Premium
Use of Spectral Character to Evaluate Soil Organic Matter
Author(s) -
Wang Chao,
Feng Mei-chen,
Yang Wu-de,
Ding Guang-wei,
Wang Hui-qin,
Li Zhi-hua,
Sun Hui,
Shi Chao-chao
Publication year - 2016
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj2015.10.0364
Subject(s) - partial least squares regression , multivariate statistics , soil test , mean squared error , soil organic matter , mathematics , linear regression , soil science , regression analysis , absorption (acoustics) , chemistry , environmental science , soil water , analytical chemistry (journal) , statistics , environmental chemistry , physics , optics
Core Ideas Reflectance spectroscopy was used to characterize soil properties. Continuum‐removed processing was utilized to achieve absorbed peaks and absorbed characters. Multiple liner regression and PLS methods were also applied and compared for assessment. Variable importance in projection was used for evaluation. To estimate soil organic matter (SOM) and evaluate the performance of a SOM prediction model with two different soil particle sizes (sieved soil and nonsieved soil), multivariate statistical analysis is applied to construct an SOM model with absorption peak parameters derived from the continuum‐removed (CR) method. In this study, 107 samples from Exp. 1 were randomly categorized into 86 samples as a calibration set and 21 samples as a validation set to construct the SOM models. The second field experiment, which included 47 samples, was also implemented to validate the spectral absorption peak parameters of SOM. The results indicated that the SOM models based on the nonsieved soil ( R 2 > 0.692; RMSE < 6.018) had better performance than the sieved soil ( R 2 < 0.627; RMSE > 6.732). The most accurate SOM model was based on the multiple liner regression (MLR) method. The SOM models constructed with the partial least squares regression (PLSR) method were stable and also showed a moderate prediction for both nonsieved soil and sieved soil. Moreover, the important absorption peak parameters (trend slopes [TS] at 2410 nm, spectral absorption index [SAI] at 1320 nm, and TS at 662 nm for nonsieved; SAI at 892 nm, TS at 2345 nm, and absorption depth [H] at 669 nm for sieved soil) were determined to be related to SOM by using the PLSR method. Therefore, the combination of multivariate statistical methods (MLR and PLSR) provides a foundation to estimate SOM.