AErosol model RObustness and Sensitivity study for improved climate and air quality prediction (AEROS)

AEROS是利兹大学、牛津大学、英国气象局和EMEP的合作项目，旨在全面评估全球气溶胶模型的性能，量化不确定性，并制定改进最新一代全球气溶胶模型的策略。气溶胶对气候有重要但非常不确定的影响（IPCC, 2007）。这种不确定性主要来自模式间差异、由于计算成本原因而对气溶胶过程进行必要的简化，以及用于模式评估的观测中的不确定性。最近已经为几个气候模式开发了复杂的“下一代”气溶胶微物理方案，旨在提高模式的真实感和提高预测的可靠性。这些模型可以解析颗粒大小和各种化学成分，并使用完整的微物理方案，包括成核、凝聚、尺寸分解沉积、云处理等。这种先进的气溶胶模型的发展创造了新的和实质性的挑战，这一建议旨在解决。首先，复杂气溶胶模型的计算需求意味着不确定性分析技术尚未常规使用，因此我们很少有信息来指导模型改进（模型因素的不确定性重要性，结构与参数不确定性的相对重要性等）。我们将使用敏感性和不确定性分析技术来确定需要改进的最重要的模型。其次，由于气溶胶模式已经消耗了气候模式运行时间的很大一部分，因此客观地评估模式的复杂性水平，以便优先考虑和优化未来的发展至关重要。以前的模型评估没有回答模型是否比需要的复杂，或者开发工作应该投资在哪里的问题。这个建议的一个重要方面是模型解释能力与复杂性的量化，这可能是依赖于尺度的。在已经很昂贵的模型中找到一个适当的复杂程度的好处将是巨大的：更多更长的模型运行，更多的气候敏感性测试，等等。第三，更复杂的模型需要对同样信息丰富的数据集进行评估。但是，大多数微物理量（如颗粒数、大小分辨成分等）只能通过飞机和地面站的相当局部的原位技术来测量。稀疏测量将模型评估的许多方面限制为案例研究，而不是在以前的评估（如AeroCom）中使用的长期平均测量。因此，目前这一代的气溶胶模型是根据一小部分可获得的微物理观测结果进行评估的。在这个项目中，我们的目标是通过利用2008年5月开展的EUCAARI和EMEP密集运动的观测结果来克服这个问题。通过综合密集的观察，我们将致力于预测数量之间的一致性，并避免在使用单个数据集时出现的补偿模型因素的问题。AeroCom国际气溶胶比对项目在记录模拟气溶胶的最新技术方面非常成功。它汇集了大多数全球气溶胶模式的观测和结果，以评估我们对全球气溶胶效应的理解。然而，在广泛的模型中建立可比较的诊断方法的困难使得很难将结果中的差异归因于特定的过程。我们的方法将在过程水平上评估模型，并首次根据微物理观察评估它们的性能。这一建议的总体结果将是改善气溶胶特性、变异性和空间分布的预测，这是准确预测气溶胶气候和空气质量影响的基本要求。

AEROS is a collaboration of the University of Leeds, Oxford University, the UK Met Office and EMEP to comprehensively assess the performance, quantify the uncertainties and develop strategies for improvements of the latest generation of global aerosol models. Aerosols have an important but very uncertain impact on climate (IPCC, 2007). The uncertainty derives primarily from inter-model differences, the necessary simplification of aerosol processes for computational cost reasons, and uncertainties in the observations used for model evaluation. Complex 'next generation' aerosol microphysics schemes have recently been developed for several climate models that are intended to enhance model realism and improve the reliability of predictions. The models resolve particle sizes and various chemical components, and use a full microphysics scheme including nucleation, coagulation, size-resolved deposition, cloud processing, etc. The development of such advanced aerosol models creates new and substantial challenges that this proposal aims to address. Firstly, the computational demands of complex aerosol models mean that techniques of uncertainty analysis have not been routinely used, so we have very little information to guide model improvement (uncertainty importance of model factors, relative importance of structural versus parameter uncertainty, etc). We will use sensitivity and uncertainty analysis techniques to identify the most important model improvements required. Secondly, because aerosol models already consume a large fraction of climate model run-time, it is vital to assess the level of model complexity objectively so as to prioritise and optimise future development. Previous model assessments have not answered the question of whether models are more or less complex than required or where development effort should be invested. An important aspect of this proposal is the quantification of model explanatory power versus complexity, which may be scale-dependent. The benefits of finding an appropriate level of complexity in an already expensive part of the model will be enormous: more and longer model runs, more climate sensitivity tests, etc. Thirdly, more complex models require evaluation against equally information-rich datasets. But most microphysical quantitites (such as particle number, size-resolved composition, etc) can only be measured with fairly localised in situ techniques from aircraft and from ground stations. The sparse measurements restrict many aspects of model evaluation to case studies rather than long-term average measurements used in previous evaluations such as AeroCom. So the present generation of aerosol models have been evaluated against a tiny fraction of available microphysics observations. In this project we aim to overcome this problem by exploiting observations from the EUCAARI and EMEP intensive campaigns conducted in May 2008. By synthesising intensive observations we will aim for consistency among predicted quantities and avoid the problem of compensating model factors that arises when single datasets are used. The AeroCom international aerosol intercomparison project has been very successful in documenting the state-of-the-art of the simulated aerosol. It has assembled observations and results from the majority of global aerosol models to assess our understanding of global aerosol effects. However, the difficulty of establishing comparable diagnostics across a wide range of models has made it difficult to attribute differences in the results to specific processes. Our approach will assess the models at the processes level and evaluate their performance against microphysics observations for the first time. The overall outcome of this proposal will be improvement in predictions of aerosol properties, variability and spatial distribution that are fundamental requirements for accurate prediction of aerosol climate and air quality effects.