木材密度的全球分布與驅動因子及其對森林碳儲量的影響The global distribution and drivers of wood density and their impact on forest carbon stocks

Mo, Crowther, Maynard, van den Hoogen, Ma, Bialic-Murphy, Liang, de-Miguel, Nabuurs, Reich, Phillips et al.｜Nature Ecology & Evolution 8(12): 2195-2212｜DOI: 10.1038/s41559-024-02564-9

狀態：AI_DRAFT_FROM_REVIEW｜分級：A｜閱讀深度：FULL_TEXT_CHECKED｜Jacky 審核：False

森林數位孿生底層致能

讀學習筆記 →

wood densityforest carbon stockaboveground biomassforest inventory plotsrandom forestglobal mappingglobal

專討核心文獻定位

[40] Ch2 · 地面量測新增

Mo et al. · 2024

整合110萬個森林樣區與10,703樹種木材密度，建立全球木材密度分布圖，顯示忽略空間變異會使生物群系碳儲量估算最多差21%

使用警示

本頁是文獻知識庫卡片，不等於可直接引用的最終查核稿。只有狀態升級為 CITABLE 後，才可直接進入論文引用候選。

為什麼納入這篇

This paper anchors the ground-measurement chapter by showing that wood density, a core parameter feeding allometric biomass equations, varies systematically across space and that ignoring this variation introduces large errors into forest carbon stock estimates.

結構式摘要｜中英文對照

研究問題	全球森林木材密度如何在空間上分布、由哪些環境因子驅動，而這種空間變異又會如何改變森林碳儲量的估算？ How is wood density distributed across global forests, which environmental factors drive that variation, and how does the spatial variation change estimates of forest carbon stocks?
資料來源	原文確認：整合來自全球森林生物多樣性倡議（GFBi）資料庫約110萬個地面森林樣區，與10,703個樹種的木材密度資料；族群層級木材密度（CWD）以各樹個體的底面積加權計算。 The text confirms the integration of about 1.1 million ground-sourced forest inventory plots from the Global Forest Biodiversity Initiative (GFBi) database with wood density data for 10,703 tree species. Community-wide mean wood density (CWD) was computed by basal-area weighting of each observed individual.
方法	原文方法確認：使用62個氣候、地形、土壤、植被與人類活動全球圖層建立隨機森林模型，採200次空間自助抽樣（樣點間距至少50公里）以降低空間自相關，十折交叉驗證全球平均R2為0.53；再以偏迴歸與隨機森林評估九個驅動變數的相對重要性；最後結合木材密度圖、活立木材積、根質量比與生物量擴展因子估算全球森林生物量。 The text confirms random forest models built with 62 global layers of climate, topography, soil, vegetation and human activity, using 200 spatially bootstrapped subsets with points at least 50 km apart to reduce spatial autocorrelation; the final 62-predictor model reached a global average R2 of 0.53 under tenfold cross-validation. Partial regression and random forest assessed the relative importance of nine driver variables. Global biomass was then estimated by combining the wood density map with live tree volume, root mass fraction and biomass expansion factors.
主要結果	原文確認：木材密度呈明顯緯度梯度，熱帶森林木材密度比寒帶最高可達高30%；裸子植物平均0.47±0.07 g cm-3，被子植物平均0.59±0.14 g cm-3。年均溫是最具影響力的驅動因子，溫度每升高1°C木材密度平均增加0.5%；土壤水分為次要主因。納入木材密度空間變異後，相較於使用固定密度模型，部分生物群系碳儲量被低估，熱帶潮濕、熱帶乾燥、熱帶莽原與地中海型森林分別平均低估12%、17%、17%與21%，溫帶針葉與寒帶森林則分別高估10%與13%。全球總樹木生物量估計為374 GtC。 The text confirms a pronounced latitudinal gradient, with tropical forest wood up to 30% denser than boreal forest wood; gymnosperms averaged 0.47±0.07 g cm-3 and angiosperms 0.59±0.14 g cm-3. Mean annual temperature was the most influential driver, with a 1°C increase associated with an average 0.5% increase in wood density; soil moisture was the next primary control. Accounting for spatial variation in wood density changed carbon stock estimates relative to a constant-density model: tropical moist, tropical dry, tropical savanna and Mediterranean forests were underestimated by 12%, 17%, 17% and 21% on average, while temperate coniferous and boreal forests were overestimated by 10% and 13%. Total tree biomass was estimated at 374 GtC.
限制	原文確認：火頻為九個變數中影響最弱者，部分因全球96%森林過去20年未經歷火災，火災長期影響可能被低估；多數模型外推離群點位於非洲莽原（取樣密度較低）；模型全球平均R2為0.53，仍有相當未解釋變異。 The text confirms that fire frequency was the least impactful of the nine variables, partly because 96% of global forests had no fire in the past 20 years, so long-term fire impacts may be underestimated; most extrapolation outliers were in African savanna regions with lower sampling density; and the global average model R2 of 0.53 leaves substantial unexplained variation.

