3.88 KB
Newer Older
1 2 3 4 5
<img src=img/ODS_logo_450px.png width=135/>  eumap library
[![GitLab license](img/apache2.svg)](./LICENSE)
[![CRAN Status Badge](](
[![PyPI version](](
Leandro Parente's avatar
Leandro Parente committed
[![Zenodo dataset](](

8 9 10 11 12 13
[Community]( |
[Documentation]( |
[Resources](demo/ |
[Release Notes](

eumap is a library to enable easier access to EU environmental maps and functions to produce and improve new value-added spatial layers.
Leandro Parente's avatar
Leandro Parente committed
It implements efficient spatial and spatiotemporal overlay, High Performance Computing, extends the Ensemble Machine Learning algorithms developed within the [mlr3]( and [scikit-learn]( framework.
15 16
eumap builds upon various existing softare, especially on GDAL, R and Python packages for spatial analysis and Machine Learning libraries.

Leandro Parente's avatar
Leandro Parente committed
17 18 19 20 21
pyeumap Workflow 

![pyeumap Workflow](img/pyeumap_workflow.png)

The workflow implemented by pyeumap 1) fills all the gaps for different remote sensing time-series, 2) does the space time overlay of point samples on several raster layers according to the reference date, 3) trains and evaluate a machine learning model, and 4) does the space time predictions for a specific target variable. These processing steps are [demonstrated ](demo/python) using a [benchmark dataset for land-cover classification]( in different areas of the EU
Leandro Parente's avatar
Leandro Parente committed
23 24 25 26 27 28 29 30 31

This image presents the output of the gap filing approach for an area located in Croatia (tile 9529). This image refers to a Landsat temporal composites for the 2010 fall season, however all the 4 seasons since 2000 were analysed to fill the gaps.

![pyeumap Workflow](img/gapfill_example.png)

This animation shows the land-cover classes for an area located in Sweden (tile 22497) according to the space time predictions. This example is a small use case that used 680 point samples, obtained in different years, to train a single model and to predict the land-cover in the region over the time.

![pyeumap Workflow](img/land_cover_predictions.gif)

32 33 34 35 36 37 38 39
Spatiotemporal Machine-Learning
In the Geo-harmonizer project, we prepare Analysis-Ready Earth Observation images from Landsat and 
Sentinel missions, then use ground observations from the European Commission projects such as [LUCAS surveys]( and CORINE and similar to 
overlay the ground observations in the spacetime cubes. From this data we create spatiotemporal regression and classification matrices (see: [sample data set]( The eumap package (python and R versions) will allow accessing this data and testing models that apply Machine Learning for predictive mapping in spacetime.

![EU LUCAS scheme](img/Scheme_LUCAS_poinst_st_overlay.png)

40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57
© Contributors, 2020. Licensed under an [Apache-2]( license.

Contribute to eumap
eumap has been developed and used by a group of active community members. Your help is very valuable to make the package better for everyone. Refer to the [Community Page](

- _Publication is pending_
- eumap is one of the deliverables of the GeoHarmonizer INEA project.

This work has received funding from the European Union's the Innovation and Networks Executive Agency (INEA) under Grant Agreement [Connecting Europe Facility (CEF) Telecom project 2018-EU-IA-0095](

<img src=img/CEF_programme_logo_650px.png width=650/>