Silage & Feed

Author

RialtoLabs - Lishia Teh

Published

Oct 2025

Silage

We compared results from the NeoSpectra NIR device (Model: Ingot Grass Silage and Big Bale Silage; Accessory: Rotator) with those from the IAS laboratory method to assess how closely the two approaches agree in measuring key silage quality indicators.

  • Overall Performance: Good to excellent agreements were observed between the NeoSpectra and IAS laboratory results for 10 of the 11 silage quality parameters.

  • Inter-Device Consistency: Results from the two NeoSpectra devices were consistent across most parameters, with higher variation noted for VFA.

  • Parameter of Concern (Ammonia-N): NH₃‑N measurements showed poor correlation with IAS data, and bias correction was ineffective. This inconsistency may be partially attributable to variability in the IAS laboratory results, as earlier tests showed a strong correlation with FBA lab (r = 0.85).

  • Recommendations:

    • Apply bias correction for pH, VFA, Ash, NDF, Protein

    • Continue Monitoring Ammonia-N with lab results to determine if a reliable trend emerges over time

NeoSpectra vs IAS
Silage Quality Measurements
Measurement Correlation1 Agreement Bias2 Bias Correction
Dry Matter (%) 0.97 Excellent -1.88 No
Lactic Acid (g/kg) 0.93 Excellent -8.82 No
pH (%) 0.91 Excellent 0.26 Required
VFA (g/kg) 0.90 Excellent 10.52 Required
Oil (%) 0.88 Excellent -0.24 No
Ash (%) 0.76 Good 1.03 Required
WSC (%) 0.73 Good 0.76 No
NDF (%) 0.63 Good 8.41 Required
D Value (%) 0.60 Good 0.39 No
Protein (%) 0.60 Good 1.49 Required
NH3 - N (%) -0.52 Poor 0.26 No
1 1.0 = perfect agreement, 0 = no relationship
2 Systematic difference between methods

Dry Matter (%)

Recommendation: No correction needed

  • Correlation: 0.969 → excellent agreement
  • Bias: -1.881 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 2.886
    • Bias-corrected: 3.292 (-14.1% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 30.5 31.6 31.3 28.9 −2.37
2 DO_B1 6 53.6 56.7 55.7 53.8 −1.88
3 DO_B2 6 32.2 33.5 32.9 36.3 3.42
4 DQ_2C 6 36.8 41.2 39.5 41.1 1.56
5 EM_1C 6 19.5 21.3 20.5 18.6 −1.89
6 EM_2C 6 21.5 23.4 22.6 22.0 −0.581
7 EM_3C 6 22.2 23.4 22.8 20.1 −2.74
8 JN_2C 6 41.6 44.7 43.2 43.8 0.614
9 JN_JT3F_2C 6 25.1 28.6 26.7 23.5 −3.23
10 JN_JT3F_3C 6 36.0 41.1 38.8 44.2 5.36
11 JN_P_2C 6 34.3 37.4 36.1 31.9 −4.17
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

Lactic Acid (g/kg)

Recommendation: No correction needed

  • Correlation: 0.927 → excellent agreement
  • Bias: -8.825 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 16.28
    • Bias-corrected: 16.542 (-1.6% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 54.1 66.5 60.0 49.1 −10.9
2 DO_B1 6 9.32 26.8 15.3 6.50 −8.83
3 DO_B2 6 39.5 48.6 43.0 26.2 −16.8
4 DQ_2C 6 17.9 25.9 22.5 15.8 −6.70
5 EM_1C 6 90.5 106 96.7 133 35.9
6 EM_2C 6 81.4 90.6 84.7 94.4 9.73
7 EM_3C 6 78.4 87.8 82.0 60.3 −21.7
8 JN_2C 6 44.7 63.0 51.9 36.7 −15.2
9 JN_JT3F_2C 6 67.2 71.6 69.8 65.6 −4.17
10 JN_JT3F_3C 6 22.7 34.6 28.3 24.9 −3.41
11 JN_P_2C 6 43.3 47.3 45.2 28.6 −16.6
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

pH (%)

