浅水湖泊生态系统稳态转换的阈值判定方法

浅水湖泊生态系统对人类干扰的反应会随着干扰力度的改变或增强而出现突然的变化,即发生稳态转换；对其机理和驱动机制的揭示将有助于对湖泊富营养化的控制及恢复.基于“多稳态”理论的稳态转换研究已广泛开展,但对浅水湖泊生态系统稳态转换的驱动机制结论各异,采用的阈值判定方法相差很大,主要有实验观测、模型模拟和统计分析3种.实验观测多关注少数特定指标,指标筛选过程复杂且工作量大；模型模拟虽能从较为全面的尺度上理解生态系统稳态变化的特征和主要机理过程,但在模型误差和不确定性的处理等问题上尚存在不足；统计分析方法基于对长时间序列数据的统计变化规律分析,用以判断或者预警稳态转换现象的发生,是目前最为常用的方法.目前稳态转换领域的研究大都是对已发生的稳态转换进行机制分析或过程反演,对未来预测与预警的问题仍然亟需加强.

Survey on threshold detection methods of regime shift in shallow lake ecosystem

Lake ecosystems can undergo catastrophic regime shifts when there are abrupt state changes from one state to another. Since its appearance, it has occurred in many other complex systems such as socio-economic systems and climate system. There are no obvious omens before regime shift occurs, namely the process of regime shift is abrupt and catastrophic. Regime shift has been observed worldwide for the lake ecosystem. It is thus becoming crucial to examine the driving mechanisms behind the regime shift in the lake ecosystem management, not only for predicting how a lake will respond to undesirable human disturbances, but also for helping lake managers determine the level of efforts for lake restoration. Previously, there were 3 types of methods used to detect and identify the threshold of regime shifts. Field observation method was growing in the past years through monitoring large-scale abrupt changes occurred in the field, however, it can only target on a limited number of specific indicators and its indicator selection process is complicated. Therefore regime shifts detection or foreshadow using experimental observation is limited. Simple mechanism models are able to address the catastrophic transitions under varied control parameters as well as the nonlinearity and large external fluctuation features of ecosystems through novel early warning signals, however, they are unable to precisely detect the threshold of regime shifts due to the model deviations and uncertainties. Previous studies revealed that either reduced resilience or increased external fluctuation could turn ecosystems to an alternate stable state. The changes in asymmetry in the distribution of time-series data can be used as model-independent and reliable early warning signals for both regime shift routes, and they can be quantified by changing the variance, changing skewness recovery rate, conditional heteroscedasticity and auto-correlation. The studies proved that statistic analysis of long time series data would be the useful and common method of regime shift detection, thanks to the advantage that statistical analysis method is independent of complex mechanism of lake ecosystems as well as the fact that it is relatively easy and reliable to analyze a long time series data. Comparing with these two methods, although statistical analysis can successfully analyze the occurred regime shift, it may not forecast a regime shift before it happens. In conclusion, limited to the long-term observation complexity and difficulty selecting monitoring indicators, experimental method is not firstly preferred. As for mechanism model, despite the fact that model analysis can simulate the internal structure and mechanism of the lake ecosystem, building the model requires to have a good grasp of the complex mechanism of lake ecosystems, and the parameter estimation and model validation needs a long time scale data and uncertainty analysis is of great difficulty. Compared with above two methods, although statistical analysis can successfully analyze the occurred regime shift, it may not forecast a regime shift before it happens. Therefore, an integrated approach through combing mechanism models and statistical analysis method into a general framework was suggested in the paper in order to overcome the limitations of both approaches. The focuses of this study would be on (a) enhancing the mechanism analysis of shallow lake ecosystem, (b) addressing the uncertainties in model simulations, and (c) improving the detection methods for regime shifts in shallow lake ecosystem.