James B. Reeves

Mid-Infrared and Near-Infrared Diffuse Reflectance Spectroscopy for Soil Carbon Measurement

Diffuse reflectance spectroscopy offers a nondestructive means for measurement of C in soils based on the The ability to inventory soil C on landscapes is limited by the reflectance spectra of illuminated soil. Both the NIR ability to rapidly measure soil C. Diffuse reflectance spectroscopic (400–2500 nm) and MIR (2500–25 000 nm) region have analysis in the near-infrared (NIR, 400–2500 nm) and mid-infrared been investigated for utility in quantifying soil C (Dalal (MIR, 2500–25 000 nm) regions provides means for measurement of soil C. To assess the utility of spectroscopy for soil C analysis, we and Henry, 1986; Meyer, 1989; Janik et al., 1998; Reeves compared the ability to obtain information from these spectral regions et al., 1999; McCarty and Reeves, 2000; Reeves et al., to quantify total, organic, and inorganic C in samples representing 14 2001). The characteristics of spectra obtained in these soil series collected over a large region in the west central United regions varies markedly, with the MIR region domiStates. The soils temperature regimes ranged from thermic to frigid nated by intense vibration fundamentals, whereas the and the soil moisture regimes from udic to aridic. The soils ranged NIR region is dominated by much weaker and broader considerably in organic (0.23–98 g C kg 1 ) and inorganic C content signals from vibration overtones and combination bands. (0.0–65.4 g CO3-C kg 1 ). These soil samples were analyzed with and These divergent spectral characteristics may be exwithout an acid treatment for removal of CO3. Both spectral regions pected to have substantial influence on the ability to contained substantial information on organic and inorganic C in soils obtain quantitative information from spectral data. studied and MIR analysis substantially outperformed NIR. The supeOver the last two decades, NIR spectroscopy (NIRS) rior performance of the MIR region likely reflects higher quality of has developed as a major tool for quantitative determiinformation for soil C in this region. The spectral signature of inorganic nations of components within often complex organic C was very strong relative to soil organic C. The presence of CO3 matrices whereas MIR spectroscopy (MIRS) has been reduced ability to quantify organic C using MIR as indicated by improved ability to measure organic C in acidified soil samples. The used mainly in research for qualitative analysis involving ability of MIR spectroscopy to quantify C in diverse soils collected spectral interpretation of chemical structures. The main over a large geographic region indicated that regional calibrations reason for the exclusion of MIRS in quantitative analysis are feasible. has been the belief that quantitative analysis using the MIR region required KBr dilution because of the strong absorptions present (Perkins, 1993; Olinger and Griffiths, 1993a, 1993b). The strength of these absorptions I CO2 content of the atmosphere from ancan lead to spectral distortions and nonlinearities (Culthropogenic sources has stimulated research to assess ler,1993), and could make quantitative analysis difficult the role of terrestrial ecosystems in the global C cycle. or impossible in undiluted samples. Recent work, howThe terrestrial biosphere is an important component of ever, with a number of sample matrices including food the global C budget, but estimates of C sequestration (Downey et al., 1997; Kemsley et al., 1996; Reeves and in terrestrial ecosystems are partly constrained by the Zapf, 1998), forage (Reeves, 1994), and soil (Janik and limited ability to assess dynamics in soil C storage. AgSkjemstand, 1995; Janik et al., 1998; Reeves et al., 2001) ricultural croplands have a great potential for sequesterhas demonstrated that good quantitative measurements ing atmospheric C (Lal et al., 1998), but current technolare possible in the MIR region. These reports have ogies for monitoring soil C sequestration in terrestrial demonstrated that quantitative MIRS analysis can be ecosystems are not cost effective, or they depend on performed on neat (as is) samples with good accuracy. intensive methods. Recent work has demonstrated good ability to establish local (within-field) NIRS and MIRS calibrations for G.W. McCarty and J.B. Reeves, Environmental Quality Laboratory, soil C (Reeves et al., 1999; McCarty and Reeves, 2000; Building 007 Room 201, BARC-West, Beltsville, MD 20705; V.B. Reeves et al., 2001). The diversity of samples included Reeves III, FDA, Rockville, MD; R.F. Follett, USDA-ARS Fort Collins, CO; and J.M. Kimble, USDA-NRCS Lincoln, NE. Received Abbreviations: MIR, mid-infrared; MIRS, MIR spectroscopy; NIR, 4 Jan. 2001. *Corresponding author (mccartyg@ba.ars.usda.gov). near-infrared; NIRS, NIR spectroscopy; PLS, partial least squares; RMSD, root mean squared deviation; SD, standard deviation. Published in Soil Sci. Soc. Am. J. 66:640–646 (2002). MCCARTY ET AL.: INFRARED DIFFUSE REFLECTANCE SPECTROSCOPY 641 Fig. 1. Geographic location of the 14 sampling sites within the west central United States. of soil carbonates involved addition of 100 mL of 0.33 M in these evaluations was limited to a few agricultural H3PO4 to 5 to 6 g of soil and shaking for 1 h. The procedure fields, and a question remained concerning the ability was repeated until the pH of the soil solution remained within to establish broader calibrations across diverse soil types. 