GHADA

GHADA is a Geo-Harmonized Atmospheric Dataset for Aerosol optical depth at 550 nm. It contains daily estimations of AOD550 over Europe with 1km spatial resolution. GHADA was built based on the MODIS MCD19A2 product with 1km spatial resolution which was used as the target variable in the ML models and modelled AOD data from Copernicus Atmosphere Monitoring Service (CAMS) with 80km spatial resolution that used to overcome the high percentage of gaps found in the prementioned product.

Three stages were applied to generate GHADA: first, we applied a simple average for all available pixels that passed the quality assurance criteria (QACloudMask = Clear and QAAdjacencyMask = Clear) of the MCD19A2 product.

Second, we created an Extra-trees machine learning model for every year between 2018-2020 to predict the missing AOD values in the MCD19A2 using modelled AOD with enhanced spatiotemporal information.

Finally, we filled MODIS MCD19A2 gaps with the predicted AOD by merging the outputs from stages one and two.

Daily AOD data for each year between 2018 – 2020 during the MODIS satellite overpassing (10:30 a.m. – 1:30 p.m.) are included in this dataset.

Table 1. Results of the space-time extremely randomized models used to predict the missing AOD in the MODIS MCD19A2 product for each year of the study period.

Year

R-squared (%)

RMSE

MAE

2018

        92.5

0.02

0.012

2019

        92.9

0.019

0.012

2020

        94.1

0.016

0.010

GHADA was validate using AERONET observations and showed a good result with ~84% of the samples falling within the expected error envelopes defined for AOD products.