利用高光谱估算博斯腾湖西岸湖滨绿洲土壤有机质含量  

Estimation of soil organic matter content in lakeside oasis on the west coast of Bosten Lake by hyperspectral method

作　　者：牛芳鹏 李新国[1,2] 靳万贵 赵慧[1,2] 麦麦提吐尔逊·艾则孜 NIU Fang-peng;LI Xin-guo;JIN Wan-gui;ZHAO Hui;Mamattursun·Eziz(College of Geographic Sciences and Tourism,Xinjiang Normal University,Urumqi Xinjiang 830054;Xinjiang Key Laboratory of Lake Environment and Resources in Arid Zone,Xinjiang Normal University,Urumqi Xinjiang 830054)

机构地区：[1]新疆师范大学地理科学与旅游学院,新疆乌鲁木齐830054 [2]新疆干旱区湖泊环境与资源重点实验室,新疆乌鲁木齐830054

出　　处：《中国土壤与肥料》2021年第1期9-16,共8页Soil and Fertilizer Sciences in China

基　　金：国家自然科学基金项目(41661047,U2003301)。

摘　　要：以新疆博斯腾湖西岸湖滨绿洲为研究区,利用实测的土壤有机质含量与高光谱数据,通过多元逐步回归与偏最小二乘回归法分别构建反演土壤有机质含量估算模型。结果表明:(1)研究区土壤有机质含量变化范围为5.09~44.00 g·kg^(-1),均值为16.87 g·kg^(-1),变异系数为44.69%,呈中等变异;土壤有机质含量与土壤光谱反射率呈极显著负相关(P<0.01),相关系数为0.09<|r|<0.42;(2)通过显著性检验(P<0.01)的波段主要集中在590~687、758~892、1003~1092和1171~1297 nm 4个波段,反射率(1/R)′变换下相关性最高,相关系数|r|为0.50(P<0.01);(3)研究区的土壤有机质含量高光谱估算模型为Y=30428.37X677+12738.78X775+2894.02X865+11589.35X885+5.56,建模集和验证集的决定系数(R2)分别为0.83和0.82,均方根误差(RMSE)分别为4.01和2.64 g·kg^(-1),验证集统计量F=116.41(P<0.01),相对分析误差(RPD)=2.30,预测能力较好。Taking the lakeside oasis in the western lakeside of Bosten Lake in Xinjiang as the research area,based on the measured soil organic matter content and hyperspectral data,a multivariate stepwise regression and partial least squares regression method were used to construct an inverse soil organic matter content estimation model.The results showed that:(1)The range of soil organic matter content in the study area ranged from 5.09 to 44.00 g·kg^(-1),the average value was 16.87 g·kg^(-1),and the coefficient of variation was 44.69%,which showed moderate variability.Soil organic matter content was significantly negatively correlated(P<0.01)with soil spectral reflectance,and the correlation coefficient was 0.09<|r|<0.42;(2)The bands that passed the significance test(P<0.01)were mainly concentrated in four bands:590~687,758~892,1003~1092 and 1171~1297 nm.The reflectance(1/R)′transform had the highest correlation.The correlation coefficient|r|was 0.50(P<0.01);(3)The hyperspectral estimation model of soil organic matter content in the study area was Y=30428.37 X677+12738.78 X775+2894.02 X865+11589.35 X885+5.56,the coefficient of determination of the modeling set and the verification set were 0.83 and 0.82,the root mean square error were 4.01 and 2.64 g·kg^(-1),respectively,and the verification set statistic F=116.41(P<0.01),RPD=2.30.The prediction ability was good.

关 键 词：土壤有机质含量 高光谱数据 估算模型 偏最小二乘回归 湖滨绿洲 

分 类 号：S153.621[农业科学—土壤学]