Skip to content

Earth Observation guide

This guide aims to help researchers to work with Earth Observation (EO) data using CSC's computing resources. The purpose of this guide is to give an overview of available options, so it would be easier to decide if CSC has suitable services for your EO research. It helps you find the right data and tools for raster data based EO tasks. This guide focuses on spaceborne platforms. However, many tools and concepts also apply to airborne platforms. If you are interested in the fundamentals of EO, please check the resources and further reading section.

What are the benefits of using EO data?

  • Possibility to observe wide areas at same time
  • Non-intrusive
  • Same sensor for different parts of the world, easy to compare different areas
  • Time series to see changes during different seasons and years
  • Cost-efficient

Raster data format

Most EO data is available in raster format. The most common file formats are GeoTiff and GeoJPEG2000.

Why should I use CSC computing resources for EO?

For working with EO data in general, there are three main options:

1) EO specific services, which provide both data and ready-to-use processing environments. Usually these give better user experience and efficiency, but the services might be limited in computing power, available tools and options for adding own data or tools. These might have usage fees. Examples are Copernicus Data Space Ecosystem, Google Earth Engine and Microsoft Planetary Computer.

2) Cloud services with access to EO data. Practically, the data is often stored in object-storage and can be accessed as independent service. They provide general computing services, such as virtual machines, to which EO tools need to be installed by the end-user. These options usually have some fees, mainly for processing and storage. The data download may be free of charge or have a small cost, depending on the amount of data needed. One example is Amazon Web Services; also the Microsoft Planetary Computer somewhat fit this category.

3) Own computing environment - PC, local cluster, virtual machines. Data needs to be downloaded and all tools must be installed to this system. On the other hand, it gives more freedom to select the tools and set-up. Usually this does not cause any extra costs, but the computing power is usually rather limited.

CSC services do not fit well in this categorization, as they provide some features from all of these. CSC computing services provide a lot of computing power and storage space, and they are free of charge for Finnish researchers for academic or educational use.

At CSC, EO data can be processed and analyzed using a supercomputer, for example supercomputer Puhti, or a virtual machine in the cPouta cloud service. Puhti's computing capacity can hardly be compared to any other EO service, in both available processing power and amount of memory. Both Puhti and cPouta have also GPU resources, which are especially useful for large simulations and deep learning use cases.

Puhti has a lot of pre-installed applications, so it is a ready-to-use environment. cPouta virtual machines are similar to commercial cloud services, where all set-up and installations are done by the end-user. In general, both services only support Linux software.

At CSC, some Finnish EO datasets are available for direct use. In many cases, however, downloading EO data from other services (see list of EO data download services) is a required step of the process. Puhti and cPouta provide local storage of ~1-20 Tb. For more storage space, Allas object storage can be used.

Using CSC computing services requires basic Linux skills and ability to use some scripting language (for example Python, R, Julia) or command-line tools. In addition, supercomputers and virtual machines require you to understand some specific concepts, so it takes a few hours to get started. The Puhti web interface makes the start considerably easier, providing a desktop environment in the web browser, which enables the use of tools with Graphical User Interfaces (GUI) and also tools like R Studio and JupyterLab for an easy start with R, Python and Julia.

What data do I need?

When starting a task that requires EO data, there are multiple factors to consider. The decision on what are the most important factors depends heavily on the task and the resources available. The following list summarizes what one needs to consider when defining the data needs:

  • Sensor: Different sensors cover different intervals of the electromagnetic (EM) spectrum and with that show different properties of the observed areas, they can be active or passive:
    • Multispectral: multiple intervals around the visible spectrum of the EM are observed at the same time
    • Hyperspectral: more but usually shorter intervals of the EM are observed at the same time
    • RADAR (Radio Detection and Ranging), SAR (Synthetic Aperture Radar), active sensing in the microwave/radio frequencies of the EM spectrum
    • LiDAR (Light Detection and Ranging), using a laser as energy source in the optical part of the EM spectrum
    • Note that depending on the wavelengths observed, clouds, ground conditions and atmospheric artifacts may result in data gaps
  • Resolution
    • Temporal: when and how often a certain area is revisited
    • Spatial: the area on the ground that each pixel covers, determining the size of the smallest possible feature that can be detected
    • Spectral: the area of the electromagnetic spectrum that is observed and spectral width of each band provided
    • Radiometric: number of bits used to represent the energy recorded (bit-depth)
  • Costs:
    • Some EO data is freely available as open data
    • Some commercial datasets might be possible to get for free/less for research
  • Preprocessing level
    • Raw data - can have different levels and often need to be processed before it can be used for reliable analysis
    • Different levels of preprocessed data - make sure you are aware of what kind of preprocessing has been performed on your data
    • Analysis ready data (ARD)
    • Mosaics
  • User experience and knowledge
    • Appropriate background knowledge required for many tasks
    • ARD is "ready to go", but be aware of what preprocessing has been performed on your data

