Circle-A: Parametrizing Convection in the Hard Grey Zone: Modelling the Interaction of Turbulent Cloud processes with Explicit Cloud Dynamics.

“无政府状态是秩序之母”，Proudhon声称，他指的是无政府主义运动。许多人可能会质疑这是否适用于人类行为，但它肯定适用于大气。地球上一些最极端的天气与深对流暴雨有关，深对流暴雨是由水蒸气变成液态云水或冰时释放的热量和浮力所驱动的。这些风暴一开始是小的湍流涡流，后来发展成猛烈的雷暴，其中最严重的通过不断再生的过程来保护它们的存在。在适当的条件下，风暴可以组织成更大的结构，如覆盖数百公里并持续数小时的暴风线。最极端、最猛烈、最美丽的有组织对流是热带气旋。然而，这个规模越来越大的组织在一定程度上仍然受到发生在最小动荡尺度上的过程的控制。如果我们在用来预测天气和气候的模型中表现得很差，我们就会严重影响他们预测的准确性。不断增长的计算机能力使我们能够运行更高分辨率的模型，这些模型开始以某种现实的方式表示单个云，但我们远未直接解决控制云的大小、强度、数量等的过程。这个项目的目的是在实用的天气预报模式和气候模式中大幅改进我们表示这些过程对云增长和消散的影响的方式。我们可以通过从物理基本方程的理论推导到目前为止，但这样做会带来更多悬而未决的问题，例如，在只有有限的较大尺度(例如几公里)的平均性质信息的情况下，我们如何预测云中的水的分布，云中的水对湍流内的运动高度敏感？我们可以通过根据比例和对称性等论点形成关于依赖关系的假设来“结束”这个问题，但这些假设需要测试和校准。作为项目的一部分，我们计划对受控的、理想化的对流进行一些雄心勃勃的参考模拟，以提供数据来检验这些假设。我们还建议开发一套湍流简化表示的层次结构，专门针对具有成本效益的深对流模拟。具体地说，我们计划实施三个方案：1.劳斯莱斯方案，尽可能少的近似。2.高度简化的方案，类似于常用的方案，但对关键参数进行了全新的分析。一个新的方案可以从1开始追踪，但基于1的简化，专门针对深对流云的表示，专注于垂直运动和产生的凝结与水平运动分开。高分辨率数据的分析方法将与观测数据几乎相同，但我们也计划利用高分辨率模式，这是一种永远不能通过观测实现的方式；我们计划在运行时直接向较低分辨率的模拟提供“真相”的信息，无论是湍流通量还是平均变量。这可以通过多种方式来完成，以了解哪些术语在驱动解析的流程中最重要。一种令人兴奋和新颖的方法是使用数据同化研究中心最近开发的技术来使用“观测”(在我们的情况下，参考高分辨率模拟)来客观地确定哪些参数化最能代表模式中的“缺失”项，而忽略这些参数会导致模式与观测背道而驰。

"Anarchy is the mother of Order", claimed Proudhon, referring to the Anarchist movement. Many might question whether this is true of human behaviour, but it certainly is of the atmosphere. Some of the most extreme weather on earth is associated with deep convective rain storms driven by the heat released and buoyancy generated when water vapour turns into liquid cloud water or ice. These storms start out as small turbulent eddies which grow into violent thunderstorms, the most severe of which preserve their existence by a process of continual regeneration. In the right conditions, storms can organise into much larger structures such as squall lines covering hundreds of kilometres and lasting many hours. The most extreme, violent and beautiful example of organised convection is the tropical cyclone. This ever-larger scale organisation nevertheless remains, in part, controlled by processes occurring at the smallest turbulent scales. If we represent these processes poorly in models used to predict weather and climate, we severely compromise the accuracy of their predictions. Ever-increasing computer power has given us the ability to run higher resolution models that are beginning to represent individual clouds with some realism, but we are far from directly resolving those processes that control the size, intensity, number etc. of clouds. This project is aimed at substantially improving the way we represent the effect of these processes on cloud growth and dissipation in practicable weather forecast models and climate models.We can get so far by theoretical derivation from the fundamental equations of physics, but doing so raises more unanswered questions, such as how can we predict the distribution of water in a cloud, which is highly sensitive to motions within turbulent eddies, given only limited information of average properties over larger scales (e.g. a few km)? We can 'close' the problem by forming hypotheses about dependencies based upon arguments such as scaling and symmetry, but these need testing and calibrating. As part of the project we plan to run a number of ambitious reference simulations of controlled, idealized, convective flows, to provide data to test these hypotheses. We also propose developing a hierarchy of simplified representations of the turbulent flow directed specifically at cost effective modelling of deep convection. In particular, we plan to implement three schemes:1. A 'Rolls-Royce' scheme, with as little approximation as possible. 2. A highly simplified scheme similar to those commonly used but and with a fresh analysis of key parameters.3. A new scheme traceable from 1 but based on a simplification of 1 directed specifically at the representation of deep convective clouds, focussing on vertical motion and resulting condensation separately from horizontal motion.The high resolution data will be analysed much the same way that observational data would be if available, but we also plan to make use of high-resolution models is a way which could never be achieved through observation; we plan to provide information about the 'truth' directly to lower resolution simulations as they run, either turbulent fluxes or mean variables. This can be done in a number of ways to learn about which terms are most important in driving the resolved flow. One exciting and novel way is to use techniques recently developed in the Data Assimilation Research Centre to use 'observations' (in our case, reference high-resolution simulations) to determine objectively which parametrizations best represent the 'missing' terms in a model, the omission of which lead the model to diverge from the observations.