Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels
The immense problem of missing satellite aerosol retrievals (Aerosol Optical Depth, (AOD)) detrimentally affects the prediction ability of ground-level PM<sub>2.5</sub> concentrations and may lead to unavoidable biases. An appropriate missing-imputation method has not been well developed...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-09-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/12/18/3008 |
| _version_ | 1797553614547320832 |
|---|---|
| author | Zhao-Yue Chen Jie-Qi Jin Rong Zhang Tian-Hao Zhang Jin-Jian Chen Jun Yang Chun-Quan Ou Yuming Guo |
| author_facet | Zhao-Yue Chen Jie-Qi Jin Rong Zhang Tian-Hao Zhang Jin-Jian Chen Jun Yang Chun-Quan Ou Yuming Guo |
| author_sort | Zhao-Yue Chen |
| collection | DOAJ |
| description | The immense problem of missing satellite aerosol retrievals (Aerosol Optical Depth, (AOD)) detrimentally affects the prediction ability of ground-level PM<sub>2.5</sub> concentrations and may lead to unavoidable biases. An appropriate missing-imputation method has not been well developed to date. This study developed a two-stage approach (AOD-imputation stage and PM<sub>2.5</sub>-prediction stage) to predict short-term PM<sub>2.5</sub> exposure in mainland China from 2013–2018. At the AOD-imputation stage, geostatistical methods and machine learning (ML) algorithms were examined to interpolate 1 km satellite aerosol retrievals. At the PM<sub>2.5</sub>-prediction stage, the daily levels of PM<sub>2.5</sub> were predicted at a resolution of 1 km, based on interpolated AOD and meteorological data. The statistical performances of the different interpolation methods were comprehensively compared at each stage. The original coverage of retrieved AOD was 15.46% on average. For the AOD-imputation stage, ML methods produced a higher coverage (98.64%) of AOD than geostatistical methods (21.43–87.31%). Among ML algorithms, random forest (RF) or extreme gradient boosted (XG-interpolated) AOD produced better interpolated quality (CV R<sup>2</sup> = 0.89 and 0.85) than other algorithms (0.49–0.78), but XGBoost required only 15% of the computing time of RF. For the PM<sub>2.5</sub> predicted stage, neither RF-AOD nor XG-AOD could guarantee higher accuracy in PM<sub>2.5</sub> estimations (CV R<sup>2</sup> = 0.88 (RF or XG-AOD) compared to 0.85 (original)), or more stable spatial and temporal extrapolation (spatial, (temporal) CV R<sup>2</sup> = 0.83 (0.83), 0.82 (0.82), and 0.65 (0.61) for RF, XG, and original). For the AOD-imputation stage, the missing-filled efficiency depended more on external information, while the missing-filled accuracy relied more on model structure. For the PM<sub>2.5</sub> predicted stage, efficient AOD interpolation (or the ability to eliminate the missing data) was a precondition for the stable spatial and temporal extrapolation, while the quality of interpolated AOD showed less significant improvements. It was found that XG-AOD is a better choice to estimate daily PM<sub>2.5</sub> exposure in health assessments. |
| first_indexed | 2024-03-10T16:19:02Z |
| format | Article |
| id | doaj.art-dcb0101edc4d4829b10143eea6fd9ee6 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T16:19:02Z |
| publishDate | 2020-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-dcb0101edc4d4829b10143eea6fd9ee62023-11-20T13:50:46ZengMDPI AGRemote Sensing2072-42922020-09-011218300810.3390/rs12183008Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> LevelsZhao-Yue Chen0Jie-Qi Jin1Rong Zhang2Tian-Hao Zhang3Jin-Jian Chen4Jun Yang5Chun-Quan Ou6Yuming Guo7Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Melbourne, VIC 3004, AustraliaState Key Laboratory of Organ Failure Research, Department of Biostatistics, Guangdong Provincial Key Laboratory of Tropical Disease Research, School of Public Health, Southern Medical University, Guangzhou 510515, ChinaDepartment of Urban Planning and Design, The University of Hong Kong, Pokfulam, Hong KongState Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, ChinaDepartment of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Melbourne, VIC 3004, AustraliaInstitute for Environmental and Climate Research, Jinan University, Guangzhou 511443, ChinaState Key Laboratory of Organ Failure Research, Department of Biostatistics, Guangdong Provincial