Improving Spatial Coverage for Aqua MODIS AOD using NDVI-Based Multi-Temporal Regression Analysis

Abstract

Abstract:The Moderate Resolution Imaging Spectroradiometer (MODIS) provides widespread Aerosol Optical Depth (AOD) datasets for climatological and environmental health research. Since MODIS AOD clearly lacks coverage in orbit-scanning gaps and cloud obscuration, some applications will benefit from data recovery using multi-temporal AOD. Aimed at qualitatively describing the relationship between multi-temporal AOD, AOD loadings and Normalized Difference Vegetation Index (NDVI) have been considered based on the mechanism of satellite AOD retrieval. Accordingly, the NDVI-based Weighted Linear Regression (NWLR) has been proposed to recover AOD by synthetically weighing AOD similarity, spatial proximity, and NDVI similarity. To evaluate the performance of AOD recovery, simulated experiments applying gap and window masks were conducted in South Asia and Beijing, respectively. The evaluation results demonstrated that the linear regression R 2 achieved 0.8 and the absolute relative errors remained steady. Further validation was conducted between the recovered and actual AODs using 56 Aerosol Robotic Network (AERONET) sites in East and South Asia from 2013 to 2015, which demonstrated that over 41% of recovered AODs fell within the expected error (EE) envelope. Additional validation conducted in South Asia and Beijing showed that recovery by NWLR did not expand satellite-derived AOD errors, and the accuracy of recovered AOD was consistent with the accuracy of the original Aqua MODIS Deep Blue (DB) AOD. The recovery results illustrated that AOD coverage was improved in most regions, especially in North China, Mongolia, and South Asia, which could provide better support in aerosol spatio-temporal analysis and aerosol data assimilation.

Mapping Seasonal High-Resolution PM2.5 Concentrations with Spatiotemporal Bagged-Tree Model across China

“…Given cloud effect and MODIS aerosol retrieval method, a large number of AOD data are missing in the study area. According to the mechanism of retrieving aerosol loadings from satellite-based sensors, some researchers considered the relationship between AOD loadings and NDVI, comprehensively weighing the spatial proximity, AOD and NDVI similarity, to recover AOD [30,31]. The VIIRS IP AOD at 550 nm can provide a reliable dataset with a high resolution (750 m) [32].…”

LGHAP: the Long-term Gap-free High-resolution Air Pollutant concentration dataset, derived via tensor-flow-based multimodal data fusion

“…In general, merging AOD data acquired from diverse instruments and/or platforms is the most popular approach to improve AOD spatial coverage (Sogacheva et al, 2020). Statistical methods such as linear regression (Bai et al, 2019a;Wang et al, 2019;Zhang et al, 2017), inversed variance weighting Ma et al, 2016;Sogacheva et al, 2020), and maximum likelihood estimation (Xu et al, 2015) are often applied to account for systematic bias among different datasets. Data fusion methods such as Bayesian maximum entropy can be applied to blend AOD products with different resolutions (Tang et al, 2016;.…”

Improved 1-km-Resolution Hourly Estimates of Aerosol Optical Depth Using Conditional Generative Adversarial Networks

“…Usually, most methods for merging multi-sensor AOD products are mainly spatial and temporal interpolation, which utilize the neighborhood pixel values of an AOD product to fill in the missing values of other types of AOD products at the same locations [9]. For instance, according to the relationship of group AOD pixel values at the same geographic locations from different satellite sensors, researchers developed polynomial regression models [10], maximum likelihood estimation models [11], least square estimation models [12], optimal interpolation [13], and some simplified merge schemes [14]. Additionally, there are some geostatistical methods, including the universal kriging method [15], the geostatistical inverse modeling [16], and the spatial statistical data fusion [17].…”