Survey on threshold detection methods of regime shift in shallow lake ecosystem
LI Yuzhao,LIU Yong,ZHAO Lei,ZOU Rui,WANG Cuiyu,GUO Huaicheng
DOI: https://doi.org/10.5846/stxb201203140340
2013-01-01
Abstract: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.