Identifying contributors to PM2.5 simulation biases of chemical transport model using fully connected neural networks

The processed data and codes in the study are included. 

Source data:

The training and testing dataset is composed of observed and simulated data of pollutants and meteorology in the BTH and YRD regions in the whole year of 2015. The processed datasets used for training are named as "dataset_BTH" and "dataset_YRD" in the folder.

• The hourly observed pollution data are from China National Urban Air Quality Real-time Release Platform of the National Environmental Monitoring Station
• The hourly simulated pollutants data comes from the output of WRF-CMAQv5.2 (spatial resolution of 27 km).
• Meteorological observation data is provided by China Meteorological Data Service Centre
• The meteorological simulation data comes from the simulation results of the WRF model

Codes:

• preprocessing of raw CMAQ data, observed pollution data and meteorological data
• bulid and train process of fully connected neural networks
• calculation of correlation between variables
• feature selection method
• contribution analysis