relatingromanrings_caa2025

https://github.com/ronaldvisser/relatingromanrings_caa2025

Keywords

Repository

README.Rmd

---
bibliography: references.bib
output: github_document
---

# Relating Roman Rings: an open and reproducible approach to understanding provenance patterns of wood using networks

[![DOI](https://zenodo.org/badge/974772798.svg)](https://doi.org/10.5281/zenodo.15342407)

Author: Ronald Visser

Presentation presented at the CAA 2025 in Athens ()

## Abstract

Wood was, is and will be on of the most important resources. While wood is one of the most sustainable and strongest building materials, wood is also prone to decay and therefore often lacking in archaeological contexts. In the Roman period many objects and structures were made of wood, such as roads, bridges, ships, buildings, chairs, boxes and buckets. However, these are only found in very dry or very wet conditions. The lower Rhine area is such an environment, with high ground water levels and a high sedimentation rate, resulting in excellent preservation conditions for organic materials, such as wood. Over the past decades, wood has been excavated and part of this has been subjected to dendrochronological and dendroarchaeological research. This resulted in a large dataset of tree-ring material. While the find location is often known, the provenance of the trees, i.e. the growth region, is often unknown and has to be reconstructed. Determining the provenance of wood based on dendrochronological research is termed dendroprovenance. This is often based on matching tree-ring patterns with existing regional and/or site chronologies using various software packages. This process does not always comply with the aims of Open Science, since not all software is open, sometimes resulting in workflows that are not fully reproducible. This paper will address the following central research question: How to determine the provenance of wood using dendroarchaeology in a reproducible and open method?

Before being able to address the question in detail the data needs to be described, explaining the life history of the data set, what temporal and spatial patterns are present in the dataset. These patterns define the suitability of the data for further analyses and this will be discussed in the presentation.

Dendroprovenance is an important field of research aiming to determine the provenance of wood. A new method has been developed using network analyses to assess the provenance [@visser2022; @visser2021]. Dendroprovenance is based on the statistical relations between tree-ring series or chronologies and a strong statistical relation between series assumes similar growth patterns and therefore a similar provenance. As argued elsewhere [@visser2021], methods such as clustering are not suitable for this, due to the varying temporal and spatial distribution of the sometimes sparse data. The use of a network for studying the relations between a tree-rings series for determining the provenance solves many issues and makes the method also more transparent. Networks are created by using the pair-wise similarity of tree-ring patterns to build the edges, while the nodes represent the find location or the wood sample(s). The neighbours of a node in the network reflects the similarity of the growth patterns and therefore the trees probably grew in the same region. However, since wood could have been moved and transported in the past, various factors need to be taken in to consideration when interpreting the patterns, such as context, spatial location, growth patterns [@visser2021, 244]. These networks have led to new insights in the Roman wood economy and has also shown that wood was only transported over long distances in specific projects, such as the road along the Roman *limes* or to construct Roman barges [@visser2022]. Local wood procurement was the rule, and long-distance transport remained the exception [@visser2025].

The analyses and method were published as open as possible and all data and source code was made available upon request, but to advance the openness and reproducibility, a package for R was developed [@visser2024]. This package is reviewed and published with rOpenSci: . Building a package ensured that the methodology became more generally applicable. At the same time, lessons were learned, especially how to share a method in a reproducible and universal way ().

This presentation will deal with the issues in the dataset and how network analyses using open software resulted in a reproducible workflow leading to new insights into the wood economy in the Roman period. It will also address some of the challenges of making research completely open by sharing scripts and a package, and how network science is an essential aspect thereof.

## References

Owner

GitHub Events