Towards nation-wide individual tree carbon sink and biodiversity mapping utilizing high performance computing

There exists an urgent need towards forest value chain optimization on societal level. Effective policy making is required to reduce effects of global climate change and or to improve yields in forestry. Unfortunately, these two ecosystem services are competing and even conflicting with each other. To succeed in their joint optimization, precise carbon intake and water balance estimates in different biomes are crucial and require new tools for the task.

Forest data are typically collected with forest inventories. These inventories provide the fundamental information for all decision-making in society and industry that are relevant to human interventions, including harvest planning. Nordic countries have long performed forest inventories on a national level to estimate the country-wide forest averages with area-based inventories (ABA) and at stand level relying on airborne laser scanning (ALS). Current ABA techniques are limited in their spatial resolution and can be improved by focusing on individual tree level. Individual tree level mapping allows to focus not only on the wood material volumes, but also to their quality and health. Computation of this information, especially over wide geographic areas, is a significant computational and data management challenge and requires high-performance computing.

Our goal is to develop the missing mapping technology and demonstrate this in Finland where we will automatically count and characterize all five billion dominant and co-dominant forest trees. We will do this by merging already existing laser scanning technologies on different scales, by developing novel methods as needed, and finally implementing them in the EuroHPC LUMI supercomputer. Each tree will be individually segmented and imputed with wood quality information. All trees and their parameters are collected into the “Metsäkanta” database for interactive mapping and Digital Twinning applications.

We will further enhance the database by computing the CO₂ sink potential for each detected tree. The CO₂ sink potential is modelled from dense spatiotemporal in-situ laser scanning references collected from individual trees. The results are then imputed to all trees. The outcome of these efforts will be a Digital Forest Replicate (DFR) at individual tree level. The DFR combines the information of individual tree wood quality, growth potential, and near real-time carbon sink reporting. This allows improved country-level carbon stock estimates.