文献类型：期刊文献

中文题名：结合Sentinel-2和机载激光雷达数据的南亚热带山地森林树种多样性估测

英文题名：Estimating tree species diversity in a south subtropical mountainous forest using Sentinel-2 and airborne laser scanning data

作者：刘晴[1,2] 李世明[1,2] 张浩芫[1,2] 齐志勇[3] 张译[1,2] 张虎[1,2]

第一作者：刘晴

机构：[1]中国林业科学研究院资源信息研究所,北京100091;[2]中国林业国家林业和草原局林业遥感与信息技术重点实验室,北京100091;[3]北京大学地球与空间科学学院遥感与地理信息系统研究所,北京100871

年份：2026

卷号：46

期号：3

起止页码：1289-1300

中文期刊名：生态学报

外文期刊名：Acta Ecologica Sinica

收录：;北大核心:【北大核心2023】；

基金：国家重点研发计划(2023YFB3905801);中央级公益性科研院所基本科研业务费专项资金项目(CAFYBB2022ZB00105)。

语种：中文

中文关键词：树种多样性;Sentinel-2;机载激光雷达;光谱异质性;高度异质性;XGBoost

外文关键词：tree species diversity;Sentinel-2;airborne laser scanning;spectrum heterogeneity;height heterogeneity;XGBoost

分类号：S718.5

摘要：估测大面积树种多样性的空间分布对于生物多样性评估、森林资源的可持续管理和保护至关重要。当前结合植被结构信息和气候特征以量化亚热带森林树种多样性的相关研究较少。以云南省普洱市的南亚热带山地森林为研究对象,利用Sentinel-2影像和机载激光雷达(ALS)数据,根据光谱异质性和高度异质性假说,通过XGBoost算法对树种多样性进行建模预测。结果表明:(1)干湿季组合提升了树种多样性制图精度,Shannon-Wiener多样性指数在湿季预测效果更好(R^(2)=0.606,RMSE=0.565),Simpson多样性指数则在干季表现更佳(R^(2)=0.578,RMSE=0.238);(2)光谱特征对物种丰富度预测有价值(R^(2)=0.738,RMSE=2.448),LiDAR特征在预测Shannon-Wiener多样性指数上更具优势(R^(2)=0.718,RMSE=0.199),二者结合提高了Simpson多样性指数的预测能力(R^(2)=0.801,RMSE=0.189);(3)红边波段、Rao′s Q、变异系数、纹理特征和冠层特征(CCm)在制图中具有较高的应用价值;(4)研究区内树种多样性及其空间分布是生态位分化、气候、地形与森林管理等因素的综合结果。研究为估测南亚热带山地森林的树种多样性提供了一个方法流程,有助于当地的森林经营管理和生物多样性保护。
Estimating spatial patterns of tree species diversity(TSD)across extensive regions is fundamental for biodiversity assessment as well as the sustainable management and conservation of forest resources.However,relatively few studies have quantified TSD in subtropical forests by integrating both vegetation structural data and climatic variables.Based on Spectral Variability Hypothesis(SVH)and Height Variation Hypothesis(HVH),we modeled and predicted TSD through Extreme Gradient Boosting(XGBoost)machine learning algorithm in subtropical montane forests in Pu'er,Yunnan,with Sentinel-2 imagery and aerial laser scanning(ALS)data.Our results showed that:(1)the combination of spectral features based on dry and wet seasons imagery improved the mapping accuracy of TSD.The Shannon-Wiener diversity index in the wet season outperformed that in the dry season(R^(2)=0.606,RMSE=0.565),while the Simpson diversity index behaved better in the dry season(R^(2)=0.578,RMSE=0.238).(2)Spectral features showed an advantage on predicting species richness(R^(2)=0.738,RMSE=2.448),and LiDAR features were beneficial for the Shannon-Wiener diversity index(R^(2)=0.718,RMSE=0.199).Combining both especially enhanced the prediction of the Simpson diversity index(R^(2)=0.801,RMSE=0.189).(3)The red-edge band,Rao's Q,coefficient of variation,texture features and canopy characteristic(CCm)were identified as highly valuable predictors for mapping TSD.(4)TSD and its spatial distribution in this area are the comprehensive result of niche differentiation,climatic conditions,topography,and forest management practices.Our findings provide a framework for assessing and monitoring TSD in subtropical montane forests and offer practical insights for local forest management and biodiversity conservation.