Adaptive turbulence modelling to improve high-impact weather forecasts in next generation atmospheric models

高影响天气事件往往是极端局地化的，因此精确的空间分辨率对于这类事件的准确预测至关重要。然而，随着气象模式向亚公里网格间距移动，数值天气预报（NWP）可能陷入僵局。尽管最近的研究表明，随着水平分辨率的提高，对暴雨事件的模拟有了一些改进，但它也揭示了重大的挑战，因为这种改进并不像预期的那样明显，而且对未解决的湍流长度尺度的处理非常敏感。这些未解运动对应于边界层湍流和云发展的主要尺度，并与其临近的环境混合。亚千米分辨率的子网格运动参数化背后的基本假设似乎需要重新审视。提议的研究项目旨在通过下一代亚公里NWP模式中更物理和动态的子网格尺度表示，为预测深度对流和随后的强降雨提供一个逐步变化的能力。这将通过根据已分解的流场动态推导亚网格混合方案中的湍流长度尺度来实现，而不是事先静态地指定它们。动力学方法将首先在一个理想的框架中使用，通过诊断边界层和云层中不同长度尺度的混合，来提高对大气边界层与深层对流之间耦合的理解。该方法可以进一步深入了解云环境混合，研究对流发展不同阶段湍流混合的影响，并确定湍流输送与天气扰动之间的反馈。然后，一阶动态方案将用于预测，并将在亚公里分辨率下与静态常规方法对比再现对流发展时进行评估。下一步，我将开发一种新颖的、尺度感知的、流动自适应的动态子网格参数化方法，通过使用已解决的尺度来确定子网格混合的强度，从而更好地表示未解决的尺度。本文将利用亚网格湍流输运守恒方程，通过重建网格尺度附近的解析场来动态计算亚网格湍流混合长度，为亚网格运动提供更准确的表示。因此，该方案将是自包含的，具有最小的可调闭包参数。新方法将在高亚千米分辨率的运行NWP模式中进行测试，以验证模式动力学在弱和强天气强迫的实际案例研究中明确解决深层对流问题的能力。这种新方法有可能改善天气预报，使气象中心能够向决策者和公众提供更准确的高影响天气预报，同时为大气科学的进一步研究提供依据。

High-impact weather events are often extremely localised therefore refined spatial resolution is essential for the accurate prediction of such events. However, Numerical Weather Prediction (NWP) might have hit a stalemate as meteorological models move towards the sub-kilometre grid spacing. Even though recent research has shown some improvements in the simulation of heavy rainfall events with increasing horizontal resolution, it has also revealed significant challenges as this improvement is not as pronounced as expected and very sensitive to the treatment of the unresolved turbulence length scales. Those unresolved motions correspond to the dominant scales of boundary layer turbulence and cloud development and mixing with its imminent environment. It seems that the fundamental assumptions behind the parametrization of sub-grid motions at sub-kilometre resolutions need to be revisited.The proposed fellowship aims to provide a step-change in capabilities for forecasting deep convection and subsequent heavy rainfall, through a more physical and dynamic representation of the sub-grid scales in the next generation sub-kilometric NWP models. This will be achieved by dynamically deriving the turbulence length scales in the sub-grid mixing scheme depending on the resolved flow field rather than statically specifying them beforehand. The dynamic method will be first used in an idealised framework to improve the understanding of the coupling between the atmospheric boundary layer with deep convection, by diagnosing the different length scales of mixing in the boundary and cloud layer. This approach can provide further insight on cloud-environment mixing to study the impact of turbulent mixing at the different stages of convection development and identify the feedback between turbulent transport and the synoptic disturbances. A first-order dynamic scheme will then be used prognostically and will be assessed in reproducing convection development against static conventional methods at sub-kilometre resolutions.As a next step, I will develop a novel, scale-aware and flow-adaptive dynamic sub-grid parametrization approach to better represent the unresolved scales by using the resolved scales to determine the intensity of the sub-grid mixing. It will utilise the conservation equations for sub-grid turbulent transport, to provide a more accurate representation of sub-grid motions, through reconstructing the resolved field near the grid scale to dynamically calculate the sub-grid turbulence mixing lengths. Hence, the scheme will be self-contained with minimum tuneable closure parameters. The new approach will be tested in an operational NWP model at very high, sub-kilometre resolutions to validate the ability of model dynamics to explicitly resolve deep convection in realistic case studies, under weak and strong synoptic forcing. This new method, has the potential to improve weather forecasting by enabling weather centres to provide more accurate forecasts of high-impact weather to policy makers and the general public while providing grounds for further research in atmospheric science.