Black snow in Prokopyevsk

The snow cover is not always white. Sometimes it is orange, sometimes it is black. The images below were captured by Sentinel-2 in Prokopyevsk, Russia.

According to the Siberian Times the deposition of black dust on snow in February 2019 was due to the failure of a filtering system in a coal processing plant. Looking at the picture below, I have the feeling that this kind of event was not exceptional in Prokopyevsk this winter...

Pictures from Kemerovo region by Orlovprklife, Willravilov, Typical Kemerovo

Source: Siberian Times 15 Feb 2019. Pictures from Kemerovo region by Orlovprklife, Willravilov, Typical Kemerovo

Thanks to François Tuzet for pointing this to me!

Near-real time analysis of the 2018-2019 snow season in the Pyrenees and the Alps from satellite data

Here in southwest France ski lovers did not really enjoy the beginning of the snow season... But how does it compare to the previous years? Using Sentinel-2 and Landsat-8 data, we computed the snow cover duration since September 01 until January 20 for the past three snow seasons in the Alps and Pyrenees.

Snow cover duration (in days) from 01 September of year N-1 to 20 January of year N

Continue reading

Two billion pixels to check your next ski destination

More exactly: two maps of 934'343'100 pixels!

We [1] have processed 6205 Sentinel-2 images and 593 Landsat-8 images to compute the annual snow cover duration in the Alps and the Pyrenees at 20 m resolution for hydrological years 2016-2017 and 2017-2018. The snow cover duration (or snow persistence) is defined as the total number of days with snow on the ground over a hydrological year (from 01 September to 31 August). We also added the ski runs from the great OpenSnowMap project.


Continue reading

The Khumbu Icefall by Venµs

This is a time-lapse of all clear-sky images captured by Venµs over the Khumbu Icefall near Mount Everest since November 2017 (one year of data, 51 images at 5 m resolution).

Khumbu Icefall by Venµs

Here I used the Level-1C products (i.e. without atmospheric correction) because the Level-2A products are provided at a lower resolution (10 m). Anyway, the atmosphere is rather thin in this area..

To make this animation (without the date annotation to simplify):
1) download
python ./ -l 'Nepal' -c VENUS -a config_theia.cfg -d 2017-11-01 -f 2018-12-01 --level 'LEVEL1C'
2) unzip
mkdir -p ../VENUS
parallel unzip -d ../VENUS ::: $(find . -name "VENUS*zip")

3) export as natural color pictures
cd ../VENUS
mkdir -p VIS
parallel gdal_translate -srcwin 4118 3132 1058 770 -of JPEG -b 7 -b 4 -b 3 -scale 0 800 0 255 -ot byte {} VIS/{/.}.jpg ::: $(find . -name VE*[0-9].DBL.TIF)

4) animate with imagemagick
convert -delay 10 VIS/*jpg anim.gif

Snow cover duration from 01 Sep 2017 to 31 Aug 2018 in the Alps and Pyrenees


We are thinking to distribute a new product that would provide the snow cover duration in the Alps and the Pyrenees over a hydrological year (snow persistence map). Here is a preview of a "beta" version that is only 100 m resolution for now.

Snow cover duration from 01 Sep 2017 to 31 Aug 2018. Pixels with less than 60 days were masked out.

A few words on the method (see also here): this map is generated after a linear interpolation at the daily time step of every available Theia snow product. Here we have selected 26 Sentinel-2 tiles, which represents 3189 snow maps for the study period, that is 96 billion pixels to process.
With the accumulation of Sentinel-2 data it becomes possible to look at interannual variability. Here is for example a visual comparison of the snow cover duration calculated from January to May for three years in the Pyrenees (31TCH):

 Snow cover duration 31TCH

Snow Duration from Jan 01 to May 31 in the Pyrenees (tile 31TCH)

We can see longer snow durations for the year 2018, which was an exceptional year!

Durée d'enneigement dans les Alpes et les Pyrénées du 01/09/2017 au 31/08/2018


Nous sommes en train de penser à un nouveau produit qui donnerait la durée d'enneigement dans les Alpes et les Pyrénées au cours de la dernière année hydrologique. Voici ci-dessous un aperçu d'une version "beta" (artisanale) qui n'est qu'à 100 m de résolution pour l'instant.

Durée d'enneigement du 01/09/2017 au 31/08/2018.

Durée d'enneigement du 01/09/2017 au 31/08/2018. Les pixels avec moins de 60 jours ont été masqués.

Quelques mots sur la méthode (voir aussi ici) : cette carte est générée après une interpolation linéaire au pas de temps journalier de toutes les cartes d'enneigement Theia disponibles. Ici nous avons sélectionné 26 tuiles Sentinel-2, ce qui représente 3189 cartes d'enneigement pour la période considérée, soit 96 milliards de pixels à traiter.
Avec l'accumulation des données Sentinel-2 il devient possible de comparer les années entre elles. Voici par exemple une comparaison visuelle de la durée d'enneigement calculée de janvier à mai pour trois années dans les Pyrénées (31TCH):

Durée d'enneigement 31TCH

Durée d'enneigement du 01 janvier au 31 mai sur la tuile 31TCH

On voit ressortir des durées d'enneigement élevées pour l'année 2018 qui fut une année exceptionnelle !
Un grand merci à Germain Salgues de Magellium pour avoir produit toutes les données rapidement ce qui nous a permis de montrer cette carte au séminaire Theia à Montpellier dès le 18 octobre ! Et merci à Michel Le Page pour la mise en ligne.

Sentinel-2 captured a jökulhlaup in Afghanistan

In the Landslide blog Dave Petley has analyzed Planet images of the Pashgor debris flow in Afghanistan (here and here). Here I used two Sentinel-2 images (before and after the event) to show the path of the debris flow from the high mountain area to the Panjshir Valley. Sentinel-2 images have a lower spatial resolution than Planet images but they have a larger swath and the near-infrared channel is useful to highlight the water-rich surfaces (dark blue) and the vegetation (red). Also, Sentinel-2 images are free to use for everyone.

According to the experts this event can be called a jökulhlaup since it was due to the abrupt collapse of a supraglacial lake, i.e. a lake formed on the surface of a glacier, in this case a debris-covered glacier. The debris flow (a mix of water and debris) has traveled 13 km from the source to the deposit area where it has dammed the Panjshir river.