区生态系统多功能性的关系，选取能够反映中国旱区生态系统多稳态特性的遥感指标；分析了中国旱区生态系统稳态数目、边界阈值以刻画其多稳态特征，辨识生态系统稳态转换的早期预警信号；通过干湿状况动态和植被变化识别生态系统稳态转换临界点及其临界特征；进而探究稳态转换前后水蚀变化规律，以揭示稳态转换对生态系统水蚀风险的影响。主要结论如下：

（1）遥感地表反照率与地面生态系统多功能数据具有显著相关性。植被指数与地下净初级生产力、植物物种丰富度呈显著正相关，然而植被指数与土壤微生物多样性无显著相关性，并且植被指数与土壤多功能性的相关性小于反照率与土壤多功能性的相关性。地表反照率与地下净初级生产力、植物物种丰富度、土壤微生物多样性和土壤多功能性均具有显著相关性。其中，短波和可见光反照率与地下净初级生产力、植物物种丰富度呈显著负相关，近红外反照率与土壤微生物多样性呈显著正相关。此外，黑空短波反照率与土壤多功能性无显著相关关系，而三种白空反照率均与土壤多功能性呈显著负相关，因此白空反照率可以更好的反映中国旱区生态系统多功能性。

（2）中国旱区生态系统存在两种稳态，分别为低反照率态和高反照率态。反照率的多稳态分析结果呈现一致性规律：两种稳态在干旱度为0.58-0.82同时出现（即双稳态共现区），并在干旱度0.69时发生稳态转换。在双稳态共现区内，当干旱度高于0.69时，生态系统开始由低反照率态转换为高反照率态，植被覆盖度迅速下降，此时系统倾向于从森林/灌木/草地混合植被和稀疏植被退化为无树草地。中国旱区生态系统退化的早期预警信号表现为干旱水平位于0.69-0.82、高反照率态的势能值小于低反照率态势能值。

（3）中国旱区生态系统在1997-1999年期间跨越临界阈值发生了稳态转换，临界点后植被弹性和恢复力均减小。1997-1999年为中国旱区干湿状况临界点。旱区西部由变湿转为变干趋势，东部变湿趋势加剧。临界点前、后，标准化降水蒸散发指数均值自西向东由增大转为减小趋势。与研究区干湿状况的突变现象一致，植被系统在临界点附近也发生了显著变化。临界点前、后，NDVI、LAI和NPP三种植被指数均发生显著变化，其中45%的区域呈现植被绿度、覆盖度和生产力的不同步变化。同时，在临界点之后，生态系统弹性和恢复力均呈下降趋势。

（4）中国旱区水蚀状况在稳态转换前（1985-1995 年）、后（2000-2018年）呈现相反的变化趋势。稳态转换前、后水蚀状况呈现不同的变化趋势：稳态转换前旱区水蚀面积减少，侵蚀强度降低；稳态转换后水蚀面积增加，侵蚀强度增加。稳态转换后期，中国旱区水蚀风险增加15%，降雨侵蚀力（R因子）呈增加趋势，植被覆盖和管理因子（C因子）呈减小趋势。大部分地区R因子的变化率大于20%，而C因子变化率通常低于20%，R因子比C因子的变化更为显著。R因子为水蚀动态的主要贡献因子，除东南区域外，其余区域的R因子贡献率均大于C因子贡献率。但C因子贡献率呈增加趋势的区域面积大于R因子贡献率增加的区域面积，植被动态对于水蚀风险变化产生愈加重要的影响作用。

