NIR spectra of Hawaii avocados
Data files
May 16, 2024 version files 3.13 MB
-
Felix_2nd_ave.xlsx
-
OO_All199_ave_SNV.xlsx
-
README.md
Abstract
Avocados are an important economic crop of Hawaii, contributing to approximately 3% of all avocados grown in the United States. To export Hawaii-grown avocados, growers must follow strict United States Department of Agriculture Animal and Plant Health Inspection Service (USDA-APHIS) regulations. Currently, only the Sharwil variety can be exported relying on a systems approach, which allows fruit to be exported without quarantine treatment; treatments that can negatively impact the quality of avocados. However, for the systems approach to be applied, Hawaii avocado growers must positively identify the avocados variety as Sharwil with APHIS prior to export. Currently, variety identification relies on physical characteristics, which can be erroneous and subjective, and has been disputed by growers. Once the fruit is harvested, variety identification is difficult. While molecular markers can be used through DNA extraction from the skin, the process leaves the fruit unmarketable. This study evaluated the feasibility of using near-infrared spectroscopy to non-destructively discriminate between different Hawaii-grown avocado varieties, such as Sharwil, Beshore, and Yamagata, Nishikawa, and Greengold, and to positively identify Sharwil from the other varieties mentioned above. The classifiers built using a bench-top system achieved 95% total classification rates for both discriminating the varieties from one another and positively identifying Sharwil while the classifier built using a handheld spectrometer achieved 96% and 96.7% total classification rates for discriminating the varieties from one another and positively identifying Sharwil, respectively. Results from chemometric methods and chemical analysis suggested that water and lipid were key contributors to the performance of classifiers. The positive results demonstrate the feasibility of NIR spectroscopy for discriminating different avocado varieties as well as authenticating Sharwil. To develop robust and stable models for the growers, distributors, and regulators in Hawaii, more varieties and additional seasons should continue to be added.
README: NIR spectra of Hawaii avocados
https://doi.org/10.5061/dryad.98sf7m0rj
The dataset contains two XLSX files; OO_All199_ave_SNV is spectra of three varieties of Hawaii avocado in 2022, collected using a bench-top near-infrared spectrometer, Felix_2nd_ave is spectra of five varieties of Hawaii avocado in 2023, collected using a handheld near-infrared spectrometer.
Description of the data and file structure
OO_All199_ave_SNV
This is the common wide format for spectral data, showing absorbance at each wavelength for each sample
- Variety Avocado varieties: Beshore, Sharwil, or Yamagata
- Sample ID Internal reference ID for each fruit
- 1000.087 - 2400.404 (Column C to Column QE) wavelengths in nm
Felix_2nd_ave
This is also in the common wide format for spectral data, showing absorbance at each wavelength for each sample
- Harvest Date Date each avocado was harvested
- NIR Date Date each avocado was measured using the handheld near-infrared spectrometer
- Variety Avocado varieties
- SW Sharwil
- GG Greengold
- N Nishikawa
- CB non-commercial variety, colloquially known as Cannon ball
- SS non-commercial variety, unnamed seedling of Sharwil
- Days in Room Temp Days since the fruits were removed from a cold room to ripen in room temperature
- Sample ID Internal reference ID for each fruit
- W729nm - W975nm (Column F to Column CJ) wavelengths in nm
Software
The files can be opened and analyzed using any common statistical software.
Methods
Avocado samples
931 avocados were acquired through the only packing house on the Island of Hawaii, Kane Plantation Avocados in Honaunau, from June 2022 through June 2023. Avocados were kept between 4°C and 6°C until one to eight days before spectral measurements. The selection of avocado varieties was based on seasonal availability. Sharwil (n=537) avocados were collected in June, July, and September of 2022, as well as January and March of 2023. Beshore (n=52) avocados were collected between June and July of 2022. Yamagata (n=32) avocados were collected in June and July of 2022. Nishikawa (n=97) avocados were collected in March 2023 and Greengold (n=129) avocados were collected in February and March of 2023. Two other non-commercial varieties were also collected in January and March 2023; one is colloquially known as Cannon ball (n=43) for its shape and size and the other is an unnamed seedling of Sharwil (n=41).
The spectra data was collected and processed using the below methods:
-
A high-power tungsten halogen light source (Ocean Insight, Orlando, FL, USA) and a bench-top NIR spectrometer (NIRQuest+2.5, Ocean Insight, Orlando, FL, USA) with an InGaAs detector were used to measure adsorption of incident light on avocado samples at wavelengths from 900 to 2500 nm. Optical fibers were used to couple the light source, sample holder, and spectrometer. After ripening in room temperature for three days, 115 Sharwil, 52 Beshore, and 32 Yamagata from 2022 were measured at three random locations along the equatorial region using a desktop application (Ocean View 2.0, Ocean Insight, Orlando, FL, USA) with 50 ms integration time and a Boxcar width of 3 (i.e. 7-point averaging). Each fruit was measured at three random locations along the equatorial region. Average spectrum per fruit was calculated between 1000 and 2400 nm, truncating the shortest and the longest 100 nm due to sensor limitations. Standard Normal Variate (SNV) preprocess was applied to the raw spectra to remove additive effects and multiplicative scatter.
-
Subsequent experiments were carried out using a handheld NIR spectrometer (F-750, Felix Instruments, Camas, WA, USA) measuring 2nd derivative spectra between 729 nm and 975 nm. Spectra of additional 422 Sharwil, 129 Greengold, 97 Nishikawa, 43 Cannon ball, and 41 unnamed Sharwil seedling from 2023 were collected at three random locations along the equatorial region after ripening in room temperature for one, three, six, seven, and/or eight days. Average spectrum per fruit was calculated and no additional spectral preprocessing was applied.