Fish meal analysis by near-infrared spectroscopy

Significance of the topic

Near-infrared spectroscopy (NIRS) offers a rapid, reagent-free alternative to conventional wet-chemical assays for routine quality control of fish meal. Fish meal is a nutritionally valuable feed ingredient (rich in protein, omega-3s, minerals), and rapid multiparameter analysis (protein, fat, moisture, ash, fiber) directly supports production control, supplier screening and in-process decisions. Replacing hours-long laboratory workflows with second-scale NIR measurements improves throughput, reduces costs, and enhances laboratory safety.

Objectives and overview of the study

The application note documents development and validation of NIR prediction models for key fish meal parameters using the OMNIS NIR Analyzer Solid. The main goals were to demonstrate feasibility of rapid quantification of protein, fat, moisture, ash and fiber, and to compare expected routine performance (prediction errors) with conventional reference methods.

Methodology

• Sample set: over 1,700 fish meal powder samples from multiple suppliers were measured to capture broad variation in composition and supplier-specific differences.
• Spectral acquisition: samples were measured in reflection mode with a large sample cup; spectra were recorded at five distinct sample positions and averaged to reduce effects of non-homogeneity and improve representativeness.
• Calibration and validation: models were built using the OMNIS Software. An independent external validation set comprising 25% of the data was used to assess prediction performance. Figures of merit reported include R2, standard error of calibration (SEC), standard error of cross-validation (SECV) and standard error of prediction (SEP).
• Reference chemistry: conventional ISO-based wet methods were used as laboratory reference values for model development (Kjeldahl for protein, hydrolysis/extraction for fat, loss on drying for moisture, acid-alkali digestion for fiber, furnace ash for ash).

Used instrumentation

• OMNIS NIR Analyzer Solid (Metrohm) — near-infrared spectrometer for solids and viscous samples; measurement time for solids stated as less than 10 seconds.
• Large cup OMNIS NIR, 100 mm — reflection sample vessel with resistant glass bottom for powders and granulates; used with a large holder that permits defined positioning and rotation.
• OMNIS Software (Quant Development and stand‑alone licensing options) — used for spectral acquisition, model development and validation. A pre-calibration for fish meal (protein, fat, moisture, fiber, ash) is also available.

Main results and discussion

The NIR models showed strong predictive performance for most parameters, with some variation in robustness between analytes:

• Protein: R2 = 0.973; SEC = 1.25% (absolute), SECV = 1.27%, SEP = 1.64%. This indicates very good predictive ability suitable for QC decision-making.
• Fat: R2 = 0.902; SEC = 0.72%, SECV = 0.73%, SEP = 0.80%. Strong performance; low absolute error relative to typical fat ranges in fish meal.
• Moisture: R2 = 0.955; SEC = 0.27%, SECV = 0.28%, SEP = 0.35%. Excellent agreement with reference loss-on-drying and very low prediction error.
• Ash: R2 = 0.900; SEC = 0.78%, SECV = 0.80%, SEP = 1.10%. Good predictive quality for mineral (ash) content.
• Fiber: R2 = 0.647; SEC = 0.21%, SECV = 0.22%, SEP = 0.21%. Although prediction error (SEP) is small in absolute terms, the low R2 indicates limited explained variance—likely due to a narrow concentration range, complex matrix interferences or weaker NIR absorption features for fiber in this matrix.

General discussion points:

• High sample numbers and multi-position averaging improved model robustness to sample inhomogeneity and supplier variability.
• Excellent results for protein, fat and moisture support replacing or supplementing wet-chemical assays in routine QC; ash prediction is also suitable for many applications.
• Fiber modeling is comparatively weak—users should treat fiber results with caution or rely on periodic wet-chemistry checks for fiber if critical.
• SEP and SECV values provide realistic expectations for day-to-day QC performance; routine monitoring and occasional recalibration with new reference data is recommended to maintain accuracy over time.

Benefits and practical applications of the method

• Massive time savings: NIR spectra acquisition takes seconds per sample versus several hours for standard ISO wet-chemistry methods (typical reported times: protein 2–4 h; fat 7–10 h; moisture ~5 h; fiber 4–7 h; ash 3–6 h).
• Higher sample throughput enables frequent in-process checks and rapid supplier screening, improving production control and reducing inventory risk.
• No reagents or hazardous chemicals required, lowering operating costs and improving lab safety and environmental footprint.
• Compatibility with automation and integration into production analytics and LIMS—OMNIS supports automation links including titration and other analyses.

Future trends and potential applications

• Model transfer and standardization across sites: improved procedures for transferring calibrations between instruments and sites will broaden deployment in multi-site operations.
• Advanced chemometrics and machine learning: nonlinear regression, domain adaptation and transfer learning can improve predictive power for challenging constituents (e.g., fiber) or when matrix variability increases.
• Online and at-line deployment: integration of NIR probes or conveyor-mounted systems can enable real-time process control in feed production lines.
• Miniaturization and handheld NIR: portable instruments will support field-based supplier checks and rapid receiving inspections at docks and silos.
• Data integration: combined use with digital laboratory platforms, predictive maintenance and supply-chain analytics will create value beyond single-sample QC.

Conclusion

This study demonstrates that NIRS, implemented with the OMNIS NIR Analyzer Solid and supported by a large, representative calibration set, provides rapid and reliable quantification of protein, fat, moisture and ash in fish meal, enabling seconds-scale multiparameter screening for QC. Fiber prediction was less robust and will require additional development or complementary reference testing. Overall, NIR is a practical tool that reduces analysis time from hours to seconds, lowers costs and increases throughput for routine fish meal quality control.

References

• ISO 5983 — Protein determination in animal feeding stuffs (Kjeldahl method).
• ISO 6492 — Determination of fat content (hydrolysis and extraction) in feeding stuffs.
• ISO 6496 — Determination of moisture and other volatile matter in feeding stuffs (loss on drying).
• ISO 6865 — Determination of crude fibre in feeding stuffs (acid-alkali digestion).
• ISO 5984 — Determination of ash in feeding stuffs (incineration in a furnace).