Recommendation: Apply bias correction

  • Correlation: 0.906 → excellent agreement
  • Bias: 0.261 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 0.387
    • Bias-corrected: 0.202 (47.6% improvement)
sampleName
NeoSpectra
IAS diff2 NeoSpectra_corrected
n1 min max Average
1 BK 6 3.89 4.01 3.94 4.10 0.155 4.21
2 DO_B1 6 4.62 4.84 4.73 5.30 0.572 4.99
3 DO_B2 6 4.24 4.35 4.29 4.50 0.211 4.55
4 DQ_2C 6 4.26 4.42 4.38 4.60 0.223 4.64
5 EM_1C 6 3.57 3.76 3.68 4.00 0.318 3.94
6 EM_2C 6 3.81 3.89 3.86 4.10 0.237 4.12
7 EM_3C 6 3.74 3.89 3.83 4.30 0.470 4.09
8 JN_2C 6 3.98 4.28 4.14 4.60 0.457 4.40
9 JN_JT3F_2C 6 3.94 4.09 4.02 4.20 0.179 4.28
10 JN_JT3F_3C 6 4.17 4.41 4.29 5.00 0.715 4.55
11 JN_P_2C 6 4.06 4.22 4.14 4.40 0.261 4.40
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

VFA (g/kg)

Recommendation: Apply bias correction

  • Correlation: 0.903 → excellent agreement
  • Bias: 10.518 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 10.608
    • Bias-corrected: 7.93 (25.2% improvement)
sampleName
NeoSpectra
IAS diff2 NeoSpectra_corrected
n1 min max Average
1 BK 6 7.45 21.3 14.4 24.9 10.5 24.9
2 DO_B1 6 0 0 0 12.1 12.1 10.5
3 DO_B2 6 0.332 15.9 9.37 15.9 6.53 19.9
4 DQ_2C 6 0 16.6 7.77 22.2 14.4 18.3
5 EM_1C 6 25.2 44.7 34.6 38.3 3.74 45.1
6 EM_2C 6 33.4 49.4 39.7 29.1 −10.6 50.3
7 EM_3C 6 28.8 40.6 34.8 39.0 4.20 45.3
8 JN_2C 6 0 2.28 0.657 15.3 14.6 11.2
9 JN_JT3F_2C 6 12.3 30.2 21.4 26.5 5.06 32.0
10 JN_JT3F_3C 6 0 8.35 4.06 16.3 12.2 14.6
11 JN_P_2C 6 0 19.0 8.63 22.6 14.0 19.2
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

Oil (%)

Recommendation: No correction needed

  • Correlation: 0.875 → excellent agreement
  • Bias: -0.245 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 0.414
    • Bias-corrected: 0.344 (16.9% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 3.32 3.49 3.40 2.70 −0.699
2 DO_B1 6 3.61 3.87 3.75 3.50 −0.245
3 DO_B2 6 3.63 3.76 3.70 3.70 0.00374
4 DQ_2C 6 2.85 3.30 3.02 2.30 −0.717
5 EM_1C 6 3.83 3.99 3.93 3.40 −0.530
6 EM_2C 6 3.77 3.90 3.83 4.10 0.270
7 EM_3C 6 3.67 3.76 3.71 3.80 0.0863
8 JN_2C 6 3.61 3.69 3.65 3.40 −0.245
9 JN_JT3F_2C 6 3.68 3.87 3.77 3.80 0.0321
10 JN_JT3F_3C 6 3.38 3.46 3.42 3.30 −0.115
11 JN_P_2C 6 3.28 3.38 3.32 2.70 −0.619
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

Ash (%)

Recommendation: Apply bias correction

  • Correlation: 0.76 → good agreement
  • Bias: 1.034 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 1.307
    • Bias-corrected: 0.917 (29.8% improvement)
sampleName
NeoSpectra
IAS diff2 NeoSpectra_corrected
n1 min max Average
1 BK 6 7.14 7.75 7.40 9.00 1.60 8.43
2 DO_B1 6 6.99 7.66 7.37 8.40 1.03 8.40
3 DO_B2 6 7.48 7.82 7.67 7.90 0.233 8.70
4 DQ_2C 6 6.37 6.73 6.52 5.90 −0.616 7.55
5 EM_1C 6 8.47 8.90 8.67 11.3 2.63 9.71
6 EM_2C 6 8.32 8.57 8.42 8.80 0.381 9.45
7 EM_3C 6 8.20 8.39 8.29 10.0 1.71 9.32
8 JN_2C 6 7.24 7.64 7.49 8.30 0.814 8.52
9 JN_JT3F_2C 6 8.02 8.42 8.23 8.30 0.0733 9.26
10 JN_JT3F_3C 6 6.30 6.55 6.41 7.70 1.29 7.45
11 JN_P_2C 6 6.67 6.95 6.85 8.40 1.55 7.88
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

WSC (%)

