Using Hyperspectral Imaging to Assess Salt Stress Tolerance in Crops

Resonon has reliable, well-calibrated sensors, along with great customer service. They are always approachable and helpful.

Dr. Ali Moghimi, University of California, Davis

3,000 km^2
Farmland lost annually to salinity200,000+ km^2
Additional farmland each year with lowered production via salinity

Challenge

Salinity stress occurs when excessive salt builds up in soil and interferes with water absorption and nutrient uptake in crops. Over time, salinity stress leads to decreased crop productivity.

Some regions of the world are facing rapid increases in soil salinity, which is rendering large swaths of farmland unusable. With less farmland, yield potential decreases and food insecurity increases.

Selecting salt-tolerant crop varieties is one promising approach to addressing this problem; however, there needs to be a more sensitive and less laborious method to assess and screen for salt tolerance.

Solution

Hyperspectral imaging (HSI) is well suited to investigate the internal structures, composition, and functions of plants. To date, HSI has garnered attention for plant phenotyping because it allows investigators to observe how plants are impacted by environmental variables.

Researchers at the University of Minnesota set out to assess and quantitatively rank salt tolerance in wheat using HSI and image analysis.

The investigators sought to rank four different wheat lines based on their stress tolerance, establish a method for quantitative ranking (rather than qualitative), and evaluate the feasibility of precision ranking.

Application Overview

To develop a method for analyzing hyperspectral images, four bread wheat lines, with varying levels of salt tolerance, were assayed in a hydroponic system with control and salt treatments. A push broom hyperspectral camera and glider gear were used for image acquisition 24 hours after salt application. Once collected, images were calibrated to account for non-uniformities. From there, investigators proceeded to run three methods to analyze hyperspectral images of wheat lines: area under the normalized reflectance distance (NRD) curve, minimum difference of pair assignments (MDPA), and posterior probability.

Equipment Used

The experimental setup included a push broom (i.e., along-track scanner) hyperspectral camera, Resonon’s Pika II (now the Pika L), for image acquisition. The camera scanned over 240 spectral channels ranging from 400 to 900 nm with a spectral resolution of about 2.1 nm. In the along-track direction, the number of pixels was set to 2,000 so both the control group and the salt treatment group were captured in a single image. During image processing, radiometric calibration was performed using Resonon’s Spectronon software.

NRD (Normalized Reflectance Difference) for four different wheat lines. The NRD represents the change in reflectance of the salt-stressed wheat compared to the control group wheat for the four different wheat lines tested, over each wavelength sampled. A lower NRD value signifies a wheat line that is more tolerant to salt stress.

In the figure above, NRD (Normalized Reflectance Difference) for four different wheat lines is plotted. The NRD represents the change in reflectance over each wavelength sampled of the salt-stressed wheat compared to the control group for the four different lines tested. A lower NRD value signifies a wheat line that is more tolerant to salt stress.

Results

This method demonstrates a more feasible means of ranking salt tolerance in individual wheat lines (from a labor and cost perspective).
The investigators were able to illustrate quantitative, non-invasive means of phenotyping plants using machine learning and Hyperspectral Imaging.

Download the article, download the hyperspectral datacubes used in the study, or see all precision agricultural application examples.