# README

11.09.2026  
**Antti Arola, Timo H. Virtanen, Antti Lipponen**  
Finnish Meteorological Institute (FMI)

This README describes the data and code package associated with the manuscript:

**Retrieval-dependent sampling leads to biased satellite estimates of aerosol–cloud interactions**

by Antti Arola, Timo H. Virtanen, Antti Lipponen, Pekka Kolmonen, Tom Goren, Goutam Choudhury, Edward Gryspeerdt, Vishnu Nair, David Painemal, Hannes Keernik, Velle Toll, and Harri Kokkola, submitted to *Science Advances* in 2026.

## Overview

This repository contains data and analysis code used to investigate retrieval-dependent sampling effects in satellite-based aerosol-cloud interaction studies. The datasets are derived primarily from NASA MODIS observations, and the included scripts reproduce the analyses and figures described in the associated manuscript.

The repository is organized into three independent subsets, each corresponding to a specific analysis and figure set in the manuscript. Detailed descriptions of the data, processing methods, software requirements, and source datasets are provided in the README file within each subset.

## Repository contents

### 1. ML (Machine Learning)

Machine-learning and climatological analyses used to generate:

- Figure 2
- Figures S6-S8

See the README file within the **ML** subset for details.

### 2. Bengal

MODIS-based aerosol-cloud analysis over the Bay of Bengal used to generate:

- Figure 4

See the README file within the **Bengal** subset for details.

### 3. Kilauea

MODIS-based analysis of cloud retrieval populations near Hawaii used to generate:

- Figure S2

See the README file within the **Kilauea** subset for details.

## Access and licence

**Access right:** Open Access  
**Software licence:** MIT License

## Funding

This work was supported by:

- European Union's Horizon Europe Programme
- Research Council of Finland

## Citation

If you use these data or codes, please cite the associated manuscript and this data record.

## Contact

For questions concerning the data, code, or analysis workflow, please contact the manuscript authors.