SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Photonics West 2018 | Call for Papers

SPIE Defense + Commercial Sensing 2018 | Call for Papers




Print PageEmail PageView PDF

Remote Sensing

Prediction of soil content using near-infrared spectroscopy

Near-infrared spectroscopy can be used to accurately predict certain soil properties, making it a valuable tool in precision farming.
22 May 2006, SPIE Newsroom. DOI: 10.1117/2.1200604.0164

Modern technological developments in positioning, sensing, and control systems have opened a new era in which many traditional agricultural practices are being left behind. Replacing them are ‘precision farming’ techniques that manage variability within a field by applying agronomic inputs in the right place, at the right time, and in the right quantity to reduce the environmental impact of crop production. Visible and near-infrared (NIR) spectroscopy has emerged as one of these techniques, enabling the rapid and nondestructive analytical correlation of diffusely reflected NIR radiation with chemical and physical properties of soil components.

NIR spectroscopy has been used for assessing grain and soil qualities1–5 and has proven to be a rapid, convenient means of analyzing many soil constituents at the same time. The goals of this study were to analyze the potential of NIR spectroscopy to estimate soil nitrogen content (N), phosphorus content (P), potassium content (K), organic matter content (OM), and pH and to combine these predictions with geographic information systems (GIS).

The site where we performed our experiments is located in Hangzhou, China. The soil type was classified as loamy mixed active thermic aeric endoqualf according to Zhejiang soil classification.6 The test field boundary was defined using a Trimble AgGPS 132 global positioning system receiver. A total of 165 soil samples were collected at a depth of 20 centimeters, then divided into two groups. Group A was tested for N, P, K, OM, and pH using traditional methods; group B was analyzed with a visible and NIR spectrometer (ASD, 350-2500nm). Group B samples were air dried and sieved with a 2mm sieve, then set in petri dishes. Reflectance spectra were taken over the central area of each dish, recoded and checked visually, then averaged and exported to the multivariate analysis software (The Unscrambler 8.0). Representative mean absorbance (log 1/R) soil spectra of 10 samples are shown in Figure 1.

Figure 1. Representative mean absorbance soil spectra of 10 samples.

The 165 samples were also divided into two sets for another purpose. Set I, with 135 samples, was used to develop a calibration model, with the equations developed using principal components analysis (PCA) and partial least squares (PLS). The best calibration was deemed to be the one with lowest root mean square error of prediction (RMSEP), the lowest standard error of prediction (SEP), the lowest standard deviation, and the highest coefficient of correlation (r). Each calibration equation developed from set I was then used to predict the constituent values for the independent spectra in set II. For each, the NIR-predicted values for set II were correlated with their measured values. Calibration statistics for each prediction model are shown in Table 1.

Table 1. Calibration statistics for N, P, K, OM and pH (click to enlarge)

PCA reveals that the first principal component (PC1) accounts for the largest possible amount of variance in the data, while subsequent components account for successively less of the variance, showing no particular trend with each constituent composition. Non-samples were removed from the prediction.

Figure 2. Score plot of the 135 samples on the first two principal components.


The prediction set demonstrated that NIR spectroscopy predicts N, OM, and pH, but not P and K in this soil. Statistics for predicted N, P, K, OM, and pH are shown in Figure 3.

Figure 3. Correlation between measured and predicted values of N, P, K, OM, and pH.

Predicted and measured N, P, K, OM, and pH spatial distribution was interpolated by ArcView GIS 3.1 software using the spline method, which creates a smooth curve restrained by specified points. Spatial distribution maps are shown in Figure 4.

Figure 4. Spatial distribution maps of predicted and measured N, P, K, OM, and pH.

NIR spectroscopy has the potential to accurately predict N, OM, and pH in this soil. Further investigation is required to try to make successful P and K predictions.

The slight variation between the maps obtained for measured and predicted N, OM, and pH, as well as the speed, easy operation, and portability of NIR instruments make NIR spectroscopy an appealing technology for precision-farming field monitoring.

Yong He
College of Biosystems Engineering and Food Science, Zhejiang University
Hangzhou, Zhejiang
Professor Yong He is Vice Dean of the College of Biosystems Engineering and Food Science at Zhejiang University. He has served as a vice director of the High Education Guidance Committee of Agriculture Engineering at the Ministry of Education for the People's Republic of China. He received his Ph.D. from Zhejiang University in 1998 and was a visiting professor and advanced visiting scholar at the University of Illinois at Urbana-Champaign and Tokyo University of Agriculture and Technology (1998–2002). His research and teaching interests include spatial technology, precision agriculture, operational research, agriculture mechanization, artificial intelligence, and expert systems.
Haiyan Song
College of Biosystems Engineering and Food Science, Zhejiang University
Hangzhou, Zhejiang