0.2 pH unit of that of the original acid solution (Follett et al., The purpose of this study was to compare the abilities 1997; Follett and Pruessner, 2000). These acidified soil samples of MIRS and NIRS to measure total, organic, and inorwere oven dried at 60 C, ground to pass a 180m screen ganic C in a highly diverse set of soils and to assess opening, and analyzed for C by dry combustion. Follett and feasibility of establishing regional diffuse reflectance Pruessner (2000) reported that acidification removed soil inorcalibrations for soil C. ganic C (carbonates), but little or no organic C. However, they did caution that for some soils, acidification may remove neutral sugars and possibly other soluble organic compounds MATERIALS AND METHODS and the significance of this influence needs further investigation. Soil Collection and Conventional Analyses The 273 samples used in this study were soil profile samples Infrared Spectroscopy collected as described by Follett et al. (2001) from 14 geographically diverse locations in the central United States (Fig. Samples were scanned in the MIR from 4000 to 400 cm 1 1). Soil temperature regimes ranged from thermic to frigid (2500–25 000 nm) at 4 cm 1 resolution with 64 coadded scans and soil moisture regimes from udic to aridic. From each per spectra, on a DigiLab FTS-60 Fourier transform spectromlocation, the soil samples were collected from adjacent parcels eter (Bio-Rad, Randolph, MA) equipped with standard DRIFT of land under crop production, native vegetation (never cultioptics under purge and with a custom fabricated sample transvated), and conservation reserve program (CRP) manageport which allowed a 50 by 2 mm sample to be scanned ment. The soils were sampled to a depth of 200 cm by genetic (Reeves, 1996). Samples of ground soil were placed in the horizons with the surface layer sampled at 0 to 5, 5 to 10, and sample cell without sample dilution and no precautions were 10 to 25 cm (bottom of the Ap for cultivated soils). Before used to avoid specular reflection. Log-transformed reflectance analyses, soil samples were air dried, mixed, sieved, and data was used in analysis. Near infrared spectra were obtained ground by a roller mill (180m mesh size). Soil C analyses using a NIRSystems model 6500 scanning monochromator were performed by dry combustion (1500 C) on a Carlo Erba (Foss-NIRSystems, Silver Spring, MD). Samples were scanned C/N analyzer (Haake Buchler Instruments Inc., Saddle Brook, from 1100 to 2498 nm (PbS detector) using a rotating cup. NJ ). Total soil C was determined on unamended soil samples Data were collected every 2 nm (700 data points per spectra) and organic soil C was determined on acidified soil samples. at a resolution of 10 nm. Inorganic soil C was determined by difference between total and organic soil C. The acidification procedure for removal Statistical Analysis Descriptive statistics on soil properties were performed us1 Trade and company names are included for the benefit of the ing SAS data analysis software (SAS, 1988),and analyses of reader and do not imply endorsement or preferential treatment of the product by the authors or the USDA. NIRS and MIRS spectral were performed by Partial least 642 SOIL SCI. SOC. AM. J., VOL. 66, MARCH–APRIL 2002 Table 1. Location, soil series, texture, and classification of soils studied. Location Map symbol† Soil series Texture Taxonomic classification Akron, CO COS Weld silt loam Fine-loamy, smectitic, mesic Aridic Argiustolls Indianola, IA IAS Macksburg silty clay loam Fine, smectitic, mesic, Aquic Argiudolls Dorothy, MN DOS Radium loamy sand Sandy, mixed, frigid, Oxyaquic Hapludolls Glencoe, MN GCS Nicollet clay loam Fine-loamy, mixed, superactive, mesic Aquic Hapludolls Roseau, MN ROS Percy loam Coarse-loamy, mixed, superactive, frigid Typic Calciaquolls Columbia, MO MOS Mexico silt loam Fine, smectitic, mesic Aeric Vertic Epiaqualfs Sidney, MT MTS Bryant loam Fine-silty, mixed, superactive, frigid Typic Haplustolls Lincoln, NE NES Crete silt loam Fine, smectitic, mesic Pachic Argiustolls Mandan, ND MDS Farnuf loam Fine-loamy, mixed superactive, frigid Typic Argiustolls Medina, ND MES Barnes loam Fine-loamy, mixed, superactive, frigid Calcic Hapludolls Boley, OK BOS Stephenville loamy fine sand Fine-loamy, siliceous, active, thermic Ultic Haplustalfs Vinson, OK VIS Madge loam Fine-loamy, mixed, superactive, thermic Typic Argiustolls Bushland, TX BLS Pullman clay loam Fine, mixed, superactive, thermic Torrertic Paleustolls Dalhart, TX DHS Dallam fine sandy loam Fine-loamy, mixed, mesic Ari

