CONUS forest carbon under climate x harvest x disturbance: a yield-curve factorial (2025-2125)

Projections of conterminous United States (CONUS) forest aboveground carbon from 2025 to 2125 under a full factorial of climate (RCP4.5, RCP8.5), harvest (five removal intensities), and natural disturbance, on an FIA-anchored yield-curve engine run per state and summed to CONUS on the OSC Cardinal cluster. Disturbance uses a regrowth-aware recursion validated against Monte Carlo; the disturbance-timing variance band is about 0.1 percent at CONUS scale. Version 1.3.0 replaces the v1.1 LCMS state-ratio arm with a true per-cell LCMS disturbance arm. LCMS disturbance rates were derived per forest-type-group x EPA Level III ecoregion cell from the CBM/libcbm LCMS event files (per-cell disturbed area) divided by a 30 m zonal forest-area denominator from the aligned FIA forest-type and EPA ecoregion grids, then crosswalked to the yield-curve cell keys (97 percent of cells matched) and anchored to the CONUS LCMS rate. Files: lcms_percell_rate.csv (per prov|ft LCMS annual rate), per_cell_area.csv (30 m zonal forest area), lcms_arm_compare.csv (CONUS comparison of the disturbance arms). Disturbance carbon effect, CONUS No_harvest 2125 (penalty vs no-disturbance, Tg C): TreeMap source 5,386; LCMS per-cell 6,651; pool-consistent (CBM live-vs-dead matched) 612. The aboveground source bracket is 5,386 (TreeMap) to 6,651 (LCMS); the pool-consistent central estimate (612) keeps CONUS a sink. The robust, method-independent result is the direction (disturbance lowers live carbon, reserve carries the most exposed live carbon); the absolute sink-to-source flip is the aggressive aboveground end and is not the central result. All released values are state- or CONUS-level aggregates; no FIA true plot coordinates are included.