DIFFUSE REFLECTANCE SPECTROSCOPY FOR SOIL CHARACTERIZATION

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: HATCH
• Reporting Frequency: Annual
• Accession No.: 0200128
• Project Start Date: Oct 1, 2004
• Project End Date: Sep 30, 2007
• Program Code: [(N/A)]- (N/A)

Recipient Organization
MONTANA STATE UNIVERSITY
(N/A)
BOZEMAN,MT 59717

Performing Department
LAND RESOURCES AND ENVIRONMENTAL SCIENCES

Non Technical Summary
Standard soil characterization techniques are time-consuming and expensive. Statistically rigorous soil mapping, precision agriculture and soil C monitoring and measurement for carbon sequestration all require a more rapid and inexpensive method of soil characterization. The purpose of this study is to develop and refine predictive soil characterization models based on VNIR diffuse reflectance.

Animal Health Component

30%

Research Effort Categories

Basic

40%

Applied

30%

Developmental

30%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
101	0110	2040	40%
101	0110	2061	30%
101	7299	2090	30%

Knowledge Area
101 - Appraisal of Soil Resources;

Subject Of Investigation
0110 - Soil; 7299 - Research equipment and methods, general/other;

Field Of Science
2090 - Statistics, econometrics, and biometrics; 2040 - Mineralogy; 2061 - Pedology;

Keywords

soil mapping

soil characterization

spectroscopy

precision farming

carbon

carbon cycle

sequestrants

infrared reflectance

near infrared radiation

measurement

data processing

soil properties

soils

soil mineralogy

appraisal

wheat

soil types

soil nutrients

soil organic matter

statistical analysis

soil surveys

soil samples

soil classification

calibration

validation

Goals / Objectives
a) Develop statistical models to predict the type and amount of soil C in Montana wheat-growing soils based on Visual and Near-Infrared (VNIR) diffuse reflectance measurements. i) Compare in-situ and lab-based soil VNIR reflectance measurements to determine the feasibility of using VNIR in the field to rapidly assay soil organic and inorganic C. Develop spectral processing techniques to correct for soil moisture effects. ii) Develop statistical models using lab-based VNIR reflectance to estimate the relative abundance of Soil Organic Matter (SOM) functional fractions for soils from Montana and research sites throughout the western states. b) Develop VNIR-based statistical models to comprehensively characterize and classify soil materials, using soil samples from the USDA-NRCS National Soil Survey Center archives.

Project Methods
All spectral reflectance scans for this project will be obtained using an ASD Fieldspec Pro FR spectroradiometer which measures reflectance in the Visible and Near Infrared (VNIR) spectral range (350-2500 nm). Using smoothing splines, 1st derivative reflectance values will be obtained at 10 nm intervals. Several predictive modeling approaches will be compared, including chemometrics (PLS regression), data mining (Treenet), and mixture modeling. The ultimate goal will be to construct robust, regression-based models in R using only spectral information that can be directly and consistently related to soil properties of interest. The precision of diffuse reflectance spectroscopy (DRS) soil carbon will be determined for a range of soil materials from in situ soils to oven-dried and milled samples. By determining the analytical precision of each technique can we make an informed cost-benefit analysis of best way to proceed with future research on the application of diffuse reflectance spectroscopy to carbon monitoring and carbon credits. Traditional laboratory assays will serve to calibrate spectral models of soil organic and inorganic C on a mass basis, with texture also determined to test for particle size effects with total C and N will be determined using a Leco CNS-2000 carbon-nitrogen analyzer. Inorganic C will be determined with the Modified Pressure-Calcimeter method (Sherrod et al., 2002), using apparatus available in the department. Using bleach and acid treatments, as well as sedimentation, the spectral curves of soil constituents (SOC, POM, IC and clays) will be isolated in the laboratory. Mixture modeling will then be employed to construct predictive models of soil composition. Additional soil fractions, obtained from Dr. Johan Six (UC-Davis) and Dr. Cambardella (Soil Tilth Lab, Ames) will also be analyzed with a similar procedure to test the stability of the spectral model for soils across the great plains and western states. A total of 4204 diverse and well-characterized soil samples from 4204 pedons stored in the U.S. Soil Survey archives have been previously scanned using our spectroradiometer. Working with Dr. Shepherd at the World Agroforestry Centre in Nairobi, we will compare chemometric and data mining approaches to empirical modeling. More theoretical approaches like continuum removal and mixture modeling approaches will also be compared to develop more robust characterization models.

