Evaluating the value of raster cell-level information in multi-objective forest planning

Abstract In remote sensing-based forest inventories, forest information is typically predicted at the grid cell level. For operational use this information is aggregated into larger management units. The aggregation simplifies the planning process leading to a loss of fine-scale spatial information and potentially reducing decision quality. Using individual cells as management units could preserve quality of the information but remains impractical for implementation. To capture the benefits of cell-level data while maintaining operational feasibility, careful grouping of cells may be useful. We used a value of information (VoI) analysis to assess the potential of appropriately grouping cells in the planning process. We compared the use of cell-level data grouped into alternative management units (segments) without data aggregation against stands with aggregated data. The VoI was assessed using a multi-objective optimization model incorporating economic and ecological objectives, using varied targets provided by a decision-maker. We generated a wide range of possible segmentation options, allowing the model to select the optimal configuration for each specific problem. We also tested the use of conventional stand delineation with preserving the cell information. The results show that using cells as the basic units for simulation and optimization with either segment or stand boundaries consistently improved the quality of the decisions, when comparing to the stands with aggregated data. The results highlight that alternative segmentation generated the largest improvement to management outcomes. However, simply using the conventional stands boundaries with cell-level simulations provided meaningful improvement. This approach enables decision-makers to evaluate segmentation approaches, and vary the approach based on the management objectives.