Get reliable rangeland science

NEAR INFRARED SPECTROSCOPY FOR THE PREDICTION OF RANGELAND SOIL CHARACTERISTICS.
Author
Tolleson, Douglas R.
Boutton, Thomas
Walker, John W.
Mushinski, Ryan
Zhou, Yong
Casola, Claudio
Publisher
Society for Range Management
Publication Year
2017
Body

Prescribed fire and grazing are two common rangeland management practices in the Edwards Plateau region of Texas. Rapid, reliable, and cost effective assessment of soil chemical and physical properties will facilitate their application in research and management. An experiment was conducted to determine the ability of near infrared spectroscopy (NIRS) to quantify percent sand, silt, clay, organic carbon (OC), ?13 C, total carbon (TC), ?15 N, total nitrogen (TN), and total phosphorus (TP) on savanna rangelands that were: 1) burned versus unburned, 2) moderately versus heavily grazed, and 3) within oak, juniper, or grass vegetation. Soil samples (N = 36) were collected from the top 10cm and sieved with a 2mm screen, milled to 1mm particle size and composited (approximately 10g) from triplicate samples prior to chemical analyses. Spectra (400-2500nm) were collected via NIRS. Calibrations for each component were developed using multiple partial least squares regression and evaluated using multiple coefficient of determination (RSQ) and standard error of cross validation (SECV). RSQ was > 0.9 for all components except silt (0.73) and sand (0.76). SECV ranged from 0.01 (TP) to 4.36 (clay). A series of �leave one group out� validations were performed for all group combinations; for example, a calibration developed from 27 samples from heavily grazed, moderately grazed and burned treatment groups were used to predict 9 control group samples, etc. The simple coefficient of determination, standard error of prediction, and slope were: OC (0.92 � 0.02; 1.07 � 0.15; 0.90 � 0.12), TC (0.90 � 0.04; 1.11 � 0.23; 0.96 � 0.13), and TN (0.87 � 0.03; 0.09 � 0.02; 1.08 � 0.1 4). �The effectiveness of these validations for other components varied. With larger and more diverse calibrations, NIRS should be capable of predicting important chemical and physical properties in rangeland soils with different land use histories.

Language
English
Resource Type
Text
Document Type
Conference Proceedings
Conference Name
SRM St. George, UT
Collection
SRM Annual Meeting Abstracts