Introduction

The POPCORN project provides an algorithm that improves the accuracy of satellite data products, combining conventional retrieval algorithms and machine learning . It is a general approach that can be applied to any satellite data product, provided traning data are available and suitable.

In this case the POPCORN approach is implemented to improve the accuracy of the Sentinel-3 Synergy 300-meter resolution aerosol optical depth(AOD)

The Code

The POPCORN processing for the Sentinel-3 use case is implemented in a single Python script that can be run directly in the CLI (Command Line Interpreter) after providing the required data and arguments.

In particular, the script requires the following Sentinel-3 data products available (in zip format) under a directory defined by the user.

• SY_2_SYN____
• OL_1_ERR____
• SL_1_RBT____

The complete details about the project and the related code are available in the GitLab repository

A POPCORN running example

According with the instructions given in the POPCORN repository, it is possible to run the processing on the data products already available in the current POPCORN directory of Jupyter:

! python /home/jovyan/work/S3POPCORN/S3POPCORN.py /home/jovyan/work/S3POPCORN /home/jovyan/work/S3POPCORN S3A_SY_2_SYN____20220115T095432_20220115T095732_20220117T032715_0179_081_022_2160_LN2_O_NT_002.zip

Note of the AI4EO KMT mantainer: this POPCORN running example aims to show only the complete integration of the POPCORN Python script in the Jupyter environment. For a matter of data availability the data products involved in the processing are different from the ones that should be used to obtain a proper result.