Key Findings

發現	證據	確定性
Wood density follows a clear latitudinal gradient, with tropical forest wood up to 30% denser than boreal forest wood, driven primarily by mean annual temperature and soil moisture.	Original abstract and Results: tropical wood up to 30% denser than boreal; a 1°C temperature rise correlates with ~0.5% higher wood density; soil moisture is a primary hydrothermal control.	checked_against_original_txt
Ignoring spatial variation in wood density introduces biome-level carbon stock errors of up to about 21%.	Original Results: constant-density model underestimates tropical moist/dry/savanna and Mediterranean forests by 12%, 17%, 17% and 21%, and overestimates temperate coniferous and boreal forests by 10% and 13%.	checked_against_original_txt

Key Figures and Tables

公開網站原則：未確認授權前，不直接複製原文圖表；優先使用自製圖表導讀或重繪圖。

項目	內容	關鍵數字	Jacky 判讀	重用策略
Fig. 1, Fig. 4, Fig. 5	Fig. 1 shows observed wood density across plots, gymnosperms vs angiosperms, forest types and biomes; Fig. 4 ranks the nine driver variables; Fig. 5 shows the effect of spatially explicit wood density on biomass and carbon stock estimates by biome.	1.1 million inventory plots; 10,703 tree species; gymnosperm mean 0.47 g cm-3, angiosperm mean 0.59 g cm-3; global model R2 = 0.53; total tree biomass 374 GtC; biome carbon stock differences up to 21%.	Use as evidence that a ground-derived parameter like wood density carries large, spatially structured uncertainty that any biomass or FDT carbon layer must absorb rather than treat as a constant.	Figures are CC BY 4.0 and may be reused with attribution; prefer a self-drawn simplified version focused on the 21% carbon error message.

Extracted Evidence Table

可支撐主張	指標或結果	原文位置	可引用	備註
Spatially explicit wood density materially changes forest carbon stock estimates.	Biome-level carbon stock differences of up to 21% (e.g., Mediterranean forests underestimated by ~21% under a constant-density assumption).	Original text, section 'Wood density and global biomass estimates', p.2199; Fig. 5.	True	Cite as evidence for wood-density-driven uncertainty in ground/allometric carbon estimation; the 374 GtC total is a global estimate, not a Taiwan-specific figure.

Critical Appraisal

Strengths

Exceptionally large empirical base (1.1 million plots, 10,703 species).
Directly quantifies how a single ground parameter propagates into carbon stock error.
Open access CC BY 4.0 with public data on Zenodo.

Weaknesses

Global average model R2 of 0.53 leaves substantial unexplained variation.
Lower sampling density in some regions (e.g., African savanna) yields extrapolation outliers.
Fire impact likely underestimated due to short observation window.

Validation quality	strong for a global statistical model; tenfold cross-validation and spatial bootstrapping reported
Transferability to Taiwan	medium; the principle that wood density variation drives carbon error transfers, but Taiwan-specific densities require local inventory data
Risk of overclaiming	Do not present the 374 GtC global biomass or the 21% biome error as Taiwan figures; they are global/biome-level results.

與 Jacky 博論 / Review 的用途

博士論文	Supports the dissertation argument that ground-measured parameters such as wood density are a key, spatially structured source of uncertainty that an FDT carbon layer must model explicitly.
TJFS Review	Strengthens the TJFS review's ground-measurement chapter by quantifying how allometric inputs propagate into carbon stock error.
可引用句候選	2024 年，Mo 等人發表的文獻中指出，將木材密度的空間變異納入考量後，生物群系層級的森林碳儲量估算最多會出現約 21% 的差異，凸顯地面參數對碳估算精度的關鍵性。
不可用來主張	Do not use this paper as a Taiwan-specific carbon stock figure or as evidence about LiDAR or satellite retrieval accuracy.

授權與圖表重用

Article license	CC-BY-4.0
Figure reuse policy	REUSE_ALLOWED_WITH_ATTRIBUTION_CC_BY_4.0
Notes	Open Access article licensed under Creative Commons Attribution 4.0 International (CC BY 4.0); figures may be reused with appropriate credit per the article's licence statement.

待查核清單

Optionally extract Supplementary Table 2/3 biome-level CWD values into a structured appendix.
Prepare a self-drawn figure highlighting the up-to-21% carbon stock error message.
Cross-check Taiwan-relevant wood density references for the local angle.