Modeling Marine Atmospheric PM2.5 Concentrations over the Bohai Sea Using a Deep Neural Network: Spatiotemporal Variations under China's Emission Reduction Policies
Description
This dataset supports the research titled "Modeling Marine Atmospheric PM2.5 Concentrations over the Bohai Sea Using a Deep Neural Network: Spatiotemporal Variations under China's Emission Reduction Policies", with a focus on atmospheric PM2.5 concentrations over the Bohai Sea (37°N - 41°N, 117°E - 122°E) during 2015–2023. It includes raw observational data, model input/output data, and model files, all aimed at ensuring the reproducibility of research results related to deep learning-based marine atmospheric pollutant monitoring and emission reduction policy assessment. (1) Model Input Data Data were sourced from the China National Environmental Monitoring Center (PM2.5; hourly), MERRA-2 reanalysis dataset (AOD; hourly, 0.5°×0.625°), and ERA5 reanalysis dataset (T, DEW, U, V; hourly, 0.25°×0.25°). Covering the period 2015–2023, the data were processed via 24-hour averaging and spatial interpolation to form the final dataset. (2) Model Output Data This consists of estimated atmospheric PM2.5 concentrations over the Bohai Sea output by the PM2.5-DNN model, with a spatiotemporal resolution of daily and 0.1°×0.1°. (3) Datou Mountain Station Observational Data PM2.5 observational data provided by the Shandong Provincial Environmental Monitoring Center, used for model validation and comparison. The temporal resolution is daily, covering the periods 2019–2020 and 2022–2023. (4) Test Set Data Randomly selected 10% of the PM2.5 data from the model input, used for model validation. (5) Shipborne Observation Data Collected during aerosol sampling campaigns aboard the 'Dongfanghong 2' scientific research vessel in July 2016 and September 2017. These campaigns utilized KC-1000 high-volume samplers (Qingdao Laoshan Electronics Co., Ltd.), equipped with AN-200 Andersen impactors and Whatman 41 filters, to collect total suspended particles (TSP) and PM2.5. (6) Reanalysis Comparison Data PM2.5 data derived from the EAC4 and MERRA-2 reanalysis datasets. The original temporal resolution is hourly, with original spatial resolutions of 0.75°×0.75° (EAC4) and 0.5°×0.625° (MERRA-2), covering 2015–2023. The data were processed via 24-hour averaging and spatial interpolation to form the final dataset. (7) WRF-CMAQ Comparison Data Data obtained from the WRF-CMAQ model outputs. The original temporal resolution is hourly, with an original spatial resolution of 12 km. The data were processed via 24-hour averaging and spatial interpolation to form the final dataset. (8) Model A Python-based model program built on the TensorFlow + Keras framework. Full Names and Units of Variables Involved in the Above Data: - AOD: Aerosol Optical Depth - T: Air Temperature (K) - DEW: Dew Point Temperature (K) - U/V: U/V-Component of the Wind (m s⁻¹) - PM2.5 (μg m⁻³)
Files
Institutions
- Qingdao University of Science and Technology
- Ocean University of China