研究目的
To develop a soil spectral allocation model that considers the effect of soil moisture for the purpose of eliminating the effect of soil moisture on soil reflectance, thereby improving the universality of soil spectral allocation, especially for field soils.
研究成果
The study concludes that it is feasible to build soil spectral allocation models that are not affected by soil moisture, with the DT model considering the effect of soil moisture identified as the optimum model. The accuracy of soil allocation was significantly improved, suggesting that such models can be of considerable help in soil classification, especially for field soils.
研究不足
The study focuses on the effect of soil moisture on soil spectral allocation but does not consider other factors such as different crop residue levels, sand and stone contents, and impurities of cattle manure that can also affect the soil in field environments.
1：Experimental Design and Method Selection
The study involved re-sampling topsoil spectral curves of four typical soils from the Songnen Plain in Northeast China to 10-nm intervals and converting them to first-derivative spectral curves and continuum removal curves. Spectral feature parameters were extracted from continuum removal curves in the visible-near infrared (VNIR) range (350–2500 nm), with the range of 430–2400 nm used to build soil allocation models to reduce the effect of noise.
2：Sample Selection and Data Sources
A total of 135 samples (0–20 cm) of Phaeozems, Chernozems, Cambisols, and Arenosols were collected in 2014. Soil samples were ground, dried, and passed through a 2-mm screen prior to spectral analysis. One sample of each air-dried soil class (53 samples in total) was selected to prepare soil samples with different soil moisture.
3：List of Experimental Equipment and Materials
Air-dried samples for each of the soil samples were scanned in a dark room using an ASD FieldSpec?3 portable spectroradiometer with a spectral range of 350–2500 nm. Soil samples were placed in a Petri dish (12 cm in diameter and 1.8 cm deep) and the surface of the soil was scraped with a ruler. A 50 W halogen lamp with a zenith angle of 30° at a distance of 100 cm from the samples was used as the light source.
4：Experimental Procedures and Operational Workflow
The dark current was removed before the test and the spectroradiometer was calibrated with a white panel of known reflectance. Each soil sample was tested with 10 spectral curves, which were averaged to obtain the reflectance of the sample. Spectral reflectance of 188 samples were smoothed for reducing noise, resampled at 10-nm intervals and converted to the first-derivative and CR.
5：Data Analysis Methods
PCA was applied to extract the spectral information of 188 mixed samples and 53 samples with different soil moistures. The CV was calculated from PCR and PCFD of the 53 samples with different soil moistures to assess the effect of soil moisture. DT, MLR, and MLPNN models were built and the Kappa coefficient was used for accuracy evaluation.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
