A machine-learning hybrid-classification method for stratification of multidecadal beach dynamics

Coastal areas are one of the most threatened natural systems in world. Environmental beach indicators, such as erosion and deposition rates exposed beaches Andalusia (640 km), were calculated using upper limit active profile detailed orthophotos (1:2500) for periods 1956–1977, 1977–2001 2001–2011. A hybrid classification method, both supervised unsupervised, based on machine-learning (ML) techniques was then applied to model response dynamics this 55-year period. The use a K-means technique allowed stratification into four groups that have responded similarly terms coastline mobility erosion/deposition patterns. Furthermore, application regression tree (CART) results helped identify threshold values erosional depositional period characterises each cluster or stratum, enabling correct 1415 out 1509 (93.77%).