The Art and Science Behind NIR Calibration

Innovation in agricultural testing is about more than just speed; it’s about confidence in every result. Our use of Near Infrared Spectroscopy (NIR) is one example of a technology that we have invested in to allow us to analyse soil and feed samples quickly, and accurately.  

The power of NIR lies not only in speed and efficiency but also in the strength of the calibration behind it. At Hill Labs, that strength comes from decades of data, rigorous validation, and a continuous commitment to accuracy, ensuring farmers and agronomists can make decisions with confidence.  

What exactly is NIR, how does it work and what do we do to provide accurate results for New Zealand soils?   

Understanding NIR  

Near infra-red refers to the region of light just beyond the visible spectrum, between 700 and 3000 nanometres in wavelength. This region corresponds to the energy of molecular vibrations in organic materials. Different molecules absorb NIR light in distinct ways, depending on their chemical structure.  

Organic compounds containing O–H, C–H, N–H, and C–O bonds absorb NIR radiation strongly.  This strong absorbance makes NIR a valuable technique for analysing organic materials such as plants, feed, and soils with high organic matter. In contrast, NIR is less effective for testing inorganic compounds, which do not contain these absorbent functional bonds. 

The resulting absorbance spectra is interpreted using advanced mathematical models (known as chemometrics) to determine the chemical composition of a sample.  

The Hill Labs Approach  

At Hill Labs, we build our calibration models in-house, using samples from New Zealand. Rather than using one large, static model, we apply a local weighted regression approach. This starts with collecting thousands of spectra each representing the light signature of a sample and pairing those with reference chemistry values. By combining these two data sets, we’re able to create robust calibration models that underpin every NIR result we deliver.   

The calibration process is much like any other analytical method. By comparing a wide range of samples with known analyte values and their corresponding spectra, we teach the model how to predict the chemistry of a new sample.  

While this sounds straightforward, building these calibrations is both a science and an art. There are many ways to build a model, but what matters most is how well it performs and how reliably it predicts results for new, real-world samples.  

We have used New Zealand sourced samples for the creation of all our libraries. This is particularly important for soil samples given the unique properties of New Zealand soils.  

From Light to Laboratory Results   

We use a local weighted regression approach. This method searches our vast database to find the several hundred samples most similar to the new sample being analysed. These nearest neighbours are used to build a mini calibration specifically for that sample, with greater weight given to the most similar cases.  

This approach is known as Local Weighted Modelling (LWM) and provides flexibility and precision, ensuring each new sample is interpreted in the most relevant context.  

One of the key challenges with NIR analysis is ensuring that a new sample truly fits within the calibration model. Any powdered sample can be placed in an NIR instrument and will produce a number, but that doesn’t automatically mean the result is meaningful. A valid calibration requires enough variation across both the analyte and the matrix (the sample type) to confirm that the new sample is comparable to those the model was built from.  

For all of our calibrations, we’ve worked to clearly define the matrix, validate it extensively, and confirm that it’s fit-for-purpose. Because every organic material has its own unique spectrum, some samples may still fall outside the calibration range. In those cases, we have robust checks to detect and flag outliers before results are released.   

As with any indirect lab method, NIR analyses may have a small trade-off in accuracy compared to the reference method. Hill Labs has rigorous internal quality processes to minimise this.  

Looking Ahead  

We are committed to providing New Zealand relevant results and continuing this innovative use of NIR for soil testing. Our current NIR capability for soils covers the key soil nitrogen and carbon tests as well as total sulphur, extractable organic sulphur and ASC. We are actively exploring new parameters that can be reliably measured using NIR, expanding our range of fast and accurate testing options available to New Zealand’s agricultural sector.