The Development and Application of Machine Learning in Atmospheric Environment Studies

Machine learning (ML) plays an important role in atmospheric environment prediction, having been widely applied in atmospheric science with significant progress in algorithms and hardware. In this paper, we present a brief overview of the development of ML models as well as their application to atmospheric environment studies. ML model performance is then compared based on the main air pollutants (i.e., PM2.5, O3, and NO2) and model type. Moreover, we identify the key driving variables for ML models in predicting particulate matter (PM) pollutants by quantitative statistics. Additionally, a case study for wet nitrogen deposition estimation is carried out based on ML models. Finally, the prospects of ML for atmospheric prediction are discussed.

Using deep learning to nowcast the spatial coverage of convection from Himawari-8 satellite data

Predicting the timing and location of thunderstorms (“convection”) allows for preventive actions that can save both lives and property. We have applied U-nets, a deep-learning-based type of neural network, to forecast convection on a grid at lead times up to 120 minutes. The goal is to make skillful forecasts with only present and past satellite data as predictors. Specifically, predictors are multispectral brightness-temperature images from the Himawari-8 satellite, while targets (ground truth) are provided by weather radars in Taiwan. U-nets are becoming popular in atmospheric science due to their advantages for gridded prediction. Furthermore, we use three novel approaches to advance U-nets in atmospheric science. First, we compare three architectures – vanilla, temporal, and U-net++ – and find that vanilla U-nets are best for this task. Second, we train U-nets with the fractions skill score, which is spatially aware, as the loss function. Third, because we do not have adequate ground truth over the full Himawari-8 domain, we train the U-nets with small radar-centered patches, then apply trained U-nets to the full domain. Also, we find that the best predictions are given by U-nets trained with satellite data from multiple lag times, not only the present. We evaluate U-nets in detail – by time of day, month, and geographic location – and compare to persistence models. The U-nets outperform persistence at lead times ≥ 60 minutes, and at all lead times the U-nets provide a more realistic climatology than persistence. Our code is available publicly.

Measuring electrical properties of the lower troposphere using enhanced meteorological radiosondes

In atmospheric science, measurements above the surface have long been obtained by carrying instrument packages, radiosondes, aloft using balloons. Whilst occasionally used for research, most radiosondes – around one thousand are released daily – only generate data for routine weather forecasting. If meteorological radiosondes are modified to carry additional sensors, of either mass-produced commercial heritage or designed for a specific scientific application, a wide range of new measurements becomes possible. Development of add-on devices for standard radiosondes, whilst retaining the core meteorological use, is described here. Combining diverse sensors on a single radiosonde helps interpretation of findings, and yields economy of equipment, consumables and effort. A self-configuring system has been developed to allow different sensors to be easily combined, enhancing existing weather balloons and providing an emergency monitoring capability for airborne hazards. This research programme was originally pursued to investigate electrical properties of extensive layer clouds, and has expanded to include a wide range of balloon-carried sensors for solar radiation, cloud, turbulence, volcanic ash, radioactivity and space weather. For the layer cloud charge application, multiple soundings in both hemispheres have established that charging of extensive layer clouds is widespread, and likely to be a global phenomenon. This paper summarises the Christiaan Huygens medal lecture given at the 2021 European Geoscience Union meeting.

Opinion: The germicidal effect of ambient air (open-air factor) revisited

The term open-air factor (OAF) was coined following microbiological research in the 1960s and 1970s which established that rural air had powerful germicidal properties and attributed this to Criegee intermediates formed in the reaction of ozone with alkenes. We have re-evaluated those early experiments applying the current state of knowledge of ozone–alkene reactions. Contrary to previous speculation, neither Criegee intermediates nor the HO radicals formed in their decomposition are directly responsible for the germicidal activity attributed to the OAF. We identify other potential candidates, which are formed in ozone–alkene reactions and have known (and likely) germicidal properties, but the compounds responsible for the OAF remain a mystery. There has been very little research into the OAF since the 1970s, and this effect seems to have been largely forgotten. In this opinion piece we remind the community of the germicidal open-air factor. Given the current global pandemic spread by an airborne pathogen, understanding the natural germicidal effects of ambient air, solving the mystery of the open-air factor and determining how this effect can be used to improve human welfare should be a high priority for the atmospheric science community.

How Undergraduate Students Learn Atmospheric Science: Characterizing the Current Body of Research

Educators can enrich their teaching with best practices, share resources, and contribute to the growing atmospheric science education research community by reading and participating in the scholarship of teaching and learning in atmospheric science. This body of scholarship has grown, particularly over the past fifteen years, and is now a sizable literature base that documents and exemplifies numerous teaching innovations in undergraduate atmospheric science education. This literature base benefits the entire atmospheric science community because graduates of atmospheric science programs are better prepared to enter the workforce. This literature base has not yet been examined, however, to see how well the evidence supports education practices in the atmospheric science education literature. In this study, we characterized that evidence and show that the majority of papers we reviewed share education innovations with anecdotal or correlational evidence of effectiveness. While providing useful practitioner knowledge and preliminary evidence of the effectiveness of numerous innovative teaching practices, opportunities exist for increasing readers’ confidence that the innovations caused the learning gains. Additional studies would also help move conclusions toward generalizability across academic institutions and student populations. We make recommendations for advancing atmospheric science education research and encourage atmospheric science educators to actively use the growing body of education literature as well as contribute to advancing atmospheric science education research.

Data-driven global weather predictions at high resolutions

Society has benefited enormously from the continuous advancement in numerical weather prediction that has occurred over many decades driven by a combination of outstanding scientific, computational and technological breakthroughs. Here, we demonstrate that data-driven methods are now positioned to contribute to the next wave of major advances in atmospheric science. We show that data-driven models can predict important meteorological quantities of interest to society such as global high resolution precipitation fields (0.25°) and can deliver accurate forecasts of the future state of the atmosphere without prior knowledge of the laws of physics and chemistry. We also show how these data-driven methods can be scaled to run on supercomputers with up to 1024 modern graphics processing units and beyond resulting in rapid training of data-driven models, thus supporting a cycle of rapid research and innovation. Taken together, these two results illustrate the significant potential of data-driven methods to advance atmospheric science and operational weather forecasting.