Collaborative Research: Consensus on Climate Predication by Adaptive Synchronization of Models

该项目的最终目标是融合气候模型，因为它们在21世纪的模拟中运行，从而形成对气候变化细节的共识。拟议的融合“多模型”应该比任何单独的模型或其输出的任何平均值更可靠，因为它将在每个时间点使用每个模型的最佳预测特征。模型的融合是基于混沌系统通常观察到的当它们仅通过几个或多个变量连接时的同步趋势。混沌同步作为一种将持续观测的数据同化到运行模型中的观点已经被提出，有效地使模型与现实同步。在这里，模型将相互吸收信息。由于同步方案通常可以扩展到同步参数和状态，因此可以动态地调整连接不同模型中不同变量对的连接系数。在一系列日益复杂的模型中进行实验后，共识方案最终将被应用于融合三个全复杂的全球气候模型，通过使用20世纪数据进行训练，调整连接系数，自动选择每个模型的最佳预测特征。该项目将通过探索同步理论和发展对如何在实践中最好地进行同步的理解，为实现同步气候模型的目标迈出第一步。这个项目更广泛的影响是提供了一个基于自适应同步的共识形成的例子，它可以应用于任何情况下，使用一系列替代模型来表示任何类型的正在进行的物理过程。特别是，任何新的气候模式都可以通过将其纳入培训计划而加入共识，从而使气候变化预测具有更高的客观性，从而产生更大的影响。

The ultimate goal of this project is to fuse climate models, as they run in a 21st century simulation, so as to form a consensus on the details of climate change. The proposed fused "multi-model" should be more reliable than any of the separate models or any average of their outputs, since it would use the best predictive features of each model at each point in time.The fusion of models is based on the commonly observed tendency of chaotic systems to synchronize when they are connected through only a few of many variables. Synchronization of chaos has been advanced as a view of data assimilation for ongoing observations into a running model, effectively synchronizing the model with reality. Here, the models will assimilate information from each other. Since synchronization schemes can generally be extended to synchronize parameters as well as states, the connection coefficients linking different pairs of variables in different pairs of models can be dynamically adapted. After a series of experiments in a hierarchy of increasingly complex models, the consensus scheme would ultimately be applied to fuse three full-complexity global climate models, with the best predictive features of each selected automatically by adapting the connection coefficients in a training run using 20th century data.This project will take the first steps towards this goal of synchronized climate models, by exploring the theory of synchronization and developing an understanding of how synchronization may best be carried out in practice.Broader impacts of this project are in providing an example of adaptive synchronization-based consensus formation that can be applied in any situation where a collection of alternative models is used to represent an ongoing physical process of any sort. In particular, any new climate model could be added to the consensus by including it in the training scheme, thus giving climate change predictions a higher level of objectivity, with greater resulting impact.