Back
Jul
09
2018

Latvian team takes top 5 position in global space competition - with an idea born from a hackthon

  • Image provided by:

Most hackathon teams dissolve after the 48-hour sprint. Not in the case of SpaceTree, which went on to take 5th place at global space competition ActInSpace, with over 500 teams from around the world. 

The SpaceTree team has returned from the ActInSpace, a global space competition, where they competed with their idea of forest monitoring via satellite, which was initially developed in the Garage48 space hackathon. They went on to participate in the ActInSpace competition, which is hosted by the French Space Agency. Of 500 participants from 5 continents, the Latvian team ended up in the top 5. 

What is the solution? It uses satellite data to monitor forests and determine their characteristics, such as age, height, species, crown size, and more. The current prototype can only determine species, but the next prototype will be able to determine the rest. Founder Nauris Durbe emphasizes that the team has also been developing a solution for data flow, which ensures data reception from the satellites, which is an achievement that is on par with determining forest species. "The process isn't simple, it requires specific algorithms. Afterwards the data is harvested and then "pasted" together." says Nauris.

The idea was born when founder Nauris Durbe, data engineer at the Latvian telecommunications company Lattelecom, was invited to participate in a creative workshop organized by "Latvia's State Forests" to find solutions to industry problems with the help of technology. There Nauris presented his team's idea, which both he and Latvia's State Forests representatives liked. 

Not long afterwards, the hackathon organization Garage48 was hosting a space-themed hackathon in Riga. There Nauris met Gundega Dekena, who teamed up with him to work on SpaceTree during the hackathon, along with other participants. The idea won the hackathon, and though there were more team members, only three decided to remain in the team afterwards - Nauris, Gundega, and Karls Tõnissoo.

After the successful hackathon, the team decided to apply for the next challenge - the ActInSpace competition. They chose to apply for the regional rounds in both Latvian and Estonia, because only the Estonian regional round had mentors from the European Cosmos Agency. "To everyone's surprise, we won both. The organizers had no other choice but to create a joint Latvian-Estonian team to proceed to the final in France. Our team was joined by others, who were to pitch our idea. Our presenter had the heaviest load in France. All of the semi-finalists were called into a room, where there were five envelopes containing the name of the top 5 finalists in a random order. Each would have to then pitch their ideas. Our name was in the fifth envelope, yet we didn't win the competition." says Nauris. 

 

SpaceTree uses Copernicus satellites, which provide data free of charge. Other satellite images may be better quality, however they require payment. "The satellite divides the earth into pixels. For our needs, one pixel represents 10x10 meters. We try to tell our machine learning model that one pixel contains trees of certain species, age, or height. Latvia's State Forest helped us quite a bit, because they provided us with information about what can be found in certain pixes. The machine learning model remembers that. Afterwards we show it pixels that we don't have information for, to test if it learned it correctly. Now we've gotten so far in the development of the project that we can determine what's in just about any pixel." 

"European Union institutions have invested billions to launch these satellites. As a result, they want to prove that this money was not wasted For that reason there are many grants available that support new product development that use the satellite data. We've written an application for the Copernicus Masters program, which specifically supports startups. So our work will continue." - Nauris Durbe, cofounder of SpaceTree.