面向多模式跨尺度涌现的复杂生态系统可预测建模方法研究

Complex system oriented theoretical and modeling exploration of ecological processes prediction will open up a new research approach to cognition of complex ecosystem internal structure and evolution mechanism. Furthermore, it would have important significance for revealing the complex multi-scale effects of ecological processes, understanding the effects of habitat fragmentation, species extinction and invasion as well as environmental pollution on the ecological stability. However, based on the complex feature of ecological system such as high dimensional nonlinear, all kinds of closed loop feedback, self-organizing and adaptive evolution, how to effectively predict its behavior, there is a lack of specific research around the world. The aim of this research is that from complex adaptive system perspective instead of traditional modeling method in statistical theory and differential equation, with the combination of ecology, complex adaptive system, artificial intelligence and information theory, to decode multi-scale mechanism and hidden dependencies of ecological process as the basis, to design and organization model objects based on multi-agent technology, to seek the emerging model structure solution and optimization strategy based on pattern information, to achieve the goal of ecosystem prediction under future unknown and complex environment. This research would be able to provide significance insights and methods for the transition of ecological prediction from "physical model" and " phenomenological model" to " intelligent emergence model ".

面向复杂系统的生态过程可预测理论与建模研究将为认知生态系统复杂的内部结构组织与演化动力学机理开辟全新的研究途径，对于系统研究生态过程的复杂多尺度效应，理解生境破碎化、物种灭绝与入侵、环境污染等对生态稳定性的影响具有重要意义。然而，针对生态系统的高维非线性、各类闭环反馈、自组织和自适应特征，如何对其演化进行有效预测，目前国内外尚缺乏具体研究。本项目有机结合生态学、复杂适应系统、人工智能和信息论等多学科知识，改变以微分方程和统计学理论为基础的传统建模思路，以复杂适应系统为视角，以解码生态过程的复杂模式为基础，通过模式信息揭示生态过程的跨尺度作用机制和依赖关系，结合多Agent技术设计和组织模型对象，寻求基于涌现的模型结构求解和优化策略，形成在未知复杂环境下生态系统状态和过程的预测方法和建模理论，为实现生态过程预测从"物理过程模型"和"现象解释模型"向"智能涌现模型"的转变提供全新研究思路。

近年来由于人类的过度活动，环境污染、生境破碎化和栖息地消失等不断冲击生态系统的稳定性。而生态系统与人类社会的生存发展密切相关，如何预测生态系统在环境冲击下的演化路径是一个相当复杂和困难的科学问题。因为生态系统本身的高维非线性、各类闭环反馈、自组织和自适应等复杂特征，使得传统的建模理论与方法难以描述生态系统内部的组织结构和演化动力学机理。具体难点包括如何理解生态系统的复杂模式并建立与模型结构的有机关联，如何使模型有效地表达生态系统的涌现关系并进行优化选型等。本项目以生态模式数据为研究对象，研究了模式的分类结构、表达方式、耦合与解耦理论，建立了基于范式的模式描述体系；并将模式及其在不同组织层次和时空尺度的依赖关系形式化作为POM模型的建模依据，建立对象与可变环境的关联；在此基础上，通过模式信息熵与涌现层次的配准优化模型结构的选型，以典型农业生态系统实测数据为基础验证了模式分析、模型构建、模型选型和预测理论及方法的有效性，为复杂生态系统的可预测模型研究开拓了全新的研究思路和方法，为生态系统管理、决策和科学研究提供了实用的建模工具。