Search for: All records

Abstract Atmospheric processes involve both space and time. Thus, humans looking at atmospheric imagery can often spot important signals in an animated loop of an image sequence not apparent in an individual (static) image. Utilizing such signals with automated algorithms requires the ability to identify complex spatiotemporal patterns in image sequences. That is a very challenging task due to the endless possibilities of patterns in both space and time. Here, we review different concepts and techniques that are useful to extract spatiotemporal signals from meteorological image sequences to expand the effectiveness of AI algorithms for classification and prediction tasks. We first present two applications that motivate the need for these approaches in meteorology, namely the detection of convection from satellite imagery and solar forecasting. Then we provide an overview of concepts and techniques that are helpful for the interpretation of meteorological image sequences, such as (a) feature engineering methods using (i) meteorological knowledge, (ii) classic image processing, (iii) harmonic analysis, and (iv) topological data analysis; (b) ways to use convolutional neural networks for this purpose with emphasis on discussing different convolution filters (2D/3D/LSTM-convolution); and (c) a brief survey of several other concepts, including the concept of “attention” in neural networks and its utility for the interpretation of image sequences and strategies from self-supervised and transfer learning to reduce the need for large labeled datasets. We hope that presenting an overview of these tools—many of which are not new but underutilized in this context—will accelerate progress in this area. 

Abstract. An ability to accurately detect convective regions isessential for initializing models for short-term precipitation forecasts.Radar data are commonly used to detect convection, but radars that providehigh-temporal-resolution data are mostly available over land, and the qualityof the data tends to degrade over mountainous regions. On the other hand,geostationary satellite data are available nearly anywhere and in near-realtime. Current operational geostationary satellites, the GeostationaryOperational Environmental Satellite-16 (GOES-16) and Satellite-17, provide high-spatial- and high-temporal-resolution data but only of cloud top properties; 1 min data, however, allow us to observe convection from visible andinfrared data even without vertical information of the convective system.Existing detection algorithms using visible and infrared data look forstatic features of convective clouds such as overshooting top or lumpy cloudtop surface or cloud growth that occurs over periods of 30 min to anhour. This study represents a proof of concept that artificial intelligence(AI) is able, when given high-spatial- and high-temporal-resolution data fromGOES-16, to learn physical properties of convective clouds and automate thedetection process. A neural network model with convolutional layers is proposed to identifyconvection from the high-temporal resolution GOES-16 data. The model takesfive temporal images from channel 2 (0.65 µm) and 14 (11.2 µm) asinputs and produces a map of convective regions. In order to provideproducts comparable to the radar products, it is trained against Multi-RadarMulti-Sensor (MRMS), which is a radar-based product that uses a rathersophisticated method to classify precipitation types. Two channels fromGOES-16, each related to cloud optical depth (channel 2) and cloud topheight (channel 14), are expected to best represent features of convectiveclouds: high reflectance, lumpy cloud top surface, and low cloud toptemperature. The model has correctly learned those features of convectiveclouds and resulted in a reasonably low false alarm ratio (FAR) and highprobability of detection (POD). However, FAR and POD can vary depending onthe threshold, and a proper threshold needs to be chosen based on thepurpose.