Uncertainty analysis and detection of structural changes in environmental systems dynamics for a better understanding of climatic and anthropogenic risks

数字技术的进步以及更快的处理速度促使人们使用复杂的环境模型进行研究、设计和规划。如今，强大的计算机代码可以用来描述几乎任何感兴趣的环境过程。随着全球变暖灾难性影响的证据以惊人的速度增长，目前的做法是使用这些模型在使用观测数据进行校准后进行影响评估。然而，一致性的问题很少被考虑：在当前气候下表现相对较好的模型是否会在气候变化时提供可信的结果？在水文学中，几乎所有类别的模型都已被使用，不幸的是，很少有人关注所选环境模型在未来气候条件下的适用性。在进行此类研究时，应极其谨慎地使用黑盒模型，因为众所周知，黑盒模型的参数包含了大量与校准条件有关的信息。在这种情况下，分水岭对气候反应的一致性的隐含假设是站不住脚的。原则上，后一种说法在较小程度上适用于物理驱动的和概念性模型，这些模型总是涉及一些参数化和校准。由于全球变暖正在以前所未有的方式推动地球气候，因此需要洞察一个特定的模型如何在未来的气候中发挥作用。

Progress in numerical technologies, as well as greater processing speed, has prompted the use of complex environmental models for research, design and planning purposes. Nowadays, powerful computer codes are available to describe nearly any environmental process of interest. As evidence for catastrophic impact of global warming grows at an alarming rate, the current practice is to use these models for impacts estimation after calibration using observed data. However, the issue of consistency is rarely considered: will a model that is performing relatively well in the current climate provide trustworthy results in a changed climate? In hydrology, nearly all categories of models have been used, unfortunately with very little attention if any to the suitability of the chosen environmental model under future climatic conditions. Black-box models should be used with extreme caution for such studies, since it is well known that the parameters of black-box models capture a lot of information related to the calibration conditions. In this case, the implicit assumption of consistency in a watershed's response to climate does not hold. The latter statement applies, in principle at a lesser extent, to physics-driven and conceptual models, which always involve some parameterisation and calibration. Since global warming is driving earth's climate in an unprecedented manner, insight is needed as to how a particular model can be expected to perform in the future climate.