湿地土壤全氮和全磷含量高光谱模型研究

Abstract

PDF HTML阅读 XML下载 导出引用 引用提醒 湿地土壤全氮和全磷含量高光谱模型研究 DOI: 10.5846/stxb201501230186 作者: 作者单位: 辽宁师范大学海洋经济与可持续发展研究中心,辽宁师范大学海洋经济与可持续发展研究中心 作者简介: 通讯作者: 中图分类号: 基金项目: 国家自然科学基金项目（41271421）；教育部人文社会科学研究规划基金项目（14YJA630064） Estimating the total nitrogen and total phosphorus content of wetland soils using hyperspectral models Author: Affiliation: Center for Studies of Marine Economy and Sustainable Development,Liaoning Normal University,Center for Studies of Marine Economy and Sustainable Development,Liaoning Normal University Fund Project: 摘要 | 图/表 | 访问统计 | 参考文献 | 相似文献 | 引证文献 | 资源附件 | 文章评论 摘要:氮磷是湿地生态系统土壤中的重要营养元素，其对湿地植被生长、湿地生态系统生产力、区域富营养化变化、湿地环境生态净化功能等具有重要的影响作用。研究氮磷营养物质在湿地土壤中的分布变化特征，对湿地生态系统评估、恢复和管理具有重要的意义。以中国高纬度地区面积最大的滨海芦苇湿地——盘锦湿地为研究区，采用不同建模方法（再抽样多元逐步回归模型bootstrap SMLR和再抽样偏最小二乘回归模型bootstrap PLSR）和光谱变换技术（包络线去除CR、光谱一阶微分FD和光谱倒数的对数LR），分别建立了湿地土壤全氮和全磷含量的估算模型。基于湿地土壤实测光谱，模拟高光谱Hyperion数据和多光谱TM数据，在此基础上进行湿地土壤营养元素含量估算。对比所建反演模型的估算精度，探讨高光谱遥感技术对湿地土壤营养元素组分的估算能力和适用性。研究结果表明：bootstrap PLSR相比于bootstrap SMLR建模方法，其对研究区湿地土壤全氮和全磷含量的估算获得了较高精度；对盘锦湿地土壤全氮含量的估算，最高估算精度产生于CR光谱变换技术结合bootstrap PLSR建模；对湿地土壤全磷含量的估算，最高估算精度产生于原光谱数据结合bootstrap PLSR建模；模拟高光谱数据Hyperion对湿地土壤全氮和全磷含量的估算精度均高于模拟多光谱数据TM，模拟Hyperion的估算精度更接近于实测光谱的估算精度。 Abstract:Nitrogen and phosphorus in wetland soils are important limiting nutrients for plant growth, maximum photosynthetic rate and capacity, and net primary productivity. They have been found to be significantly involved in the estuarine eutrophication and environmental purification of wetland ecosystem. The research may be focused on distributing and changing characteristics of nitrogen and phosphorus in wetland soils. It is important for the evaluation, restoration, and management of wetland ecosystems. Our study area is located in the Panjin wetland (40°45'-41°10'N, 121°45'-122°00'E), which is a part of the Shuangtaihekou National Nature Reserve Administration. Panjin wetland is the largest coastal reed wetland situated in the high-latitude areas of China. The laboratory measurements of total nitrogen, total phosphorus, and spectral reflectance for surface soil samples had been conducted. Different modeling methods, such as bootstrap stepwise multiple linear regression (SMLR), bootstrap partial least square regression (PLSR), and spectral transformation techniques, such as continuum removal (CR), first difference derivative (FD), and log transformed spectra (LR), were used to develop the estimation models of total nitrogen and total phosphorous in wetland soils. Based on the simulated hyperspectral Hyperion data and multispectral Thematic Mapper (TM) data of the wetland soils, soil nitrogen and phosphorous contents were estimated, respectively. Subsequently, the estimated accuracies of the developed models were compared, and thus, the ability and suitability of estimating nitrogen and phosphorous components in wetland soils using hyperspectral technologies were explored. The results indicated that bootstrap PLSR achieved higher accuracies of estimating the total nitrogen and total phosphorous content of wetland soils in the study area than did bootstrap SMLR. The spectral transformed technique of CR used in combination with the modeling method of bootstrap PLSR yielded the highest estimation accuracy for the prediction of the total nitrogen content of soils collected from Panjin wetland. The original spectral data combined with bootstrap PLSR produced the highest estimation accuracy to predict the total phosphorous content in wetland soils. Simulated hyperspectral Hyperion data attained higher accuracies of estimating total nitrogen and total phosphorous in wetland soils compared to simulated multispectral TM data. The estimation accuracies of the simulated Hyperion were closer to those of the measured spectra. The estimation accuracy of the total nitrogen content achieved from the measured spectra, simulated hyperspectral Hyperion, and multispectral TM were all higher than those of the total phosphorous content of the same soils. 参考文献 相似文献 引证文献