AI-Driven Material Analysis: Overcoming Specialist Bottlenecks in Archaeobotany and Provenance
As artificial intelligence matures, its application in archaeology is shifting from macro-scale remote sensing (like LiDAR) to micro-scale laboratory analysis of physical artifacts. Two major peer-reviewed studies published in late 2025 and 2026 demonstrate how machine learning and explainable AI (XAI) are resolving "specialist bottlenecks" in archaeobotany and reconstructing prehistoric trade networks with unprecedented accuracy.
1. Automated Archaeobotany: The APSNet Seed Classification System
In a study published on June 22, 2026, in npj Heritage Science, researchers from Shandong University and Lingnan University introduced APSNet, the world's first large-scale AI system dedicated to identifying ancient plant seeds.
The Bottleneck
Reconstructing ancient diets, agricultural practices, and human-environment interactions relies heavily on identifying charred plant seeds excavated from archaeological sites. Traditionally, this has required archaeobotanists to manually examine each seed under a microscope—a highly specialized, time-consuming process that is difficult to scale.
The AI Solution
To automate this preliminary screening, the research team:
- Created the Ancient Plant Seed Image Classification (APS) dataset, compiling 8,340 images across 17 genus- or species-level seed categories excavated from 18 archaeological sites in China spanning 5,000 years.
- Designed APSNet, a deep learning framework that incorporates seed scale information via a Size Perception and Embedding (SPE) module and utilizes an Asynchronous Decoupled Decoding (ADD) architecture to learn fine-grained spatial and channel features.
- Achieved 90.2% classification accuracy, demonstrating that AI can reliably handle the physical damage, warping, and carbonization typical of ancient botanical remains.
The project was funded by public scientific bodies, including the National Natural Science Foundation of China, the National Cultural Heritage Administration, and the Hong Kong Research Grants Council (RGC).
2. Geochemical Fingerprinting: The VORTEX Gemstone Provenance Framework
In another major study published in the Journal of Archaeological Science, a multidisciplinary team of Spanish and Portuguese researchers developed VORTEX (Variscite Origin Recognition Technology X-ray based) to trace the prehistoric trade of "green gemstones" (variscite) across Western Europe between the 6th and 2nd millennium BC.
The Methodology
The team compiled the largest geochemical database of its kind, featuring portable X-ray fluorescence (pXRF) measurements of 1,800 geological samples and 571 archaeological artifacts. They trained a random forest machine learning algorithm to analyze subtle chemical variations and recognize the unique "geochemical footprint" of individual prehistoric mines.
Key Breakthroughs
- 95% Predictive Accuracy: The model can determine the exact geological origin of prehistoric variscite beads with 95% accuracy.
- Rewriting History: The AI's classifications revealed that the mines of Gavà (Barcelona, Spain) and Aliste (Zamora, Spain) were the primary Western European production centers. It also showed that variscite found in Brittany (France) likely traveled via trans-Pyrenean land routes, rather than the maritime routes previously assumed by archaeologists.
- Explainable AI (XAI): Rather than operating as a "black box," the system uses XAI techniques to explicitly show which chemical elements were decisive in each classification, ensuring scientific transparency.
- Open Science Integration: Under the European OpenAIRE program, the entire database has been uploaded to the Zenodo open repository (operated by CERN) to allow global researchers to conduct their own independent analyses.
The VORTEX project was led by the University of Lisbon in collaboration with the Milá y Fontanals Institution for Research in the Humanities (IMF-CSIC), the University of Seville, and the University of Alcalá.