Infrared Reflectance Spectroscopy (NIRS)

Infrared Reflectance Spectroscopy (NIRS) is becoming a promising technique in soil analyses. Several studies have demonstrated the ability of NIRS, for rapid and non-destructive technique to quantify soil carbon and nitrogen in different ecosystems. This study had used this technique for measuring the soil carbon and nitrogen content under different types of vegetation cover from three different climatic regions (semi arid, dry sub humid and mist sub humid) in India. The effects of different soil moisture levels on predicting equations were developed to quantify soil carbon and nitrogen. Totally 180 soil samples were used for developing equations (calibration data n=136), validation of the equations (n=31) and evaluate the effects of soil moisture on predicting equations (n=13). Soil carbon and nitrogen content was successfully predicted (R2= 0.90 for carbon and R2= 0.85 for nitrogen) by the equations developed. The standard error of prediction (SEP) standard error of prediction corrected for bias SEP (C) and bias for predicting equations of carbon and nitrogen were 0.73, 0.73, 0.04 and 0.07, 0.07, 0.005 respectively. The root mean square error (RMSE) and ratio performance deviation (RPD) for the validation of predicted equation of carbon and nitrogen was 0.83, 2.83 and 0.01, 6.98 respectively. Our results of soil moisture experiments showed that when dry sample calibrations were applied to the moistened soil samples, the prediction error is increasing manifold. Further study should test the robustness of NIRS prediction of soil carbon and nitrogen in other climatic regions of India and extend this study to predict other soil properties.

Key words: Soil carbon, Soil nitrogen, NIRS, chemometric analysis, India.

1. Introduction
Understanding the rate and storage of carbon and nitrogen in soils with respect to land use changes and land management activities, particularly on soil fertility is required an obvious measurement of soil carbon and nitrogen in regional to the global level at frequent time intervals. Analysing the soil samples at frequent time intervals by conventional methods such as tri-titrometric, dry combustion and other chemical based methods are expensive and time consuming. Recently the diffuse reflectance spectroscopy has been widely used for the rapid quantification of carbon, nitrogen and other properties in soils (Viscarra Rossel et al., 2006; Rinnan and Rinnan, 2007; Cambule et al., 2012). Near infrared reflectance spectroscopy (NIRS) is a physical, non-destructive, rapid, low cost and environmentally good technique of soil analysis (Stenberg et al., 2010; Nicotia et al., 2011). Replacing the conventional laboratory methods with NIRS would eventually reduce the laboratory costs (Viscarra Rossel et al., 2006, Mouazen et al., 2007). This technique is needed a consideration since highly advanced laboratory/portable NIRS been used to allow in situ determination of the main soil properties such as soil organic carbon, soil organic matter, nitrogen, soil pH, cation exchange capacity (CEC) and moisture (Sudduth et al., 1993; Malley et al., 2000; Viscarra Rossel et al., 2006; Zornoza et al., 2008 Nocita et al., 2013).
NIRS is an indirect analytical technique that characterises the materials according to their reflectance wavelengths normally ranging between 800 nm and 2500 nm. Spectral signatures of the materials are defined by their reflectance (R) or absorbance [log 1/R] as a function of wavelengths in the electromagnetic spectrum. Near Infrared (NIR) spectra is a result of bending, stretching, twisting and scissoring of the chemical bonds, C-H, N-H, S-H and C=O, found in the organic materials and soil particles (sand, silt and clay) that absorb light in the near infrared region (Viscarra Rossel et al., 2006; Stenberg et al., 2010). The resulting spectra do not contain distinct/sharp peaks due to the overlapping absorption of soil constituents in the mid-infrared region. Since the complex mixture of spectrum of the soil constituents impedes a direct interpretation. Thus, it requires the development of calibrations that relate the content of soil constituents, for example, carbon and nitrogen content in soil, to the spectral information by using sophisticated statistical tools such as principal component analysis (PCA), principal component regression (PCR) and partial least-square regression (PLSR) and so on (Stenberg et al., 2010; Viscarra Rossel, 2008; Viscarra Rossel and Webster, 2011). For example, quantitative analysis of soil properties with NIRS normally involves multivariate statistical techniques that consist of regression method used to create predictive models (Viscarra Rossel, 2008). The present study made an attempt to quantify the soil carbon and nitrogen in three major climatic regions of India by using NIRS. The objectives of this study were 1) to assess the efficacy of NIRS to predict the soil carbon and nitrogen content in three major climatic regions of India and 2) to evaluate the effects of different soil moisture content on prediction formula of carbon and nitrogen.

