Reconstructing land use change from archival records - Using text mining, large language models and GIS to reconstruct land use change processes in former Dutch colonies

Abstract

Quantifying historical land use change (LUC) can help estimate historical contributions to greenhouse gas emissions and biodiversity loss. Additionally, understanding LUC in former colonies can help fairly distribute the remaining carbon budget over colonising and colonised countries. Historical LUC is often quantified using a modelling approach and historical data is only sporadically used. In contrast, this thesis uses solely historical documents and assesses the archives of the Dutch East India Company (Vereenigde Oost-Indische Compagnie, VOC) to find information on land use. Additionally, a text mining methodology is developed for the identification of such information from transcribed archives of the VOC. A total of 14 lists concerning land use were identified with the text mining methodology. For Ceylon (modern-day Sri Lanka) a land use reconstruction is made using one of these lists. The list is georeferenced onto a contemporary map allowing 34% of the locations in the list to be connected to a location on a map. Additionally, two historical maps on land use were georeferenced, offering another data source for the estimation of land use. By identifying more lists and improving the georeferencing procedure, future research can disentangle the drivers for LUC and make land use estimates for all the colonies of the VOC. This thesis offers the foundations for such an approach and is a starting point for improving our understanding of historical LUC.