摘要
【目的】在Google Earth Engine(GEE)云平台上借助其强大的计算和数据存储能力,融合多源遥感数据对森林林龄进行遥感反演和动态监测。【方法】融合2017—2023年间Sentinel-1、Sentinel-2及高程数据,通过随机森林(Random forest,RF)分类获取土地覆盖信息,并进一步提取森林的分布和面积,同时构建时间序列植被指数来准确提取森林变化区域。基于森林资源清查数据和融合的多源遥感数据,在GEE上构建RF回归、分类回归树(Cart)以及梯度提升回归树(Gradient tree boost,GTB)3种回归模型,用于杉木组、马尾松组、毛竹林、硬阔叶树组以及其他类树种组的2018年林龄遥感反演,并估算出2017年和2023年的林龄信息,以揭示林龄和龄组在2017—2023年的动态变化情况。【结果】1)2017—2023年,研究区森林面积的整体变化总计113.93 km^(2),此间森林的减少和更新并存,其空间分布特征呈现出明显的区域差异。具体而言,森林面积变化多发生于靠近城区和低海拔地区,且靠近城区的森林面积减少往往不再恢复至森林;2)在5种不同树种组构建的3种模型中,RF回归模型的林龄反演结果最佳,平均R^(2)为0.845,平均RMSE为5.32 a,其中毛竹林反演精度最高,R^(2)为0.863,RMSE为2.411 a;3)2017—2023年,研究区林龄在40 a以下的森林由54.59%减少至51.06%,其中龄组变化最显著为杉木组成熟林,面积增加了38.88%。【结论】在GEE上融合多源遥感数据进行林龄反演和动态监测具有重要的应用潜力,本研究结果可为使用云平台及哨兵系列卫星数据对森林资源长时间序列的林龄反演和动态监测的应用提供参考和借鉴。
【Objective】To fuse multi-source remote sensing data for inversion and dynamic monitoring of forest stand age on the Google Earth Engine (GEE) cloud platform with the help of its powerful computing and data storage capabilities.【Method】The land cover information was obtained through the fusion of Sentinel-1,Sentinel-2,and elevation data from 2017 to 2023 using the random forest (RF) classification method.Additionally,the distribution and area of forests were further extracted.Meanwhile,time series vegetation indices were constructed to accurately identify areas of forest change.Based on forest inventory data and the fusion of multi-source remote sensing data,three regression models were developed on GEE:RF,classification and regression trees (CART),and Gradient tree boosting (GTB).These models were applied to estimate the forest age in 2018 for different tree species groups,such as Chinese fir,masson pine,moso bamboo,hardwood,and other tree species groups.The estimated forest ages in 2017 and 2023 were also obtained to reveal the dynamic changes in forest age and age groups from 2017 to 2023.【Result】1) From 2017 to 2023,the overall change in forest area in the study area amounted to 113.93 km2.During this period,forest reduction and regeneration coexisted,and the spatial distribution exhibited distinct regional variations.Specifically,forest area changes were more prominent in areas near urban centers and lower altitudes,with forest area reductions near urban areas often not recovering to a forested state;2) Among the three models constructed for five different tree species groups,the RF regression model produced the best results for forest age estimation.It achieved an average R~2 of 0.845 and an average RMSE of 5.32 a,with bamboo forests exhibiting the highest accuracy (R~2=0.863,RMSE=2.411 a);3) From 2017 to 2023,forests ages below 40 years in the study area decreased from 54.59% to 51.06%.The most significant age group change was observed in mature pine forests,with an increase of 38.88% in their a
作者
陈馨
孙玉军
丁志丹
CHEN Xin;SUN Yujun;DING Zhidan(State Forestry&Grassland Administration Key Laboratory of Forest Resources&Environmental Management,Beijing Forestry University,Beijing 100083,China)
出处
《中南林业科技大学学报》
CAS
CSCD
北大核心
2024年第6期19-29,共11页
Journal of Central South University of Forestry & Technology
基金
国家自然科学基金项目(31870620)
林业科学技术推广项目([2019]06)
中央高校基本科研业务费专项(PTYX202307)。