Near infrared reflectance spectroscopy for the analysis of agricultural soils

The objective of this work was to investigate the usefulness of near infrared (NIR) reflectance spectroscopy in determining: (i) various constituents (total N, total C, active N, biomass and mineralisable N, and pH), (ii) parameters (soil source, depth from which sample was obtained, type of tillage used) and (iii) rate of application of NH4NO3 fertiliser) of low organic matter soils. A NIRSystems model 6250 spectrometer was used to scan soil samples (n = 179) obtained from experimental plots at two locations with three replicate plots under plow and no till practices at each location with three rates of NH4NO3 for each plot (2 × 3 × 2 × 3). For each of these, samples were taken from five depths for a total of 2 × 3 × 2 × 3 × 5 or 180 samples (one sample lost). The results demonstrated that NIR reflectance spectroscopy can be successfully used to determine some compositional parameters of low organic matter soils (particularly total C and total N). It is also apparent that for non-biological parameters (excluding soil type as reflected by source) such as the depth from which the sample was obtained, the rate of application of NH4NO3 fertiliser and the form of tillage used, that NIR reflectance spectroscopy is not very useful, unless a very limited set of samples is used (i.e. single tillage and location). For other determinations, such as pH, biomass N and active N, the results may be useful depending on the exact needs in question. Finally, from the results presented here, NIR reflectance spectroscopy was not successful in determining soil N mineralisable in 21 days.

The potential of diffuse reflectance spectroscopy for the determination of carbon inventories in soils

Investigations have shown that near- and mid-infrared reflectance spectroscopy can accurately determine organic-C in soil. Efforts have also demonstrated that both can differentiate between organic and inorganic-C in soils, but the mid-infrared produces more accurate calibrations. Nevertheless, the greatest benefit would come with in situ determinations where factors such as particle size, sample heterogeneity and moisture can be important. While the variations in large (> 20 mesh) particle size can adversely effect calibration accuracy, efforts have demonstrated that the scanning of larger amounts of sample can overcome this, but the effects of moisture have not been fully explored. While under in situ conditions C distribution and sample heterogeneity are a problem for any analytical method, the rapid analysis possible with spectroscopic techniques will allow many more samples to be analyzed. In conclusion, near- and mid-infrared spectroscopy have great potential for providing the C values needed for C sequestration studies.

COMPARISON OF NEAR INFRARED AND MID INFRARED DIFFUSE REFLECTANCE SPECTROSCOPY FOR FIELD-SCALE MEASUREMENT OF SOIL FERTILITY PARAMETERS

Data-intensive technologies such as precision agriculture require new approaches for acquisition of soil data on landscapes. We compared the ability of near infrared (NIR; 400-2500 nm) and mid infrared (MIR; 2,500-25,000 nm) spectroscopy for field-scale acquisition of soil fertility parameters. Samples were obtained in a grid pattern (25 m spacing) from the surface (0-10 cm) and sub-surface (10-30 cm) samples collected at 272 locations. Samples were analyzed for organic C and total N, texture (clay, silt, and sand), soil pH, and Mehlich I extractable Ca, K, Mg, and P. We found that chemometric analyses NIR and MIR provided good calibrations for organic carbon, total N, and soil texture. To varying degrees of precision, these regions also calibrated for pH and exchangeable Ca, Mg, and K. Exchangeable P did not form useful calibrations in either spectral region. In all cases, MIR calibrations were better than those formed in the NIR region. Test of calibrations based on one-third of the samples was used to predict the remaining samples. This demonstrated the strategy of developing field-scale calibrations for the spectral regions by chemical analysis of a small sub-set of samples for the prediction of large numbers of samples. This approach can be used to accurately map the spatial distribution of soil properties within agricultural landscapes. These studies demonstrate the utility of infrared spectral approaches for generating the spatial soil properties data needed to implement precision agriculture technology.

