Assessing forest vegetation and fire simulation model performance after the Cold Springs wildfire, Washington USA

Given that resource managers rely on computer simulation models when it is difficult or expensive to obtain vital information directly, it is important to evaluate how well a particular model satisfies applications for which it is designed. The Forest Vegetation Simulator (FVS) is used widely for forest management in the US, and its scope and complexity continue to increase. This paper focuses on the accuracy of estimates made by the Fire and Fuels Extension (FFE-FVS) predictions through comparisons between model outputs and measured post-fire conditions for the Cold Springs wildfire and on the sensitivity of model outputs to weather, disease, and fuel inputs. For each set of projected, pre-fire stand conditions, a fire was simulated that approximated the actual conditions of the Cold Springs wildfire as recorded by local portable weather stations. We also simulated a fire using model default values. From the simulated post-fire conditions, values of tree mortality and fuel loads were obtained for comparison to post-fire, observed values. We designed eight scenarios to evaluate how model output changed with varying input values for three parameter sets of interest: fire weather, disease, and fuels. All of the tested model outputs displayed some sensitivity to alternative model inputs. Our results indicate that tree mortality and fuels were most sensitive to whether actual or default weather was used and least sensitive to whether or not disease data were included as model inputs. The performance of FFE-FVS for estimating total surface fuels was better for the scenarios using actual weather data than for the scenarios using default weather data. It was rare that the model could predict fine fuels or litter. Our results suggest that using site-specific information over model default values could significantly improve the accuracy of simulated values.

► The accuracy of FFE-FVS predictions was evaluated across suites of model outputs. ► The sensitivity of FVS-FFE model was examined for weather, disease, and fuel inputs. ► Multiple model outputs were evaluated simultaneously. ► Tree mortality and fuels estimates were most sensitive to fire weather inputs. ► Total surface fuel estimates were better for scenarios with actual weather data.