Near-infrared (NIR) spectroscopy has become the industry-standard technique for rapid, non-destructive protein analysis in dairy products, delivering results in seconds without chemical reagents. This article provides a comprehensive technical overview of NIR protein determination across liquid milks, milk powders, and fermented dairy products, covering measurement principles, calibration methodology, applicable standards, and practical applications in quality control and process monitoring.
Technical Overview
Near-infrared spectroscopy is a secondary analytical technique that measures the absorption of electromagnetic radiation in the wavelength region between approximately 780 nm and 2520 nm (12,500 to 4,000 cm⁻¹). In this spectral region, protein molecules produce characteristic absorption signals arising from overtone and combination bands of N–H, C–H, and O–H bond vibrations. For dairy protein analysis, the most important absorption features include the N–H stretch first overtone near 1500–1570 nm, the N–H stretch second overtone near 1030 nm, and the amide-related combination bands between 2050 and 2180 nm.
Because individual NIR absorption bands are broad and heavily overlapped, direct measurement of protein concentration is not possible from a single wavelength. Instead, multivariate chemometric techniques—most commonly partial least squares (PLS) regression—are used to extract quantitative protein information from the full spectrum. The NIR instrument must therefore be calibrated against a primary reference method, typically Kjeldahl or Dumas combustion nitrogen determination, using a representative set of dairy samples spanning the expected concentration range and natural matrix variability.
The principal advantages of NIR protein analysis in dairy applications include analysis times of less than one minute per sample, minimal or no sample preparation, no hazardous chemical reagents, simultaneous determination of fat, lactose, total solids, and moisture alongside protein, and excellent suitability for at-line, in-line, and online process deployment. These characteristics make NIR complementary to classical wet chemistry: reference methods establish traceability, while NIR provides the speed and throughput required for modern dairy production. International guidance for the application of NIR spectrometry in dairy products is provided in ISO 21543, while broader methodological guidance is given in ISO 12099.
Testing Samples
Liquid Dairy Products
Liquid matrices include raw bovine milk collected at farm or reception level, pasteurized whole and skimmed milk, UHT-processed milk, reconstituted milk, flavored and formulated milk beverages, cream, and condensed milk. Liquid samples are typically measured in transflectance or transmission mode using a temperature-controlled cuvette, since both the viscosity and the NIR water absorption background of milk are strongly temperature-dependent. Standard practice is to equilibrate samples to approximately 40 °C and homogenize before measurement to stabilize the fat emulsion and ensure reproducible spectral acquisition.
Milk Powders and Powdered Ingredients
Powdered dairy matrices include whole milk powder, skimmed milk powder, whey powder, whey protein concentrate and isolate, milk protein concentrate, casein and caseinates, lactose-reduced powders, and infant formula base powders. Powders are analyzed by diffuse reflectance, either in rotating cups that average out packing heterogeneity or through glass vials in automated sampling modules. Because particle size and bulk density influence reflectance spectra, calibration sets must include the natural variation in powder morphology produced by different spray-drying and agglomeration conditions.
Fermented and Ripened Dairy Products
Fermented matrices include set and stirred yogurt, drinking yogurt, fermented milk drinks, fresh cheese, quark, and ripened cheese. These products present higher viscosity, larger particle size, and pH-dependent protein hydration, all of which broaden spectral features. Samples are usually warmed, homogenized, and presented in cups with a consistent optical path to reduce scattering artifacts. Ripened cheese requires grating or maceration to a uniform consistency before measurement, and separate calibrations are generally recommended for each product family.
Testing Items
The core analytical deliverable of NIR dairy testing is rapid quantification of protein content, together with related compositional parameters that are determined simultaneously from the same spectrum. Typical measurement items include:
- Total (crude) protein content, expressed as nitrogen multiplied by the dairy conversion factor 6.38, in g/100 g or g/100 mL
- True protein content, obtained when non-protein nitrogen is accounted for through the reference calibration design
- Casein and soluble whey protein fractions, measurable in standardized matrices with dedicated calibrations
- Moisture or total solids, critical for powder products and yield calculations
- Fat and lactose content, determined concurrently for full nutritional characterization
- Milk urea nitrogen as an ancillary parameter in raw milk screening programs
- Protein-related process indicators such as protein standardization compliance and powder blend uniformity
Typical working ranges cover approximately 2.8–4.2 g/100 mL protein in liquid milk, 30–40 g/100 g in whole milk powder, and up to 80–90 g/100 g in whey and milk protein concentrates. Reporting follows the applicable national compositional standards, such as GB 19301 for raw milk and GB 19644 for milk powder, or the corresponding Codex and ISO product standards.
Testing Methods
Reference (Primary) Methods
NIR calibrations are anchored to internationally recognized reference procedures. Total nitrogen in milk and milk products is determined by the Kjeldahl method according to ISO 8968 (Part 1 and related parts) or by Dumas combustion according to ISO 14891. In Chinese regulatory practice, the equivalent procedure is GB 5009.5 for determination of protein in foods. Protein is calculated as nitrogen content multiplied by the factor 6.38 for dairy matrices. The quality of the NIR result can never exceed the quality of the reference data; duplicate or triplicate reference analysis, balanced sample selection, and control of laboratory bias are therefore essential during calibration development.
Instrumentation and Spectral Acquisition
Commercially available platforms include Fourier-transform NIR spectrometers, scanning grating monochromator systems, and diode-array or fixed-filter analyzers designed for high-throughput dairy reception. Liquid samples are measured in thermostatted transmission or transflectance cells with path lengths around 0.5–1.0 mm to manage the intense water absorption. Powders are measured by diffuse reflectance with gold or ceramic internal references. Daily instrument performance verification uses certified check samples and photometric noise standards, and wavelength accuracy is confirmed with polystyrene or rare-earth oxide filters.
Calibration Development and Validation
A robust calibration is built from several hundred representative samples covering seasonal, regional, breed, and processing variation. Spectral pretreatments such as standard normal variate (SNV), multiplicative scatter correction (MSC), and Savitzky–Golay derivatives are applied to reduce scatter and baseline effects. PLS regression models are then validated using independent sample sets, with performance judged against the statistics recommended in ISO 12099 and ISO 21543: coefficient of determination (R²), standard error of prediction (SEP), bias, slope, and the ratio of performance to deviation (RPD). Routine laboratories typically achieve protein prediction errors in the range of 0.03–0.06 g/100 mL for liquid milk and 0.3–0.5 g/100 g for milk powders, adequate for screening, process control, and labeling verification.
Quality Control and Model Maintenance
Continued validity is maintained through control charts on check samples, periodic slope and bias adjustment against new reference data, outlier monitoring using Mahalanobis distance and global-H statistics, and standardized calibration transfer procedures when instruments are replaced. Guidance on proficiency testing and ongoing verification can be aligned with ISO 12099, and accredited third-party laboratories supporting dairy producers routinely participate in interlaboratory comparison schemes to demonstrate equivalence between NIR screening results and reference chemistry.
Applications
NIR protein analysis supports nearly every stage of the dairy value chain, from farm-level milk collection to finished-product release. Major application areas include:
- Raw milk reception and payment testing, where rapid protein, fat, and solids determination underpins fair milk pricing and supplier grading
- Process standardization, in which real-time protein values guide the dosing of skimmed milk powder or milk protein concentrate during fat and protein standardization
- Evaporation and spray-drying control, where inline NIR monitoring of total solids and powder protein improves energy efficiency and end-point accuracy
- Finished-product quality control and nutritional label verification for pasteurized milk, UHT milk, milk powders, and infant formula
- Blend uniformity and batch consistency checks for protein powders and nutritional ingredients
- Screening support for authenticity programs, where unexpected spectral residuals can flag adulteration with non-protein nitrogen sources for confirmatory testing
- Animal nutrition and herd management, using milk urea nitrogen and protein trends to optimize feeding strategies
- Research and product development, enabling high-throughput compositional profiling of pilot-scale and experimental dairy formulations
When results are intended for regulatory release or labeling purposes, laboratories typically operate NIR as a validated screening method backed by periodic confirmatory analysis using Kjeldahl or Dumas reference procedures performed under an ISO/IEC 17025-accredited quality system.
FAQ
How accurate is NIR protein analysis compared with Kjeldahl?
NIR is a secondary method calibrated against Kjeldahl or Dumas reference data. Well-maintained calibrations typically predict milk protein within approximately 0.03–0.06 g/100 mL of the reference value. The technique is highly reliable for screening and process control, but confirmatory reference analysis remains the basis for legal traceability of absolute values.
Is sample preparation required?
Preparation is minimal. Liquid milk is warmed to about 40 °C and gently homogenized before presentation in a cuvette; powders are simply loaded into cups or vials at consistent fill and compaction; cheese is homogenized or grated to a uniform consistency. Temperature control is the single most important preparatory variable for liquids.
Can one calibration be used for all dairy products?
No. Matrix-specific calibrations are strongly recommended because fat content, particle size, hydration state, and pH alter the spectral background. Separate models for liquid milk, milk powder, yogurt, and cheese deliver markedly better accuracy than a single global model, although localized models can be merged using chemometric networking techniques.
How often must NIR calibrations be updated?
Calibrations should be monitored continuously with control charts and updated whenever systematic bias, new seasonal variation, formulation changes, or new suppliers introduce spectral population shifts. Most laboratories perform slope and bias corrections quarterly and full database expansion annually, in line with the maintenance principles of ISO 12099.
Is NIR accepted by regulators for product release?
NIR is widely accepted as a validated in-house method for routine screening and process control, and many screening results are used for internal release decisions. For official compositional compliance and labeling disputes, reference nitrogen methods under ISO 8968, ISO 14891, or GB 5009.5 remain the definitive procedures, often performed by accredited third-party laboratories.
Can NIR detect protein adulteration in dairy products?
NIR can flag many adulteration scenarios, such as addition of nitrogen-rich compounds or substitution of milk solids with non-dairy fillers, through spectral outlier detection. Any suspicious result should be escalated to targeted confirmatory methods, since NIR is fundamentally a quantification tool rather than a trace-contaminant detector.
