EAGER: An International, Dedicated High-End Collaborative Project to Revolutionize Climate Modeling

目前的全球气候模式必须参数化大气对流，因为现有的专用计算资源还不足以在足够精确的分辨率下运行模式，以明确模拟对流。该项目将探索为超高分辨率实验提供专用高端计算支持是否能够真正加速气候模拟领域的进展。这个方法将是三管齐下的：1。最先进的高分辨率数值天气预报模式将在多年时间尺度上运行，以评估高分辨率对系统误差的影响。具体而言，全球数值天气预报模式将以10-15公里的分辨率运行，并将分析以高分辨率模拟天气统计（包括恶劣天气）的能力的提高。这种整合不仅与气候变化问题相关，而且对数值天气预报中心指导其未来的业务策略，以满足月预报和季节预报模式对高分辨率模式的需求具有重要意义。时间片气候变化模拟将使用现有的气候模型，以比通常使用的高得多的分辨率运行。专门的计算资源将用于运行最先进的10-15公里分辨率的全球模型，以模拟20世纪和21世纪的情况。将对这些运行进行分析，重点是温室气体对天气、极端事件和水文循环统计数据变化的影响，并与缩小到区域尺度的100-150公里分辨率的相同运行进行比较。允许对流的大气模式能够解析大气中的云系统（4-8公里网格）和海洋中的高能涡流（10公里网格），将用于评估解析这些过程对季节性气候模拟的影响。特别是，全球云系统解析大气模式将在两层、区域耦合模式下使用，以产生一系列的后预测，与使用传统分辨率模式做出的后预测进行比较。预测的表面热通量、动量通量和淡水通量将用于驱动一个漩涡分辨率海洋模型，以确定模拟的海洋气候是否与使用较粗分辨率通量模拟的海洋气候有显著不同。一个国际专家小组已经成立，以提供模型，对实验设计提出建议，并努力运行和评估模型模拟。该研究具有很强的变革性和影响计算地球流体动力学未来方向的潜力。这项研究是探索性的，工作处于早期阶段，尚未在符合气候过程物理和动力学的时间和空间尺度上进行测试。因此，这是一项高风险、高回报的研究。更广泛的影响：国际天气和气候模型界在2008年世界模型峰会（WMS）上聚集在一起，达成共识，认为通过开发统一天气和气候预测问题的无缝预测方法，彻底改变数值模型在气候预测中的应用的时机已经成熟。其基本原理包括：a.认识到当前的气候模型在模拟地球气候的显著特征方面已经达到了一个平台期；减少对未来气候变化预估的不确定性的社会需求；对未来30年可能预测的气候变化，特别是极端事件的变化进行更大的空间区分的补充需求；一个假设可以解决大气、海洋和陆地表面的重要过程，以及它们之间的相互作用，就像天气预报模型中已经出现的那样，可以极大地提高模型的保真度。

Atmospheric convection must be parameterized in present day global climate models because available dedicated computational resources have not been sufficient to run the models at fine enough resolution to simulate convection explicitly. This project will explore whether dedicated high-end computing support for ultra high resolution experiments can truly accelerate progress in the area of climate modeling. The approach will be three-pronged:1. State-of-the-art high-resolution numerical weather prediction models will be run on multi-year timescales to assess the impact of high resolution on systematic error. Specifically, global numerical weather prediction models will be run at 10-15 km resolution with 20th century forcings and analysis will be performed on the improved ability to simulate the statistics of weather, including severe weather, with high resolution. As well as being relevant for the climate change problem, such integrations will be of importance to numerical weather prediction centers in guiding their future operational strategies on the need for high resolution models in monthly and seasonal forecast mode.2. Time-slice climate change simulations will be made using available climate models run at strikingly higher resolution than is typically used. Dedicated computational resources will be used to run state-of-the-art global models at 10-15 km resolution to produce simulations for the 20th and the 21st centuries. Analysis of these runs will be done, focusing on the impact of greenhouse gases on changes in the statistics of weather, extreme events and the hydrologic cycle, in comparison with identical runs made at 100-150 km resolution, downscaled to regional scales.3. Convection-permitting atmospheric models capable of resolving cloud systems in the atmosphere (4-8 km grids) and energetic eddies in the ocean (10 km grids) will be used to evaluate the impact of resolving these processes on simulation of seasonal climate. In particular, global cloud-system-resolving atmospheric models will be used in two-tier, regionally-coupled mode to produce a series of hindcasts to be compared with hindcasts made using conventional resolution models. The predicted surface fluxes of heat, momentum and fresh water will be used to drive an eddy-resolving ocean model, to determine if the simulated ocean climate is significantly different from that simulated using coarser resolution fluxes.An international team of experts has formed to contribute models, advice on experimental design, and effort to run and evaluate the model simulations. The research has strong potential to be transformative and influence the future directions of computational geo-fluid-dynamics. The research is exploratory, the work is in its early stages and untested on the time and spatial scales consistent with climate process physics and dynamics. It is therefore, high-risk-high payoff research. Broader Impact: The international weather and climate modeling community came together in 2008 at the World Modeling Summit (WMS) to reach a consensus that the time is ripe to revolutionize the application of numerical models for prediction of climate through the development of seamless prediction methodologies which unify the weather and climate forecast problems. The rationale includes: a. a recognition that the climate models of the current generation have reached a plateau in their ability to simulate salient features of Earth's climate,b. a societal demand for reducing the uncertainty in projections of climate change in the future,c. a complementary demand for greater spatial discrimination in the climate changes that may be anticipated in the next 30 years, especially concerning changes in extreme events, andd. a hypothesis that resolving important processes in the atmosphere and ocean and at the land surface, as well as interactions among them, as is already the case in weather prediction models, can dramatically improve the fidelity of the models.