• Home
  • Applications
  • An analytic approach to evaluate Near InfraRed (NIR) data from plants under stress conditions

An analytic approach to evaluate Near InfraRed (NIR) data from plants under stress conditions


Reference publication:
Vello E, Tomita A, Diallo AO, Bureau TE. A Comprehensive Approach to Assess Arabidopsis Survival Phenotype in Water-Limited Condition Using a Non-invasive High-Throughput Phenomics Platform. Front Plant Sci. 2015 Dec 15;6:1101

1) Definition of percentiles and quartiles

Figure 1: Bell-shaped curve. The central tendency, the middle is the median, 50th percentile. 25% to the left is the 25th percentile, the first quartile (Q1). 25% to the right of the median is the 75th percentile, the third quartile (Q3). The interquartile range goes from Q1 to Q3 and makes up 50% of the area under the curve. The largest value is the 100th percentile.

2) Distribution of pixel intensity of NIR reflected light from control and stressed Arabidopsis plants

Figure 2: Left panel: Visible (VIS) and NIR images of Arabidopsis plants at the rosette stage. Right panel: Distribution of NIR pixel intensities for two “digital plants”, one under treatment 1 (well-watered) and one under treatment 2 (drought). The green and red vertical lines intersect the corresponding I75 (or Q3) values for each distribution.

3) Example of data table with I75 values for different plant lines, treatments and days after sowing (DAS)

Each line in the table corresponds to I75 (or Q3) values calculated from the NIR pixel distribution for each sample (1 to 30) at different times after sowing (DAS); n=7 (biological replicates).

4) Calculations:

§ The calculations are based on the method described by Vello et al., 2015

§ The I75 value has been selected for analysis because it detects early changes in NIR intensity signals (Vello et al.; 2015).

i) The 75th percentile (I75) of the distribution of NIR pixel intensities is calculated for each “digital plant” (i.e. segmented plant in NIR image).

ii) A “Base” I75 value (I75B) is selected for each sample as follow:

*Select an I75 NIR intensity value for each sample that is consistently repeated during the first days of measurement (before a treatment is applied or before changes in NIR intensity are observed).

*For example,

- I75 NIR intensities for sample #1 at days 1,2,3,4…20:
• Day1: 60
• Day2: 80
• Day3: 81
• Day4: 80
•…..(start drought)
• Day20: 110

*Choose 80 (Day2) as the I75B value. The value of 80 is the lowest consistent value during the first days of measurements.

iii) The Index for NIR intensity is calculated as: I75/I75B x 100

iv) For example,

- Index values for sample #1:

• Day1: - (not used)
• Day2: 100
• Day3: 101
• Day4: 100
•…Day20: 137

v) Depending on the researcher and experiment, a cutoff value can be set to classify plants as stressed/not-stressed, healthy/non-healthy, etc*.

* See also “Important notes” at the end of the document.

5) Representation of the results (example)

Figure 3: I75 (Q3) NIR intensity index. Means and standard errors of the indexes based on the 75th percentile (i.e. third quartile; Q3) of the NIR intensity for each line and measurement. Water-limited (A,C) and well-watered (B,D) conditions for “experimental protocol 1” (A,B) and“experimental protocol 2”(C,D; DR: drought; WW:well-watered). WT: wild type (control), gtl1-5 and drs1 correspond to different Arabidopsis lines (see reference paper for specific details).

Important notes:

  • If NIR light reflected from plant tissue is to be correlated with water content, the researcher needs to first generate a calibration curve using an independent method for water content estimation (e.g., shoot dry versus fresh weight).

  • The calibration curve needs to be generated using the subject crop and defined experimental conditions (identical to those planned to be applied in the high-throughput digital phenotyping experiment).

  • The calibration curve can then be used for calibration of NIR data coming from the high-throughput digital phenotyping experiment.

  • An appropriate statistical method to evaluate the results needs to be selected and applied by the researcher. This method can vary and depends on the hypothesis, experimental setup, size of the experiment (e.g., number of replicates, number of treatments, number of plant lines), etc.

Gustavo Bonaventure

Application Scientist

Enable JavaScript to view protected content.