Some widely used EO datasets

Name Max resolution, m Revisit time, days Years of operation Open data
ESA, Sentinel-2 10-60 5 2015-> Yes
NASA, Landsat 15-120 8 1972-> Yes
ESA, Proba-V 100-1000 1-2 2013-> Yes
Airbus, Spot 1.5 - 1986-> No
Planet, several satellites 0.5-5 - 2009-> No*
DigitalGlobe, WorldView 0.3-30 - 1997-> No
Airbus, Pleiades 0.3-0.5 - 2012-> No
NASA, MODIS 250-100 1-2 1999-> Yes
NASA, EO-1 10-30 - 2000-2017 Yes
Radar, SAR
ESA, Sentinel 1 5 6 2014-> Yes
ESA, Radarsat 1-100 24 1995-> Yes
TanDEM-X/TERRASAR-X 0.25-40 - 2010-> No
ICEYE 0.5-2.5 1 2018-> No
LiDAR Footprint size
NASA, ICESat2 13 91 2019-> Yes
NADA, GEDI 25 - 2018-> Yes

* See Planets page for education and research for limited, non-commercial access to PlanetScope and RapidEye imagery.

EO database

Database of all EO missions and instrument information can be found in the CEOS EO handbook database. See also EOReader band mapping graphics for an overview of observed wavelength intervals for different optical sensors.

Where can I find the data?

Commercial datasets are usually available from data provider, while open datasets may be available in different processing stages from different services. Where possible, it might be a good idea to check processing options close to the data, for direct access or faster download. While graphical browse and download services can provide a good overview of the data and are easy to use, the download of huge amounts of data gets considerably easier using a bulk downloader or download API (Application Programming Interface).


Many data providers provide a Spatio Temporal Asset Catalog (STAC) of their datasets. These catalogs help in finding available data based on time and location with the possibility for multiple additional filters, such as cloud cover and resolution. The STAC Index provides a nice overview of available catalogs from all over the world. The STAC Index page includes many resources for learning and utilizing STAC. Finnish data is available from Paituli STAC. Check out also CSC's examples for utilizing STAC from Python and STAC from R.

EO data at CSC

Some Finnish EO datasets are available locally at CSC. Paituli STAC includes all raster data available at CSC.

EO data download services

SYKE/FMI, Finnish image mosaics : Sentinel-1, Sentinel-2 and Landsat mosaics, also index mosaics. For several time periods per year. These are included in Paituli STAC

ESA Copernicus Data Space Ecosystem provides worldwide main Sentinel products, see below for more information.

FinHub covers Finland and the Baltics and offers Sentinel-2 L1C (but not L2A) and Sentinel 1 SLC, GRD and OCN products. No STAC. sentinelsat Python package is suitable for downloading data from FinHub, see CSC FinHub sentinelsat example.

USGS EarthExplorer huge datastore with focus on US data, but also worldwide Landsat datasets. USGS is the main provider of the new Landsat Collection 2 data. Landsat Collection 2 STAC

NASA Earthdata provides among many others harmonized Landsat 8 and Sentinel-2 dataset. NASA STAC

Amazon Web Service (AWS) open EO data is a collection of worldwide EO datasets provided by different organizations, including Landsat and Sentinel. Some of the data can be downloaded only on "requestor pays" basis. Currently, Sentinel-2 L2A Cloud-optimized Geotiffs by Element 84 are available for free, inc. STAC.

Microsoft planetary computer provides a STAC of all available data, which includes Sentinel, Landsat, MODIS.

Google Cloud Storage open EO data, including Sentinel-2 L1C and Landsat Collection 1 data. Data can be downloaded for example with FORCE.

