Implementing a multi-criteria GIS analysis and predictive modelling to locate Upper Palaeolithic decorated caves in the Périgord noir, France

Abstract: Geographical Information Systems are a powerful tool for the processing of geographical data and can be used to implement predictive modelling for the purpose of archaeological research. The study presented here proposes a predictive model based on multi-criteria evaluation, and aims at analysing the parameters of known Upper Palaeolithic decorated caves in the Périgord noir, in southwest France. The main objective was to show that known decorated caves have enough parameters in common to create a predictive model that would describe zones of possible decorated caves’ occurrences in the area. A hybrid methodology was developed to extract significant pieces of information from the selected variables in order to design the predictive model. Elevation, slope, aspect, and proximity to water were selected as the four variables to build the model. They were processed individually and were standardised using a numerical rating scale. Each criterion was weighted individually through the rating methodology in order to be aggregated using weighted linear combination. The model created showed that areas close to water streams were more likely to contain undiscovered decorated caves from the Upper Palaeolithic period. The validation statistic Kvamme Gain was finally used to assess the model and the value that was calculated showed that a large proportion of known caves were contained in a relatively restricted area, confirming the model’s predictions.