在天⽓预报中应⽤机器学习
原⽂发表于 2017年7⽉21⽇ ，是由英国⽓象信息部门（Met Office Informatics Lab, UK）发表的。
Authors list ：Rachel Prudden, Niall Robinson, Alberto Arribas , Charles Ewen
In the 1950s, there was a revolution in weather forecasting. Advances in technology made it possible to simulate the atmosphere using dynamical models, quickly and accurately enough to be ud for operational forecasts. Dynamical models are now a central part of weather forecasting. Starting from basic physical laws, they make it possible to predict events
such as storms before they have even begun to form.
⼆⼗世纪五⼗年代，天⽓预报有了⾰命性的变化。技术进步使我们可以使⽤模式来模拟⼤⽓运动，这种⽅法在预报业务中是快速⽽准确的。模式直到现在仍是天⽓预报的核⼼。通过基本的物理学原理，模式可以在暴风⾬形成之前便做出预测。
A crucial challenge in the coming decade will be the integration of direct physical simulations on the one hand, and data-driven approaches on the other. Such a hybrid approach holds many opportunities for weather forecasting, as well as countless other fields.
小楷字体
未来⼗年的⼀个关键挑战将是直接物理模拟与数据驱动⽅式融合应⽤。这种混合⽅式为天⽓预报以及⽆数其他领域带来许多机会（可能性）。承认近义词
From model to outcomes 从模式到结果
Localisation and super-resolution (downscaling) 局地和超⾼分辨率（降尺度）
Links to the real world 与其他领域结合
幼儿行为观察记录Operational weather models are usually run at a resolution of between 1km and 10km, that is, everything within the same square kilometer is reprented by a single grid cell. This resolution is fine enough to capture a wide range of phenomena, but will obviously be unable to capture very localid details.
马卡龙是什么我的假期生活手抄报⽬前业务运⾏的天⽓模式的空间分辨率在1公⾥和10公⾥之间，这意味着在这个分辨率⽹格内只有⼀个值。这个分辨率对于⼀个⼤尺度的天⽓现象是够⽤的，但是对于⼀些局地性的天⽓却是不够的。
It may be possible to perform this kind of localisation using models trained on historical data, providing a mapping between the large-scale predictions of the simulation and the small-scale effects. This is an area of active rearch which could make forecasts more uful for day-to-day acti
vities.
可以尝试使⽤历史数据训练的模型（机器学习的⽅法）来预测局地效应，之后建⽴⼀个⼤尺度模型预测与⼩规模效应之间的映射关系。此类研究现在⾮常活跃，有助于提升天⽓预测对⽇常活动的价值。
As well as predicting weather at finer scales, similar techniques could help to link weather forecasts with their broader impacts. Many things are affected by the weather, either directly or indirectly; the include traffic, hayfever, flight delays, and hospital admissions. While some effects may not be easy to simulate, using data-driven models could help to provide advance warning of significant impacts.
除了在更细微的尺度上预测天⽓，类似的技术可以帮助将天⽓预报与更⼴泛的领域联系起来。许多事情直接或间接地受到天⽓的影响，包括交通、花粉过敏、飞⾏延误和住院率，这些事情不容易通过模型来推理，但可以使⽤数据驱动的模型来预测进⽽提供预警。
Emulation
Faster components (emulation) 局部加速
Hybrid models 混合模式
Once a machine learning model has been trained, it is often much faster to run than a full simulation. This is the motivation for a technique called model emulation. The idea is to build a fast statistical model which cloly approximates a far more expensive simulation. Emulators are already being applied to problems such as climate nsitivity. An area of current interest is using the same tools to speed up some components of the weather model.
机器学习模型⼀旦被建⽴，通常是要⽐完整的数值模拟⼯程要快。可以使⽤⼀种模式仿真（model emulation）的⽅法，建⽴⼀个⾮常接近于数值模式的统计学模型，这种⽅法已经应⽤于⽓候敏感性研究。现在⽐较热的领域是使⽤机器学习⼯具加速天⽓模式的部分 组件。
There are some aspects of weather prediction which require a full physical simulation; this is what lets you predict unen events with confidence. Other places this is not possible or even justified, and a statistical approximation may be the best you can do. This cond ca is where emulation can be uful in operational forecasting.
天⽓预测中的⼀些场景是需要通过⼤⽓物理模式来实现，但有些场景使⽤模式却是不可能或不合理的，这些场景下使⽤统计学趋近是最好的选择，模式仿真（model emulation）在预报业务中会有效果。
变种生物Beyond emulators, there is broader potential for hybrid models with both learned and simulated components. Such models would combine data-driven and physically-driven approaches. For example, it may be possible to adapt statistical components of the model to the local terrain, bad on previous obrvations.
除了模式仿真（model emulation），建⽴融合机器学习与数值模拟的混合模式也是⾮常有潜⼒的。这种混合模型可以融合数据驱动和物理驱动两种⽅法。⽐如，在局地地形对天⽓影响⽅⾯，可以基于前期观测的结果训练模型，融合到数值模式中。
Descriptive learning 描述学习
Finding features 特征识别
Exploring and summarising 信息汇总
An area where machine learning has made dramatic progress is feature detection. You can e examples of this in apps which not only detect your face, but add glass and a moustache in real-time.
图表类型祝星歌词机器学习取得了显着进步的⼀个领域是特征检测。⼀些基于机器学习的应⽤程序不仅可以检测到您的
脸部，还可以实时在脸上添加眼镜和胡⼦。
There is currently a lot of interest in applying similar methods to hazard detection, especially to storm tracking. Trained experts are able to recogni storms and trace their paths from weather imagery; in principle there is no reason an
algorithm could not learn to do the same.
⽬前有很多研究在使⽤类似的⽅法做灾害监测，特别是风暴跟踪。训练有素的专家能够识别风暴，并从天⽓图像中追踪路径，理论上算法也可以做得到。
Another application could address the challenges pod by data volume and complexity when dealing with data from physical simulations. The fields output by such models are highly multidimensional; making n of them is a complex task, requiring many “screens” of information. An algorithm which could summari the salient features and bring them to the forecaster’s attention would help streamline this task.
预报员在使⽤观测数据和数值预报结果时，需要处理⼤量的多维度的数据，理解这些数据是⼀项复杂的⼯作，经常需要切换多个屏幕来查阅信息。通过算法可以⾃动识别这些数据中的关键信息，然后汇总到预报员的桌⾯，从⽽简化这项⼯作。
Summary 总结
Exploring combinations of machine learning and numerical simulation is an area of great interest and promi for the Met Office. Not only does it offer an advance in scientific capability, but the challenges arising from the attempt could drive new rearch in the field of machine learning. This article has given an outline of a few rearch directions within meteorology, but a similar story holds across a range of scientific disciplines.
探索机器学习和数值模拟的组合是 Met Office ⾮常感兴趣且抱有期望的领域。它不仅促进了预报能⼒的进步，⽽且可能会推动机器学习领域的新研究。本⽂概述了⽓象学中的⼀些研究⽅向，在其他科学学科中，机器学习的应⽤的⽅向与本⽂所述类似。
