期刊
CATENA
卷 219, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.catena.2022.106603
关键词
Soil organic carbon density; Airborne hyperspectral; Bare soil spectral index; Deep neural network; Digital soil mapping
资金
- Key Project of Department of Edu-cation of Guangdong Province [2020ZDZX1052]
- Basic Research Key Program of Science, Technology and Innovation Commission of Shenzhen [JCYJ20210324120209027]
The study focuses on the digital soil mapping of Soil Organic Carbon Density (SOCD) in agricultural land using different spectral data, including laboratory, airborne, and Sentinel 2 spectra. The results show that laboratory spectra perform the best in SOCD prediction, followed by airborne and Sentinel 2 spectra. The study also compared the performance of bare soil spectral indices (BSSIs) and vegetation indices (VIs) in SOCD prediction and found that BSSIs have higher accuracy. Among the three prediction models used, the deep neural network (DNN) model shows the best performance in digital soil mapping of SOCD.
Soil organic carbon density (SOCD) is an important parameter of agricultural soils and is useful for the improvement of environment and agricultural production. Proximal and remote sensing techniques are effective methods for digital soil mapping of SOCD. The current study used three types of spectral data, including labo-ratory proximal spectra, airborne hyperspectral and Sentinel 2 multispectral images, to predict SOCD in an agricultural land. Bare soil spectral indices (BSSIs) were developed to predict SOCD and compared with the published vegetation indices (VIs). With 45 soil samples, the partial least square regression (PLSR), back prop-agation neural network (BPNN) and deep neural network (DNN) prediction models were established to map SOCD. The results showed that the laboratory spectra (R-2 = 0.70-0.80) had the best performance of SOCD prediction, followed by the airborne (R-2 = 0.43-0.81) and Sentinel 2 (R-2 = 0.14-0.57) spectra. The SOCD maps derived from airborne and Sentinel 2 images had similar spatial distribution trends. The BSSIs (R-2 = 0.24-0.81) obtained higher accuracy than the VIs (R-2 = 0.14-0.74) in SOCD prediction. Moreover, DNN model was the best for digital soil mapping of SOCD among three prediction models. This study offered an effective approach for mapping SOCD in bare soil areas.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据