期刊
SURVEYS IN GEOPHYSICS
卷 35, 期 2, 页码 395-413出版社
SPRINGER
DOI: 10.1007/s10712-013-9249-8
关键词
Pedotransfer functions; Loessial soil; Artificial neural network; Multiple regression; Topography
资金
- National Natural Science Foundation of China [41101204, 41071156]
- State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau [K318009902-1308]
- China Postdoctoral Science Foundation [2012T50130]
Soil bulk density (BD) is a key soil physical property that may affect the transport of water and solutes and is essential to estimate soil carbon/nutrients reserves. However, BD data are often lacking in soil databases due to the challenge of directly measuring BD, which is considered to be labor intensive, time consuming, and expensive especially for the lower layers of deep soils such as those of the Chinese Loess Plateau region. We determined the factors that were closely correlated with BD at the regional scale and developed a robust pedotransfer function (PTF) for BD by measuring BD and potentially related soil and environmental factors at 748 selected sites across the Loess Plateau of China (620,000 km(2)) at which we collected undisturbed and disturbed soil samples from two soil layers (0-5 and 20-25 cm). Regional BD values were normally distributed and demonstrated weak spatial variation (CV = 12 %). Pearson's correlation and stepwise multiple linear regression analyses identified silt content, slope gradient (SG), soil organic carbon content (SOC), clay content, slope aspect (SA), and altitude as the factors that were closely correlated with BD and that explained 25.8, 6.3, 5.8, 1.4, 0.3, and 0.3 % of the BD variation, respectively. Based on these closely correlated variables, a reasonably robust PTF was developed for BD using multiple linear regression, which performed equally with the artificial neural network method in the current study. The inclusion of topographic factors significantly improved the predictive capability of the BD PTF and in which SG was an important input variable that could be used in place of SA and altitude without compromising its capability for predicting BD. Thus, the developed PTF with only four input variables (clay, silt, SOC, SG), including their common transformations and interactive terms, predicted BD with reasonable accuracy and is thus useful for most applications on the Loess Plateau of China. More attention should be given to the role of topography when developing PTFs for BD prediction. Testing of the developed PTF for use in other loess regions in the world is required.
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