Customer stories
The war on bark beetles: protecting forests using machine learning & satellite imagery
Industry:
Forestry
Business type:
Forestry monitoring and protection
Overview
Did you know that one of Earth’s biggest environmental threats is smaller than a grain of rice?
Bark beetles are devouring the world’s spruce forests at an alarming rate. Climate change is accelerating the problem, calling for new tactics to defeat the pest. Read on to learn how Waldstolz combines forestry expertise with Planet Skysat imagery and Picterra’s geospatial machine learning platform to defend spruce trees from widespread outbreaks.
Achievements
The bark beetle problem you never knew we had
If a bark beetle infestation takes root, forests that have been built up over generations can disappear in a matter of months. What’s worse: overlooking just one infected spruce tree can lead to a widespread infestation in just a few weeks. And the bark beetle problem is escalating quickly across the world. Due to rising temperatures and dry summers, the pest has better conditions and longer seasons to proliferate. They’re now multiplying an estimated 2 to 3 times faster within a season than before.
Because they tend to prey on distressed or weak trees first i.e., those that have been damaged by wildfire, droughts, or storms, Europe’s particularly dry summers have created a thriving environment for bark beetles. But make no mistake: as populations grow, they are also known to overtake perfectly healthy trees, which is why early detection is critical.
The impact of spruce mortality
To stop an infestation, you cannot simply cut out the affected area. Each infected tree needs to be completely removed from the forest. Losing spruce forests at this scale has several implications, for landowners and citizens of the world alike:
Combatting bark beetles from behind a desk
But there is light at the end of the tunnel. German-founded Waldstolz has launched a war on bark beetles using state-of-the-art technology. They realized that if forest owners could detect infestations before they spread, they could prevent major spruce loss and salvage timber value.
Waldstolz’ forest monitoring solution uses a combination of satellite imagery and geospatial machine learning to spot early outbreaks from behind their desks. Waldstolz’ clients no longer need to send hundreds of rangers into the field every two weeks to look for sawdust on the ground (an early sign that the millimeters-sized bark beetles have made themselves at home in a tree). Not only is this process impractical and laborious, it is also prone to human error since walking difficult terrain makes it easy to overlook potentially infected trees.
The technology is built on detecting infestation through satellite imagery. When trees are infected, they turn pink, making them possible to spot from above (much easier than scanning the ground for sawdust, right?). Working with 50 cm resolution imagery from Planet’s Skysat satellites, they regularly survey their clients’ parcels during peak bark beetle season (April to August). Using Picterra’s machine learning platform, Waldstolz has built detectors that recognize inklings of infestations in the images that would be impossible for the human eye to spot.
We used to manually review images of clients’ parcels to find the pink areas that indicated an infestation. With machine learning, we only have to train the model once on how infestation appears in satellite images. Then we can run it on all imagery time and time again. In fact, the algorithm gets even more accurate the more examples we feed it! It enables our small team to serve more clients and in turn, help our clients protect more woodland.
When Waldstolz’ Picterra detectors find traces of an infestation, they alert their client’s forest managers through email and SMS. The GPS-marked areas of interest are exported to the clients’ operations system so a crew can be sent out to inspect the trees. Alternatively, they refer the work to a forest monitoring company to carry out the work.
Partnering with Planet & Picterra to form a geospatial defense
Waldstolz has been working with Planet data since the start. After running several tests, they determined that Planet’s Skysat 50cm resolution imagery was optimal for their use case for a number of reasons:
- The very high spatial resolution makes it possible to detect infestations on a single-tree basis.
- Planet’s flexible tasking allows them to stay responsive to clients’ needs and task satellites for specific requirements.
- The possibility of edge parameters of the recording, such as the position of the sun, provides ideal conditions to start the image analysis.
However, Waldstolz still needed a way to increase their efficiency in analyzing the bark beetle nests. Their intuition told them that artificial intelligence could potentially be a solution to their problem. But they did not have the resources or time to build up internal data science capabilities; they needed a solution fast. That’s when they found Picterra’s geospatial machine learning platform. A team of foresters and geomaticians with some data science knowledge, their team was drawn to the fact that they could use Picterra to benefit from the advantages of neural networks without needing to become experts themselves. With Picterra’s no-code UI, they could get started immediately.
The quantitative impact of preventing bark beetle outbreaks
Waldstolz has monitored over 5,000 hectares of German spruce forest to date. Using satellite imagery and machine learning detectors, they’ve prevented 109 bark beetle outbreaks. Assuming an annual spread of two generations, they have salvaged over 43,000 trees or 74 hectares of spruce forest. If these trees had died, Earth would have suffered 400 tons of extra carbon dioxide in the atmosphere and 3,170 tons more air pollutants.
They have saved their clients roughly €200,000 in projected losses due to wood quality degradation.
With the capabilities of Plane’s satellites, regions around the world can leverage their services to obtain useful imagery to combat infestations. For example, more than 5 million hectares were affected by bark beetles in the western US between 1997 and 2010. We can’t wait to enable companies with geospatial machine learning to help save the world’s spruce forests!
