Libraries, archives, and museums (LAMs) have digitized — and are still digitizing — vast collections of scientific materials, ranging from botanical records to early climate observations, opening new opportunities for a more nuanced understanding of environmental change. Europe's natural history museums alone house data on more than 1.5 billion plants, animals, and minerals, collected across the globe over the last four hundred years.
The computational study of such materials — created by Carl Linnaeus and other naturalists — can help enrich present-day understanding of scientific knowledge as it is created across environmental sciences. With advances in AI methods such as language and computer vision technologies, historians and other researchers can now analyze large, heterogeneous datasets to uncover long-term patterns and changes in environmental and other forms of knowledge production. However, these domains pose challenges to AI methods since the data is generally more prone to noise, and to organisational and interpretation challenges.
At the same time, new data streams are becoming available, such as drone footage and sensor and tag data for monitoring biodiversity. Integrating historical and contemporary data, across different modalities and fidelities, is a major challenge for knowledge representation — and one that must be solved to build world models that support resilient responses to natural hazard problems.
The workshop is organised around two core challenges — making historical and contemporary data speak to each other, and doing so honestly. We identify three intersecting topic areas:
