黑土区田块尺度土壤有机质含量遥感反演模型  

Soil organic matter content inversion model with remote sensing image in field scale of blacksoil area

作　　者：刘焕军[1,2] 潘越 窦欣 张新乐[1] 邱政超 徐梦园[1] 谢雅慧[1] 王楠 Liu Huanjun;Pan Yue;Dou Xin;Zhang Xinle;Qiu Zhengchao;Xu Mengyuan;Xie Yahui;Wang Nan(College of Resources and Environmental Sciences, Northeast Agricultural University, Harbin 150030, China;Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, Changchun 130012, China)

机构地区：[1]东北农业大学资源与环境学院,哈尔滨150030 [2]中国科学院东北地理与农业生态研究所,长春130012

出　　处：《农业工程学报》2018年第1期127-133,共7页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金项目(41671438;41501357);"中国科学院东北地理与农业生态研究所"引进优秀人才项目

摘　　要：为了对田块尺度土壤有机质进行空间反演并提高模型精度和稳定性,该文以黑龙江省黑土带41.3 hm^2田块为例,获取2016年5月中下旬两期(受限于拍摄周期和天气原因而选择不同卫星影像,2016年5月17日Landsat 8影像和5月25日Sentinel-2A影像)裸土时期遥感影像和4 m分辨率DEM数据;分析单期影像与土壤有机质(soil organic matter,SOM)的关系,两期影像所包含的土壤含水量变化信息与地形因素对SOM预测模型精度的影响,建立基于BP神经网络的SOM遥感反演模型。结果表明:该田块内SOM含量差异较大;利用单期影像预测SOM时,基于红波段和785~899 nm波段建立的预测模型精度(建模均方根误差RMSE 1.033,检验RMSE 1.079)和稳定性(建模决定系数R2 0.677,检验R20.644)较高;两期影像时,基于红波段和1 570~1 650 nm波段建立的预测模型精度(建模RMSE 0.855,检验RMSE 0.898)和稳定性(建模R2 0.792,检验R2 0.797)显著提高;在两期影像模型基础上,加入地形因子作为输入量,模型精度(建模RMSE 0.492,检验RMSE 0.499)和稳定性(建模R2 0.917,检验R2 0.928)进一步提高。研究成果可为土壤碳库估算和农田精准施肥提供理论与技术支持。In this paper,an agricultural field of41.3hm2in the black soil region of Heilongjiang Province was selected as the study area,2phases of remote sensing images during the second half of May(Landsat8image on May17th and Sentinel-2A image on May25th)and4m resolution DEM data were used as the basis research data,and the spatial distribution of soil organic matter(SOM)in the field scale was investigated.Through the analysis of the relationship between remote sensing image reflectance,spectral index and organic matter,the comparison of the difference information between2soil spectral reflectance curves of2different images for characterizing soil moisture,and the analysis of the relationship between terrain factors and SOM spatial distribution,this paper established the SOM predicting model by BP(back propagation)neural network.Comparing and analyzing the models'accuracy by using3kinds of modeling methods,this paper took the optimal model for mapping spatial distribution of SOM in study area and analyzed the spatial distribution of SOM through the inversion map.Results showed that the spatial differences of field SOM were significant.From west to east in field,with the terrain increasing,the difference of spatial distribution of SOM increased.The spatial distribution of SOM was affected by slope,aspect and slope position;the SOM content on the0-3°slope area was significantly higher than that of other slope areas;the content of SOM on shady slope was slightly higher than that of the sunny slope;the content of SOM decreased from the bottom of the slope to the sunny slope,and the content of SOM increased from the sunny slope to the top of the slope;the content of SOM decreased from the top of the slope to shady slope,and the content of SOM increased from the shady slope to the bottom of the slope.The3-5bands of Landsat8image and3,4,8bands of Sentinel-2A image can be used as the main reference bands to inverse SOM;the5-7bands of Landsat8image and8,11,12bands of Sentinel-2A image can be used as the characterization band of

关 键 词：遥感 模型 预测 土壤有机质 多期遥感影像 土壤含水量 地形因子 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] S127[自动化与计算机技术—控制科学与工程]