Convective Gusts Nowcasting Based on Radar Reflectivity and a Deep Learning Algorithm

Abstract

Abstract. Convective wind gusts (CGs) are usually related to thunderstorms, and they may cause great structural damage and serious hazards, such as train derailment, service interruption, and building collapse. Due to the small-scale and nonstationary nature of CGs, reliable CGs nowcasting with high spatial and temporal resolutions has remained unattainable. In this study, a novel nowcasting model based on deep learning – namely, CGsNet – is developed for 0–2 h of quantitative CGs nowcasting, first achieving minute-kilometer-level forecasts. CGsNet is a physics-constrained model established by training on large corpora of average surface wind speed (ASWS) and radar observations, it can produce realistic and spatiotemporally consistent ASWS predictions in CGs events. By combining the gust factor (1.77, the ratio of the observed peak wind gust speed (PWGS) to the ASWS) with the ASWS predictions, the PWGS forecasts are estimated with a spatial resolution of 0.01° × 0.01° and a 6-minute temporal resolution. CGsNet is shown to be effective, and it has an essential advantage in learning the spatiotemporal features of CGs. In addition, quantitative evaluation experiments indicate that CGsNet exhibits higher generalization performance for CGs than the traditional nowcasting method based on numerical weather prediction models. CGs nowcasting technology can be applied to provide real-time quantitative CGs forecasts and alerts the damaging wind events in meteorological services.

Image rain removal based on multi-stage optimized residual network

“…Traditional image rain streak removal algorithms [1] are usually based on predefined models, and the predefined models have limited ability to detect real rain streaks and cannot recover image details. Deep learning based methods [2] improve the ability to remove raindrop streaks, but there are some problems, most of the existing deep raindrop streak removal methods focus only on the underlying raindrop streak removal task, and there are problems such as the inability to completely remove the raindrop streaks, and the complexity of the network model. Therefore, the algorithm proposed in this paper, based on residual network [3] and convolutional neural network [4] , adopts two kinds of mapping in terms of data and combines the dense connection module to complete the high-dimensional feature extraction, and based on the physical properties of real rainfall images, combines the motion recovery structure of monocular sequence images and the uncertainty estimation of depth information, and realizes the image de-raining in a real case by using the rain map dataset.…”