2. Materials and Methods
2.1 Experimental site details
For this study, about 180 soil samples were collected from the selected site in 1) semi arid (27??07'6"N to 28??32'10.6"N and 77??10'33.7"E to 77??33'9"E) 2) Dry-sub humid (30??17'03.97"N to 30??39'52.56"N and 78?? 19'46.6"E to 78??59'30.37"E) and 3) Moist sub- humid (33??23'32.41"N to 33??33'41.00"N and 74??18'38.42"E to 74??34'52.00"E) region under different types of vegetation cover, India. The details of these climatic regions are described elsewhere (Raju et al., 2013). The annual rainfall was 500-750 mm, 750-1000 mm and 1000-1250 mm in semiarid, dry sub humid and moist sub humid region respectively (Raju et al., 2013). Soils in these different climatic regions are dominated by Aridisols, Entisols and Incebtisols (FAO/USDA soil map of the world). In semiarid region, soils were collected from the sites where the vegetation cover is primarily occupied by Acacia nilotica, Mitragyna parviflora, and intermixed with Prosopis juliflora. In dry and moist sub humid region, the soil were collected from the sites under the vegetation cover of 1) mixed cover (Shorea robusta (Sal), Bombax ceiba, Dalbergia sissoo, Anogessius latifolia, Holoptelea integrifolia, Adina cordifolia, Mallotus philippensis and Lagerstroemia parviflora) 2) Pinus roxburgi plant species cover 3) mixed cover (Quercus sps (Oak), Rhdodendron sps. Alnus sps) and Pasture/grazing land (located in ~2400 msl).
2.2 Soil sampling and reference data, chemical analysis
Soils were collected from the selected sites with 10 cm depth interval up to a depth of 30 cm. Soils were collected from the sites by trench method and by using an AMS professional soil sampling kit. The collected soil samples were air dried and passed through a 2 mm sieve. After that the soil samples were ultimately grounded by mortar and pestle. The grounded samples were used for determination of total soil carbon and nitrogen by dry combustion using a vario-micro cube CHNS analyzer (Germany).
2.3 NIRS analysis and chemometric treatment
All the collected soil samples, uniformly grounded in mortar and pestle, were scanned in a FOSS NIRS system 5000 working in reflectance mode between 1100 and 2498 nm at 2nm intervals. Ring cup has been used for all the measurements. The resulting spectrum of each sample is the average of 32 scans recorded as absorbance (log 1/R). The WinISI III project manager ver. 1.61software, installed in the computer, was used to collect the spectrum and chemometric anaylsis. About 136 soil samples were used as 'calibration data set', 31 soil samples were used for 'cross validation' and 13 soil samples used for the experiment on the effect of soil moisture on predicting equations. The collected spectra were corrected by using a scatter corrected using standard normal variate and detrending (SNVD) of the Win ISI software (Co??teaux et al., 2003; Nduwamungu et al., 2009) in the computer system. The most appropriate mathematical treatment, i.e. 1,4,4,1 was used. In 1,4,4,1 the first digit is the number of the derivative; the second is the gap over which the derivative is calculated; the third is the number of data points in a running average or smoothing, and the fourth is the second smoothing (Brunet et al., 2007; Fuentes et al., 2012) .
The Win ISI software included the CENTER algorithm which carries out a principal component analysis of spectral data and calculated the Mahalanobis distance of each soil sample from the average spectrum. It provides the details about the distance between one sample and the average sample in a set of spectra (Perez-Marin et al., 2007). Samples were not included with an H-value >3 for developing equations. Calibrations were developed for predicting the soil carbon and nitrogen by using the WinISI III project manager ver. 1.61 chemometric software. The spectral region between 1100 and 2498 nm at 2 nm intervals was selected to perform the calibration models. Prediction equations were obtained by using a modified partial least square (MPLS) regression method of the Win ISI software. The quality of developed calibration equations was checked in the validation stage. The validation sample (n=31) sets also collected from the sites which have been not used for developing equations. During validation the NIRS predicted values from the equation are regressed against the laboratory reference values. The standard error of prediction (SEP) which is the standard deviation (SD) of difference between NIR reflectance and reference values and that should be calculated on real independent data set.
The model accuracy was evaluated by root mean squared error of calibration (RMSE) of the cross validation predictions and R2 of soil carbon and soil nitrogen (Cambule et al., 2012).
RMSE = 2
Where Yi is the measured value, yi is predicted value and N is the number of data.
The prediction capacity of the calibrated models was evaluated with the ratio-performance deviation (RPD) a capacity parameter defined as the relationship between the standard deviation of the analysed data (SD reference) and the RMSE of validation data set (Chang et al., 2001; Chang and Laird, 2002; Waiser et al. 2007).
RPD = Standard deviation of analysed data / RMSE

