Near-Infrared Spectroscopy for soil analysis
Our population is growing rapidly, creating a much bigger need for food, while climate change and poor soil management has already affected large areas around the world. As a result, the need for techniques and instruments that allow rapid field soil monitoring has never been greater. Unfortunately, soil sampling and laboratory analysis are often too time-, cost- and labor- intensive in order to serve the current needs, however the great precision is still a prerequisite.
Over the past few decades, research teams have focused their attention on the use of near-infrared diffuse reflectance spectroscopy for measuring soil constituents. In the mid-1990s, and early 2000s, the field began to pick up speed, initiating and bringing to the spotlight significant developments of technology and techniques. Great amount of work, by numerous researchers around the world, contributed to what is now the core of infrared spectroscopy for soil science. Some of the major game-changing studies were conducted by Ben-Dor and Banin (1995), Viscarra Rossel and McBratney (1998), Shepherd and Walsh (2002), among others. Nowadays, there are still many scientists and developers using the same principles, and building on them, continuously improving the methods.
Basic soil composition, or more precisely, soil organic matter, soil texture and soil clay mineralogy have been in the core of most infrared spectroscopy research for soils. Of course, nutrient availability, soil structure, soil microbial activity and soil fertility have also been a major subject of interest over the past two decades.
One of the greatest advantages of using near-infrared spectroscopy for soil analysis is the simple (or better yet, almost non-existent), hazard-free sample preparation. What is more, the time it takes for a measurement to be taken is just a few seconds, and this measurement can be done both in the lab and in the field. Various soil properties can be estimated from the spectra signal (also referred to as a soil spectral signature or absorption spectral curve) of the measured soil.
Picture 1: The Near Infrared spectrometer of the SoilCares Scanner.
Spectrometers are the instruments used to obtain the spectral signature of the soil. These instruments can be either passive, using the sun as a source of light, or active with inbuilt light source. The radiation from the source triggers vibrations of the molecular bonds of the soil constituents, allowing them to absorb light differently. The result is an absorption curve with highly characteristic shape that is used for soil analysis and property predictions.
This curve is a representation of the frequencies at which light absorbs, and it is typically measured in percentage reflectance or absorbance. Various environmental factors, together with the chemical matrix, will determine the absorption pattern across the wavelength range. The wavelengths at which most light is absorbed corresponds to particular soil constituents or property. This information is used to determine soil characteristics by relating the absorption to concentrations. The factors that have the highest influence on absorption are moisture, temperature and the chemical composition of the soil.
Figure 1. Spectral reflectance curves for silt loam soil at various moisture contents (Source: Bowers and Hanks, 1965)
This spectral region, however, can provide crucial information on numerous organic and inorganic soil constituents. All absorptions that are observed in the NIR region between 780 and 2500 nm are a product of the OH, NH, CH, SO and CO vibrations showing a large dipole moment. In this wavelength region, we can also observe the influence of clay mineralogy, due to the bend of the OH-metal bend and the stretch combinations of O-H bond from clay bound water (Stenberg et al., 2010) . Other properties that can be observed in the NIR spectrum are moisture content and carbonates (Viscarra Rossel et al., 2006a).
Essentially what we measure is the interaction between the IR radiation and the particular molecular combinations of the original fundamental vibrations originating in the Mid Infrared (MIR) region. Within the near infrared region (NIR), soil spectra is characterized by relatively few broad absorption features in comparison to the mid-infrared (MIR) part of the spectrum. The reason is that the bands are mostly broad and often overlapping. This, in turn, makes NIR spectra more challenging to interpret.
For example, figure 1 shows typical soil spectral signatures, derived from samples with different moisture content. As we can see, the higher the moisture, the lower the overall reflectance is. However, we can be even more specific. The wavelengths associated with moisture content are 1400nm and 1900nm. With higher moisture content, the absorption at these wavelengths becomes bigger too. In other words, the lower the reflectance value at these wavelengths is, the higher the moisture content.
Figure 2. The different optical sections of the electromagnetic spectral range. (Source: Canada Research Chair in Multipolar Infrared Vision)
The example with moisture content is one of the most simple, as the relationship between reflectance values and this property are direct. However, for many soil properties, we often need to establish indirect relationships in order to predict concentrations from NIR spectra (Luleva et al, 2011). The most common method to translate spectra into meaningful numbers is by using chemometrics. This is the mean to extract information from spectra, by statistically relating it to chemical data. The result is prediction models for chemical elements, compounds or properties that translate the spectral curve into useful values with various degrees of accuracy. Individual properties can be predicted directly, as shown above, or indirectly, using these models. The higher the number of samples, the more accurate our prediction models are.
In summary NIR is a powerful and convenient analytical tool. Although the information that can be obtained from the MIR region is often much more comprehensive, NIR measuring spectrometers are much cheaper. NIR measurements require almost no sample preparation, do not demand the use of chemicals, and are fast and easy to take. The only drawback is the complicated conversion of useful information from the spectral data recorded. However, an extensive and appropriate database and the dedicated software, which uses machine learning and chemometrics (available at SoilCares) eliminate this hurdle. Of course, the right methods should be coupled with the right data acquisition tool- we recommend, the SoilCares Scanner.