Application of near-infrared spectroscopy for determination of nutrient contents in manure

Ye, Wenyu
Major Professor
Jeffery C. Lorimor
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Agricultural and Biosystems Engineering

111 solid poultry layer, 95 solid poultry broiler litter, 39 swine solid hoop, 72 beef cattle, 85 swine slurry, and 88 swine liquid lagoon manure samples were collected from farms in three states. Samples were analyzed for total nitrogen (TN), ammonia nitrogen (NH3-N), phosphorus (P), and potassium (K) and total solids (TS) by traditional laboratory methods. The strong relationships of TN, NH3-N, P and K with TS were only found for K in swine lagoon manure and for P in swine slurry. Further, poor prediction performance was found in regression lines even with a strong relationship by using the ratio of standard deviation/standard error of prediction (RPD). These manure samples were also used to develop NIRS calibration models by using different mathematical data pretreatments. The R2 of the one-out cross validation for TS, TN, and NH3-N are between 0.80 and 0.97 for all manure samples. The R2 of the one-out cross validation for minerals range from 0.71 to 0.81, 0.50 to 0.78, 0.74 to 0.94, 0.66 to 0.91, 0.73 to 0.91, and 0.70 to 0.90 in poultry solid layer manure, poultry broiler litter, swine solid hoop, beef cattle, swine liquid lagoon, and swine slurry manure samples, respectively. The RPD values indicate NIRS can predict several minerals in certain manure samples, for example, P in beef cattle, swine lagoon, and swine slurry manure samples. Seventy-two beef cattle feedlot manure samples were scanned through the Foss NIRSystems 6500 (master) and the Foss NIRSystems 5000 (slave) instruments. Three standardization methods including cloning, direct standardization, and piece-wise direct standardization were used to transfer the master equations to slave spectra. The 72-sample data set was split into a 30-sample standardization set to generate standardized files and a 42-sample prediction set to test the accuracy of different standardization methods. The standard error of difference (SED) of the standardized slave spectra was much less than the corresponding SED of the unstandardized slave spectra. It was concluded that NIRS can determine TS, VS, TN, NH3-N, and possibly some minerals in manures rapidly and accurately, and the transfer of manure sample calibrations between instruments is possible.