Variational Data Assimilatin with the NMC Spectral Model

数值天气预报（NWF）的一个潜在改进策略是及时纳入大量新的观测资料。这些将从多普勒雷达网络和其他自动化仪器系统中获得，这些系统将在本十年的剩余时间内上线。随着计算机数据处理能力的同步增长，这种机会将会增加。同样重要的是，应用数学的最新发展，特别是最优控制理论。通过对控制进化系统行为的方程（或计算算法）的数学分析，有可能确定关键值，其调整将使系统的不良特性最小化。然而，随着控制方程变得复杂，分析变得极其困难，即使成功地进行了分析，仍然存在快速和廉价地实现其结果以在实践中有用的问题。就NWF而言，该程序可应用于调整开始预报的初始气象场的问题，以便使数值模式的预测与预报期早期的测量之间的差异最小化。这是通过在短时间内反复进行试验预测来实现的，同时反复微调初始字段，直到找到所需的最小误差。然后使用调整后的初始场在整个预测期内进行预测。pi关注的模型是美国国家气象中心（NMC）的主力业务天气预报模型。在NWF刚刚结束的NSF-NMC联合项目的资助下，他们成功地构建了一套算法（“伴随模型”），通过在试验预测期间交替使用伴随模型和预测模型来补充NMC计算机程序，可以显著减少迭代次数。这已被NMC发展主任描述为“最重要的成就”。建议的工作是通过进一步分析这两个模型的性质来利用这一突破，以实现理论发现的成功操作实施。这可以提高天气预报的准确性，达到或超过欧洲中期天气预报中心的水平，并增加国家对下一代天气观测网络的投资回报。

One tactic for potential improvement of Numerical Weather Forecasting (NWF) is the timely incorporation of large bodies of new observational data. These will become available from networks of Doppler radars and other automated instrumental systems that are to come on line throughout the remainder of this decade. The oppor- tunity will be enhanced with the concurrent growth of computer data-handling power. Also, and just as important, will be recent developments in applied mathematics, particularly in the theory of optimal control. By the mathematical analysis of the equations (or computational algorithms) governing the behaviour of evolving systems, it is possible to identify key values whose adjustment will minimize undesirable properties of the system. However, the analysis becomes extremely difficult as the governing equations become complex, and even if it is successfully carried out, there remains the problem of implementing its results quickly and cheaply enough to be useful in practice. In the case of NWF, the procedure can be applied to the problem of adjusting the initial meteorologi- cal fields that start a forecast so that the differences between the predictions of a numerical model and measurements coming in early in the forecast period are minimized. This is done by making repeated trial predictions for a short period, while simulteneously and repeatedly fine tuning the initial fields until the desired minimum error is found. The adjusted initial field is then used to carry out the prediction over the full forecast period. The model of concern to the PIs is the workhorse operational weather prediction model of the US National Meteorological Center (NMC). In a grant just ending under the NSF-NMC Joint Program in NWF they have successfuly constructed a set of algorithms (the "adjoint model") that complement the NMC computer program in such a way that by alternately using the adjoint and prediction models during the trial prediction period the number of iterations can be significantly reduced. This has been characterized by the Director of Development of NMC as a "most significant achievement". The proposed work is to exploit the breakthrough by further analysis of the properties of the two models in order to achieve successful operational implementation of the theoretical findings. This could enhance the accuracy of the weather forecasts to levels equal to or exceeding those achieved at the European Center for Medium Range Weather Forecasting, and increase the return on the national investment in the next-generation weather observing network.