| Hyperspectral estimation of soil organic carbon and its influencing factors in arid oasis |
|
View Fulltext View/Add Comment Download reader |
| |
| DOI：10.7606/j.issn.1000-7601.2018.05.29 |
| Key Words: soil organic carbon hyperspectral spatial distribution prediction environmental factors |
|
| Hits: 975 |
| Download times: 502 |
| Abstract: |
| Soil organic carbon (SOC) concentration is a useful soil property which can guide agricultural applications of chemical inputs.Soil organic carbon is also an important factor affecting regional carbon budget.It is impending to develope more time and cost-efficient methodologies for SOC analysis.To reveal the distribution pattern of SOC and its influencing factors in arid area, the experiment was conducted in the delta oasis between the Weigan-Kuqa River Oasis .The SOC content was predicted by VNIR-PLSR model using kriging. The results showed that the SOC value in the oasis varies from 0.172 to 17.376 g·kg-1 among the soil samples. The value of the coefficient of variation was between 35%~53%.The most important wavelengths for SOC prediction were 550nm and 650 nm in the visible, and 780nm, 818nm, 866nm, 1423nm, 1733nm, 2005 nm and 2172 nm in the near-infrared region. Second-order differential transformation was better than the first-order differential transformation. VNIR-PLSR regression model determination coefficient (R2) was higher than 0.853 and all residual prediction deviation (RPD) was higher than 1.5.The distribution pattern of SOC in each layer was similar. The SOC accumulation in top soil and the content in the oasis was significantly higher than which in the oasis periphery. Soil salt content was the most important environmental factor for the content of SOC. Quantitative spectral analysis of SOC by vis-NIR reflectance spectroscopy was feasible. The spectrum transformed with second-order differential could accurately predict SOC content.The model with partial least squares regression model had higher prediction accuracy and better robustness. |
|
|
|