This virtual issue brings together articles that discuss the need for high resolution observations and the use of opportunistic data in numerical weather prediction (NWP). The issue was assembled to complement the Royal Meteorological Society's “Big data assimilation workshop” held in the autumn of 2019 (https://www.rmets.org/event/assimilation-big-data).

Over recent decades advances in computing techniques and power have allowed the resolution of NWP models to increase. As a result, convection‐permitting NWP is now feasible and commonplace at most operational forecasting centres. The first article in the collection, Clark et al. (2016), provides a summary of operational convection‐permitting NWP models and their benefits. The authors show that these high resolution NWP systems provide a new forecasting capability with considerable improvements in short range forecasts; compared with coarser resolution NWP models, convection‐permitting models are able to produce more realistic convection and provide more skilful rainfall forecasts. The paper also provides a summary of data assimilation techniques and observation types that are required for convection‐permitting NWP.

The accuracy of any forecast is inherently linked with the accuracy of the initial conditions used to start the forecast. Therefore, it is important that the initial conditions are kept close to reality by updating the previous forecast with appropriate observations that capture the most recent atmospheric state. To match the increased resolutions of the convection‐permitting models there is a need for observations that contain information on spatial and temporal scales at an appropriately high resolution (Gustafsson et al., 2018). The use of these high resolution observations is not restricted to the assimilation step of a NWP system; they are also required for the statistical post‐processing of model outputs and in the forecast verification process.

The current conventional network does provide some observations at sufficiently high resolution for convective scale NWP. Until recently only a fraction of these high resolution observations were assimilated, although recent work has improved their use in assimilation (e.g. Geer et al., 2019; Simonin et al., 2019). Despite such improvements, as we move to even higher resolution models, the resolution of the observations will also need to increase. Of course the operational network of conventional observations will continue to provide the gold standard of atmospheric observations; however, to meet the very high resolution requirements it may be necessary to supplement conventional observations with those from more unconventional sources.

Unconventional data may be obtained from a number of sources in a number of ways. It may be possible to derive observations from information collected for non‐meteorological purposes. Alternatively, information can be provided by the general public, either through targeted campaigns or collected without the routine input of the collector, for example via private automatic weather stations (AWSs) or a smartphone “app”. Not only do these unconventional data provide information at higher spatial and temporal resolution than is often quantified by conventional observations, but they may also provide information in locations, such as individual urban streets, where conventional data are often unrepresentative of the local situation. An additional benefit is the low cost of the observing infrastructure and network. As a result, operational NWP centres are showing interest in these unconventional observations (Hintz et al., 2019). However, before these unconventional observations can be used in NWP an assessment of the observations’ quality and usefulness is imperative. Furthermore, due to the typically low quality but large quantity of unconventional data, the development of appropriate observation processing will be crucial. The remaining papers included in this virtual issue discuss different opportunistic datasets, their quality, processing that can be implemented to improve the data, and the potential benefits of the data for NWP.

The work of Cornes et al. (2020) considers the potential of citizen‐science thermometer measurements collected using the Met Office Weather Observations Website (WOW). It is known that such data are prone to biases, although the metadata required to correct such biases are typically unavailable. Therefore, the authors develop a statistical correction for short wave radiation bias for AWS temperature data across the Netherlands. Using this correction it is shown that data add local detail to the previously existing temperature field. They conclude that the information provided by the corrected data could be used to supplement conventional measurements.

The work of Mirza et al. (2019) performs an assessment of observations that can be derived from aircraft mode‐select enhanced tracking data (ModeS EHS reports). The authors show that application of quality control and an additional data smoothing must be applied to the high spatial and temporal resolution data to reduce the uncertainty in derived temperature observations to an acceptable level. The application of the data smoothing results in vertical profiles that may be useful in operational meteorology, to study the evolution of temperature inversions above 1,000 m. However, the uncertainty in the smoothed temperature data is still too large for assimilation. The ModeS data discussed in Mirza et al. (2019) provide the desired spatial and temporal resolutions required for convection‐permitting NWP and, although the temperature data are not sufficiently precise, ModeS winds are now one of the opportunistic observations assimilated by operational NWP centres (e.g. Gustafsson et al., 2018; Milan et al., 2020).

