Validation of Forest Vegetation Simulator Model Finds Overprediction of Carbon Growth in California

Abstract

Using regression-based, bootstrapped equivalence tests, and remeasured inventory plot data from thousands of plots across California, we found that the Forest Vegetation Simulator (FVS), as typically used out-of-the-box, overpredicts carbon sequestration in live trees that remain alive ten years later by 27%, on average. We found FVS growth prediction sensitive to forest type and FVS variant, with the largest overpredictions occurring in stands within the North Coast variant, growing on the lowest site class, having ages that are unknown or between 50 and 100 years, and that are within governmentally designated reserved areas or on national forests. Direction and magnitude of errors are related to the stand attributes; these relationships point the way towards opportunities to improve the underlying growth models or calibrate the system to improve prediction accuracy. Our findings suggest that forest managers relying on out-of-the-box FVS growth models to forecast carbon sequestration implications of their management of California forests will obtain estimates that overstate the carbon that can be sequestered under light-touch or caretaker management, potentially leading to management decisions that fail to deliver the expected carbon sequestration benefits—a failure that could take a long time to recognize.