The dryland ecosystem is very sensitive to climate change and human activities because of the limitation of environmental conditions such as lack of water resources and low vegetation coverage. The ecological environment is very fragile. Under the global changes, China's dryland ecosystem is facing unprecedented pressure. The ecosystem is in rapid change, and the nonlinear change of ecosystem structure and function is easy to lead to the regime shifts in ecosystem, which affects the regional ecological security and sustainable development. In the process of regime shift, ecosystem structure is reorganized and its feedback regulation mechanism changes, which changes ecosystem risk response process. Soil water erosion is a typical degradation risk in China’s dryland and is also the key control object of many ecological projects. Therefore, it is very important for response and management of ecosystem abrupt changes to explore the characteristics of alternative stable states and regime shifts, define ecosystem thresholds and detect early warning of ecosystem changes, and discuss the influence of regime shift on water erosion. Therefore, we systematically evaluated the relationships between 3 vegetation indices, 6 remote sensing surface albedo indices and the multi-function of dryland ecosystem, in order to select remote sensing index that can reflect the multi-stable-state characteristics of dryland ecosystem in China. To characterize the alternative stable states, this study detected the number of stable states and their thresholds in China's dryland ecosystem under the theoretical framework of alternative stable states and developed the early warning signals for regime shifts. And then we identified ecological state tipping points and their changing characteristics through dynamic in dry-wet conditions and vegetation indices. Finally, this study analyzed the water erosion changes before and after regime shift to reveal the impact of regime shift on water erosion risk. The main conclusions are as follows:

（1）     The remote sensing surface albedo has significant correlation with the ecosystem multifunctionality. There was a significant positive correlation between vegetation indices and subsurface net primary productivity and plant species richness, but there was no significant correlation between vegetation indices and soil microbial diversity, and the correlation between vegetation index and soil multifunctionality was smaller than that between albedo and soil multifunctionality. Surface albedo was significantly correlated with subsurface net primary productivity, plant species richness, soil microbial diversity and soil multifunctionality. The shortwave and visible albedo were significantly negatively correlated with net primary productivity and plant species richness, while the near infrared albedo was significantly positively correlated with soil microbial diversity. In addition, the shortwave black-sky albedo has no significant correlation with soil multifunctionality, while the three white-sky albedo indices have significant negative correlation with soil multifunctionality. Therefore, white-sky albedo can better reflect the ecosystem multifunctionality in China’s dryland.

（2）     There are two stable states in China’s dryland, those are low albedo state and high albedo state. The remote sensing surface albedo had significant correlation with the multifunctional data of ecosystem. The potential analysis results of albedo are consistent: the two stable states appeared simultaneously at the aridity level of 0.58 to 0.82 (i.e., the bistable zone), and discontinuous mutation appeared at the aridity level of 0.69. In the bistable zone, when the aridity level was higher than 0.69, the ecosystem began to change from low albedo state to high albedo state, and the vegetation coverage decreased rapidly, at this time, forest/shrub/grassland mixed vegetation and sparse vegetation tended to degenerate to treeless grassland.

（3）     During 1997-1999, the China’s dryland ecosystem experienced a regime shift across the critical threshold, after which the vegetation resilience and recovery decreased. The period of 1997-1999 was the tipping point of the dry and wet conditions of the dryland ecosystem in China. The wet trend changed to a dry trend in the west, and the wetness changed to aggravate in the east. The mean value of SPEI showed opposite longitude variation before and after the tipping point, and SPEI changed from increasing to decreasing with the increase of longitude. Consistent with the abrupt change of the dry and wet conditions in the study area, the vegetation system also changed significantly near the tipping point. The vegetation indexes of NDVI, LAI and NPP changed significantly before and after the tipping point. It is worth noting that there were still up to 45% of the grid areas showing the asynchronous change of vegetation greenness, coverage and productivity. At the same time, after the tipping point, the resilience and recovery of the ecosystem showed a decreasing trend.

（4）     The soil water erosion condition in China’s dryland showed an opposite trend before (1985-1995) and after (2000-2018) the regime shift. The water erosion area and the erosion intensity decreased in the typical area of the dryland before the regime shift, and the water erosion area and the erosion intensity increased after the regime shift. After the regime shift, the water erosion risk increased by 15%. Rainfall erosivity (R factor) showed an increasing trend, while vegetation cover and management (C factor) showed a decreasing trend. In most areas, the percent change of R factor is greater than 20%, while the percent change of C factor is usually lower than 20%. The change rate of R factor is more significant than that of C factor. R factor is the main contributing factor of water erosion dynamics, except the southeast region, the contribution of R factor is greater than that of C factor. However, the area where the contribution of C factor increased was larger than that of R factor. Vegetation dynamics had a more important effect on the change of water erosion risk.