Deep learning for Aerosol Forecasting

TL;DR

This paper develops a deep learning model combining CNN with MERRA-2 reanalysis data to improve aerosol optical depth predictions, validated against ground measurements, addressing biases in traditional reanalysis datasets.

Contribution

Introduces a hybrid CNN-based model that enhances aerosol optical depth estimation by integrating reanalysis data with ground truth measurements.

Findings

The hybrid CNN model outperforms MERRA-2 alone in AOD prediction.

Validated against AERONET data, showing improved accuracy.

Addresses biases in reanalysis datasets for aerosol forecasting.

Abstract

Reanalysis datasets combining numerical physics models and limited observations to generate a synthesised estimate of variables in an Earth system, are prone to biases against ground truth. Biases identified with the NASA Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) aerosol optical depth (AOD) dataset, against the Aerosol Robotic Network (AERONET) ground measurements in previous studies, motivated the development of a deep learning based AOD prediction model globally. This study combines a convolutional neural network (CNN) with MERRA-2, tested against all AERONET sites. The new hybrid CNN-based model provides better estimates validated versus AERONET ground truth, than only using MERRA-2 reanalysis.