Quantifying the contribution of vegetation layer uncertainty to simulated forest fuels treatment placement

Abstract

Background Western US forests face rising high-severity fire risk motivating increased investment in forest management. The high cost of treatments and vast areas needing work require efficient resource allocation. Forest landscape models can help scale up analysis and support management, but underlying uncertainties can limit their utility. Aims We examined uncertainty in the initial forest input layer, quantifying the error and determining how much field data are needed to produce outputs suitable for informing management. Methods We created eight initial forest layers using various sample sizes and selection methods, ran 50-year simulations and compared outputs. We also resampled model outputs from 100 to 500 m resolution and evaluated the effects. Key results Stratified sampling stabilized errors and allowed using smaller sample sizes relative to random sampling, and coarser resolution resampling reduced errors while retaining key information. Combining these strategies produced similar outputs using all and only 25% of the data. Conclusions Combining stratified sampling with output coarsening reduces data demands while maintaining analytical rigor, enabling more actionable guidance for management planning. Implications This work advances the operational use of landscape models as a decision-support tool, allowing managers to scale up planning by integrating model outputs with expert knowledge and reducing reliance on labor-intensive field inventories.