Advances in technology also create the potential to gather opportunistic data from the public. The work of Hintz et al. (2019) considers if pressures obtained from smartphones have the potential to be assimilated in convection‐permitting models. The complete chain of data acquisition, processing, quality controlling and assimilation is described. The smartphone pressure data considered are collected via the Danish Meteorological Institute Weather App, which also performs some of the data processing reducing the computation and storage required at the NWP centre. Case studies show that the smartphone pressures can replicate distinct weather patterns. Furthermore, the benefit of assimilating the crowd‐sourced pressures is highlighted by reduced bias in the forecast surface pressure.

Opportunistic data have uses beyond assimilation; the high resolution nature makes them beneficial for the analysis of past events. The work of Clark et al. (2018) considers if “user‐contributed” observations are of sufficient quality to examine past thunderstorm events, and hence may be of use in the forecasting process. The authors created gridded datasets of surface wind, pressure and temperature using filtered and corrected data from private AWSs. These gridded fields were in agreement with radar data and conformed with conceptual models of supercells. Additional crowd‐sourced hail reports and eyewitness photographs were shown to provide further insight into the storm structure. The results presented suggest that data from private AWSs are useful for examining past events and may provide benefit in NWP.

The papers brought together in this virtual issue both highlight the need for very high resolution observations and show that unconventional data, either crowd‐sourced or obtained from non‐meteorological sources, can provide such information. The papers of Clark et al. (2018), Mirza et al. (2019), Hintz et al. (2019) and Cornes et al. (2020) all show that careful quality control and processing of unconventional data are important to ensure that biases and the uncertainty are reduced to a level sufficient for use in NWP. With appropriate quality control and processing, opportunistic observations can accurately represent small scale meteorological phenomena and are now beginning to be used in NWP to complement conventional observations.

这个虚拟问题汇集了一些文章，这些文章讨论了对高分辨率观测的需求以及在数值天气预报（NWP）中使用机会数据的问题。本期杂志的发行是对皇家气象协会于2019年秋季举行的“大数据同化研讨会”的补充（https://www.rmets.org/event/assimilation-big-data）。

在最近的几十年中，计算技术和功能的进步使得NWP模型的分辨率得以提高。因此，在大多数业务预报中心，允许对流的NWP现在是可行且司空见惯的。集合中的第一篇文章，Clark等。（2016年），概述了允许对流的NWP模型及其优势。作者表明，这些高分辨率的NWP系统提供了一种新的预测功能，并且对短程预报有很大的改进。与较粗糙分辨率的NWP模型相比，对流允许模型能够产生更逼真的对流并提供更熟练的降雨预报。本文还概述了允许对流的NWP所需的数据同化技术和观测类型。

任何预测的准确性都与用于启动预测的初始条件的准确性有内在联系。因此，重要的是，通过使用捕获最新大气状态的适当观测值更新先前的预测，使初始条件与实际情况保持接近。为了匹配对流允许模型增加的分辨率，需要以适当的高分辨率包含时空尺度信息的观测结果（Gustafsson et al。，2018）。这些高分辨率观测的使用不限于NWP系统的同化步骤；模型输出的统计后处理和预测验证过程中也需要它们。

当前的常规网络确实提供了对流尺度NWP足够高分辨率的一些观测结果。直到最近，这些高分辨率观测中只有一小部分被同化了，尽管最近的工作已经改善了它们在同化中的使用（例如Geer等人，2019 ; Simonin等人，2019）。尽管有了这些改进，但随着我们转向甚至更高分辨率的模型，观测的分辨率也将需要提高。当然，常规观测的业务网络将继续提供大气观测的黄金标准；然而，为了满足非常高分辨率的要求，可能有必要用来自非常规来源的常规观测来补充常规观测。

