Megan Stevens, a physics student from Southampton University, has been analysing pasture conditions in Kyrgyzstan using satellite data as part of her summer internship with eOsphere. The SIBELIUs team believe that the Kyrgyz Data Cube will be useful for many applications and understanding pasture conditions will be very beneficial for supporting the herding community, who are facing many challenges. The Kyrgyz Data Cube contains current and historical satellite data and derived environmental products covering the whole country, with a resolution of up to 10 metres.
Megan’s first task was to create an updated karagana map of the Suusamyr Valley. The Suusamyr Valley provides important summer pasture for many livestock herders and karagana is a bush with spiky leaves that cannot be eaten by livestock. Over recent years, herders have reported that karagana is spreading and reducing the pasture available for their livestock. The aim of Megan’s work was not necessarily to produce a perfect new product, but instead it was more to demonstrate what can be achieved using satellite data by following some simple logical steps.
Megan’s new karagana map of the Suusamyr Valley is based on Sentinel-2 data acquired in 2020 and shows where the karagana has increased since the last map was made, which was based on data from 2014. Overall, the two maps show a net increase of karagana from 6.3% of the area of the Suusamyr Valley to 7.1% between 2014 and 2020. However, it is evident from studying the two maps that different levels of precision were applied in their creation, so these figures are not directly comparable.
To create the karagana map, Megan first downloaded red, green, blue (RGB) satellite data from July 2020 from the Data Cube in a GeoTiff format so that it could be imported to the QGIS system. QGIS is a free, open-source geographic information system that supports viewing, editing, and analysis of geospatial data. The RGB data provides an image similar to a visual photograph with 10 metre pixels. Using the previous karagana map, it was possible to see roughly what karagana looks like in the satellite image. In general, it has a darker green colour compared to the surrounding pasture. However, the darker green areas were also similar to regions that had previously been identified as forest. Megan assumed that the previous forest regions were correct and that the forests hadn’t changed much since 2014.
Megan imported the 2020 RGB Sentinel-2 satellite data into QGIS, together with the previous karagana map. Next, she used the pencil icon in QGIS to edit a new Shapefile in the Add Polygon feature based on the latest satellite data. A key aim was to establish a consistent level of detail in the new map, so that all significantly sized regions of karagana were identified, but without requiring every single pixel to be classified, which would have taken a very long time. Megan explains:
“I determined Karagana mainly by the colour. It was very dark and was often easily distinguished from the lighter green (though still dark) of forest shrubbery surrounding it. It was more difficult to determine whether or not something was Karagana when there were dark green patches surrounded by brown, as it was difficult to see clearly how dark a green it was when not directly next to other shades of green.
It can be quite clearly seen that Karagana grows along rivers so if there were dark green patches without any visible surrounding water it was more likely that it was not actually Karagana. There were many dark patches that looked like Karagana that were actually the shadows of clouds. To rule these out I either looked for the cloud that would have made the shadow, as these were white patches nearby of the same shape, or I looked at a raster image collected from a different month to see if it was still there or not, as moving clouds won’t show up in different months’ images.”
Once the Karagana map was complete, Megan exported the new Shapefile and calculated the percentage of the Suusamyr valley that was covered in Karagana and how this had changed since 2014. Using the approach described here, Megan spent about 25 hours creating the new karagana map. Prior to this summer, Megan had no previous experience working with satellite data or with GIS systems, which shows that these techniques should be accessible to a wide range of people with an interest in pasture and agriculture.