Betting on cosmic rays

At some point, a light bulb went on in someone’s brain: hydrogen atoms on the ground are mainly found in water so, what if we measured the “returning” neutrons to derive information about the presence of water in the soil? The wetter the soil, the less energy would return to Space....

High-resolution satellite images have proven to be indispensable in completing the picture and providing support for a hydrological model of the region.
The Eurac Research team then processed, or in the technical jargon of remote sensing, conducted “image classification” of the satellite images to create increasingly accurate maps.
“Starting with data from the Sentinel 2 satellite , we developed an algorithm that would classify the image by providing a snow cover fraction map. This kind of map shows the percentage of an area covered by snow for each pixel in the image,” explain the Eurac Research representatives Ludovica De Gregorio and Giovanni Cuozzo. “The algorithm we used is based on artificial intelligence techniques, and by ‘training’ the model we improved upon Copernicus’s snow cover fraction estimate ̶ the reference map made by the European Union's Earth observation program.”

_{Left: the original image sent by Sentinel 2. This is an optical image, albeit with distorted colors, of the Tyrol and South Tyrol area. An optical image means that it was captured by optical sensors, with a technology similar to that of cameras although obviously more sophisticated. The other big family of satellites that are used to capture images are radar sensors. Centre: the image processed by the Eurac Research team. In yellow the values of snow cover fraction ̶ the areas covered by snow. Right: The map developed by the European Union’s Copernicus Earth observation program, taken as a reference. The comparison shows how the Copernicus product underestimates the shaded areas – the darker parts of the first image and reports them as snow-free. The algorithm developed by Eurac Research was able to detect the presence of snow even in shaded areas.}

Very high-resolution images, detailing objects up to two meters apart, were then used to check the validity of the generated map and were also processed and used as a comparison for assessing the actual snow cover. The ultra-high-resolution images are from the WorldView satellite.

_{High-resolution satellite image, with classified version on the right. The image was acquired on November 27, 2020 in the test area around the Dresdner Hut near the Stubai Glacier in Tyrol.}

The research team used additional satellite images as an aid for estimating the extent of moisture on the ground, in particular they chose Sentinel-1 microwave and Landsat-8 optical images which had also been classified with artificial intelligence-based algorithms.

_{Map of soil moisture on September 18, 2021. The color blue shows soil moisture. Red lines mark North Tyrol, South Tyrol, and East Tyrol. It can be seen that in the westernmost valleys and Lower Adige, except for the Isarco/Eisack river line, light shades of blue which correspond to drier areas, predominate.}

“The research world and even government departments are increasingly investing in these hybrid forms of monitoring and on cosmic ray probes to better monitor the natural resources of the land,” De Gregorio and Cuozzo explain. “In addition, our maps, due to their ability to display high-resolution topographic features of the terrain will also be useful in choosing the ideal locations to place sensors.”