Effect of stage of maturity of alfalfa and orchardgrass on in situ dry matter and crude protein degradability and amino acid composition

Abstract Four maturities of alfalfa (early bud, early, mid and full bloom) and orchardgrass (vegetative, early head, full head and anthesis) were used to test the effect of stage of maturity on in situ rumen dry matter, crude protein degradation and amino acid composition. Two rumen fistulated Holstein cows were used in a 2 × 2 Latin square design with forage species as treatments. Cows were fed a basal diet consisting of 33% corn silage, 17% alfalfa haylage and 50% concentrate on a dry matter basis. Forages were incubated in situ in polyester bags at eight different times (0–96 h) in the rumen. Mean effective degradabilities for both crude protein (82.3 vs. 74.5%) and dry matter (67.6 vs. 63.4%) were higher for alfalfa than for orchardgrass, respectively. Crude protein degradability decreased from 84.8 to 80.4% in alfalfa and from 78.0 to 69.6% in orchardgrass with increasing stage of maturity. Corresponding changes in effective dry matter degradability with increasing stage of maturity declined from 72.9 to 61.9% in alfalfa and from 68.6 to 56.1% in orchardgrass. Most of the depression in rumen degradability of both crude protein and dry matter with increasing stage of maturity could be explained by increases in the indigestible fraction of the forage as rates of digestion were not readily affected by stage of maturity. Individual amino acids declined with increasing stage of maturity of forage as crude protein declined but there was no change in the relative proportions of individual amino acids due to stage of maturity within forage.

Quantitative analysis of agricultural soils using near infrared reflectance spectroscopy and a fibre-optic probe

The objective of this work was to investigate the usefulness of near infrared (NIR) reflectance spectroscopy in conjunction with a fibre-optic probe for determining various constituents (total N, organic C, active N, biomass and mineralisable N and pH) in agricultural soils. A NIRSystems model 6500 spectrometer equipped with a fibre-optic reflectance probe was used to scan soil samples (n = 180) obtained from experimental plots at two locations with three replicate plots under plow and no till practices at each location with three rates of NH4NO3 for each plot (2 × 3 × 2 × 3 = 36). For each of these, samples were taken from five depths for a total of 2 × 3 × 2 × 3 × 5 or 180 samples. Optimal calibrations were achieved using first derivative spectra, and only data from 1100 to 2300 nm, with every 20 data points averaged from 1900 to 2300 nm. Compared to results achieved using a spinning cup, the use of the probe resulted in more concentration outliers (up to 6% with probe and none with the spinning cup), the reason for which is unknown. Otherwise, the final calibration results were quite similar to those achieved using a spinning cup to obtain spectra, with calibration (outliers removed) R2 and RMSD/Mean of 0.96, 6.6; 0.95, 6.5; 0.88, 15.3; and 0.80, 19.3; for organic C, total N, active N and biomass N, respectively. The R2 and RMSD for pH were 0.80 and 4.5 pH units, respectively. In summary, the work presented here demonstrated that NIR spectroscopy, based on data obtained using a fibre-optic probe, can be successfully used to determine compositional parameters of agricultural soils (particularly organic C and total N). However, there appeared to be a greater problem with outliers when using a fibre-optic probe than was true when obtaining spectra using a spinning cup.

Can Near or Mid‐Infrared Diffuse Reflectance Spectroscopy Be Used to Determine Soil Carbon Pools?

Abstract The objective of this study was to compare mid‐infrared (MIR) an near‐infrared (NIR) spectroscopy (MIRS and NIRS, respectively) not only to measure soil carbon content, but also to measure key soil organic C (SOC) fractions and the δ13C in a highly diverse set of soils while also assessing the feasibility of establishing regional diffuse reflectance calibrations for these fractions. Two hundred and thirty‐seven soil samples were collected from 14 sites in 10 western states (CO, IA, MN, MO, MT, ND, NE, NM, OK, TX). Two subsets of these were examined for a variety of C measures by conventional assays and NIRS and MIRS. Biomass C and N, soil inorganic C (SIC), SOC, total C, identifiable plant material (IPM) (20× magnifying glass), the ratio of SOC to the silt+clay content, and total N were available for 185 samples. Mineral‐associated C fraction, δ13C of the mineral associated C, δ13C of SOC, percentage C in the mineral‐associated C fraction, particulate organic matter, and percentage C in the particulate organic matter were available for 114 samples. NIR spectra (64 co‐added scans) from 400 to 2498 nm (10‐nm resolution with data collected every 2 nm) were obtained using a rotating sample cup and an NIRSystems model 6500 scanning monochromator. MIR diffuse reflectance spectra from 4000 to 400 cm−1 (2500 to 25,000 nm) were obtained on non‐KBr diluted samples using a custom‐made sample transport and a Digilab FTS‐60 Fourier transform spectrometer (4‐cm−1 resolution with 64 co‐added scans). Partial least squares regression was used with a one‐out cross validation to develop calibrations for the various analytes using NIR and MIR spectra. Results demonstrated that accurate calibrations for a wide variety of soil C measures, including measures of δ13C, are feasible using MIR spectra. Similar efforts using NIR spectra indicated that although NIR spectrometers may be capable of scanning larger amounts of samples, the results are generally not as good as achieved using MIR spectra.

