Application of LSTM for short term fog forecasting based on meteorological elements

Abstract

Fog is the main weather phenomenon that causes low visibility, which makes traffic and outdoor work extremely dangerous. In this paper, we propose a novel LSTM framework for short-term fog forecasting. The proposed network framework consists of an LSTM network and fully connected layer. In order to make the proposed LSTM framework work, the meteorological element observation data returned hourly is transferred into time series data. Based on these time series data, four datasets are created for short-term fog forecasting. In order to evaluate the proposed LSTM framework, we conduct comprehensive experiments with different machine learning algorithms. Compared with K-Nearest Neighbor (KNN), AdaBoost and convolutional neural network (CNN) algorithms, the experimental results show that the proposed LSTM framework achieves best prediction performance in four evaluation criteria. Especially in TS-Score, the proposed LSTM framework achieve 1.1%, 11%, 3%, and 11% higher performance than the best traditional machine learning algorithm.