遇见数据集

CONUS-HCS: a continental harvest decision-surface product (240 m annual harvest probability, intensity, and silvicultural type, 2024)

收藏
Zenodo2026-06-11 更新2026-06-12 收录
官方服务:

资源简介:

Continental harvest decision surfaces for the 48 conterminous United States at 240 m (EPSG:5070), produced by the Continental U.S. Harvest Classification System (CONUS-HCS), a four-stage Bayesian hierarchical harvest decision module fit on the full USDA FIA remeasured plot-pair record (452,887 pairs) and projected via TreeMap 2022 tile substitution. Layers (Cloud-Optimized GeoTIFFs, anchor year 2024): annual partial-harvest probability; annual stand-replacement probability; posterior standard-deviation layers; partial and clearcut harvest intensity (fraction of basal area removed); modal partial and clearcut silvicultural class with per-pixel confidence; a composite expected annual basal-area-removed fraction; and a modeling support-tier layer. What changed in v5.0.0 (coverage correction): v4 was silently empty over nine states — all of New England (Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, Rhode Island, Vermont) and the coastal Pacific Northwest (Oregon, Washington) — because the climatic-water-deficit covariate was missing there and failed the render's complete-case filter. The model always covered these states; only the rasters omitted them. The underlying TerraClimate PET/AET inputs were re-derived from the intact source netCDFs, the resulting deficit was calibrated to the model's training scale (holdout R-squared 0.959), and the full stack was re-rendered. Coverage is now complete across CONUS. There was no model refit: coefficients, method, color breaks, and all previously covered states are unchanged from v4. Interpretation note: the stand-replacement layer captures an FIA forest-condition transition over the remeasurement interval, not silvicultural clearcutting; it anti-correlates with Hansen Global Forest Change at the state scale (Spearman -0.53) and must not be benchmarked against TPO clearcut bands or satellite forest-loss products. Plot-level discrimination is modest (AUC ~0.59-0.61); the product is calibrated for area-aggregate annual projection (held-out ECE 0.0235). Designed for ingestion into forest projection systems (FVS, LANDIS-II, CBM-CFS3, the RPA framework) and the PERSEUS forest intelligence platform. Companion product: the 30 m CONUS standing-timber-value rasters (doi:10.5281/zenodo.20515241).

提供机构:
Zenodo
创建时间:
2026-06-11
二维码
社区交流群
二维码
科研交流群
商业服务