The data is:

1. Ground-truth data from air quality monitoring stations in the continental part of Lombardy and Veneto regions.
2. Remote sensing data of NO2 from Sentinel-5P TROPOMI, precipitation from CHIRPS and land surface temperature from NOAA datasets, all from Google Earth Engine (GEE).
3. Participants don't have to use all the parameters provided, but are encouraged to create new ones from the existing ones, which could help to improve the performance of the model.

Note: Participants will be provided with two datasets. One for modelling which includes ground-truth and remote sensing data, and the second and the second which includes only remote sensing data and will be used to evaluate the developed model.

Update: Participants are welcome to use elevation data from SRTM and geopy (Python) or elevatr (R) libraries in building their models.

Winning solutions need to submit:

1. A model file developed in the challenge that can be used to estimate surface NO2 levels.
2. An evaluation dataset file with an added column of estimated surface values of NO2 concentrations by the developed model.
3. A technical document describing the whole process of developing your model (from pre-processing to evaluation).
4. The code used to process the data and train the model, together with the final modelling dataset you used.

Note: we will evaluate the procedure using the technical document; the submission will not be accepted if the methodology is evaluated as unrepeatable. The submitted code is limited to R, Python and/or GEE JavaScript.