Key Laboratory of Tropical Disease Research, School of Public Health, Southern Medical University, Guangzhou 510515, ChinaDepartment of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Melbourne, VIC 3004, AustraliaThe immense problem of missing satellite aerosol retrievals (Aerosol Optical Depth, (AOD)) detrimentally affects the prediction ability of ground-level PM<sub>2.5</sub> concentrations and may lead to unavoidable biases. An appropriate missing-imputation method has not been well developed to date. This study developed a two-stage approach (AOD-imputation stage and PM<sub>2.5</sub>-prediction stage) to predict short-term PM<sub>2.5</sub> exposure in mainland China from 2013–2018. At the AOD-imputation stage, geostatistical methods and machine learning (ML) algorithms were examined to interpolate 1 km satellite aerosol retrievals. At the PM<sub>2.5</sub>-prediction stage, the daily levels of PM<sub>2.5</sub> were predicted at a resolution of 1 km, based on interpolated AOD and meteorological data. The statistical performances of the different interpolation methods were comprehensively compared at each stage. The original coverage of retrieved AOD was 15.46% on average. For the AOD-imputation stage, ML methods produced a higher coverage (98.64%) of AOD than geostatistical methods (21.43–87.31%). Among ML algorithms, random forest (RF) or extreme gradient boosted (XG-interpolated) AOD produced better interpolated quality (CV R<sup>2</sup> = 0.89 and 0.85) than other algorithms (0.49–0.78), but XGBoost required only 15% of the computing time of RF. For the PM<sub>2.5</sub> predicted stage, neither RF-AOD nor XG-AOD could guarantee higher accuracy in PM<sub>2.5</sub> estimations (CV R<sup>2</sup> = 0.88 (RF or XG-AOD) compared to 0.85 (original)), or more stable spatial and temporal extrapolation (spatial, (temporal) CV R<sup>2</sup> = 0.83 (0.83), 0.82 (0.82), and 0.65 (0.61) for RF, XG, and original). For the AOD-imputation stage, the missing-filled efficiency depended more on external information, while the missing-filled accuracy relied more on model structure. For the PM<sub>2.5</sub> predicted stage, efficient AOD interpolation (or the ability to eliminate the missing data) was a precondition for the stable spatial and temporal extrapolation, while the quality of interpolated AOD showed less significant improvements. It was found that XG-AOD is a better choice to estimate daily PM<sub>2.5</sub> exposure in health assessments.https://www.mdpi.com/2072-4292/12/18/3008machine learningaerosol optical depthmissing replacementshort-termPM<sub>2.5</sub> |
| spellingShingle | Zhao-Yue Chen Jie-Qi Jin Rong Zhang Tian-Hao Zhang Jin-Jian Chen Jun Yang Chun-Quan Ou Yuming Guo Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels Remote Sensing machine learning aerosol optical depth missing replacement short-term PM<sub>2.5</sub> |
| title | Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels |
| title_full | Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels |
| title_fullStr | Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels |
| title_full_unstemmed | Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels |
| title_short | Comparison of Different Missing-Imputation Methods for MAIAC (Multiangle Implementation of Atmospheric Correction) AOD in Estimating Daily PM<sub>2.5</sub> Levels |
| title_sort | comparison of different missing imputation methods for maiac multiangle implementation of atmospheric correction aod in estimating daily pm sub 2 5 sub levels |
| topic | machine learning aerosol optical depth missing replacement short-term PM<sub>2.5</sub> |
| url | https://www.mdpi.com/2072-4292/12/18/3008 |
| work_keys_str_mv | AT zhaoyuechen comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT jieqijin comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT rongzhang comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT tianhaozhang comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT jinjianchen comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT junyang comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT chunquanou comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels AT yumingguo comparisonofdifferentmissingimputationmethodsformaiacmultiangleimplementationofatmosphericcorrectionaodinestimatingdailypmsub25sublevels |