Model Integration Techniques for Sustainability Decision-Making

地球系统模式提供了物理气候系统的详细模拟，以及它如何对人类活动（如温室气体排放）作出反应。然而，目前的建模能力与决策者的需求之间存在很大差距，决策者必须制定基于证据的气候变化应对措施。为了缩小这一差距，我们需要将不同的计算模型结合在一起，以推理跨越自然科学和社会科学多个专业领域的系统相互作用，我们需要更好的工具来表征模型中不确定性的本质以及这种不确定性对决策的影响。本研究将探讨如何构建、验证和使用“系统的系统”的综合模型，为在气候变化的世界中建立有弹性、可持续的社区的挑战提供决策支持。这将需要探索自然和社会系统之间的相互依赖关系，以及对复杂动态行为进行建模和推理的能力，并整合由不同专家小组构建的模型组件，每个小组都有自己的假设和局限性。首先，我将比较气候科学家使用的模型整合技术（我已经深入研究过）与其他领域采用的方法。。例如，气候模型开始于对大气和海洋热力学的模拟，但已经扩展到包括与其他系统的相互作用，如冰盖、植被和土壤、大气化学和人类活动。同样，城市规划中使用的模型必须考虑到土地使用、人口、交通、能源、水、废物等方面的相互作用。对城市系统建模的初步评估表明，综合建模目前很少见，但由于不同城市系统之间的行为相互作用会影响城市整体的复原力和可持续性，因此越来越需要综合建模来支持可持续发展的长期决策。该研究将集中于用于模型集成的软件工具和技术，以及科学家围绕测试和部署集成模型的实践，包括向模型和模型结果的下游用户传达不确定性。为了应对这一挑战，我将采用一种混合方法，使用多个案例研究来提供一种比较和对比的方法，行动研究与现有的建模社区合作来解决具体的挑战，并进行试点研究来评估新的解决方案。本研究的长期目标是为可持续发展的循证决策奠定更坚实的基础，并帮助克服跨学科问题的学科障碍和研究孤岛。一个指导性的假设是，模型整合倡议可以加速向跨学科思维的转变，因为它们为发现和解决不同学科背景的科学家之间在假设、术语和方法上的差异提供了一个焦点。从这个意义上说，计算模型作为科学理论的详细、明确的描述，并将它们整合起来，促进了更深层次的跨学科理解的发展。这项工作的实例应用包括：减缓和适应气候变化的政策制定、可持续城市的城市规划以及可持续农业。本研究的预期结果包括集成技术的类型学和科学模型的软件架构，集成模型的科学验证的最佳实践的识别，以及从长远来看，用于系统的系统预测建模的软件框架。

Earth system models provide detailed simulations of the physical climate system and how it responds to human activities, such as emission of greenhouse gases. However, there is a large gap between current modelling capabilities, and the needs of decision-makers who must develop evidence-based responses to climate change. To close this gap, we need the ability to bring together disparate computational models to reason about systems interactions that span multiple areas of expertise across the natural and social sciences, and we need better tools to characterization the nature of uncertainty in the models and the implications of this uncertainty for decision-making. The proposed research will investigate how to construct, validate and use integrated models of systems-of-systems to provide decision support for the challenge of building resilient, sustainable communities for a climate changed world. This will require an exploration of inter-dependencies across natural and social systems, and the ability to model and reason about complex dynamical behaviour and integrate model components built by diverse groups of experts, each with its own assumptions and limitations. I I I will begin by comparing the model integration techniques used by climate scientists (which I have already studied in depth) with approaches taken in other areas. For example, climate modelling began with simulations of the thermodynamics of atmosphere and oceans, but has expanded to incorporate interactions with other systems, such as ice sheets, vegetation and soils, atmospheric chemistry, and human activities. Similarly, the models used in urban planning must take into account the interaction of land use, demographics, transport, energy, water, waste, etc. An initial assessment of urban systems modelling indicates that integrated modelling is currently rare, but increasingly in demand to support longer term decision-making for sustainability, as the interaction of behaviours across different urban systems affect the resilience and sustainability of the city as a whole. The study will focus on the software tools and techniques used for model integration, and the scientists' practices around testing and deploying integrated models, including the communication of uncertainty to downstream users of models and model results.To tackle this challenge, I will apply a mixed-methods approach, using multiple case studies to provide a compare-and-contrast approach, action research to work with existing modelling communities to tackle specific challenges, and pilot studies to evaluate new solutions. The long term aim of this research is build a stronger foundation for evidence-based decision-making for sustainability, and to help overcome disciplinary barriers and research silos for trans-disciplinary problems. A guiding hypothesis is that model integration initiatives can accelerate the move to trans-disciplinary thinking, as they provide a focus for the process of discovering and resolving differences in assumptions, terminology, and methodology between scientists with different disciplinary backgrounds. In this sense, computational models act as detailed, explicit descriptions of scientific theories, and integrating them stimulates the development of a deeper cross-disciplinary understanding.Example applications for this work include policymaking for mitigation and adaptation to climate change, urban planning for sustainable cities, and sustainable agriculture. Expected outcomes of this research include a typology of integration technologies and software architectures for scientific models, identification of best practices for scientific validation of integrated models, and, in the longer term, software frameworks for predictive modelling of systems-of-systems.