Welcome to visit Scientia Silvae Sinicae，Today is

Scientia Silvae Sinicae ›› 2025, Vol. 61 ›› Issue (9): 22-38.doi: 10.11707/j.1001-7488.LYKX20240472

• Research papers • Previous Articles    

Estimation of Forest Aboveground Biomass Based on Non-Wall-to-Wall Remote Sensing Auxiliary Data

Liuyuan Huang1,2,Yan Zheng1,2,*(),Xinjie Cheng1,2,Weisheng Zeng3,Qing Xu4,5,Zhengyang Hou1,2   

  1. 1. Key Laboratory for Silviculture and Conservation of Ministry of Education, Beijing Forestry University　Beijing 100083
    2. Ecological Observation and Research Station of Heilongjiang Sanjiang Plain Wetlands, National Forestry and Grassland Administration　Shuangyashan 518000
    3. Academy of Forestry Inventory and Planning, National Forestry and Grassland Administration　Beijing 100714
    4. Institute of Resources and Environment, International Centre for Bamboo and Rattan　Beijing 100102
    5. Key Laboratory of National Forestry and Grassland Administration/Beijing for Bamboo & Rattan Science and Technology　Beijing 100102
  • Received:2024-08-06 Online:2025-09-25 Published:2025-10-10
  • Contact: Yan Zheng E-mail:houzhengyang@bjfu.edu.cn

RichHTML

2

PDF (PC)

28

Abstract:

Objective: Aiming at the problems of small sample size and non-wall-to-wall remote sensing auxiliary data when generating annual estimates in forest inventory (such as forest aboveground biomass, AGB), based on the hybrid estimation method and Monte Carlo simulation method, the actual inference ability of the hybrid estimation method to solve this problem and the response law of the method to the sample size were explored, the inventory sample size was optimized, and the relationship between the two types of statistical inference methods (based on wall-to-wall remote sensing auxiliary data and based on non-wall-to-wall remote sensing auxiliary data) was clarified to provide a methodological theoretical reference for generating annual estimates in forest inventory. Method: The hybrid estimation theory was used to break through the bottleneck of improving sampling accuracy under the conditions of small sample and non-wall-to-wall remote sensing auxiliary data. Based on the complete data of the 9thnational forest inventory in Genhe Forestry Bureau, different sample sizes of plots m1 and remote sensing auxiliary data sample sizes m2 were designed to simulate the small sample size and non-wall-to-wall remote sensing auxiliary data. The conventional model-based and design-based statistical inference methods were introduced for comparison to explore the inference efficiency of the hybrid estimator and its response to the sample size. At the same time, based on the variance analysis of the Hybrid estimator, the mathematical relationship between the Hybrid estimator and the CMB estimator is revealed from the variation rules of the variance components and the formula connection. Result: 1） When the sampling intensity of the Hybrid estimator is 100%, the total variance is equal to the model variance, which is consistent with the inference results of the CMB estimator. 2） When sample plot $ S_1 $ is complete, the variance calculated by the model-based method is only about 51.95% of the variance based on the design method, and the model-based inference accuracy is improved by 2.04% compared with the design-based inference accuracy. 3） When the Hybrid estimator infers the estimates of AGB in the study area, if the sample size m1 remains the same, the inference accuracy of the Hybrid estimator is basically stable after m2 is about 12 800 (0.127%), and the inference accuracy has an inverted J-shaped nonlinear relationship with the sample size of the sample plot and auxiliary data. Considering the survey cost and the stability of the results, the best cost-effective sample size for inferring the estimates of AGB is about $ {m}_{1}=40 \left(41\%\right) $ and $ m_2=12\ 800\left(0.127\%\right) $. 4） The Hybrid estimator is based on about 1/5 sample plots of the 9th National Forest Inventory of Genhe Forestry Bureau and 0.127% of the non-wall-to-wall free remote sensing auxiliary data samples, and has achieved an inference accuracy of nearly 90% (87.16%). Conclusion: 1） The CMB estimator can be regarded as a special case of the Hybrid estimator. In actual forest inventory, the method of inferring estimates can be adjusted according to accuracy requirements and cost constraints. 2） In the Hybrid estimator, the variance component derived from the model dominates the uncertainty of inference, which means that increasing the measured sample size, using high-precision auxiliary data (such as lidar data), and optimizing modeling methods will be powerful means to reduce its inference uncertainty. 3） Based on the inverted J-shaped nonlinear relationship between the Hybrid estimator inference accuracy and the sample size of sample plots and auxiliary data, the cost-effective sample size for Genhe Forestry Bureau to infer the annual forest resource estimates is determined to be $ {m}_{1}=40 \left(41\%\right) $, $ m_2=12\ 800\left(0.127\%\right) $. Compared with traditional design-based and model-based statistical inference, it has significant improvements in accuracy and cost, and can better meet the needs of county's annual forest management inventory. 4） For the national forest annual monitoring system, the Hybrid estimator reaches nearly 90% (87.16%) of the inference accuracy in the small-scale area (Genhe Forestry Bureau), based on about 1/5 of the field samples and 0.127% of non-wall-to-wall remote sensing auxiliary data. The inference accuracy can be further improved by improving the quality of auxiliary data and optimizing modeling methods. It has the potential to generating annual estimates innational forest inventory based on annual field plots.

Key words: hybrid estimator, statistical inference, non-wall-to-wall remote sensing auxiliary data, forest inventory, forest aboveground biomass (AGB)

CLC Number: