Dynamical and microphysical evolution of convective storms (DYMECS)

英国一系列潮湿的夏天提醒我们雷暴可能引发的毁灭性洪水，从2004年8月博斯卡斯尔洪水的标志性图像到2007年反复发生的洪水事件，导致格洛斯特郡的城镇没有自来水长达17天，根据随后的《皮特评论》，造成了超过30亿英镑的损失。显然，我们迫切需要提高我们预测这些风暴的能力，从几个小时到几天。大多数用于全球天气预报的计算机模型都将大气划分为几十公里宽的盒子。由于对流阵雨云和雷暴通常只有2到10公里宽，这些模式没有机会模拟单独的云;相反，他们必须试图估计每个盒子内的云集合对地面降雨的影响。这就是所谓的“对流参数化”，并且非常容易出错。然而，近年来，计算机能力不断提高，气象局目前运行着分辨率最高的国家天气预报模型之一，其模型在整个英国的水平网格尺寸为1.5公里。在这种分辨率下，它几乎能够模拟单个云中的空气流动，并且不需要对流参数化。虽然有证据表明，这提高了预报的准确性，但很明显，在这种分辨率下模拟的对流阵雨云和雷暴的性质仍然存在非常明显的缺陷。我们需要的是对英国大量阵雨云和雷暴的非常详细的观测，以及每一片云是如何生长和衰减的，以此来测试和改进这些高分辨率的模型。在这个项目中，我们将使用汉普郡的Chilbolton天气雷达获得这些信息，该雷达能够以非常高的分辨率测量风暴的各种有用特性（3D结构，表面降雨率，冰雹的发生，云上部的冰量，风暴内的气流和湍流水平）。该项目的一个特别创新的方面是，我们将开发自动控制雷达的软件，以便它可以跟踪个别雷暴的演变。在包括2011年和2012年夏季在内的18个月期间，雷达将在40个合适的日子运行，以获得可能发生的数百场风暴的详细信息。这一独特的数据集将用于评估预报模型中风暴的演变，其详细程度和条件范围前所未有。我们将使用不同的模式配置（例如，不同的方式来描述云冰粒和液滴的生长和相互作用最终形成雨的方式）来模拟模型，以确定是什么限制了该模型中阵雨云和雷暴的真实性，从而如何改进预报。这些发现将适用于世界各地的其他模型。我们将研究风暴在模型和现实中演变的详细方式，特别是一场风暴中的气流如何合谋引发另一场风暴，以及是什么导致风暴在一个地方持续下雨，因为这种行为通常是造成洪水的原因。这个项目的另一个目的是测试对流参数化中的一些假设。虽然对流参数化不像显式模拟单个云那样准确，但为了进行100年的气候预测，我们没有足够的计算机能力来完成整个世界所需的1.5公里盒子。我们详细的数据集将帮助我们确保这些参数化中关于云的大小和性质的假设是现实的，可能会提高气候预测的准确性。

A series of wet summers in the UK have reminded us of the devastating floods that can be caused by thunderstorms, from the iconic images of the Boscastle flood in August 2004 through to the repeated flooding events in 2007 that led to towns in Gloucestershire without mains water for up to 17 days and according to the subsequent Pitt Review, caused in excess of 3 billion pounds worth of damage. There is clearly an urgent need to improve our ability to forecast these storms from a few hours to a few days ahead. Most computer models used for weather forecasting worldwide divide the atmosphere up into boxes several tens of kilometers across. Since convective shower clouds and thunderstorms are typically only between 2 and 10 km across, these models have no chance to simulate individual clouds; rather they must try to estimate the effect of an ensemble of clouds within each box on surface rainfall. This is known as a 'convection parameterization' and is very error prone. However, there has been a continued increase in computer power in recent years, and the Met Office currently runs one of the highest resolution national weather forecast model operationally, its model having a horizontal grid-box size of 1.5 km over the whole of the UK. At this resolution it is just about able to simulate the air flows in individual clouds, and a convection parameterization is not needed. While there is evidence that this has improved the accuracy of forecasts, it is clear that there are still very significant shortcomings in the nature of convective shower clouds and thunderstorms simulated at this resolution. What we need are very detailed observations of a large number of shower clouds and thunderstorms over the UK, and how each cloud grows and decays, with which to test and improve these high resolution models. In this project we will obtain this information using the Chilbolton weather radar in Hampshire, which is able to measure all kinds of useful properties of storms at very high resolution (3D structure, surface rain rate, the occurrence of hail, the amount of ice in the upper parts of the cloud, the airflows within the storm and the levels of turbulence). A particularly innovative aspect to the project is that we will develop software to control the radar automatically, so that it can track individual thunderstorms as they evolve. Detailed information on potentially hundreds of storms will be obtained by operating the radar on 40 suitable days over an 18-month period, including the summers of 2011 and 2012. This unique dataset will be used to evaluate the evolution of storms in the forecast model in a level of detail and a range of conditions that has never been achieved before. We will rerun the model with different model configurations (e.g. different ways to describe the ways that cloud ice particles and liquid droplets grow and interact to eventually form rain), in order to determine what it is that limits the realism of shower clouds and thunderstorms in this model, and hence how to improve forecasts. These findings will be applicable to other models worldwide. We will examine the detailed way that storms evolve in the model and in reality, particularly how the airflows in one storm conspire to initiate another storm, and what it is that causes storms to rain persistently in one place, since it is often this behaviour that is responsible for flooding. A further aim of this project will be to test a number of the assumptions that are made in convection parameterizations. Although convection parameterization is not as accurate as simulating individual clouds explicitly, for making 100-year climate forecasts we do not have the computer power to do this with the 1.5-km boxes that would be required over the entire world. Our detailed dataset will help us to ensure that the assumptions about the size and properties of clouds in these parameterizations are realistic, potentially improving the accuracy of climate forecasts.