Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
8893640 | CATENA | 2018 | 14 Pages |
Abstract
Soil organic carbon (SOC) is an important soil property relating to soil formation, structure, and water-holding capacity. Remote sensing has been used for predicting SOC but remains a challenge due to the complex and indirect relationship between SOC and remote sensing variables. In this research we explored the approaches of predicting SOC distribution in a hickory plantation region using random forest (RF) and geographically weighted regression (GWR) based on Landsat and ancillary data, analyzed SOC spatial distribution and dynamic change between 2008 and 2013 through a thresholding approach, and examined major factors that resulted in SOC degradation in the young and mature hickory plantations using a logistic regression. The results showed that RF outperformed GWR in the prediction of SOC and provided stable and reliable SOC predictions with root mean squared errors of 4.6â¯gâ¯kgâ1 in 2008 and 4.4â¯gâ¯kgâ1 in 2013. A large area of hickory plantation was experiencing SOC decrease. The analysis of major factors causing SOC degradation indicated that steep slope and high proportion of silt component in soil resulted in SOC decrease in the young hickory plantations, and high elevation, high proportion of silt component in the soil, and the increase of soil fraction in the ground cover led to SOC decrease in the mature hickory plantations. This research provides valuable approaches to spatially predict SOC and identify major factors driving SOC degradation, which will be useful for adopting better measures to improve management of hickory plantations.
Keywords
Related Topics
Physical Sciences and Engineering
Earth and Planetary Sciences
Earth-Surface Processes
Authors
Wei Lu, Dengsheng Lu, Guangxing Wang, Jiasen Wu, Jianqin Huang, Guiying Li,