Recommendation: No correction needed

  • Correlation: 0.728 → good agreement
  • Bias: 0.756 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 1.605
    • Bias-corrected: 1.389 (13.4% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 2.52 2.90 2.68 1.70 −0.976
2 DO_B1 6 5.76 6.02 5.90 7.50 1.60
3 DO_B2 6 3.10 3.51 3.32 3.50 0.184
4 DQ_2C 6 4.68 5.16 4.97 4.70 −0.267
5 EM_1C 6 0.190 0.758 0.519 4.10 3.58
6 EM_2C 6 0.930 1.30 1.12 3.30 2.18
7 EM_3C 6 0.999 1.25 1.11 1.20 0.0892
8 JN_2C 6 4.25 4.75 4.50 6.60 2.10
9 JN_JT3F_2C 6 1.80 2.19 1.94 2.70 0.756
10 JN_JT3F_3C 6 3.71 4.52 4.13 5.50 1.37
11 JN_P_2C 6 3.57 3.83 3.69 3.20 −0.491
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

NDF (%)

Recommendation: Apply bias correction

  • Correlation: 0.634 → good agreement
  • Bias: 8.406 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 8.537
    • Bias-corrected: 4.021 (52.9% improvement)
sampleName
NeoSpectra
IAS diff2 NeoSpectra_corrected
n1 min max Average
1 BK 6 49.4 52.3 51.1 62.1 11.0 59.5
2 DO_B1 6 44.1 46.9 45.5 51.4 5.94 53.9
3 DO_B2 6 47.6 49.8 48.6 51.8 3.23 57.0
4 DQ_2C 6 51.2 54.8 52.7 62.9 10.2 61.1
5 EM_1C 6 39.9 41.5 40.8 49.2 8.41 49.2
6 EM_2C 6 41.4 43.5 42.3 52.7 10.4 50.7
7 EM_3C 6 43.0 44.5 43.8 57.4 13.6 52.2
8 JN_2C 6 46.5 49.0 47.2 50.1 2.86 55.6
9 JN_JT3F_2C 6 44.1 46.4 44.9 55.8 10.9 53.3
10 JN_JT3F_3C 6 53.1 56.2 54.1 55.8 1.73 62.5
11 JN_P_2C 6 51.7 54.9 52.6 58.6 5.96 61.0
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

D Value (%)

Recommendation: No correction needed

  • Correlation: 0.605 → good agreement
  • Bias: 0.387 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 3.446
    • Bias-corrected: 3.675 (-6.6% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 59.3 62.2 60.6 61.0 0.387
2 DO_B1 6 63.6 66.8 64.3 70.0 5.66
3 DO_B2 6 62.9 65.0 63.8 66.0 2.24
4 DQ_2C 6 59.7 60.9 60.1 54.0 −6.11
5 EM_1C 6 69.7 71.6 70.7 66.0 −4.71
6 EM_2C 6 68.4 70.6 69.8 66.0 −3.75
7 EM_3C 6 66.4 68.5 67.4 65.0 −2.42
8 JN_2C 6 64.3 65.6 64.7 67.0 2.33
9 JN_JT3F_2C 6 65.4 66.4 66.0 65.0 −0.993
10 JN_JT3F_3C 6 58.1 60.5 59.8 62.0 2.22
11 JN_P_2C 6 59.7 61.0 60.4 62.0 1.61
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

Protein (%)

Recommendation: Apply bias correction

  • Correlation: 0.602 → good agreement
  • Bias: 1.487 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 2.705
    • Bias-corrected: 2.156 (20.3% improvement)
sampleName
NeoSpectra
IAS diff2 NeoSpectra_corrected
n1 min max Average
1 BK 6 11.7 12.9 12.2 14.7 2.55 13.6
2 DO_B1 6 14.0 15.1 14.7 15.8 1.10 16.2
3 DO_B2 6 13.4 14.1 13.8 14.5 0.727 15.3
4 DQ_2C 6 10.8 11.6 11.2 8.50 −2.68 12.7
5 EM_1C 6 14.1 14.8 14.5 19.4 4.92 16.0
6 EM_2C 6 13.9 14.4 14.1 15.6 1.49 15.6
7 EM_3C 6 13.4 14.1 13.7 16.8 3.06 15.2
8 JN_2C 6 13.2 14.1 13.8 13.8 0.0482 15.2
9 JN_JT3F_2C 6 13.7 14.4 14.0 15.3 1.27 15.5
10 JN_JT3F_3C 6 11.3 11.8 11.6 16.2 4.64 13.0
11 JN_P_2C 6 11.2 11.7 11.5 14.0 2.49 13.0
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

NH3 - N (%)

