Developing Calibration Model and Validation of Mobile Near-infrared Spectroscopy to Predict Nutritive Values of Selected Feed Resources in Ethiopia Mulugeta Walelegne1, 3, Fentahun Meheret1, Melkamu Derseh2, Alan Duncan2, 4, Mesfin Dejene3 1Bahir Dar University, College of Agriculture and Environmental Sciences, P.O. Box 5501, Bahir Dar, Ethiopia; 2International Livestock Research Institute (ILRI), Ethiopia, P.O. Box 5689, Addis Ababa, Ethiopia; 3Ethiopian Institute of Agricultural Research, Holeta Agricultural Research Center, P.O. Box .31, Holeta, Ethiopia; 4University of Edinburgh, P.O. Box 3671, Scotland, United Kingdom. The nutrient composition analysis of feeds is an essential prerequisite for animal feeding practices but it is limited by high cost of analytical services and lack of laboratory infrastructures. Hence, this study aimed to develop a calibration model and validation of Tellspec mobile NIRS for predicting the nutritive values of selected feed resources in Ethiopia. A total of 142 oilseed cake and 140 natural pasture hay samples were purposively collected from central Ethiopia. A factorial arrangement of two factors (three numbers of scans and two feed forms) with a completely randomized design and three replications was used for each feed type. The spectra data of oilseed cake and natural pasture hay samples in the form of intact and ground at one, two, and ten scans were collected from the 900-1700 nm range. The raw spectra data were imported to window integrated information system (WINISI) software and calibration models were developed for Dry matter (DM), Nitrogen (N), Neutral Detergent Fiber (NDF), Acid Detergent Fiber (ADF), Acid Detergent Lignin (ADL), and Ash content of oilseed cake and natural pasture hay samples. All calibration and validation statistics data obtained from (WINISI) software were subjected to a general linear model procedure of SAS version 9.0, and analyzed the effect of feed forms and number of scans on the performance of calibrations equations. The result showed that the effect of number of scans, feed forms, and their interactions had a significant (p<0.001) effect on oilseed cake and natural pasture hay calibration and validation statistics. Among the calibration models developed for oilseed cake samples, a robust and excellent model was obtained from the interaction between ground and ten scans with the coefficient of determination of calibration (R2cal) and validation (R2val) values for N (R2cal=0.95; R2val=0.91), NDF (R2cal=0.98; R2val =0.92), ADF (R2cal=0.96; R2val=0.90), ADL (R2cal=0.93; R2val=0.94), IVOMD (R2cal =0.93; R2val=0.86), and Ash (R2cal=0.89; R2val=0.77) and intermediate for DM (R2cal =0.82; R2val=0.78) and Ash (R2cal=0.89; R2val=0.77). The R2cal and R2val obtained from the interaction between intact oilseed cake samples with ten scans for DM (R2cal=0.73; R2val=0.70), N (R2cal=0.94; R2val=0.90), NDF (R2cal=0.97; R2val=0.85), ADF (R2cal=0.92; R2val=0.87), ADL (R2cal=0.93; R2val=0.84), IVOMD (R2cal=0.94; R2val=0.80), and Ash (R2cal=0.79; R2val=0.81). Similarly, among the prediction equation developed for different chemical constituents in natural pasture hay samples, the best model were obtained from the interaction between ground samples and tens scans for the prediction of IVOMD (R2cal 0.89; R2val 0.81), ADL (R2cal =0.91; R2val=0.74), ADF (R2cal=0.85; R2val=0.71), NDF (R2cal=0.85; R2val=0.68),and lowest values were recorded for N (R2cal=0.82; R2val=0.55), Ash (R2cal=0.80;R2val=0.56), and DM (R2cal =0.69; R2val=0.51) were in the order of their predictive ability for unknown samples. Based on the overall evaluation, the accuracy of the developed oilseed cake equation model for prediction of the chemical composition of intact and ground samples by using the Tellspec mobile NIRS instrument was best. In contrast, the current natural pasture hay equation model needs further research work for most of the parameters. In general, the developed prediction equations for oilseed cake and natural pasture hay samples for several parameters can be used as a base and continuous improvement of these models through incorporating several factors should be implemented on a regular basis. Additionally, demonstration of Tellspec as alternative technology for different stake holders should be applied to ensure animal feed quality assurance at on-farm and feed market systems in the livestock sector.