The forestry sector gets its own supercomputer

Last autumn, the first specially designed supercomputer for forestry research was launched. The Computational Forestry Lab (CFL) at Umeå University can handle enormous quantities of data and can perform complex calculations. It opens up entirely new opportunities for the development of AI solutions for the forestry industry.

The Computational Forestry Lab (CFL) supercomputer has been built to meet the rapidly growing demand for high-performance computing, and advanced AI, in forestry research. It will enable the processing of large volumes of data from airborne laser scanning, from satellites and from forestry machinery. It will also enable the production of synthetic data from advanced simulations, as well as the training of new AI models. The supercomputer is physically located at Umeå University’s Centre for High-Performance Computing, HPC2N. The Kempe Foundation and Mistra Digital Forest were responsible for the establishment of the facility.

- What makes CFL unique is the fact that it is tailor-made for digitalised forestry research. By pooling data and computing power, we can carry out analyses and develop models that would be impossible with our resources spread across different universities and companies, says Lucas Hedström, researcher at the Department of Physics at Umeå University.

Facilitating collaboration

The computing infrastructure is designed to suit both advanced and less experienced users. Initially, CFL will be tested by a small group of researchers before gradually being opened up to more users. Both researchers and industry partners are able to use the supercomputer to train models, for example, that identify tree species, assess timber properties and plan logging with a high degree of precision. It will also be easier to evaluate research results when new models and algorithms can be tested directly within the infrastructure.

– At CFL, models and algorithms can be evaluated against one another using a wide variety of datasets, without sharing business-critical information. One of CFL’s most important contributions is the facilitation of collaboration between researchers, companies and technology providers, says Lucas Hedström.

Providing a better basis for decision-making in forestry and in environmental work

Anneli Ågren, a researcher at SLU, uses CFL to develop high-quality maps based on laser data and AI. She is currently working on producing a new, more detailed soil type map for Sweden. Together with her colleague William Lidberg, among other things her work includes SLU's drainage ditch maps and the soil moisture map, both of which have become highly significant for the forestry industry.

– Something I have learnt over the years is that you cannot build a good AI model without robust and comprehensive training data. Pooling data and computing power at CFL enables us to produce higher-quality maps. These maps provide us with a better basis for decision-making in both forestry and environmental work, and they are important inputs in discussions about climate adaptation and the forest’s role of the forest as a carbon sink, Anneli Ågren concludes.