Recommendation: No correction needed

  • Correlation: -0.525 → poor agreement
  • Bias: 0.262 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 2.172
    • Bias-corrected: 2.194 (-1% improvement)
sampleName
NeoSpectra
IAS diff2
n1 min max Average
1 BK 6 3.29 4.92 4.10 6.10 2.00
2 DO_B1 6 4.10 6.16 5.25 1.80 −3.45
3 DO_B2 6 4.31 5.22 4.87 3.80 −1.07
4 DQ_2C 6 1.62 3.55 2.49 7.10 4.61
5 EM_1C 6 2.99 4.23 3.75 3.90 0.149
6 EM_2C 6 3.59 4.93 4.14 4.40 0.262
7 EM_3C 6 3.32 4.01 3.68 5.10 1.42
8 JN_2C 6 4.28 5.16 4.65 4.60 −0.0535
9 JN_JT3F_2C 6 3.91 4.93 4.29 6.10 1.81
10 JN_JT3F_3C 6 2.73 3.16 2.98 3.00 0.0222
11 JN_P_2C 6 2.76 3.55 3.24 6.10 2.86
1 n: number of scans from two devices
2 diff = IAS - Average (Neospectra)

Feed

We compared results from the NeoSpectra NIR device (Model: INGOT Dairy Feed; Accessory: Rotator) with the expected feed nutrient.

  • Performance: Three of the four parameters showed a high correlation. The exception was oil content, but this result is considered preliminary because of the small sample size (n = 4–5).

  • Recommendations:

    • Apply bias correction for protein and oil.

    • Collect additional data to validate these initial trends.

NeoSpectra vs Expected
Feed Quality Measurements
Measurement Correlation1 Agreement Bias2 Bias Correction
Ash (%) 0.99 Excellent 0.18 No
Protein (%) 0.85 Excellent -2.93 Required
Fibre (%) 0.63 Good -1.14 No
Oil (EE) (%) -0.76 Poor 1.11 Required
1 1.0 = perfect agreement, 0 = no relationship
2 Systematic difference between methods

Ash (%)

Recommendation: No correction needed

  • Correlation: 0.99 → excellent agreement
  • Bias: 0.184 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 0.676
    • Bias-corrected: 0.694 (-2.8% improvement)
sampleName
NeoSpectra
Expected diff2
n1 min max Average
1 5-way Mix 3 5.67 6.16 5.87 6.00 0.126
2 Baby Calf 3 5.29 6.06 5.66 5.90 0.242
3 Bull Power 4 5.27 5.72 5.47 5.20 −0.266
4 Roland 5 6.33 7.02 6.60 7.90 1.30
1 n: number of scans from two devices
2 diff = Expected - Average (Neospectra)

Protein (%)

Recommendation: Apply bias correction

  • Correlation: 0.848 → excellent agreement
  • Bias: -2.934 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 2.801
    • Bias-corrected: 0.96 (65.7% improvement)
sampleName
NeoSpectra
Expected diff2 NeoSpectra_corrected
n1 min max Average
1 5-way Mix 3 14.1 14.9 14.5 12.0 −2.46 11.5
2 Baby Calf 3 15.9 18.5 17.2 16.0 −1.23 14.3
3 Bull Power 4 15.4 16.1 15.7 12.0 −3.67 12.7
4 Dairy 14 4 17.0 17.2 17.1 14.0 −3.10 14.2
5 Roland 5 16.8 17.1 16.9 14.0 −2.93 14.0
1 n: number of scans from two devices
2 diff = Expected - Average (Neospectra)

Fibre (%)

Recommendation: No correction needed

  • Correlation: 0.626 → good agreement
  • Bias: -1.137 → no significant bias
  • Standard Error of Prediction:
    • Uncorrected: 1.089
    • Bias-corrected: 1.172 (-7.6% improvement)
sampleName
NeoSpectra
Expected diff2
n1 min max Average
1 5-way Mix 3 12.5 13.0 12.7 11.6 −1.13
2 Baby Calf 3 10.5 12.0 11.3 10.2 −1.15
3 Bull Power 4 10.3 10.6 10.5 9.30 −1.17
4 Roland 5 9.51 10.1 9.71 10.6 0.893
1 n: number of scans from two devices
2 diff = Expected - Average (Neospectra)

Oil (EE) (%)

Recommendation: Apply bias correction

  • Correlation: -0.76 → poor agreement
  • Bias: 1.113 → significant systematic bias (measurements consistently differ from reference)
  • Standard Error of Prediction:
    • Uncorrected: 1.105
    • Bias-corrected: 0.3 (72.8% improvement)
sampleName
NeoSpectra
Expected diff2 NeoSpectra_corrected
n1 min max Average
1 5-way Mix 3 2.03 2.61 2.41 3.10 0.689 3.52
2 Baby Calf 3 1.72 2.15 1.91 3.30 1.39 3.03
3 Bull Power 4 1.75 2.47 2.08 3.10 1.02 3.19
4 Roland 5 1.87 2.17 2.00 3.20 1.20 3.11
1 n: number of scans from two devices
2 diff = Expected - Average (Neospectra)