---
license: mit
task_categories:
- tabular-regression
size_categories:
- 1K<n<10K
---

# Dataset Card for MIL-QUALAIR

<!-- Provide a quick summary of the dataset. -->

The dataset has been constructed for urban air pollution forecasting in task the Milan metropolitan area and includes Sentinel-5P satellite observations, meteorological conditions, topographical features, and ground monitoring station measurements.
## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->
The dataset encompasses a compilation of various data sources, including Sentinel-5 satellite observations, Digital Elevation Model (DEM) data, land cover information, meteorological records, and ground-level measurements, spanning the period from 2018 to 2023 within the metropolitan area of Milan. It is curated to support the task of forecasting the concentrations of five major pollutants namely PM10, PM25, NO2, O3, SO2. This dataset has been utilized and introduced in the study *Urban Air Pollution Forecasting: A Machine Learning Approach Leveraging Satellite Observations and Meteorological Forecasts*.


- **Curated by:** LINKS Foundation
- **Funded by:** [UP2030](https://up2030-he.eu/) project
- **License:** MIT License

### Dataset Sources [optional]

<!-- Provide the basic links for the dataset. -->

- **Paper:**
```latex
@misc{https://doi.org/10.48550/arxiv.2405.19901,
  doi = {10.48550/ARXIV.2405.19901},
  url = {https://arxiv.org/abs/2405.19901},
  author = {Blanco,  Giacomo and Barco,  Luca and Innocenti,  Lorenzo and Rossi,  Claudio},
  keywords = {Machine Learning (cs.LG),  FOS: Computer and information sciences,  FOS: Computer and information sciences,  I.2.m; G.3},
  title = {Urban Air Pollution Forecasting: a Machine Learning Approach leveraging Satellite Observations and Meteorological Forecasts},
  publisher = {arXiv},
  year = {2024},
  copyright = {Creative Commons Attribution Non Commercial Share Alike 4.0 International}
}

```

## Uses

<!-- Address questions around how the dataset is intended to be used. -->
The dataset is intended to serve as a comprehensive resource for researchers and practitioners interested in studying urban air quality dynamics and developing pollution forecasting models. With its diverse array of environmental data sources, including Sentinel-5 satellite observations, Digital Elevation Model (DEM) data, land cover information, meteorological records, and ground-level measurements, the dataset offers rich insights into the complex interplay of factors influencing air pollution levels in the Milan metropolitan area. Researchers can utilize this dataset to investigate correlations between different environmental variables and pollutant concentrations, identify patterns and trends over time, and develop and validate predictive models for air quality forecasting.

### Direct Use

<!-- This section describes suitable use cases for the dataset. -->
Major dataset use case is for the development of air pollution forecasting models. By combining various data sources within the dataset, users can create a comprehensive feature set for each day. This aggregated feature set provides a robust foundation for predicting the levels of the five supported pollutants with greater accuracy. The repository presents each data source separately, allowing users to follow the aggregation process outlined in the associated paper or develop their own methodology tailored to specific research objectives.


## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

**Sentinel 5P**
- sentinel5.csv : Daily readings of Sentinel5P satellite bands, sampled around each station

**DEM**
- dem.tiff : 10m-resolution map of Milan metropolitan area with Digital Elevation model measurement

**Weather**
- weather.csv : Daily measurements of weather variables

**Land Cover**
- land_cover/land_cover.tiff : 10m-resolution map of Milan metropolitan area land cover classification
- land_cover/land_cover_taxonomy.json : Association between numeric class in tiff file and correspondent label
- land_cover/land_cover_mapping.json : Mapping of land cover classes as proposed in the original work

**Ground truth**
- stations.csv : Daily readings of station measurements for the five supported pollutants

## Dataset Creation


### Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->
**Sentinel 5P**
- ESA Copernicus Sentinel 5p mission

**DEM**
- Copernicus

**Weather**
- [Visual crossing Weather](https://www.visualcrossing.com/) 

**Land Cover**
- Copernicus Land Monitoring Service - Urban Atlas

**Ground truth**
- Milan open data portal

## Dataset Card Authors [optional]

Giacomo Blanco, Luca Barco, Lorenzo Innocenti and Claudio rossi

## Dataset Card Contact
giacomo.blanco@linksfoundation.com, luca.barco@linksfoundation.com, lorenzo.innocenti@linksfoundation.com, claudio.rossi@linksfoundation.com