New Model to Identify Air Pollution Sources with Machine-learning Technique
Abstract
A new concept, potential source density function (PSDF), is developed identify the source areas of ambient trace species. PSDF can show a spatial distribution of source to explain the variation of ambient trace species' concentration at a receptor site. The PSDF model uses only ambient data at a receptor site and backward trajectories arriving at a receptor site during sampling. PSDF can be quickly calculated by the theory of Gaussian process regression (GPR) and structured kernel interpolation (SKI). PSDF can quantitatively show capability to contaminate air parcel when it pass through each point and the variance of estimated values. It can also evaluate the distribution of a certain area in the concentrations at a receptor site.
In this presentation, the concept of PSDFs is described, by referring the theory of GPR. Numerical examples are provided to demonstrate the possibility of using the concept of PSDF for studying the locations of contamination sources. Also, we validate the PSDF model to reproduce the ambient data used as input. A brief comparison of the result against that of the PSCF method is also provided. 첨부파일 영역- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMEP54C..15K
- Keywords:
-
- 1824 Geomorphology: general;
- HYDROLOGY;
- 1910 Data assimilation;
- integration and fusion;
- INFORMATICS;
- 1942 Machine learning;
- INFORMATICS;
- 4217 Coastal processes;
- OCEANOGRAPHY: GENERAL