Progress 10/01/04 to 09/30/07

Outputs
OUTPUTS: This professor has left the university. PARTICIPANTS: Nothing significant to report during this reporting period. TARGET AUDIENCES: Nothing significant to report during this reporting period. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
Left univeristy without completing project

Publications

• No publications reported this period

Progress 01/01/05 to 12/31/05

Outputs
The graduate student funded on this project took a leave of absence for most of 2005 (her sister started a restaurant in town). Therefore less progress (and expenditures) was made for this year. Preliminary results show that the low levels of soil organic carbon (SOC) in Montana soils limit the applicability of visible and near-infrared (VNIR) for the determination of this parameter-regardless of whether samples are scanned in situ, dried and sieved, or milled. However, we seem to be obtaining acceptable results for inorganic carbon (IC), a constituent with more variability. Given the limitations of VNIR, we have expanded the project to look at mid-infrared (MIR) spectroscopy. All samples (300+) have been milled and scanned (i) neat; (ii) with dilution in KBr; (iii) SOC removed using bleach treatment; and (iv) SOC removed using bleach treatment and diluted in KBr. We are currently constructing models to predict SOC and IC using these MIR scans, and identifying key SOC absorption features by comparing reflectance with and without SOC removal.

Impacts
Diffuse Reflectance Spectroscopy provides a tool for rapid, non-destructive, inexpensive, in situ soil characterization for a variety of applications including site-specific management, watershed hydrologic modeling, and the monitoring of carbon sequestration in agricultural lands. The results of this research could dramatically increase the availability of soil data for environmental modeling.

Publications

• Brown, D.J., Bricklemyer, R.S. and Miller, P.R., 2005. Validation requirements for diffuse reflectance soil characterization models with a case study of VNIR soil C prediction in Montana. Geoderma, 129(3-4): 251-267. Brown, D.J., Shepherd, K.D., Walsh, M.G., Mays, M.D. and Reinsch, T.G., 2006. Global soil characterization with VNIR diffuse reflectance spectroscopy. Geoderma, (in press).

Progress 01/01/04 to 12/31/04

Outputs
We discovered that the spatial structure of calibration and validation sample sets can have a dramatic impact on model construction and validity of statistical results. To construct robust soil carbon models, a larger number of calibration samples are required than previously thought. In response to this discovery, we have constructed a global soil visible and near-infrared spectral library in collaboration with the U.S. Soil Survey program (4,000 samples). In addition to these significant findings and results: all soil coring, VNIR scans, laboratory analysis, and PLS regression modeling as proposed has been completed. Since the project is proceeding ahead of schedule we are expanding the work in the following directions: (i) replicate cores were obtained this summer (2004) with analysis to be completed in 2-3 weeks; (ii) SOC absorption features are being mapping for the VNIR range by comparing reflectance patterns with and without SOM removal; (iii) we are currently examining ways to enhance regional predictions using a global soil-spectral library. A graduate student, supported by the seed grant, is on schedule to complete her M.S. thesis in August, 2005. One paper describing the global soil-spectral library has been accepted pending moderate revisions, and we anticipate that three peer-reviewed papers will be submitted for publication in parallel with and/or subsequent to the completion of this thesis (in addition to the paper cited below).

Impacts
Diffuse Reflectance Spectroscopy provides a tool for rapid, non-destructive, inexpensive, in situ soil characterization for a variety of applications including site-specific management, watershed hydrologic modeling, and the monitoring of carbon sequestration in agricultural lands. The results of this research could dramatically increase the availability of soil data for environmental modeling.

Publications

• Brown, D.J., Bricklemyer, R.S. and Miller, P.R., 2005. Validation requirements for diffuse reflectance soil characterization models with a case study of VNIR soil C prediction in Montana. Geoderma, (in press).