To evaluate the accuracy of the model the coefficient of determinations (R2) and RPD statistics were used. According to Sayes et al. (2005), the approximate prediction on quantifications, an R2 value between 0.66 and 0.88 whereas a value between 0.80 and 0.90 reveals the best prediction. Regarding the RPD statistic values, an RPD >2.0 is considered the most accurate prediction, whereas the value between 1.4 and 2 makes moderate prediction. The RPD value <1.4 indicates a poor prediction (Chang and Laird, 2002; Zornasa et al., 2008).
2.4 Effect of soil moisture on prediction equations
To evaluate the effect of moisture on the accuracy of the NIRS prediction formula, the soil samples (n=13) were wetted evenly (up to saturation point/sticky) and the samples were scanned and weighed. Then the samples were kept in air drying at laboratory condition (25oC). Subsequently, each day (for 3 days) the samples was weighed and scanned. The calibration and Chemometrics treatment were done as mentioned above.

3. Results and discussion
3.1 Carbon and nitrogen content
There were large variations in soil carbon and nitrogen content which reflects the different vegetation cover under different climatic regions from where the samples originate. Soil carbon and nitrogen content ranged from 0.01 to 12.24 % and 0.01 to 1.03% respectively (Table 1). Table 1 shows the mean, minimum, maximum and calibration statistics of the data set of soils used for this study. The higher carbon and nitrogen content were observed in the dry and moist sub humid region soils as compared to the semi arid region soils. Among different types of vegetation cover in dry and moist sub humid region, soils under mixed cover stored higher content of carbon and nitrogen than in other types of vegetation cover in the same climatic regions. Soil moisture and temperature are the two critical factors influencing the storage of carbon (Jobbagy and Jackson, 2000; Dinakaran et al., 2014). In general, SOC storage increases while increasing precipitation, decreasing temperature and decreasing evapotranspiration (Jobbagy and Jackson, 2000). The soils in semi arid regions showed the lowest content of carbon and nitrogen. Obviously the warm temperature, the sparse nature of vegetation cover and lower rate of litter production would ultimately lead to lower carbon storage in soils.
3.2 NIRS spectra
The NIR spectra of all soil samples were similar in appearance amidst the fluctuations in the absorbance values (Fig. 1). This study observed the prominent peaks, at 1400nm, 1900nm, 2200nm and 2300nm (less prominent), in all the collected soil spectra (Fig. 1). Viscarra Rossel and Behrens (2010) reported an outline of important fundamental absorptions in NIR region and discussed in detail about the soil constituents and its band assignments. Thus the band at 1400 nm is usually associated with aliphatic C'H and O'H, while the absorbance band at 1900 nm is related with amide N'H and O'H. In the band near 2200 nm represents the groups such as phenolic O'H, amide N'H, amine N'H and aliphatic C'H (Stenberg et al., 2010). The Log 1/R values are higher in semi arid region soils while it was lowest in dry and moist sub humid soils (Fig. 2). Nevertheless the dry and moist sub humid region soils, stored more amounts of carbon and nitrogen than semi arid region soils. The absorbance values are totally depends upon the major soil constituents such as sand, silt, clay and organic matter. For example, Stenberg et al. (2010) reported that the absorbance, Log 1 /R, value was higher in clayey soil while it was very low in sandy soil. To confirm or test whether the soil texture or organic matter, i.e. humic substances influences the NIR absorbance spectra, the standard humic and fulvic acid samples (IHSS Suwannee river standards) were also scanned and analyzed. The NIR spectra of humic and fulvic acid standard is shown in Figure 3. Compared to soil spectra, the standard humic and fulvic acid spectra have reduced peak near at 1400 nm and prominent peak near at 1900 nm. This comparison may suggest that soil texture, i.e. sand, silt and clay mineral compositions have the major influence on reflectance spectra. Although, owing overtone and combinations of bands, direct quantifications of soil carbon and nitrogen from these NIR spectra is difficult. Though multivariate statistics, i.e. chemometric treatments were applied to discriminate the response of soil carbon and nitrogen from spectral characteristics.
3.3 Prediction of carbon and nitrogen
The multivariate calibration statistics is being widely used to analyze the diffuse reflectance spectra (Viscarra Rossel, 2007; Viscarra Rossel and Lark, 2009; Mouazen et al., 2010; Viscarra Rossel and Behrens, 2010). The present study used the modified partial least square (MPLS) regression method of the Win ISI software. The MPLS is more stable and accurate than the other standard partial least squares algorithm for agriculture applications (Shenk and Westerhaus, 1995). Increasing a wide range of samples in the calibration data set generally causes a reduction in calibration accuracy (Brunet et al., 2007; Peltre et al., 2011). However, in this study the soil carbon and nitrogen content, were in wide range, was successfully predicted (R2= 0.90 for carbon and R2= 0.85 for nitrogen) by the equations developed by NIRS (Table 1 and Figs. 4 & 5). The standard error of prediction (SEP), standard error of prediction corrected for bias SEP (C) and bias for predicting equations of carbon and nitrogen was 0.73, 0.73, 0.04 and 0.07, 0.07, 0.005, respectively for calibration data set (Table 1). Soil carbon and nitrogen were strongly correlated with soil reflectance as often reported in literatures (Chang et al., 2001; Chodak et al., 2004; Zomoza et al. 2008; Peltre et al., 2011). For the validation data set, the root mean square error (RMSE) and ratio performance deviation (RPD) for the validation of predicted equation of carbon and nitrogen was 0.83, 2.06 and 0.01, 6.98 respectively (Table 2 and Fig. 7). RPD is considered as the principal indicator to determine the prediction capacity of calibration models (Chang et al., 2001; Stenberg et al., 2010). According to Nduwamungu et al. (2009) the very reliable, reliable and less reliable model should have R2 >0.9 and RPD >3, R2 0.7-0.9 and RPD 1.75-3 and R2<0.7 and RPD <1.75 respectively. Though in this study the R2 and RPD values for carbon and nitrogen were >0.7 and >2 are clearly indicating the acceptance of this model for prediction of these properties in three climatic region soils of India.
3.4 Effect of moisture on prediction equations
Low quality predictions were observed when dry calibration equations were applied to predict the soil carbon and nitrogen content (Table 3). These results are in accordance with the study of laboratory conditions (Minasny et al. 2009; Nocita et al., 2013; Viscarra Rossel et al., 2006). As reported earlier the prediction model errors, increasing with increasing moisture contents in the soil samples (Minancy et al., 2011; Nocita et al., 2013). Moistened soils have the highest peak at 1400 nm and 1900 nm as compared to air dried samples (Fig. 8). An earlier study reported that the strong absorptions near at 1412 nm and 1908 nm in the spectra of both groups indicate the occurrence of water bound in the interlayer lattice (Bishop et al., 1994). Similarly the absorbance peak near at 1135, 1380, 1455 and 1915 nm could be due to the absorption of water molecules i.e. O-H (Viscarra Rossel and Behrens, 2010; Stenberg et al., 2010). The present study observed the highest peak near at 1400 and 1900nm in moistened soil samples. This peak is reduced gradually when the moisture content in soil samples was decreased (Fig. 8). This clearly indicates that water molecules/moisture in the soil samples affecting the NIR absorbance spectra and affecting the prediction equations. The similar observations were noticed in previous studies (Nocita et al., 2013; Minancy et al., 2011).
4. Conclusion
The present study shows that NIR spectra of three climatic region soils under different types of vegetation cover in India with subject to chemometrics treatment can be used to predict soil carbon and nitrogen content in future. Though the NIRS is cost effective and environmentally friendly technology with no intemperance of chemicals. The efficacy of NIRS on the prediction of soil carbon and nitrogen content in three climatic regions is good. This study proposes that soil texture, i.e. sand, silt and clay mineral compositions have the major influence on NIR reflectance spectra. When dry sample calibrations were applied to the moistened soil samples, the prediction error is increasing manifold. Thus further studies would be recommended to analyse the efficacy of the prediction of other soil properties with different moisture levels in other climatic regions.

Source: Essay UK - http://www.essay.uk.com/free-essays/science/infrared-reflectance-spectroscopy.php



About this resource

This Science essay was submitted to us by a student in order to help you with your studies.


Search our content:


  • Download this page
  • Print this page
  • Search again

  • Word count:

    This page has approximately words.


    Share:


    Cite:

    If you use part of this page in your own work, you need to provide a citation, as follows:

    Essay UK, Infrared Reflectance Spectroscopy (NIRS). Available from: <https://www.essay.uk.com/free-essays/science/infrared-reflectance-spectroscopy.php> [22-10-19].


    More information:

    If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal: