Temperature Prediction Using the Missing Data Refinement Model Based on a Long Short-Term Memory Neural Network

Park Park,Song Song,Lee Lee,Kim Kim

doi:10.3390/atmos10110718

Abstract

In this paper, we propose a new temperature prediction model based on deep learning by using real observed weather data. To this end, a huge amount of model training data is needed, but these data should not be defective. However, there is a limitation in collecting weather data since it is not possible to measure data that have been missed. Thus, the collected data are apt to be incomplete, with random or extended gaps. Therefore, the proposed temperature prediction model is used to refine missing data in order to restore missed weather data. In addition, since temperature is seasonal, the proposed model utilizes a long short-term memory (LSTM) neural network, which is a kind of recurrent neural network known to be suitable for time-series data modeling. Furthermore, different configurations of LSTMs are investigated so that the proposed LSTM-based model can reflect the time-series traits of the temperature data. In particular, when a part of the data is detected as missing, it is restored by using the proposed model’s refinement function. After all the missing data are refined, the LSTM-based model is retrained using the refined data. Finally, the proposed LSTM-based temperature prediction model can predict the temperature through three time steps: 6, 12, and 24 h. Furthermore, the model is extended to predict 7 and 14 day future temperatures. The performance of the proposed model is measured by its root-mean-squared error (RMSE) and compared with the RMSEs of a feedforward deep neural network, a conventional LSTM neural network without any refinement function, and a mathematical model currently used by the meteorological office in Korea. Consequently, it is shown that the proposed LSTM-based model employing LSTM-refinement achieves the lowest RMSEs for 6, 12, and 24 h temperature prediction as well as for 7 and 14 day temperature prediction, compared to other DNN-based and LSTM-based models with either no refinement or linear interpolation. Moreover, the prediction accuracy of the proposed model is higher than that of the Unified Model (UM) Local Data Assimilation and Prediction System (LDAPS) for 24 h temperature predictions.

Highlights

In the last several decades, abnormal weather, such as cold weather, heavy snow, heavy rain, and drought, has been occurring more and more frequently in all parts of the world, causing humanAtmosphere 2019, 10, 718; doi:10.3390/atmos10110718 www.mdpi.com/journal/atmosphereAtmosphere 2019, 10, 718 injury and death, property damage, and health and environmental problems [1]
Meteorological institutes usually forecast prediction results based on numerical weather prediction (NWP) models that predict the weather based on current weather conditions [5]
In order to compare the performance of the proposed model, four different temperature prediction models are considered: (1) a prediction model using a deep neural network (DNN), (2) a prediction model using a conventional long short-term memory (LSTM) without any refinement function, (3) a prediction model using an LSTM with a simple gap-filling method by linearly interpolating missed data from past and future data, and (4) a unified model (UM), which is a mathematical model currently used by the meteorological office in Korea

Summary

Introduction

In the last several decades, abnormal weather, such as cold weather, heavy snow, heavy rain, and drought, has been occurring more and more frequently in all parts of the world, causing humanAtmosphere 2019, 10, 718; doi:10.3390/atmos10110718 www.mdpi.com/journal/atmosphereAtmosphere 2019, 10, 718 injury and death, property damage, and health and environmental problems [1]. Depending on the applications of the temperature forecast, a suitable model should be designed at a specific site or a region with at least 10 days’ prediction [4]. Meteorological institutes usually forecast prediction results based on numerical weather prediction (NWP) models that predict the weather based on current weather conditions [5]. NWP models target the forecasts for large geometrical areas, such as the East Asian region, and they are good at handling weather that is complexly connected with various factors that dynamically influence the subsequent day’s weather. NWP often produces biased temperatures in proportion to the increase of the local elevation and topographical complexity [6]. Predicting spatial and temporal changes in temperature for areas composed of complex topography remains a challenge to NWP models [7]

Objectives

Discussion

Conclusion