基于光谱指数的博斯腾湖西岸湖滨绿洲土壤有机碳含量估算模型  

Estimation model of soil organic carbon content in lakeside oasis on the west coast of Bosten Lake based on spectral index

作　　者：牛芳鹏 李新国[1,2] 麦麦提吐尔逊·艾则孜[1,2] 赵慧 江远东 NIU Fang-peng;LI Xin-guo;MAMATTURSUN·Eziz;ZHAO Hui;JIANG Yuan-dong(College of Geographic Sciences and Tourism,Xinjiang Normal University,Urumqi 830054,China;Xinjiang Laboratory of Lake Environment and Resources in Arid Zone,Urumqi 830054,China)

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实验室,新疆乌鲁木齐830054

出　　处：《江苏农业学报》2022年第2期414-421,共8页Jiangsu Journal of Agricultural Sciences

基　　金：国家自然科学基金项目(41661047、U2003301);新疆维吾尔自治区重点实验室开放课题(2018D04026)。

摘　　要：以博斯腾湖西岸湖滨绿洲为研究区,将野外原位高光谱实测数据和土壤有机碳(SOC)含量作为基础数据,通过对原始光谱进行4种数学变换,探索不同光谱变换形式下的弓曲差(C)、差值光谱指数(DSI)、简单比值土壤指数(RSI)、亮度光谱指数(BSI)、归一化土壤指数(NDSI)与SOC含量的关系,并建立基于随机森林法(RF)的SOC含量估算模型。结果表明:(1)研究区SOC含量主要集中在5.25~78.76 g/kg,平均值为21.82 g/kg,变异系数为69.11%,呈中等变异性;(2)在光谱数据lgR下,SOC含量与DIS指数相关系数最高,相关系数为0.80,最佳组合波段为(1758 nm,1752 nm);(3)基于不同光谱指数与弓曲差(C)建立的模型验证集精度R^(2)和RMSE分别介于0.67~0.84和5.85~8.45 g/kg,模型的RPD均在1.66以上;在基于光谱数据lg(1/R)变换下,模型的验证集R^(2)=0.82、RMSE=3.52 g/kg、RPD=3.99,可以较好地估算研究区SOC含量,为干旱半干旱地区湖滨绿洲SOC含量反演提供依据和参考。Taking the lakeside oasis on the west coast of Bosten Lake as the study area,and the in situ hyperspectral data and soil organic carbon(SOC)content were taken as the basic data.Four mathematical transformations were performed on the original spectrum to explore the relationships between bow curvature difference(C),difference spectral index(DSI),simple ratio soil index(RSI),brightness spectral index(BSI),normalized soil index(NDSI)and SOC content.A SOC content estimation model based on random forest method was established.The results indicated that the content of SOC was mainly concentrated in 5.25-78.76 g/kg,the average value was 21.82 g/kg,the coefficient of variation was 69.11%,showing moderate variability.Under the spectral data lgR,the correlation coefficient between SOC content and DIS index was the highest,the correlation coefficient was 0.80,and the best combination band was(1758 nm,1752 nm).The R^(2) and RMSE of the model validation set based on different spectral indices and C were 0.67-0.84 and 5.85-8.45 g/kg,respectively,and the RPD of the model was above 1.66.Based on the transformation of spectral data lg(1/R),the R^(2),RMSE and RPD of the model validation set were 0.82,3.52 g/kg and 3.99,respectively.The model can better estimate the SOC content in the study area,and the research results can provide the basis and reference for the inversion of SOC content in lakeside oasis in arid and semi-arid regions.

关 键 词：土壤有机碳含量 估算模型 光谱指数 随机森林 湖滨绿洲 

分 类 号：S153.6[农业科学—土壤学]