Home > Newsletters > 2019_01_23

2019_01_23

Hi there!

After running for more than two years, Sentinel Hub now processes millions of requests every day. During its successful course, this unique satellite imagery processing service has become much more than just a tool for producing astonishing images of our home planet. Find out more about how you can use it, and what we plan for Sentinel Hub's future, in this issue of our Newsletter.

We also invite you to read the latest blog post by our CEO, Grega Milčinski, presenting Sentinel Hub's Roadmap for 2019, including a new API v.2.0, custom script v.2, support for your own data, data fusion and more:

Becoming a Data Driver in the Cloud for Earth Observation


 

 
 

Land Cover Classification with eo-learn: Part 2

Learn how to do land cover classification from start to finish!

The first blog in the series presented initial procedures for accessing and organizing the data, while in the second part we focus on training a machine learning (ML) classifier and using it to make prediction of land cover. We give you an example of how to use ML algorithms with open-source Python packages. You will learn that predicting land cover is not that complex by itself, but doing it properly requires validating the ML model and performing a thorough inspection of the obtained results. In the end, we use the model to predict the land cover on a country-wide scale. To compare the results of the classification with official land use data, we have made both datasets available in Geopedia, our cloud GIS web editor.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1547467563949-eo-learn+part+2.pngA temporal stack of Sentinel-2 images of a small area in Slovenia, followed by a land cover prediction, obtained via methods presented in our blog.

Article
 

 
 

Sentinel-2 Level 2A Data Available Globally

Since December 13, 2018, the European Space Agency (ESA) has been producing operational Sentinel-2 Level-2A (L2A) products for the entire globe. These products provide Bottom of the Atmosphere (BOA) reflectance images in cartographic geometry. They are derived from the associated Level-1C (L1C) products using Sen2Cor processor.

The L2A products use the same tiling, encoding and filling structure as L1C, so working with them using Sentinel Hub is straightforward. In addition to BOA reflectance images, Sentinel Hub can be used to retrieve Aerosol Optical Thickness, Water Vapour and Scene classification maps. These additional bands have a variety of uses, for example using our statistical API to easily perform cloud detection for an area of interest.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1548149798037-Sentinel-2+L1C+and+L2A+from+2019-01-20.jpgA small town of Mahabaleshwar, with one of the few evergreen forests of India. Top of the Atmosphere L1C image (left) and associated L2A BOA image (right) as seen in EO Browser (acquired on January 20, 2019).
 

 

Global Water Resources Monitoring

We are introducing the BlueDot Water Observatory service based on the Copernicus satellite imagery that provides timely information about water levels of lakes, dams, reservoirs, wetlands and similar water bodies. The BlueDot provides both historical and current data on water levels globally, and provides an easy-to-use service on its own. It can also serve as a starting point, a reference, around which a larger, predictive analytics-oriented business focused on exploitation of satellite imagery can be built. Check our blog for further information.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1547561800922-BlueDot+Folsom+Lake+small.png

The water levels of the Folsom Lake, U.S. from January 2016 up to date (inspect the data in the BlueDot Water Observatory Dashboard).

About BlueDot
 

 

Educational Role of EO Browser

In our previous issue of the newsletter we have revealed our work on open educational resources. With a wish to spread EO knowledge and bring it closer to anyone interested, we have created an educational page. Among useful links and tools addressing various EO subjects in one place, the page features a Jupyter Notebook devoted to wildfires. While we are working on adding two new use-cases, air pollution and volcanoes, we encourage you to contribute as well. For more details check out our blog.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1547559588159-NDVI+values+of+vegetation.png NDVI values of vegetation on unburned area (left) and on burned out area (right). Check the Jupyter Notebook for details!

Go to blog

 

Evalscript v.2

We have released the Evalscript v.2. Conceptually and structurally it is very similar to the scripts used so far, but offers increased flexibility with input/output data. It includes more capabilities, such as normalization of input and output data, temporal support and outputs including multiple rasters. Follow the link below for more details and find examples of Evalscript calculations such as Leaf Area Index, fAPAR and others.

Read the Documentation on Evalscript v.2

 

News on Sentinel Hub Datasets

HCHO Added to Sentinel-5P Datasets
 Existing Setinel-5P global full archive datasets (Ozone, Nitrogen dioxide, Carbon monoxide, Aerosol, Cloud and SO2 data) have grown. You can now access the Tropospheric Formaldehyde HCHO dataset with Sentinel Hub services as well.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1548247401491-S-5P+HCHO.pngInspect it directly in EO Browser!

Extended Archive of Sentinel-1 GRD on AWS
The global archive of Sentinel-1 GRD on AWS has been extended to include all data from January 1, 2017 onward.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1548246881169-Port+of+Singapore+2017-01-11.pngOne of the busiest ports in the world, Port of Singapore, as seen on January 11, 2017 in EO Browser.

 

How Soon After Acquisition Do Data Become Available?

This is a question we receive very often as our users are eager to access new data as soon after acquisition as possible. ESA, who manages the pipeline from acquisition to publishing the L1C and L2A data, puts a lot of effort to shorten the time this process takes and has significantly improved the pipeline in the past year. They take care that after acquisition data are streamed to one of the ground stations (sometimes using lasers), then go through various stages and are finally uploaded to one of the Copernicus Hubs to be available for download.

We have analyzed last two weeks of data ingestion times to see what the average times are.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1547895575889-Statistics.jpgLeft image shows Sentinel-2 L1C, which usually takes between 2 and 9 hours. To get atmospherically corrected L2A data (right), it adds another 3-4 hours on average.

20440c99-e00d-11e7-a98f-06b2d989fe84%2F1547849513285-S1GRD.png

Sentinel-1 GRD start coming in faster, even below one hour, with almost all of it being available after 9 hours.

In order for the data to be available in Sentinel Hub, one needs to add another 5 to 45 minutes or so, as long as it usually takes to download all new data from Copernicus Hubs to one of the clouds where we operate our service.

 

Coming Events

To meet us in person and discuss your needs, feel free to send us an email to [email protected] and schedule a meeting with us at the following events:

 

January 24-25 DataSpace 2019, Glasgow, Scotland
February 19-21 Big Data from Space 2019, Munich, Germany


 

 
 
Please do not hesitate to send us any feedback. You can also use our Forum or meet us in person at the upcoming events. We look forward to hearing from you.

The Sentinel Hub team