Abstract
Accurate lightning forecast is significant for disaster prevention and reduction. However, the mainstream lightning forecast methods, which mainly rely on numerical simulations and parameterizations, can hardly cope with the spatiotemporal deviations. Meanwhile, the rapid and complex evolution of lightning regions go beyond the traditional extrapolation-based forecast methods. In this work, we propose a data-driven neural network model for hourly lightning forecast, which exploits both the numerical simulations and the recent historical lightning observations. The two kinds of data complement each other and play different roles at different stages of the forecast. The use of dual-source data greatly increases the amount of information available to improve the forecasting performance. To handle the variability of deviation patterns in numerical simulations, we introduce a channel-wise attention mechanism, which adaptively adjusts the proportion of each simulated parameter to maximize the useful information. The attention mechanism also enables the model to reveal the contribution of each simulated parameter for the forecast. Experimental results on a real-world dataset show that the proposed method outperforms several baseline methods. Ablation studies further demonstrate the effectiveness of our data fusion approach and attention module.
Highlights
Lightning has destructive power to communication systems and electrical infrastructures, and brings a serious threat to the safety of human life
To tackle the aforementioned challenges, we propose an attention-based dual-source spatiotemporal neural network (ADSNet) that aims at forecast hourly lightning occurance over the 12 hours
It leverages the strengths of conventional recurrent neural network (RNN) and Seq2Seq structure to combine the information of recent lightning observations and numerical simulations, overcoming the defect of existing methods that just rely on a single type of data
Summary
Lightning has destructive power to communication systems and electrical infrastructures, and brings a serious threat to the safety of human life. The encoder-decoder structure facilitates sequence-to-sequence (Seq2Seq) conversion between data series with different lengths [9]–[11] Since both recent lightning observations and numerical simulations can be organized in the form of spatiotemporal data grids, lightning forecasts can be implemented by deep learning methods. We propose a dual-source neural network for hourly lightning forecast It leverages the strengths of conventional RNN and Seq2Seq structure to combine the information of recent lightning observations and numerical simulations, overcoming the defect of existing methods that just rely on a single type of data. Similar design appears in [34] for travel demand prediction These works provide us with the basic ideas of proposing an appropriate model to fill the gap of deep learning in hourly lightning forecast
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call