资源简介

摘要： 近年来可见光—近红外反射光谱已被广泛应用于估算土壤全氮含量,为大范围区域土壤全氮含量获取提供了一种快速、有效的方法。基于实验室测定的三江源区146个表层土壤(0~30 cm)样品的反射光谱数据(350~2 500 nm)与全氮含量数据;利用偏最小二乘回归(PLSR)和反向传播神经网络(BPNN)两种模型方法与光谱反射率(REF)及其4种数学预处理变换相结合,分别建立分土壤类型样本和总体样本全氮估算模型;评估利用可见光—近红外光谱技术预测三江源区土壤全氮含量的能力。结果表明:BPNN模型的R 2cal R2val及验证RPD的平均值分别为0.87、0.81与2.28;而PLSR模型则相应为0.75、0.72和1.95;表明BPNN模型预测能力整体上要优于PLSR模型。BPNN与光谱各种形式的结合均具有良好、或接近良好预测全氮的能力;而PLSR与REF、倒数对数(Log(1/R))及波段深度(BD)的结合仅少部分具有良好估算能力、大部分则为粗略估算能力,一阶微分(FDR)和二阶微分(SDR)估算精度均较低,尤其是SDR(R2<0.5,RPD=1.10~1.27)均不具备估算能力。总体样本所建模型稳定性好于分土壤类型,分土壤类型建模差异性明显;此外,总体来看,BPNN模型比PLSR建模精度高、模型稳定性好,但PLSR模型可操作性强于BPNN模型。 关键词： 土壤全氮;  可见光—  近红外反射光谱;  偏最小二乘回归(PLSR);  反向传播神经网络(BPNN);  三江源区;  玉树县;  玛多县     Abstract: Visible and Near-Infrared Reflectance Spectroscopy (VNIRS) has extensively been used to estimate soil total nitrogen (TN) concentration,and can provide a rapid,convenient method for quantitatively obtaining soil TN content in a wide range of areas.In this study,we evaluated the prediction ability of Visible and Near-Infrared Reflectance Spectroscopy (VNIRS) for estimating soil TN in the Sanjiang Yuan regions of Qinghai province.Firstly,we collected about 146 surface soil samples (0~30 cm),including four soil types during the period from August 7 to 17 of 2012 in Yushu and Maduo counties;secondly,we respectively measured soil reflectance spectrum by ASD FieldSpec 4 portable spectrometer (Analytical Spectral Devices,Inc.,Boulder Colorado,2012) with the spectral range of 350～2 500 nm,and soil TN by using Vario EL Ⅲ element analyzer of ELEMENTAR Inc.in the laboratory;and then we respectively adopted PLSR and BPNN models to relate soil TN to raw spectral reflectance and its four pre-processing transformations for the overall soil samples and each soil types samples.The results showed that the average coefficients of determination(R2) of calibration and validation for BPNN are respectively 0.87 and 0.81 with the mean RPDval of 2.28,whereas those of PLSR model are 0.75,0.72 and 1.95 respectively,which suggest that BPNN has a better prediction ability than PLSR as a whole;The combination of BPNN and the raw reflectance spectrum (REF) and its all pre-processing transformations performed a good or closer good prediction ability for different and overall soil types;whereas the combination of PLSR model and REF,Log(1/R),BD produced a rough or good prediction ability for estimating TN,however,FDR and SDR with poor prediction ability,especially SDR (R2 cal<0.5,R2val<0.5,RPDval=1.10~1.27) hasnt the ability to predict soil TN;As a whole,TN estimating from the overall soil samples can produce more stability prediction accuracies than single soil types,whereas that from single soil type samples can reflect the difference among soil types;BPNN model accuracies are superior to those of PLSR model,but PLSR has stronger operability,and can show the difference among soil types,and different among transformation indicators as well as.

相关资源

1. 水质遥感监测的关键要素叶绿素a的反演算法研究进展
2. 基于无人机遥感技术的森林参数获取研究进展
3. 基于无线传感器网络的森林生态系统观测试验平台构建
4. 基于无人机激光雷达的森林冠层高度分析
5. 多源时序国产卫星影像的森林火灾动态监测
6. 基于无人机遥感技术的台风灾害倒伏绿化树木检测

下载说明

·如果您发现该资源不能下载，请通知管理员.gissky@gmail.com

·为确保下载的资源能正常使用，请使用[WinRAR v3.8]或以上版本解压本站资源,缺省解压密码www.gissky.net ，如果是压缩文件为分卷多文件，请依次下载每一个文件，并按照顺序命名为1.rar,2.rar,3.rar...，然后鼠标右击1.rar解压.

·为了保证您快速的下载速度，我们推荐您使用[网际快车]等专业工具下载.

·站内提供的资源纯属学习交流之用,如侵犯您的版权请与我们联系.