可以以多种方式从许多来源获得非常规数据。可能可以从出于非气象目的收集的信息中得出观测结果。或者，可以由公众通过有针对性的活动提供信息，也可以不通过收集器的常规输入而收集信息，例如通过私人自动气象站（AWS）或智能手机“ app”。这些非常规数据不仅以比常规观测经常量化的更高的时空分辨率提供信息，而且它们还可以在常规数据通常不能代表当地情况的位置（例如个别城市街道）提供信息。另一个好处是观察基础设施和网络的成本低。结果是，等人，2019）。但是，在将这些非常规观测值用于NWP之前，必须对观测值的质量和有用性进行评估。此外，由于通常质量低下但数量众多的非常规数据，开发适当的观察处理将至关重要。该虚拟期刊中包含的其余论文讨论了不同的机会数据集，它们的质量，可以用来改善数据的处理以及数据对NWP的潜在好处。

Cornes等人的工作。（2020年）考虑了使用大都会办公室天气观测网站（WOW）收集的公民科学温度计测量值的潜力。已知这样的数据容易产生偏差，尽管校正此类偏差所需的元数据通常不可用。因此，作者对整个荷兰的AWS温度数据的短波辐射偏差进行了统计校正。使用该校正，表明数据将局部细节添加到先前存在的温度场中。他们得出结论，由校正后的数据提供的信息可用于补充常规测量。

Mirza等人的工作。（2019）对从飞机模式选择的增强跟踪数据（ModeS EHS报告）中得出的观测结果进行评估。作者表明，必须对高空间和时间分辨率数据应用质量控制和附加的数据平滑处理，以将导出的温度观测值的不确定性降低到可接受的水平。数据平滑的应用导致垂直剖面，这可能对操作气象学有用，以研究1,000 m以上温度反演的演变。但是，平滑后的温度数据的不确定性对于同化而言仍然太大。在Mirza等人中讨论的ModeS数据。（2019）提供了允许对流NWP所需的所需空间和时间分辨率，尽管温度数据不够精确，但ModeS风现在是NWP运营中心同化的机会性观测之一（例如Gustafsson等，2018 ; Milan等）等，2020）。

科技的进步也创造了从公众那里收集机会数据的潜力。Hintz等人的工作。（2019）考虑了从智能手机获得的压力是否有可能在对流允许模型中被吸收。描述了数据采集，处理，质量控制和同化的完整链。所考虑的智能手机压力数据是通过丹麦气象学院天气应用程序收集的，该应用程序还执行一些数据处理，从而减少了NWP中心所需的计算和存储。案例研究表明，智能手机的压力可以复制不同的天气模式。此外，通过减少预测的地表压力的偏见，突出了吸收众包压力的好处。

机会数据的用途不仅仅限于同化。高分辨率的特性使它们对于分析过去的事件非常有用。克拉克等人的工作。（2018年）考虑“用户提供的”观测资料是否具有足够的质量来检查过去的雷暴事件，因此可能在预测过程中有用。作者使用来自私人AWS的经过过滤和校正的数据创建了地面风，压力和温度的网格化数据集。这些网格化的区域与雷达数据一致，并且符合超级单元的概念模型。还显示了其他群众提供的冰雹报告和目击者照片，以进一步了解风暴的结构。给出的结果表明，来自私有AWS的数据对于检查过去的事件很有用，并可能在NWP中带来好处。

围绕这个虚拟问题的论文都强调了对高分辨率观测的需求，并表明非常规数据（无论是众包还是从非气象资源获得）都可以提供此类信息。Clark等人的论文。（2018），Mirza等。（2019），Hintz等。（2019）和Cornes等。（2020年）都表明，仔细的质量控制和非常规数据的处理对于确保将偏差和不确定性降低到足以在NWP中使用的水平非常重要。通过适当的质量控制和处理，机会性观测可以准确地代表小规模的气象现象，现在已开始在NWP中用于补充常规观测。