From point clouds to forest management: Quantifying the sensitivity of a decision support framework to initialization data using close-range remote sensing

Forest management is confronted with deep uncertainties related to trajectories of future forest development, as climate change induces critical transitions in forest ecosystems. Decision support systems (DSSs) that combine climate-sensitive forest modeling with assessments of biodiversity and forest ecosystem services (BES) have the potential to systematically reduce uncertainties regarding the consequences of various climate and management pathways. However, in order to assess the reliability of DSS outputs, systematic analyses of sources of uncertainty across individual DSS components are crucial. This applies in particular to the initialization of DSSs, which remains a key challenge due to constrained data availability from traditional sources such as forest management plans and forest inventories, and thus may constitute a key source of uncertainty within DSSs. In particular, advances in close-range remote sensing, such as high-resolution LiDAR, provide detailed information on the current state and condition of forests and offer new opportunities for DSS initialization. However, the extent to which initialization with high-resolution LiDAR inventory affects DSS outputs and contributes to uncertainty remains unexplored. Therefore, this study aims to quantify the sensitivity of a DSS framework to initialization with LiDAR-based forest inventory data.

Our approach involved (1) terrestrial and airborne laser scanning (TLS/ULS) sampling, (2) initialization of the forest gap model ForClim, (3) simulation under alternative management and climate change trajectories, and (4) evaluation regarding BES. The combined ULS and TLS inventory served as reference data, from which sampling variants with different sample sizes were generated to represent varying levels of forest inventory detail. The DSS sensitivity to initial stand resolution was assessed over a 70-year simulation period under three management intensities, three climate change scenarios, and 15 stand-specific indicators, which were further aggregated into partial utilities for biodiversity and ecosystem services.

Our results revealed that low sample sizes of inventory data resulted in higher deviations from the reference simulation. This effect decreased with progressing simulation time and higher management intensity for most BES indicators. While sample size was the primary source of uncertainty in the early stages of the simulation, climate-related uncertainty increased over time. Our findings establish a 20-40 year tactical window where high-resolution initialization is the primary determinant of DSS reliability, after which climate uncertainty becomes the dominant constraint for strategic planning. Further research should aim to leverage the full potential of high-resolution LiDAR data for DSSs by extracting additional information on forest composition and state. This would enable more informed decision support for long-term forest planning under deep uncertainty and the demand for BES provision.