Spatial patterns of Early Neolithic expansion in the north-eastern Iberian Peninsula, western Mediterranean: A landscape and environmental machine learning approach

• A dual Neolithic entry in north-east Iberia via maritime and trans -Pyrenean routes is likely. • RF machine learning reveals divergent patterns of Early Neolithic settlement. • Climate variables outweighed topography in driving Early Neolithic settlement. • Central coastal and Pyrenean zones concentrated most Early Neolithic landscapes. The expansion of Neolithic involved multiple dispersal routes and regionally differentiated trajectories shaped by environmental, topographic, and cultural factors. This study investigates the Early Neolithic expansion in the north-eastern Iberian Peninsula using a spatially explicit framework that integrates geostatistical interpolation, least-cost path modelling, and machine learning–based classification. A dataset of 78 georeferenced Early Neolithic archaeological sites spanning from 5600 to 3800 cal BC was combined with reconstructed paleoclimatic and paleo-orographic variables to assess patterns of settlement establishment and movement. Empirical Bayesian kriging was used to estimate the timing of Neolithic onset at the landscape level, while least-cost path analysis evaluated potential dispersal routes. Environmental and topographic drivers of early settlement were further assessed using Random Forest classification across 134 landscape units. Results indicate a spatial concentration of Early Neolithic landscapes along the coast and within selected Pyrenean areas, supporting the feasibility of a dual entry of agriculture via both maritime and trans -Pyrenean routes. Environmental variables—particularly mean maximum temperature, temperature range, and minimum rainfall—emerged as the strongest predictors of early settlement, whereas topographic variables played a secondary role at the landscape level. These findings highlight the importance of climatic constraints and regional environmental adaptation in shaping early agricultural expansion and demonstrate the value of combining GIS-based modelling with machine learning to investigate complex non-linear agricultural dispersal processes.