Within the DODILOG project, ARVALIS is leading a pilot in France to explore the use of near-infrared spectroscopy (NIRS) as a tool for detecting insect infestations in soft wheat.
Promising Technology for Grain Storage Management
This pilot focuses on the use of the PSS 2121 spectrometer (Polytec GmbH), which makes it possible to analyse the chemical composition of grain rapidly and without damaging the samples. Beyond its technical features, this technology has strong potential to significantly transform grain storage practices.
If successfully implemented, NIRS could make it possible to detect infestations at a much earlier stage, even before they become visible or widespread. It could also enable operators to monitor large volumes of grain in real time, directly at reception or during handling operations. Compared to conventional methods, which often rely on manual sampling, this approach would provide more representative and immediate information on the sanitary status of grain batches. In addition, the possibility of integrating sensors into continuous flows, such as conveyor systems, opens the way to systematic and high-frequency monitoring.
For storage organisations, these capabilities could translate into concrete operational benefits. Faster and more reliable detection would improve quality control and help prevent the spread of infestations between batches. It would also facilitate more precise segregation of lots, allowing operators to better manage grain flows and optimise logistics. In the longer term, such tools could contribute to reducing post-harvest losses, limiting unnecessary treatments and strengthening confidence in grain quality, especially in export contexts. Ultimately, NIRS could support a shift from reactive to more proactive pest management strategies.
A first experimental phase was carried out on wheat samples from different harvest years that had been artificially infested with five species of storage insect. These samples were analysed using the spectrometer to collect spectra. The objective was to develop a calibration model capable of distinguishing between the presence and absence of five insect species.
However, the initial results show that the model is not yet sufficiently reliable. The classification model was tested on natural samples from storage organisms. Natural samples were classified almost randomly, indicating that the model could not clearly differentiate between infested and non-infested batches. These results highlight the complexity of transferring spectroscopic approaches from laboratory samples to real-world samples.
Several factors may explain these limitations, including the high variability of natural samples, the weak spectral signal associated with early infestation levels and the limited robustness of the calibration dataset.
Next Steps
Following this first phase, ARVALIS is continuing its work to improve the methodology and the performance of the models. A more detailed analysis of the collected spectra is currently underway to better understand the signals and identify potential improvements. In parallel, additional data acquired using a hyperspectral camera (FX17) are being processed, which could provide more detailed spatial and spectral information.
In the long term, the goal is to validate the improved models under real operating conditions in a port silo used for loading grain.
Looking Ahead
Although the first results do not yet meet expectations, they provide valuable insights into the challenges associated with early insect detection using spectroscopy. This pilot confirms both the complexity of the task and the potential of these technologies. The work carried out within DODILOG will contribute to developing robust and operational solutions capable of supporting grain storage operators across North-West Europe.
Camille Harel, ARVALIS