Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest
Fine particulate matter with an aerodynamic diameter less than 2.5 µm (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant=&q...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-03-01
|
| Series: | Atmosphere |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2073-4433/15/3/384 |
| _version_ | 1827307121009491968 |
|---|---|
| author | Luo Zhang Zhengqiang Li Jie Guang Yisong Xie Zheng Shi Haoran Gu Yang Zheng |
| author_facet | Luo Zhang Zhengqiang Li Jie Guang Yisong Xie Zheng Shi Haoran Gu Yang Zheng |
| author_sort | Luo Zhang |
| collection | DOAJ |
| description | Fine particulate matter with an aerodynamic diameter less than 2.5 µm (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula>) profoundly affects environmental systems, human health and economic structures. Multi-source data and advanced machine or deep-learning methods have provided a new chance for estimating the <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> concentrations at a high spatiotemporal resolution. In this paper, the Random Forest (RF) algorithm was applied to estimate hourly <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> of the North China area (Beijing–Tianjin–Hebei, BTH) based on the next-generation geostationary meteorological satellite Himawari-8/AHI (Advanced Himawari Imager) aerosol optical depth (AOD) products. To improve the estimation of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> concentration across large areas, we construct a method for co-weighting the environmental similarity and the geographical distances by using an attention mechanism so that it can efficiently characterize the influence of spatial–temporal information hidden in adjacent ground monitoring sites. In experiment results, the hourly <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> estimates are well correlated with ground measurements in BTH, with a coefficient of determination (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">R</mi><mn>2</mn></msup></semantics></math></inline-formula>) of 0.887, a root-mean-square error (RMSE) of 18.31 <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi mathvariant="sans-serif">μ</mi></semantics></math></inline-formula>g/<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">m</mi><mn>3</mn></msup></semantics></math></inline-formula>, and a mean absolute error (MAE) of 11.17 µg/<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">m</mi><mn>3</mn></msup></semantics></math></inline-formula>, indicating good model performance. In addition, this paper makes a comprehensive analysis of the effectiveness of multi-source data in the estimation process, in this way, to simplify the model structure and improve the estimation efficiency of the model while ensuring its accuracy. |
| first_indexed | 2024-04-24T18:33:53Z |
| format | Article |
| id | doaj.art-e9dffab1a829425298be29ac55abf155 |
| institution | Directory Open Access Journal |
| issn | 2073-4433 |
| language | English |
| last_indexed | 2024-04-24T18:33:53Z |
| publishDate | 2024-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Atmosphere |
| spelling | doaj.art-e9dffab1a829425298be29ac55abf1552024-03-27T13:20:52ZengMDPI AGAtmosphere2073-44332024-03-0115338410.3390/atmos15030384Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random ForestLuo Zhang0Zhengqiang Li1Jie Guang2Yisong Xie3Zheng Shi4Haoran Gu5Yang Zheng6State Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaState Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaState Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaState Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaThe Administrative Center for China’s Agenda 21, Beijing 100038, ChinaState Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaState Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, ChinaFine particulate matter with an aerodynamic diameter less than 2.5 µm (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula>) profoundly affects environmental systems, human health and economic structures. Multi-source data and advanced machine or deep-learning methods have provided a new chance for estimating the <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> concentrations at a high spatiotemporal resolution. In this paper, the Random Forest (RF) algorithm was applied to estimate hourly <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> of the North China area (Beijing–Tianjin–Hebei, BTH) based on the next-generation geostationary meteorological satellite Himawari-8/AHI (Advanced Himawari Imager) aerosol optical depth (AOD) products. To improve the estimation of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> concentration across large areas, we construct a method for co-weighting the environmental similarity and the geographical distances by using an attention mechanism so that it can efficiently characterize the influence of spatial–temporal information hidden in adjacent ground monitoring sites. In experiment results, the hourly <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi mathvariant="normal">P</mi><msub><mi mathvariant="normal">M</mi><mn>2.5</mn></msub></mrow></semantics></math></inline-formula> estimates are well correlated with ground measurements in BTH, with a coefficient of determination (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">R</mi><mn>2</mn></msup></semantics></math></inline-formula>) of 0.887, a root-mean-square error (RMSE) of 18.31 <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mi mathvariant="sans-serif">μ</mi></semantics></math></inline-formula>g/<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">m</mi><mn>3</mn></msup></semantics></math></inline-formula>, and a mean absolute error (MAE) of 11.17 µg/<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><msup><mi mathvariant="normal">m</mi><mn>3</mn></msup></semantics></math></inline-formula>, indicating good model performance. In addition, this paper makes a comprehensive analysis of the effectiveness of multi-source data in the estimation process, in this way, to simplify the model structure and improve the estimation efficiency of the model while ensuring its accuracy.https://www.mdpi.com/2073-4433/15/3/384PM<sub>2.5</sub>random forestattention mechanismspatiotemporal predictionmulti-source data |
| spellingShingle | Luo Zhang Zhengqiang Li Jie Guang Yisong Xie Zheng Shi Haoran Gu Yang Zheng Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest Atmosphere PM<sub>2.5</sub> random forest attention mechanism spatiotemporal prediction multi-source data |
| title | Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest |
| title_full | Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest |
| title_fullStr | Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest |
| title_full_unstemmed | Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest |
| title_short | Improving the Estimation of PM<sub>2.5</sub> Concentration in the North China Area by Introducing an Attention Mechanism into Random Forest |
| title_sort | improving the estimation of pm sub 2 5 sub concentration in the north china area by introducing an attention mechanism into random forest |
| topic | PM<sub>2.5</sub> random forest attention mechanism spatiotemporal prediction multi-source data |
| url | https://www.mdpi.com/2073-4433/15/3/384 |
| work_keys_str_mv | AT luozhang improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT zhengqiangli improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT jieguang improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT yisongxie improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT zhengshi improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT haorangu improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest AT yangzheng improvingtheestimationofpmsub25subconcentrationinthenorthchinaareabyintroducinganattentionmechanismintorandomforest |