Downscaling NOx emission into 1 km resolution over a typical mega-city based on POI with machine learning method
Nitrogen oxides (NO _x , NO + NO _2 ) are important air pollutants that significantly impact human health and directly contribute to the formation of ambient ozone and inorganic aerosols. High-resolution NO _x emission inventory is critical for effective pollution management, yet such data often rel...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Environmental Research Communications |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2515-7620/ade7d8 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Nitrogen oxides (NO _x , NO + NO _2 ) are important air pollutants that significantly impact human health and directly contribute to the formation of ambient ozone and inorganic aerosols. High-resolution NO _x emission inventory is critical for effective pollution management, yet such data often rely on improper proxies or require extensive preliminary work. Precise disaggregation of emissions based on the latitude-longitude coordinates of emitting facilities is crucial for constructing high-resolution emission inventories. Machine learning methods effectively analyze the association between Point of Interest (POI) and actual emission data, thereby enhancing the accuracy of downscaling process. In this study, we downscaled NO _x emissions from the transport, industry, power plant, and residence sectors in the Multi-resolution Emission Inventory for China (MEIC), originally at 0.25 degree resolution (Low-resolution Inventory, LO), into 1 km resolution (High-resolution Inventory, HI) over Hefei with machine learning that incorporates POI and multi-source remote sensing information. While total emissions in HI and LO are similar, significant spatial variations exist between them. Compared to LO, HI allocates lower emissions to the city center and higher emissions to surrounding areas, thereby providing a more precise representation of emission hotspots. We evaluated both inventories using WRF-Chem and compared the simulated results against ground-based NO _2 observations. The HI-based simulations showed better agreement with observations, with spatial correlation coefficients based on HI and LO were 0.72 and 0.19, respectively. The normalized mean bias (NMB) between simulated and observed NO _2 concentrations was −17.25% for HI and −38.68% for LO, indicating that HI-based simulations substantially reduce underestimation bias. These findings indicate that the downscaled 1 km high-resolution emission inventory provides a more accurate representation of NO _x emission distributions in Hefei. Consequently, simulations based on HI more accurately reproduce NO _2 concentrations at urban scales. |
|---|---|
| ISSN: | 2515-7620 |