Setting the morphologic quality limits enabling accurate classification of charred archaeological grape seeds

Area of Science:

• Archaeobotany
• Materials Science
• Computational Biology

Background:

• Grape pips (Vitis vinifera L. seeds) are crucial in archaeological contexts for understanding ancient agriculture.
• Charring significantly alters seed morphology, complicating identification and classification.
• Morphometric analysis and machine learning offer potential for identifying charred seeds.

Purpose of the Study:

• To investigate the morphological changes in grape pips under various charring conditions.
• To develop a method for assessing charring severity and classifying grape seeds using morphometrics and machine learning.
• To create a sorting model for archaeological grape seeds to enhance identification accuracy.

Main Methods:

• High-resolution scanning and morphometric measurements for detailed morphological analysis.
• Application of machine learning classification algorithms to charred grape seeds.
• Development of a sorting model based on length-width ratios and controlled charring data.

Main Results:

• Significant alterations in grape pip shape, length-width ratio, and crack occurrence were observed above 250°C.
• Accurate classification of grape varieties was feasible for seeds charred up to 250°C and 8 hours.
• A sorting model estimated charring temperatures, revealing drastic conditions for ~50% of archaeological seeds.

Conclusions:

• Combining machine learning with morphometric sorting effectively identifies charred grape seeds suitable for classification.
• The developed sorting model enhances the accuracy of machine learning-based identification of archaeological grape seeds.
• This approach provides a valuable tool for studying ancient viticulture and grape cultivation practices.