Combining AI technologies with satellite imagery to monitoring bark beetles

91% precision in detecting bark beetles to protect forests

Bark beetles are a serious threat to our forests and economy and manual assessments do not provide enough protection. By combining AI technologies with satellite imagery to support operational bark beetle monitoring in Austria. The goal is to detect beetles before they spread too far and to predict future beetle infestation – from space!

220113_BeatIT_heaederimage-color_Rectangle
76%

pixel-wise F1 score

95%

cluster-wise F1 score

91%

precision

100%

recall

The challenge

Protect forests from the beetle threat

Severe summer droughts lead to record levels of bark beetle-related damage. Just in 2019 in Central Europe, the estimated damage by bark beetles was 5.5 Mio cubic meters resulting in an economic loss of approximately 165 Mio € (Landwirtschaftskammer Österreich, 2019).

Currently, bark beetle monitoring in Austria is mainly done by calculating beetle populations from beetles in pheromone traps or by site visits by foresters at hot-spot areas. These methods lack information on full spatial coverage, are time-consuming and costly.

Foresters need, on the one hand, rapid spatial information on bark beetle occurrences to avoid the expansion to nearby trees. On the other hand, models that highlight risk areas and can predict future outbreaks.

220117_BeatIT_webGIS_animation

The idea

Rapid AI Prototyping to minimize processing time

For this project, our self-developed AI implementation method named “Rapid AI Prototyping” has been used, which helps to minimize coding and processing time and allows to create a new prototype or release version at any point.

The goal was to go beyond state-of-the-art in bark beetle monitoring. By combining AI technologies with satellite imagery, together with our partners FFG & Joanneum Research, the aim was to help foresters to assess the damages created by beetle infestations.

The solution

91% precision in bark beetle detection

The result is a prototype of a near real-time bark beetle monitoring system that can identify and predict bark beetle infestations and high-risk areas fast and reliable.

Our best prototype achieved a pixel-wise F1 score of 76% in detecting infestations on four hectares. The cluster-wise F1 score was even higher, reaching an incredible 95% (Recall = 100%, Precision = 91%). This means that the prototype can detect 100% infested areas correctly.

The findings of this project will support future research activities in the AI-based detection of bark beetle infestation and foster the understanding of bark beetle infestation distribution in general. In this context, in future studies, it will also help predicting short-term bark beetle infestation and respective actions for risk mitigation.

Our partners

FFG_Logo_DE_RGB_1000px
FFG

The Austrian Research Promotion Agency (Österreichische Forschungsförderungsgesellschaft) promotes innovation through applied and industrial research with the aim to strengthen the business location in Austria through targeted programs, such as the Austrian Space Applications Programme. ASAP aims on one hand to build national and international networks through multi- and bilateral projects and on the other hand to increase user communities of space technology.

joanneum
Joanneum Research

Joanneum Research analyzes remote sensing data in order to monitor the environment, develop algorithms for image analysis, process chains for data pre-processing, and generate 3D information from stereo image data.

umwelt data
Umweltdata GmbH (UWD)

Umweltdata GmbH (UMD): Umweltdata is a leading Austrian forest service provider. UWD collaborates in Beat It! by introducing and promoting the derived Products and services to their extensive business, administration and research network in the forestry domain, gathered over its more than 30 years of experience in forest remote sensing and inventory.

Reach out to us – we can help!

Interested in pushing your projects with AI?

Loading HubSpot form...

Other references