The Copernicus component is governed by Regulation (EU) No 2021/696 of the European Parliament and of the Council of 28 April 2021 establishing the Union Space Programme and the European Union Agency for the Space Programme and repealing Regulations (EU) No 912/2010, (EU) No 1285/2013 and (EU) No 377/2014 and Decision No 541/2014/EU. Within the Copernicus component, a portfolio of land monitoring activities has been delegated by the European Union to the European Environment Agency (EEA) and the DG Joint Research Centre of the European Commission.

The Copernicus land monitoring products and services are made available on a principle of full, open and free access, as established by the Commission Delegated Regulation (EU) No 1159/2013 of 12 July 2013.

Free, full and open access to the products and services of the Copernicus Land Monitoring Service is made on the conditions that:

1. When distributing or communicating Copernicus Land Monitoring Service products and services (data, software scripts, web services, user and methodological documentation and similar) to the public, users shall inform the public of the source of these products and services.

2. Where the Copernicus Land Monitoring Service products and services have been adapted or modified by the user, the user shall clearly state this.

3. Users shall make sure not to convey the impression to the public that the user's activities are officially endorsed by the European Union.

Pixel codes:

0: All non-impervious areas

1-100: 1-100% Imperviousness Density

255: Outside area

The 100m Imperviousness Density (IMD) 2021 layer is derived from the 10m Imperviousness Density (IMD) 2021 layer by aggregating pixels to 100m spatial resolution using a mean aggregation method.

Input data:

Copernicus Sentinel-2 imagery (L2A) acquired from January to December for the reference year 2021.

Pre-processing:

A data reduction algorithm was applied on of the Sentinel-2 time-series to obtain a set of band quantiles. Additionally, a combination of polynomial and harmonic functions were fitted to the time-series of Normalized Difference Vegetation Index (NDVI) to derive harmonic coefficients.

Training data:

Sentinel-2 quantiles 2018, Imperviousness Density 2018 and CLCplus Backbone 2018 layers were used for model training.

Classification and prediction:

A modified deep convolutional neural network (CNN) called U-Net was trained. The trained model was applied to the Sentinel-2 harmonics and quantiles input to predict raw Imperviousness Densities for the reference year 2021. From these predictions, areas with Imperviousness Density values greater than 90 were extracted and combined with the sealed areas (class 1) of the CLCplus Backbone 2021 layer to create a binary mask of the impervious outline for reference year 2021. The usage of this sealing mask significantly enhanced the level of detail in the impervious surface outline.

Post-processing:

Steps applied comprise: 1) Calibration using data for the reference years 2018 (Imperviousness Density, CLCplus Backbone and NDVI harmonics) and 2021 (NDVI harmonics), 2) re-calculation of the final Imperviousness Density values on the impervious pixels (class 1) of the sealing mask for 2021 and 3) pixel aggregation (100m) using an average operation.

Changes to previous release (2018):

- Introduction of a new production system and algorithm (U-Net CNN).

- Reduction of in-situ data dependency for dataset production.

- First Impervious Built-up Change (IBUC) layer (2018-2021).