ASD Goetz Instrument Student Winner Report: Identifying and Exploiting Hyperspectral Reflectance Variance when Different Plant Fertilisers are Applied

Authored by Chris Adams. Chris Adams is a PhD candidate at the Imperial College London (United Kingdom), and a 2017 ASD Goetz Instrument Student Support Program winner. ASD, a brand of Malvern Panalytical, offers two different instrument loan programs which provide ASD instruments free of charge to graduate level students in support of their novel, unconventional and/or fundamental research. The Goetz Instrument Support Program was established over ten years ago to further encourage a creative research environment within the Remote Sensing and Field Spectroscopy world-wide community.  

Identifying and Exploiting Hyperspectral Reflectance Variance when Different Plant Fertilisers are Applied

The hyperspectral reflectance patterns of leaves vary when plants are exposed to stress. An objective of our research is to identify these areas of variation and exploit them to be able to identify plants that have been subjected to certain stresses. To help with this research we were awarded an ASD Goetz Instrument Support Program grant and were graciously lent the FieldSpec^® 4, with a non-destructive plant probe with leaf clip attachment to accurately measure hyperspectral reflectance patterns.

We focused on determining whether we were able to detect small changes in the hyperspectral reflectance pattern using the instrument. We were then able to farther exploit these changes in the reflectance through further processing and artificial intelligence technology to recognize the changes in the plant leaves using multispectral data based on the gold standard hyperspectral taken with the FieldSpec 4.

We wanted to know whether we could detect a difference between the application of different fertilizers on Arabidopsis, the fertilizers were chosen as they each had a single molecule difference and a control of sterile water. The model organism of Arabidopsis was selected due to its short growth cycle. The plants were grown in controlled environment rooms, therefore all differences in the reflectance patterns could be inferred to be due to the type of fertilizer applied.

Figure 1: The Arabidopsis growing in the controlled environment room

The plants were grown for a short period, and then the different fertilizers applied. After further growth, leaves were detached at random from the Arabidopsis in each condition and imaged in midday sunlight in the South of England during late summer.

Figure 2: Image of randomly selected Arabidopsis leaves taken in the midday sun in the South of England

The FieldSpec 4 instrument loaned to us was used to determine the hyperspectral reflectance pattern of the leaves. To accomplish this, the instrument was used in a dark room so that the only light source was the instrument’s accessory plant probe bulb (the instrument had been calibrated against the white reference on the leaf clip).

We did identify differences in the mean average spectra after application of different fertilizers. We hope to relate the changes in intensity at different wavelengths to metabolites and the processing of the chemicals present in the near future; for example, the detection and presence of a sulfate ion may lead to an upregulation of secondary metabolites which are causing the change in the hyperspectral reflectance pattern as seen with the FieldSpec 4.

We further used the data to infer which regions to focus on for future imaging work. Bandpass filters were selected where the variance was present in the spectra between sample groups – this highlighted the variance that could be visibly seen. The images were then further processed and used with machine learning and artificial intelligence approaches. The result of these approaches was that the leaves were able to be sorted into the categories of fertilizer applied.

Figure 3: Mean reflectance spectra of Arabidopsis leaves from the sample group that had only water applied.

We are currently repeating the imaging work conducted in the lab out in a field trial and will use a drone (as shown in Figure 4) with the bandpass filters to image the trial and analyze the work using the same pipeline to determine whether the lab reflects the field.

Figure 4: Drone that will be used in future field trials

ASD offers two different instrument loan programs, providing instruments free of charge to graduate level students in support of their novel and unconventional research.

To learn more about ASD Instrument Student Support Programs, please click below.