钱江源-百山祖国家公园候选区钱江源园区森林蓄积量遥感估算

Remote Sensing-Based Estimation of Forest Stock Volume in the Qianjiangyuan Section of the Qianjiangyuan-Baishanzu National Park Candidate Area, Zhejiang Province of southeastern China

  • 摘要: 采用激光雷达结合地面调查的方法, 估算了钱江源-百山祖国家公园候选区钱江源园区森林蓄积量, 旨在为国家公园生态系统原真性与完整性评估及保护成效评估提供方法和案例。研究采用冠层高度和垂直结构异质性等指数量化了亚热带常绿阔叶林的冠层结构, 并结合地面样地蓄积量数据和地形因子, 建立蓄积量的回归模型, 进而估算了整个钱江源园区的蓄积量。结果表明, 冠层平均高度与蓄积量相关性最高, 1 m×1 m网格(TCH1)和2 m×2 m网格(TCH2)冠层平均高度、凹凸度及海拔高度进入了最优模型(R2=0.76, rRMSE=27.0%), 表明机载LiDAR能够准确获取复杂地形区的森林冠层结构信息, 进而有效估算亚热带常绿阔叶林区的森林蓄积量。钱江源园区森林单位面积的蓄积量主要分布在20.55—227.71 m3/hm2(0.05—0.95分位数), 平均128.50 m3/hm2, 总蓄积量约3.88×106m3。长期受严格保护的古田山、齐溪片区蓄积量显著高于保护起步较晚、以次生林和人工林为主的长虹、何田片区, 反映出保护历史与干扰强度对蓄积量空间格局的影响。该方法为国家公园森林蓄积量动态监测与保护成效评估提供了案例。

     

    Abstract: Accurately estimating forest volume is essential for assessing ecosystem productivity, carbon sequestration capacity, and habitat quality. However, in subtropical evergreen broadleaved forests—characterized by complex topography and high species diversity—traditional field-based inventories and passive remote sensing techniques often fail to adequately capture vertical structural heterogeneity. Active light detection and ranging (LiDAR) presents a promising alternative, as it penetrates the canopy to deliver detailed three-dimensional insights. Yet, the integration of multi-scale canopy metrics with terrain-derived variables remains underexplored in these rugged landscapes. In this study, we integrated airborne LiDAR data with intensive ground-based surveys to estimate forest volume across the Qianjiangyuan Section of the Qianjiangyuan-Baishanzu National Park Candidate Area. Our objective was to develop an operationally feasible approach and provide a case study to support evaluations of ecosystem integrity and conservation effectiveness. Using a stratified random design, we established 276 permanent plots (20 m×20 m) across four sub-regions that differ markedly in their protection history: from the long-term strictly protected Gutianshan and Qixi areas to the more recently managed Changhong and Hetian areas—where secondary forests and plantations predominate. Within each plot, all trees with a diameter at breast height (DBH)≥5 cm were measured, and plot-level volume was computed using species-specific allometric equations. Concurrently, airborne LiDAR—acquired during the full-leaf phenophase data—were collected at an average point density of 4 points/m2. From the normalized point cloud, we derived canopy structure indices at multiple spatial grains: mean canopy height at 1 m (TCH1) and 2 m (TCH2) resolutions, as well as vertical structure complexity. Topographic variables—including elevation, slope, aspect, and wetness index—were also extracted from a 20-m resolution digital elevation model. To select predictors and evaluate model performance, stepwise multiple linear regression with crossvalidation was employed. Canopy mean height exhibited the strongest correlation with fieldmeasured forest volume (R2=0.61 for TCH1). The optimal predictive model retained TCH1, TCH2, convexity, and elevation, yielding an adjusted R2 of 0.76 and a relative root-mean-square error (RMSE) of 27.0%. Notably, integrating both fine (1 m) and coarser (2 m) height metrics led to better predictions compared to singlescale models, indicating that multiscale structural information more effectively captures dominant crown tops and subcanopy layering. When applied across the entire park, this model estimated a total forest volume of approximately 3.88×106 m3, with perhectare values ranging from 20.55 to 227.71 m3/hm2 (5th—95th percentiles) and a mean of 128.50 m3/hm2. Spatially, the Gutianshan and Qixi areas showed significantly higher mean volumes (156.3 and 144.7 m3/hm2, respectively) than the Changhong and Hetian areas (98.2 and 104.5 m3/hm2), reflecting a clear legacy effect linked to historical disturbances intensity and the duration of protection. These findings demonstrate that a linear regression model integrating multiscale canopy height metrics and topographic variables can reliably estimate forest volume in structurally complex subtropical evergreen broadleaved forests. Furthermore, airborne LiDAR proves effective in characterizing vertical forest structure, a key determinant strongly correlated with aboveground wood stock. The proposed approach is operationally feasible for periodic monitoring, requiring only repeated LiDAR acquisitions and a modest network of calibration plots. The pronounced volumetric gradient between long-protected and recently restored zones provides quantitative evidence of conservation effectiveness, underscoring the importance of sustained protection and targeted restoration interventions in degraded subunits. This study offers a methodological template and a benchmark for forest stock assessment and conservation performance evaluation within China′s national park system, with strong potential for transferability to other mountainous protected areas sharing similar ecological characteristics.

     

/

返回文章
返回