Journal
AGROFORESTRY SYSTEMS
Volume -, Issue -, Pages -Publisher
SPRINGER
DOI: 10.1007/s10457-023-00833-3
Keywords
Coffee quality; Near-infrared spectroscopy (NIRS); Random forest; PERMANOVA
Ask authors/readers for more resources
The present study aimed to quantify the relative importance of biophysical variables affecting coffee bean quality and evaluate a near infrared spectroscopy-based model for coffee quality prediction. The study found large within farm variability, indicating the importance of coffee plant-level conditions for physical attributes and cupping scores of green coffee beans. Elevation appeared to be the key variable driving all measured coffee bean quality attributes at the farm level, while canopy cover was the most important variable at the coffee plant level.
The biophysical drivers that affect coffee quality vary within and among farms. Quantifying their relative importance is crucial for making informed decisions concerning farm management, marketability and profit for coffee farmers. The present study was designed to quantify the relative importance of biophysical variables affecting coffee bean quality within and among coffee farms and to evaluate a near infrared spectroscopy-based model to predict coffee quality. Twelve coffee plants growing under low, intermediate and dense shade were studied in twelve coffee farms across an elevational gradient (1470-2325 m asl) in Ethiopia. We found large within farm variability, demonstrating that conditions varying at the coffee plant-level are of large importance for physical attributes and cupping scores of green coffee beans. Overall, elevation appeared to be the key biophysical variable influencing all the measured coffee bean quality attributes at the farm level while canopy cover appeared to be the most important biophysical variable driving the above-mentioned coffee bean quality attributes at the coffee plant level. The biophysical variables driving coffee quality (total preliminary and specialty quality) were the same as those driving variations in the near-infrared spectroscopy data, which supports future use of this technology to assess green bean coffee quality. Most importantly, our findings show that random forest is computationally fast and robust to noise, besides having comparable prediction accuracy. Hence, it is a useful machine learning tool for regression studies and has potential for modeling linear and nonlinear multivariate calibrations. The study also confirmed that near-infrared spectroscopic-based predictions can be applied as a supplementary approach for coffee cup quality evaluations.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available