Bayesian recalibration of Forest Vegetation Simulator variants: parameter posteriors and multi-source validation (2026)

Calibrated parameter sets and validation outputs for a modernized, community-maintained fork of the USDA Forest Vegetation Simulator (FVS). Component models (diameter growth, mortality, crown ratio, height-diameter, height growth) were recalibrated against national FIA remeasurement data using Bayesian hierarchical models with species, ecoregion (EPA Level 1 to 3), and forest-type-group random effects, yielding posterior-median multiplier sets per variant. This deposit contains: (1) the calibrated configuration JSON files for the FVS regional variants (per-species diameter-growth, mortality, crown, and height-diameter multipliers); (2) validation outputs against independent Canadian National Forest Inventory (MAGPlot) remeasurement data for the Acadian (New Brunswick) and Alaska (coastal British Columbia) variants, including protocol-consistent stand remeasurement pairs and a North American CEC Level III ecoregion stratification; and (3) the cross-variant default-vs-calibrated bias comparison (variants by species by ecoregion by landowner). Acadian basal-area growth is essentially unbiased on Canadian maritime forest (R^2 = 0.88); the Alaska variant under-predicts coastal-BC growth (basal-area increment bias about -74 percent in its analog ecoregion) and tracks dry interior stands well, motivating an ecoregion-dependent growth correction.