An Improved High‐Spatial‐Resolution Aerosol Retrieval Algorithm for MODIS Images Over Land

Abstract

MODerate resolution Imaging Spectroradiometer (MODIS) data can play an important role in aerosol retrieval at the global scale due to their short revisit period and long-term observations. The operational MODIS aerosol optical depth (AOD) products are severely limited in air quality studies at the city or local scales due to their coarse spatial resolutions. Therefore, an improved aerosol retrieval algorithm for MODIS images at 1-km spatial resolution is proposed in this paper. This algorithm is based on the high-resolution aerosol retrieval algorithm with a priori land surface reflectance database support (HARLS), which was developed over bright urban areas and was subsequently modified and validated over land. For this study, an eight-day surface reflectance database and a seasonal aerosol-type database over land are constructed using the MODIS surface reflectance and aerosol products. Four typical regions (in Europe, North America, Beijing-Tianjin-Hebei, and the Sahara) with different underlying surfaces and aerosol types are selected to perform the aerosol retrieval experiments. The AOD retrievals are validated against the AERosol RObotic NETwork (AERONET) version 2 level 2.0 AOD measurements and compared with the operational MODIS AOD product at 3-km resolution (MOD04_3K). The results show that AOD retrievals adequately match with AERONET AOD measurements, with 79.56%, 72.69%, 74.71%, and 61.01% of the collections falling within the MOD04_3K expected error over land [±(0.05 + 20%)] for each of the above-listed regions, respectively. The Improved HARLS (I-HARLS) algorithm performs well overall under different surface conditions, but the data quality gradually decreases with the increase in surface reflectance. Moreover, the I-HARLS algorithm is robust with less bias and can provide more detailed aerosol spatial distributions than those provided by the MOD04_3K AOD product. These results suggest that the I-HARLS algorithm can be used for air-pollution-and climate-related studies at medium or small scales.

Improved 1 km resolution PM2.5 estimates across China using enhanced space–time extremely randomized trees

“…There are only a handful of studies examining the predictive powers of models estimating PM 2.5 concentrations in China. Comparisons show that the enhanced STET model is superior to those reported in previous studies, i.e., the two-stage model (Ma et al, 2019), the GTWR model (He and Huang, 2018), the ML + GAM model (Xue et al, 2019) and the space-time RF model (Wei et al, 2019a). The enhanced STET model has a strong predictive power and can be used to estimate historical PM 2.5 concentrations in China.…”

“…Regarding model performance, our newly developed STET model is more accurate, with higher CV R 2 values and smaller RMSE and MAE values than those from statistical regression models (Table 2), e.g., the timely structure adaptive model (TSAM; Fang et al, 2016), the Generalized Additive Model (GAM; Chen et al, 2018) model, the GWR model (Ma et al, 2014;You et al, 2016), and the geographically and temporally weighted regression model (GTWR; He and Huang, 2018). The enhanced STET model can also outperform most machine learning (ML) and deep learning approaches including the Gaussian model (Yu et al, 2017), the random forest model (Chen et al, 2018;Wei et al, 2019a), the XGBoost model (Chen et al, 2019), the GRNN and deep brief network (DBN) models (T. Li et al, 2017a, b), and some optical combined models, e.g., the Daily-GWR model (D-GWR; He and Huang, 2018), the two-stage model (He and Huang, 2018;Ma et al, 2019;Yao et al, 2019) and the ML + GAM model (Xue et al, 2019).…”

“…Statistical regression models, e.g., the multiple linear regression model, the linear mixed-effect model, the two-stage model and the geographically weighted regression (GWR) model, have been widely used for applications due to their simplicity and versatility (Gupta and Christopher, 2009;Ma et al, 2014;Xiao et al, 2017;Yao et al, 2019). Artificial intelligence models mainly involve machine learning and deep learning models, e.g., the random forest (RF; Brokamp et al, 2018;Chen et al, 2018;Wei et al, 2019a), the extreme gradient boosting model (XGBoost; Chen et al, 2019), and the back-propagation and generalized regression neural networks (BRNN and GRNN; T. Li et al, 2017a). PM 2.5 is jointly affected by numerous factors, e.g., meteorological conditions, human activities and topography, showing great spatial and temporal heterogeneities.…”

Intercomparison in spatial distributions and temporal trends derived from multi-source satellite aerosol products

“…This result is mainly due to the complex and bright underlying surfaces (e.g. desert, bare land, and urban areas), as well as intense human activities, which increase the difficulty of aerosol estimation Wei et al, , 2018aWei et al, , b, 2019d. Overall, most aerosol products overestimate the monthly AOD over North America and Oceania, while general underestimations occur over South America, Africa, and eastern Asia.…”

“…desert and urban areas). Both high surface reflectance and complex underlying surfaces increase the difficulty of aerosol retrieval Wei et al, , 2018aWei et al, , b, 2019d. For the spatial distributions over the ocean, the seasonal and annual mean AOD S values are generally lower than 0.1 in most areas, especially open seas (Fig.…”