Abstract
Landsat imagery offers remarkable potential for various applications,
including land monitoring and environmental assessment, thanks to its
high spatial resolution and over 50 years of data records.
However, the presence of
atmospheric aerosols greatly hinders the precision of land
classification and the quantitative retrieval of surface parameters.
Notably, there has been no global
retrieval of aerosol optical depth (AOD) from Landsat imagery that is
needed for atmospheric correction, among other applications.
To address this issue, this paper
presents an innovative global AOD retrieval framework for Landsat
imagery, propelled by atmospheric radiative transfer (ART) and enhanced
GeoChronoTransformers (GCT) models incorporating multidimensional
spatiotemporal sequence information and executed on the Google Earth
Engine (GEE) cloud platform. We
gathered all Landsat 8 and 9 images from their respective launch dates
(February 2013 and September 2021) up to 2022, which were used to
construct a robust ART-GCT-GEE model, and then rigorously validated the
model performance across ~470 monitoring stations over
land using diverse spatiotemporally independent methods.
Leveraging information from
multiple spectral channels, contributing to 58% according to the
SHapley Additive exPlanation (SHAP) method, our results are highly
consistent with observations (e.g., correlation coefficient = 0.863 and
root-mean-square error = 0.096),
suggesting that accurate historical and future AOD levels can be
obtained. Around 81% and 50% of
our AOD predictions meet the criteria of Moderate Resolution Imaging
Spectroradiometer (MODIS) expected errors [±(0.05+20%)] and the
Global Climate Observation System {[max(0.03, 10%)]},
respectively. Additionally, our
model is less influenced by changes in surface conditions like
topography and land cover. This
allows us to generate spatially continuous AOD distributions with highly
detailed and fine-scale information from dark to bright surfaces,
especially for densely populated urban areas and expansive deserts with
high aerosol loadings from both anthropogenic and natural sources.