ORIGINAL RESEARCH
Effect of Soil Moisture and Particle Size
on Soil Total Phosphorus Estimation
by Near-Infrared Spectroscopy
Lei Zhang1,2, Rongbiao Zhang2
More details
Hide details
1School of Electrical and Control Engineering, Henan University of Urban Construction,
Pingdingshan, 467044, China
2School of Electrical and Information Engineering, Jiangsu University,
Zhenjiang, 212013, China
Submission date: 2016-07-01
Final revision date: 2016-08-31
Acceptance date: 2016-08-31
Online publication date: 2017-01-31
Publication date: 2017-01-31
Pol. J. Environ. Stud. 2017;26(1):395-401
KEYWORDS
TOPICS
ABSTRACT
Near-infrared spectroscopy (NIRS) can detect soil total phosphorus in agricultural environments.
Considering the soil moisture and particle size on total phosphorus prediction, we applied NIRS to the
detection of soil samples with different soil moistures and particle sizes. Thus the effect of soil moisture
and particle size was analyzed quantitatively and qualitatively. The procedures to remove the effect of
soil moisture and particle size on total phosphorus prediction were also described. First, the near-infrared
reflectance spectra of soil samples with different soil moistures and particle sizes were obtained and the
absorbance values were determined. Next, the original spectra were corrected by using moisture absorbance
index (MAI) and hybrid correction to counteract the effects of soil moisture and particle size, respectively.
Absorbance of soil samples showed high correlation with soil moisture at wavelengths of 1,450 nm and
1,940 nm. MAI is a tool for normalizing the original spectral data so as to correct for soil moisture. Hybrid
correction is based on the superposition of NIR spectra, and a particle size different from that of the original
soil samples is generated. This is an effective means of correcting for the effect of soil particle size. Finally,
using the corrected absorbance values at eight wavelengths (655, 722, 1,055, 1,255, 1,467, 1,678, 1,890,
and 2,246 nm), the soil total phosphorus prediction model was built based on LS-SVM. Compared with the
model used for original spectral data, the new model exhibited higher accuracy and stability. Results showed
that MAI and hybrid correction are effective for correcting for soil moisture and soil particle size during the
prediction of soil total phosphorus.