Sofia Guerrero Moreno, MSc. Ing.
Avocado Quality (Dry Matter) Assessment with NIR: Beyond the Tool Itself
In fresh avocado systems, one of the persistent challenges is determining fruit quality at harvest, particularly dry matter content.
From a statistical perspective, conventional dry matter evaluation typically relies on destructive sampling, meaning only a small fraction of the fruit population is tested. This is understandable, but it also limits how representative the assessment can be.
Near-Infrared Spectroscopy (NIR) has become an increasingly valuable tool for estimating internal parameters such as dry matter, offering rapid, non-destructive measurement. Its appeal is clear: faster decisions, less waste, and more consistent data — when used correctly.
A common misunderstanding around NIR is the distinction between factors that affect fruit composition and factors that affect measurement accuracy.
Agronomic variables such as location, rootstock, and altitude influence dry matter content itself. However, the reliability of the NIR reading depends largely on physical and operational conditions during measurement. Dust particles (particularly silica dust), poor sensor placement on curved fruit surfaces, and suboptimal contact between the reader and the fruit can cause light scattering and inaccurate readings. Environmental conditions, such as high ambient temperatures, can also affect equipment performance during field use.
All of this assumes that appropriate predictive models and measurement modes are selected (reflectance, interactance, absorbance, and spectral derivatives). A range of modeling approaches are currently used, including PLS, MLR, k-NN, kernel-based methods, principal component-based models, and neural networks.
In practice, predictive performance depends strongly on how well models are adapted to the specific production context. Applying generalized models without local calibration can limit accuracy and lead to inconsistent interpretation. One reason for this lies in wavelength selection during model development. Hyperspectral data does not behave linearly, and multiple absorption features appear within the spectra of a single sample. These spectral peaks and curve shapes can shift depending on local growing conditions, cultivar characteristics, and fruit structure. Models developed under one context may therefore fail to capture the most informative spectral regions in another, reinforcing the need for context-specific calibration.
From a food systems perspective, tools like NIR are not just grading instruments. They are part of a broader shift toward data-informed production, where measurement supports improved harvest timing, quality consistency, and more sustainable decision-making.
Bridging production, measurement, and food quality is an area of growing importance in agri-food science, and technologies like NIR sit right at that interface.
#FoodScience #AgriTech #PostHarvest #FoodQuality #NIR #FoodSystems
29 Janvier 2026 à 20h49
Voir sur LinkedIn →
