Prediction methodology of air absorbed dose rates for Chinese cities with deep learning models

Abstract

Air absorbed dose rate is a key indicator of environmental radiation exposure. In China, automated environmental radiation monitoring systems have been established in multiple cities to continuously measure air absorbed dose rates. Nevertheless, developing effective preventive strategies based solely on data monitoring remains challenging. To address the issue, this study proposes a prediction framework for urban air absorbed dose rates based on historical data. The framework encompasses model construction, data preprocessing, outcome evaluation and prediction of future data. Specifically, three deep learning models—Long Short-Term Memory (LSTM), Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM), and Bidirectional Long Short-Term Memory (Bi-LSTM)—were employed to develop prediction methods for urban air absorbed dose rates. Given the large and complex dataset structure of air absorbed dose rates reports released by the National Nuclear Safety Administration, a Convolutional Neural Network (CNN) was utilized to extract monitoring data, significantly improving data preprocessing efficiency. Missing values were handled using Lagrange interpolation method. The results showed that the Bi-LSTM model performed best in terms of coefficient of determination (R²), mean absolute error (MAE) and root mean square error (RMSE) when predicting the air absorbed dose rates in a coastal city. When predicting the air absorbed dose rates in an inland city, the R² and RMSE indices of the Bi-LSTM model are more accurate, although the MAE value of the Bi-LSTM model is slightly higher than that of the LSTM model. To summarize, the Bi-LSTM model is more effective in predicting the air absorbed dose rates in Chinese cities.