American Society of Agricultural and Biological Engineers



Click on “Download PDF” for the PDF version or on the title for the HTML version.


If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Removing the effect of soil moisture for soil TN prediction based on a portable soil sensor

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  Paper number  131620131,  2013 Kansas City, Missouri, July 21 - July 24, 2013. (doi: http://dx.doi.org/10.13031/aim.20131620131) @2013
Authors:   Xiaofei An, Minzan Li, Lihua Zheng, Yumeng Liu, Hong Song
Keywords:   NIRS Soil TN content Effect of soil moisture EPO Precision agriculture

Abstract. Field measurement using NIRS is becoming a popular method to provide rapid and inexpensive detection of soil total nitrogen (TN) content. However, NIR is quite sensitive to external environmental conditions such as soil moisture(SM). It is a challenge to find a proper method which is insensitive to variable SM for prediction of soil TN. This paper took an External Parameter Orthogonalisation (EPO) algorithm to remove the effect of SM from NIR spectra for prediction of soil TN content. The algorithm projected all the soil spectra orthogonally to the spaces of moisture variable and non-moisture variables, hence the variation of SM can be effectively removed from the original spectra. In the algorithm, two datasets were prepared. First, projection matrix P was obtained by performing EPO on dataset (1). Subsequently, a BPNN regression model was established by the transforming spectra of dataset (1) with projection matrix P. Finally, spectra of dataset (2) was transformed using matrix P. Soil TN content was predicted with the transformed spectra. Experiments were conducted using soil at different SM in laboratory conditions. When only using BPNN, the prediction R2 value was 0.43 with a high RMSEP. However when using EPO-BPNN, the prediction R2 value reached 0.62 with a low RMSEP. It turned out that the soil TN prediction accuracy for soil samples with varying SM became similar with that for air-dried soil samples. The experiments indicated that the effect of SM could be removed successfully from the original spectra by using EPO method.

(Download PDF)    (Export to EndNotes)