Comparisons of two portable NIR spectrometers in the prediction of the chemical composition of Total Mixed Rations for buffaloes and dairy cows
Author(s)
Date Issued
2024
Type
conferenceObject
Abstract
Portable NIR spectrometers represent valid tools, providing real-time, on-farm prediction of chemical composition of feed and ration to enhance farming practices [1,2].
The aim of this work was to assess the accuracy of portable instruments, compared with standard laboratory analysis (LA), in predicting the chemical composition of Total Mixed Rations (TMR). A total of 180 TMR samples (40 for dairy cows, and 140 for buffaloes) were scanned using two commercial NIRS (NIRS_A and NIRS_B) having similar, but not identical, spectral ranges (min 902 nm, max 1680 nm), and built-in predictive calibrations for dairy cows TMR.
Laboratory analyses were carried out for dry matter (DM), crude proteins (CP), ethereal extract (EE), neutral detergent fiber (aNDF), acid detergent fiber (ADF), crude ash (ASH) and total starch (TS) [3]. To check for the accuracy of both NIRS, predicted values were regressed on LA data. Fitting (R2) and prediction errors (RMSEP) were thus calculated. NIRS and LA data were also submitted to Friedman ANOVA, and Nemenyi method pairwise comparisons was applied (XLSTAT, ver. 25.3.1.0). The study was carried out on the global dataset and then separately for both buffalo and dairy cows TMRs.
Both portable NIRS showed good predictive capacity for DM (R2=0.80 vs. 0.88, RMSEP=2.59% vs. 2.15%, for NIRS_A and NIRS_B, respectively) and fair good performance for ADF (R2=0.71 vs. 0.73, RMSEP=1.61% vs. 1.04%) and EE (R2=0.69 vs. 0.71, RMSEP=0.42% vs. 0.33%). Both NIRS have different, but not excellent, predictive capabilities for the parameters CP (R2=0.69 vs. 0.43, RMSEP=0.74% vs. 0.89%), aNDF (R2=0.54 vs. 0.68, RMSEP=2.98% vs. 1.87%) and TS (R2=0.56 vs. 0.43, RMSEp=2.13% vs. 1.72%).
When the regressions were carried out separately for buffaloes and cows TMRs, both NIRS showed a worsen predictive capacity, especially and surprisingly on TMRs for dairy cows. However, the Friedman ANOVA demonstrates that, except for ASH in dairy cow subset, the predictions from both NIRS differed from the respective LA data.
Such results and the NIRS-dependent differences observed, most probably were dictated by the sub-optimal robustness of the proprietary calibrations and/or by different optical configurations. In the application to this case study, both NIRS showed some critical issues that could be circumvented by extending the calibration sample sets and updating the calibrations. For accurate prediction of the composition of TMRs, updating built-in calibrations of both portable NIRS should be carefully considered.
[1] Evangelista C, Basiricò L, Bernabucci U. 2021. An overview on the use of near infrared spectroscopy (NIRS) on farms for the management of dairy cows. Agriculture 11: 296. doi: 10.3390/agriculture11040296.
[2] Berzaghi P, Serva L, Piombino M, Mirisola M, Benozzo F. 2005. Prediction performances of portable near infrared instruments for a farm forage analysis. Ital.J.Anim.Sci. Vol. 4 (Suppl. 3), 145-147.
[3] AOAC: Association of Official Analytical Chemists, 15th ed.; Washington, DC, USA, 1990; 17th ed.; Washington, DC, USA, 2000 and 18th ed.; Washington, DC, USA, 2005.