[目的]基于广州市增城区县域尺度的森林资源二类调查数据,构建基于空间网格化和贝叶斯优化的随机森林(Bayesian optimization-random forest,BO-RF)算法的森林碳储量估算模型,分析主要优势树种(组)的森林生物量碳储量及其空间分布特征,为增城区绿美生态建设提供参考依据.[方法]借助ArcGIS Pro 3.0建立网格,通过蓄积量-生物量转换模型和生物量-碳储量转换模型,计算网格的森林生物量碳储量,应用反距离权重插值完成估算碳储量的空间统计,结合BO-RF算法完成建模与估算评价,采用空间自相关分析和聚集热点分析,探明森林生物量碳储量的空间分布特征.[结果]利用BO-RF算法估计的划分为一亩网格(边长 25.8198 m的 1 亩正方形网格)的森林生物量碳储量的精度很好,决定系数(R2=0.984),碳储量整体性估算精度达 96.52%.森林碳储量最高的为桉树林,其次为其他软阔、阔叶混交林、其他硬阔、针阔混交林和马尾松,最低的树种(组)为杉木和针叶混交林.增城区森林生物量碳储量存在极显著的空间正相关性和高值聚集现象,碳储量高值区分布较少,以大封门林场、派潭镇中部和西南部、正果镇东南部等区域为主;碳储量低值区分布广泛,以小楼镇、中新镇等中南部区域为主.[结论]针对广州市增城区的县域尺度进行网格化,基于BO-RF算法的森林生物量碳储量估算,具有良好的精度和效率,可为县域碳储量空间分布的研究提供新的思路和方法;广州市增城区森林生物量碳储量具有明显的空间异质性和聚集特征,呈现北部和东部少部分区域高、中部和南部大部分区域都比较低的特征,未来需要调整优化北部、中部和南部等不同区域的经营措施,以精准提升森林质量.
Estimation of Forest Biomass Carbon Storage and Spatial Distribution Characteristics at the County Scale in Zengcheng District,Guangzhou
[Objective]Based on the second-class forest resource survey data at the county scale for Zengcheng District,Guangzhou,this study aimed to establish a forest biomass carbon stock estimation model using a spatial grid method and BO-RF(Bayesian optimization-random forest)algorithm.The model was used to estimate the forest biomass carbon stock of major dominant tree species(groups)and analyzed its spatial distribution characteristics,providing a reference for the eco-friendly development of Zengcheng's green and beautiful ecological infrastructure.[Methods]A grid was established using ArcGIS Pro 3.0,and forest biomass carbon stock was calculated using a volume-to-biomass conversion model and a biomass-to-carbon stock conversion model.Inverse distance weighted interpolation was applied for the spatial statistics of the estimated carbon stock.The BO-RF algorithm was used for modeling and evaluation,while spatial autocorrelation analysis and hotspot clustering analysis were employed to explore the spatial distribution characteristics of forest biomass carbon stock.[Results]The accuracy of the forest biomass carbon stock estimated using the BO-RF algorithm for the one mu grid(a one mu grid with a side length of 25.8198 m)is excellent,with a coefficient of determination(R2=0.984)and an overall estimation accuracy of 96.52%.Eucalyptus robusta had the highest carbon stock,followed by other soft broadleaved forest,broadleaved mixed forest,other hardwood forest,coniferous and broadleaved mixed forest,and Pinus massoniana.The lowest carbon stocks were found in Cunninghamia lanceolata and coniferous mixed forest.The biomass carbon stock in Zengcheng District exhibited significant spatial positive autocorrelation and clustering of high values.High carbon stock areas are scarcely distributed,mainly concentrated in regions such as Dafengmen Forest Farm,the central and southwestern parts of Paitan Town,and the southeastern part of Zhengguo Town;while low carbon stock areas are widely distributed,primarily in the central and southern regions such as Xiaolou Town and Zhongxin Town.[Conclusion]Gridding at the county scale in Zengcheng District,Guangzhou,and estimating forest biomass carbon storage based on the BO-RF algorithm demonstrate good accuracy and efficiency,offering a new approach for studying the spatial distribution of biomass carbon stocks at the county level.The biomass carbon stock in Zengcheng District shows clear spatial heterogeneity and clustering,with higher values in some northern and northeastern regions,and lower values in most of the central and southern regions.Future forest management strategies should be adjusted and optimized to enhance forest quality across northern,central,and southern regions.
forest carbon storageBO-RFspatial characteristicsspatial autocorrelationclusteringZengcheng District
万方数据
维普