Terramonitor provides pre-prosessed analysis ready Sentinel-2 data, also from Finland. It is a commercial service.

Almost all of the services provide download with web interface and bulk download via API. Most services require free self-registration.

Other geospatial datasets

To find other geospatial datasets, check out CSC open spatial dataset list.

ESA Copernicus Data Space Ecosystem

Copernicus Data Space Ecosystem (CDSE) provides the possibility to browse, visualize, download and analyze EO data. It started in late 2023 and replaced ESA's SciHub. CDSE mainly includes different Sentinel datasets, but also some complimentary datasets, inc Landsat, see full list of CDSE datasets. Note that duplicates may be available due to reprocessing with newest baselines.

CDSE data APIs and data download:

  • CDSE Browser - web interface for accessing, exploring and downloading the data.
  • CDSE Catalog APIs support 3 different options for finding suitable data: OData, OpenSearch and STAC. OData and OpenSearch provide similar functionality. In principle, also STAC has a lot of potential, because it would easier to download only needed data, for example only some bands or geographically only parts of data. It is the newest of data APIs and at the moment does not support any other search criteria than collection, time and location, so for example cloud cover filtering is not possible (yet). So use cases for Sentinel2 STAC are currently limited, but for Sentinel1 it might be more suitable.
  • CDSE S3 for high-performance parallel access and download from CDSE object storage.

Several example scripts are available for CDSE data download:

You can also read data directly from S3 with GDAL or GDAL-based tools, see CSC GDAL cloud tutorial.

These data APIs are free of charge. Different services have different limitations, see CDSE Quotas and limitations. Compared to the ESA's previous SciHub service, the number of concurrent downloads per user has increased from two to four for most APIs.

CDSE includes also OpenEO and SentinelHub services, which provide more analysis ready datasets and own download services and APIs. Both have own STAC. SentinelHub provides also OGC APIs.

How can I process EO data at CSC?

You can find information about geocomputing using CSC resources and how to get started on CSC geocomputing pages, including links to creating user accounts and all other practical information.

What to consider when choosing a software?

There is no single software perfect for every task and taste. The right software depends as much on the task to be worked on, as on the taste and skillset of the user. The following list sunmmarizes things that need to be considered when choosing a software.

  • Functionality: Does the software provide the tools you need to reach your goal?
  • Interaction type: How do you want to interact with the software?
    • Graphical User Interface (GUI)
    • Command Line Interface (CLI)
    • Scripting
  • Technical aspects:
    • Reproducibility: Does the tool provide the possibility to record work steps?
    • Supported operating systems: Can the tool be installed to the operating system available to you?
    • Automation possibility: Can the tool execution be automatized for big data processing, if needed?
    • Combination possibility: Can you combine the tool with other tools?
    • Computational efficiency: Does the tool make good use of the available computational resources (especially GPUs)?
    • Support for parallel computing or batch processing
  • Open source vs proprietary
    • Proprietary tools need licenses which may be expensive and/or limiting the use of the tool
    • FOSS (free and open source software) allows the user to inspect the source code and provide high level insights in its functionality

What applications are available on Puhti?

FORCE - Framework for Operational Radiometric Correction for Environmental monitoring. All-in-one processing engine with CLI for EO image archives. FORCE example for Puhti

GDAL (OGR) - Geospatial Data Abstraction Library. Collection of command-line tools for accessing and transforming geospatial data. It is relatively fast and requires little computational resources. GDAL supports reading data directly from the Internet or object storage. GDAL is included in many other tools for data reading and writing. GDAL example for Puhti

Julia - Puhtis Julia installation does not include any geospatial packages, but they can be installed by the user. JuliaGeo provides an overview of packages for geospatial data.

Matlab - you can run Matlab jobs on Puhti conveniently from your own computers Matlab installation.

Orfeo Toolbox (OTB) - offers a wide variety of applications from ortho-rectification or pansharpening, all the way to classification, SAR processing, and much more. Orfeo Toolbox is available as CLI, GUI and via Python interface.


  • The geoconda module provides many useful Python packages for raster data processing and analysis, such as rasterio, rasterstats, scimage, sentinelhub, xarray, boto3 and packages for working with STAC.
  • Machine learning modules provide some common machine learning frameworks, inc. for deep learning..

QGIS - open source tool with GUI for working with spatial data including limited multispectral image processing capabilities. GUI with batch processing possibility and Python interface. Used for example for visualization, map algebra and other raster processing. Many plug-ins available, for EO data processing, check out the QGIS Semi-automatic classification plugin.

R - Puhti R installation includes a lot of geospatial packages, including several useful for EO data processing, such as terra, CAST, raster, rstac and spacetime.

Sen2Cor - a command-line tool for Sentinel-2 Level 2A product generation and formatting.

Sen2mosaic - a command-line tool to download, preprocess and mosaic Sentinel-2 data.

SNAP - ESA Sentinel Application Platform. Tool for processing of Sentinel data (+ support for other data sources). GUI, CLI (Graph Processing Tool, GPT) and Python interfaces. SNAP GPT example for Puhti.

[**allas'']](../../../apps/ - tools for working with S3 storage, inc CSC Allas, CDSE S3 etc: rclone and s3cmd.

If you need further applications, you can ask CSC to install them for you.

Machine Learning with EO data

One example of the advanced usage of EO data is for machine learning. If you are interested in the topic, you can find a lot of examples from CSC machine learning with spatial data course materials. For practical guidelines, see also CSC machine learning guide

Alternative processing services

Below is a list of alternative EO processing services that might be useful, when a lot of data is required and downloading it all to CSC might not be feasible.

Google Earth Engine is a processing platform, which requires registration, but is currently free of charge for research users. It can be accessed via browser and has worldwide analysis ready data available (browse the catalog). In general, JavaScript is used on the platform, but also Python and R support exists. Check out GEE's tutorials. Note that Google Cloud Storage might be needed to export large datasets.

Microsoft planetary computer offers JupyterHub together with Dask Gateway, both CPUs and GPUs are available. It is currently available in preview.

CDSE provides also processing services, mainly via OpenEO and SentinelHub with options to bring processing close to the data. Both have free of charge options and services with a fee. They provide different APIs, which can be accessed via Python or R. Soon also On-Demand Processing.

  • Copernicus Data Workspace is a tool for managing and reviewing EO-related products, which can then be further processed or downloaded for various purposes. When products are selected for processing, you are provided with a list of processors that are capable of processing relevant data types.
  • CDSE Jupyter Notebooks provide the ability to analyze the data using Jupyter Notebooks. Each user has 10Gb of persistent space (deleted after 15 days without login) and access to 2 - 4 CPUs with 4 - 16 Gb RAM. Note that in addition to personal limits, also the total number of active users seems to be limited. It is possible to add own packages via pip. CDSE example notebooks
  • And many more, see all CDSE applications

Commercial clouds: Amazon, Google Cloud and Microsoft Azure, all provide virtual machines and other processing services, all of them have some local data, see links above.

Where can I get help?

If you are interested in using CSC services for your EO research, please make yourself familiar with the services:

You can find all the ways that you can get help from CSC specialists via CSC contact page. We are happy to help with technical problems around our services and are open for suggestions on which software should be installed to Puhti, or what kind of courses should be offered or materials/examples should be prepared. Please let us know, if you would like to add a service to this page or find anything unclear.


This guide was developed in cooperation with the Finnish Environment Institute, SYKE, as part of the Geoportti project.

Contributions welcome

If you find any mistakes or outdated links, have improvement suggestions or want to add more information about a certain topic, please add them to our Github issue for improving the EO guide, send a pull request to our CSC documentation on github or contact us via any of the ways mentioned in CSC contact page. Thank you!

Resources and further reading

If you are interested in the fundamentals of EO, take a look at these excellent resources:

  • Fundamentals of remote sensing tutorial by Canada Centre for Mapping and Earth Observation , Natural Resources Canada; an "interactive module is intended as an overview at a senior high school or early university level and touches on physics, environmental sciences, mathematics, computer sciences and geography."
  • Echoes in space - Introduction to RADAR remote sensing by the European Space Agency; "a detailed insight into the history of Radar technology, including all the basics that are needed to understand how electromagnetic waves work and a unique hands-on experience to work with Radar data in diverse application scenarios."
  • Newcomers guide to Earth Observation by the European Space Agency, "a guide to help non-experts in providing a starting point in the decision process for selecting an appropriate Earth Observation (EO) solution."
  • Earthdatascience intro to multispectral data

Further reading:

Last update: May 16, 2024