基于LandTrendr的老挝橡胶林遥感识别及扩张监测  

Remote Sensing Identification and Expansion Monitoring of Rubber Forests in Laos Based on LandTrendr

作　　者：琚旺龙 詹正豪 秦震宇 张军[2] JU Wanglong;ZHAN Zhenghao;QIN Zhenyu;ZHANG Jun(Institute of International Rivers and Eco-security,Yunnan University,Kunming,Yunnan 650500,China;School of Earth Sci-ences,Yunnan University,Kunming,Yunnan 650500,China)

机构地区：[1]云南大学国际河流与生态安全研究院,云南昆明650500 [2]云南大学地球科学学院,云南昆明650500

出　　处：《热带作物学报》2024年第8期1742-1750,共9页Chinese Journal of Tropical Crops

基　　金：国家国防科技工业局高分专项云南省政府综合治理深度应用与规模化产业化示范项目(No.89-Y50G31-9001-22/23);云南大学研究生创新人才培养项目——研究生课程教材建设质量提升计划(No.HXKC202112)。

摘　　要：天然橡胶作为世界上重要的工业原料,也是老挝农民收入的重要组成部分,受替代种植政策和橡胶价格变化的影响,准确快速地监测当地人工橡胶林对于促进橡胶产业的良性发展具有重要意义。目前大部分研究都集中在小区域的橡胶林提取,多为单时相或双时相影像的分类提取,无法得到长时间序列的橡胶林扩张特征。为探究长时间、大尺度的橡胶林种植扩张特征,本研究选取1990—2020年连续30年的Landsat时间序列影像,利用LandTrendr算法对老挝30年间橡胶林变化进行提取,并借助国际橡胶期货价格和地形数据进行时空变化分析。结果表明:经过优选后的特征组合可以更好地进行橡胶林分布提取,2020年橡胶林分布提取的总体精度为89.85%,Kappa系数为0.82,得出老挝在2020年的橡胶林总面积为32.1万hm^(2);使用LandTrendr算法和二次分类的橡胶林扩张变化监测总体精度为92.44%,Kappa系数为0.82;老挝人工橡胶林的扩张和橡胶期货市场密切相关,通过地形因素分析,得出橡胶林适宜种植在低海拔、坡度平缓且向阳的区域。本研究能够很好地在国家尺度上对橡胶林变化进行长时间的监测,为橡胶产业的发展和政策的把控提供数据支撑。Natural rubber,a crucial industrial raw material worldwide,and a significant part of the income for Laotian farmers,is influenced by alternative planting policies and fluctuations in rubber prices.Aaccurate and rapid monitoring of local artificial rubber plantation is of crucial significance for promoting the sustainable development of the rubber industry.Most research has focused on the extraction of rubber plantation information in small areas,typically utilizing single or dual-temporal image classification,which fails to capture the long-term expansion characteristics of rubber forests.To investigate the long-term,large-scale expansion features of rubber plantation,this study selected a continu-ous thirty-year Landsat time series imagery from 1990 to 2020.The LandTrendr algorithm was employed to extract changes in rubber plantation across Laos over the three decades.International rubber futures prices and terrain data were also utilized for spatiotemporal change analysis.The selected feature combination,after optimization,can better extract rubber forest distribution.The overall accuracy of rubber forest distribution extraction in 2020 was 89.85%,with a Kappa coefficient of 0.82,revealing Laos’rubber plantation area to be 321000 hectares in 2020.The overall accuracy of rubber forest expansion change monitoring using the LandTrendr algorithm and secondary classification was 92.44%,with a Kappa coefficient of 0.82.The expansion of artificial rubber forests in Laos was closely related to the rubber futures market.Through terrain factor analysis,it is deduced that rubber forests are suitable for planting in low-altitude,gently sloping,and sunny areas.This study effectively monitors changes in rubber forests at the national level over an extended period,providing valuable data support for the development of the rubber industry and policy control.

关 键 词：老挝 人工橡胶林 LandTrendr算法 LANDSAT 变化检测 

分 类 号：S31[农业科学—作物栽培与耕作技术]