Sensors & Transducers (Jun 2014)
Feasibility Study of Soil Quality Survey using Visible and Near Infrared Spectroscopy in Rice Paddy Fields in China
Abstract
Survey and monitoring of soil quality are needed to prevent soil degradation and are important for sustainable farming and food production. Conventional soil survey involves intensive soil sampling and laboratory analysis, which are time consuming and expensive. Visible and near infrared spectroscopy of soil has proved to be accurate, cheap and robust and has huge potential for survey of soil quality. To test its potential, 327 soil samples were taken from long-term paddy rice fields in four provinces in south of China and covered a wide range of soil types and texture. The samples were air-dried, ground and passed through a 2 mm sieve. They were then scanned by an ASD vis–NIR spectrometer with wavelength range from 350 to 2500 nm. Organic matter (OM), pH, total nitrogen (TN) and available nitrogen (N_av) were also measured on soil samples to build calibration models and also to validate the models’ accuracy. On the basis of the ratio of prediction deviation (RPD), which is standard deviation (SD) of prediction divided by the root mean square error of prediction (RMSEP), the accuracy of leave-one-out cross-validation of soil N_av model was classified very good (RPD=1.96) and soil OM and TN was good (RPD=1.78 and RPD=1.81, respectively). However, the model accuracy of pH was poor due to non-direct soil spectral response for soil pH in vis–NIR spectroscopy. The independent validation results showed excellent accuracy for soil N_av (RPD=3.26), good accuracy for OM and TN (RPD=1.76 and RPD=1.78) and relative poor accuracy for soil pH (RPD=1.27). This feasibility study is encouraging for the application of vis–NIR surveys of soil quality accuracy at regional and national scales; it found good to excellent accuracy for some important soil properties in quality survey.