Near infrared reflectance spectroscopy for the determination of biological activity in agricultural soils

The objective of this work was to investigate the usefulness of near infrared (NIR) reflectance spectroscopy in determining biological activity in agricultural soils. A Foss-NIRSystems model 6500 spectrometer, equipped with a spinning sample cup module, was used to scan 179 soil samples obtained from experimental plots at two locations with three replicate plots under plough and no-till practices at each location with three rates of NH4NO3 for each plot with samples taken from five depths for a total of 180 samples (one sample lost). Biological activity as measured by four enzymes (dehydrogenase, phosphatase, arylsulfatase and urease) and nitrification potential was determined by conventional methods and NIR reflectance spectroscopy. Investigations showed NIR reflectance spectroscopy to be capable of determining biological activity as reflected by the four enzymes and nitrification potential to at least some degree. With the best R2 in the range of 0.8, the results, while positive, were not as good as found previously for many other components (i.e. total C and N) in the same sample set. Efforts at simple discrimination into high, medium and low activities were not successful, and for the most part, calibrations based on subsets, such as samples from only one location, were not found to be an improvement. Correlation analysis indicated that measures of biologically-active nitrogen might be the basis for these determinations. Finally, while further research will be needed to define clearly the basis for, limitations to and usefulness of NIR reflectance spectroscopy in determining biological activity in soil samples, the results presented indicated that NIR reflectance spectroscopy might be useful for the rapid determination of such activity in cases where extreme accuracy is not required, such as spatial mapping.

The effect of nitrogen fertilizer applied to Euphorbia pulcherrima on the parasitization of Bemisia argentifolii by the parasitoid Encarsia formosa

More wasps of Encarsia formosa Gahan (Hymenoptera: Aphelinidae) were found on fertilized poinsettias, Euphorbia pulcherrima (Willd.) (Euphorbiaceae), than on non‐fertilized plants. Parasitization of Bemisia argentifolii Bellows & Perring (Homoptera: Aleyrodidae) by E. formosa was higher on plants treated with calcium nitrate than with ammonium nitrate or on control plants. In a no‐choice test, host feeding by E. formosa was higher when hosts were on fertilized plants than when hosts were on control plants. The nitrogen content of whitefly pupae reared on plants treated with ammonium nitrate was higher than those on calcium nitrate‐treated plants.

SAS® Partial Least Squares Regression for Analysis of Spectroscopic Data:

The objective was to investigate the potential of SAS® partial least squares (PLS) to perform chemometric analysis of spectroscopic data. As implemented, SAS® (Version 8) can perform PLS regression (PLSR, Type II only), principal components and reduced rank regression. While possessing several algorithms for PLSR, various cross-validation options, the ability to mean centre and variance scale data prior to PLSR analysis or for each cross-validation and various options for determining the number of factors to use, SAS® does not possess any other spectral pre-treatments routinely used in spectroscopy. A program was written using SAS® macro language to implement 1st and 2nd gap derivatives, Savitzky–Golay derivatives and smoothing, the ability to skip or average spectral data points, to correct spectra for scatter correction by either multiplicative scatter correction or standard normal variate correction with or without detrend and, finally, to mean centre all data prior to regression analysis. In addition, an F-test method for factor selection was added. These macros can be implemented alone or in differing combinations or order, and result in a summary report containing results for hundreds or thousands of different data pre-treatments. A second program implements the macros in a fixed order. Results using a set of 67 forage samples, scanned in the near infrared, demonstrated that the same results can be achieved as with commercial chemometrics packages. In conclusion, SAS® PLS, while not possessing all the data pre-treatments of standard chemometric programs, can quickly and conveniently test many different data pre-treatments resulting in a single summary results file.