Development of a CNN+LSTM Hybrid Neural Network for Daily PM2.5 Prediction

A CNN+LSTM (Convolutional Neural Network + Long Short-Term Memory) based deep hybrid neural network was established for the citywide daily PM2.5 prediction in South Korea. The structural hyperparameters of model were determined through comprehensive sensitivity tests. input features obtained from ground observations and GFS forecast. performance evaluated by comparison with 3-D CTM (three-dimensional chemistry transport model)-predicted PM2.5. newly developed estimated more accurate ambient levels compared to CTM. For example, error bias 1.51 6.46 times smaller than those 3D-CTM simulation. In addition, on IOA (Index Agreement), accuracy 1.10–1.18 higher CTM-based prediction. importance indirectly investigated sequential perturbing variables. most important meteorological atmospheric environmental geopotential height previous day obstacles current CNN+LSTM-based also discussed. promising result this study indicates that DNN-based models can be utilized as an effective tool air quality