Evaluation of clouds in climate and forecasting models using CloudSat and Calipso data.

气候变化是当今世界面临的重大挑战之一。有令人信服的证据表明，由于人类活动导致的“温室气体”（特别是二氧化碳）的不可阻挡的增加正在对气候产生变暖效应。然而，很难确定由于温室气体浓度的特定增加，未来全球平均地表温度将上升多少，而且事实上，预测结果会有三倍的变化。其中一个主要问题来自云。日常经验告诉我们，云层的存在对到达地面的太阳能量有深远的影响，同样，阴天的夜晚往往比晴朗的夜晚更温暖，这是因为云层向地面发射的红外能量几乎与向它发射的能量相平衡。完全相同的过程在大尺度上起作用，为了用任何技巧计算地表温度，我们需要准确地知道云在全球的分布情况，包括它的详细特性，如平均液滴大小。主要有两个困难。首先，用于预测气候的计算机模拟将大气分成大的网格框（通常宽200公里，深1公里），通常只使用两个数字来描述每个网格框中的云（例如云量和云水质量）。这显然是非常粗糙的。其次，用于测试模拟云的信息来自卫星上携带的成像仪，这些成像仪无法看到云的深处，无法确定它是如何垂直排列的，也无法探测到云层下面的云层。在这个项目中，我们将利用NASA将于2005年5月发射的两颗卫星所获得的令人兴奋的新数据。“云卫星”携带了一个云雷达，通过不断向下发送短脉冲无线电波来工作。云层将其中一些波散射回雷达，通过计算回波返回所需的时间和测量其强度，我们可以计算出云的高度和它含有多少水。当云卫星绕地球运行时，它将首次能够推断大气中每个高度的云特性。第二颗卫星“Calipso”携带激光雷达，原理相同，但使用的是光而不是无线电波。这个项目的关键方面是结合雷达和激光雷达来了解更多关于云的性质。雷达对大的云粒子更敏感，而激光雷达对小的云粒子更敏感，这意味着我们可以估计重要的云特性，如卷云中冰晶的大小和低空水云的毛毛雨下降率（这对确定云将持续多久很重要）。在0到-40°C之间，云可以包含小的液滴和大的冰晶；我们从地面激光雷达得到的结果表明，计算机模型对这些云的模拟效果特别差，但使用雷达和激光雷达可以很容易地区分这两种成分，并估计它们的性质。然后，我们将使用从大约一年的全球观测中提取的所有云属性，在两个世界上最好的计算机模型——英国气象局气候模型和欧洲气象中心天气预报模型——中测试这些云。这将用于突出模型的问题，并通过开发新的方法来模拟云，并根据新的观测结果再次测试它们来解决这些问题。特别有趣的是，我们将在两个模型中引入一些方案，在一个大型模型网格盒中表示水平云结构，我们将首次在全球范围内进行测试。由此产生的对模拟云的改进应该会让我们对气候变化的预测更有信心。

Climate change is one of the great challenges facing the world today. There is compelling evidence that the inexorable increase in 'greenhouse gases' (particularly carbon dioxide) due to human activity is having a warming effect on climate. However, it is very difficult to determine just how much average global surface temperature will rise in future in response to a particular increase in greenhouse gas concentration, and in fact the forecasts vary by a factor of three. One of the major problems stems from clouds. Everyday experience tells us the profound effect the presence of a cloud has on the amount of the sun's energy that reaches the ground, and similarly the fact that cloudy nights tend to be warmer than clear is because the infrared energy emitted by the surface is then nearly balanced by the energy emitted back towards it by the cloud. Exactly the same processes act on large scales, and to calculate surface temperature with any skill we need to know accurately how cloud is distributed around the globe, including its detailed properties such as average droplet size. There are two main difficulties. Firstly, the computer simulations used to predict climate split the atmosphere into large grid-boxes (typically 200 km across and up to 1 km deep), with often only two numbers being used to describe the cloud in each box (e.g. the amount of cloud and the mass of cloud water). This is clearly very crude. Secondly, the information available to test the simulated clouds is from imagers carried on satellites, which cannot see very far into a cloud to determine how it is arranged vertically or to detect one cloud layer beneath another. In this project we will make use of exciting new data from two satellites to be launched by NASA in May 2005. 'CloudSat' carries a cloud radar that works by continually sending short pulses of radio waves downwards. Clouds scatter some of these waves back to the radar, and by timing how long the echo takes to be returned and by measuring its strength, we can calculate the height of the cloud and how much water it contains. As CloudSat orbits the earth it will for the first time be able to infer cloud properties at each height in the atmosphere. The second satellite 'Calipso' carries a lidar and works on the same principle but using light rather than radio waves. The key aspect to this project is to combine the radar and lidar to learn more about the nature of the clouds. The fact that radar is more sensitive to the large cloud particles while lidar is more sensitive to the small means that we can estimate important cloud properties such as the size of ice crystals in cirrus and the rate of drizzle falling from low altitude water clouds (which is important for determining how long the cloud will persist). Between 0 and -40°C, clouds can contain both small liquid droplets and large ice crystals; our results from ground-based lidar show that these clouds are particularly badly simulated by computer models, but with radar and lidar the two components can be easily distinguished and their properties estimated. We will then use all the cloud properties extracted from around a year of global observations to test the clouds in two of the world's best computer models, the Met Office climate model and the ECMWF forecast model. This will be used to highlight problems with the models, and address them by developing new ways to simulate clouds and testing them again against the new observations. Of particular interest will be schemes shortly to be introduced in both models to represent the horizontal cloud structure in a large model grid-box which we will test around the globe for the first time. The resulting improvements in simulated clouds should give us more confidence in predictions of climate change.