パラグアイ・チャコ地域の区画別森林減少ハザード率：大規模牧場における年間伐採ダイナミクスのマルチスケール実証分析

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Provides a methodological framework for reconciling parcel-level and regional deforestation rates, useful for carbon accounting research.

🏢実務担当者:Highlights risks in REDD+ credit baselines, informing due diligence for carbon credit procurement and investment.

🏛政策担当者:Informs policy on REDD+ baseline standards and carbon market integrity, relevant for Article 6 rulemaking.

Cattle ranch expansion across the Paraguayan Chaco generates parcel-level clearing rates that appear inconsistent with regional deforestation statistics, raising concerns that REDD+ baselines constructed from proxy parcels overstate forest loss (a baseline over-crediting debate that turns on denominator consistency between parcel hazard and regional flows). Because parcels undergoing active conversion are cleared within compressed development windows of one to five years, their annualised hazard rates necessarily exceed long-run regional averages that dilute clearing across the full estate of active and inactive properties. This paper develops a deterministic scaling identity that resolves the apparent inconsistency by formally linking parcel-scale hazard intensity to regional clearing magnitudes through the legally convertible forest fraction (g) and the share of parcels actively clearing in any given year (f). From 22 cadastral parcels across nine ranch sites in the Western Region (excluding the northern UNESCO Biosphere Reserve), annualised clearing rates derived under the Verra BL-PL module (VMD0006) from Landsat imagery over 2012 to 2020 yield a sample mean of 14.99% of total parcel area per year (not percent of regional forest; 95% bootstrap CI: 9.86 to 20.68; N = 22). The distribution is right-skewed (skewness = 1.05) with substantial heterogeneity (SD = 13.15; range 0.15 to 49.67). Discrete-time hazard analysis confirms front-loaded clearing: mean hazard in Year 1 is 0.28, declining to 0.05 by Year 3 (paired t = 3.04, p = 0.006). Inverse application of the identity recovers a median implied activity share f ≈ 0.12 (95% SI ≈ 0.07–0.19), aligned with a directly measured nine-year mean f ≈ 0.121 on an expanded 46-parcel set and with the committed-conversion mean of 15.78%/yr under a D% ≥ 20 filter. A secondary forward uncertainty exercise (sampling f from the scenario band) yields mean implied deforestation of order 238,000 ha/yr (58.4% of draws in the 160,000–300,000 ha/yr historical band) and is reported as prior-sensitive, not as independent confirmation. Regional clearing rates, forest-stock inputs used in scaling, and any imputed parcel Forest% are estimates that would benefit from more detailed primary analysis. The present parcel-level work uses Chaco Vivo–linked parcels and related project documentation; it does not claim a nationally representative or wall-to-wall sample of the Paraguayan Chaco. The proxy sample is small (n = 22) and non-random. In the primary N = 22 scaling exercises f is identified as a residual; on the expanded 46-parcel universe a year-wise activity fraction is measured directly (mean 0.121) and used as a consistency check, not as a wall-to-wall regional census. Larger, more detailed sampling and tightened forest-cover measurement are reserved for a planned expansion study. The 14.99%/yr figure is not re-estimated in this paper; it is the validated output of the VMD0006 BL-PL module as applied in the Chaco Vivo Project Description under VCS Standard v4.7. This paper takes that module-derived rate as given and asks whether it can be reconciled with independently observed multi-decadal clearing records across the Paraguayan Chaco. An independent expanded-sample validation comprising 24 additional cadastral parcels drawn from fifteen separate landholders corroborates the central estimate: committed-conversion parcels from the combined 46-parcel universe yield a mean rate of 15.78%/yr [11.13, 21.03], and the directly measured nine-year activity fraction of 0.121 falls at the centre of the scaling identity’s implied range. Parcel-level Forest% for the twenty-four expanded cadastral parcels is not in the proxy archive (centroids only). Those values are temporarily imputed and flagged as estimates: one parcel from an original-sample landholder mean where the centroid owner matches an original ranch, and twenty-three parcels from the original N = 22 pool mean of 75.35% (observed range 42.39–99.16%). They are not measured 2012 forest cover; any Exp forest-dependent result is sensitivity-only until GIS Forest% is supplied in the planned expansion study.

• EarthArXiv https://eartharxiv.org/repository/object/14786/download/25954/first seen 2026-09-05 04:40:52 · last seen 2026-09-12 04:34:12