Regression on compositional covariates: assessing substrate suitability for vegetation

Investigating the relationship between vegetation cover and substrate typologies is important in habitat conservation and management. We focus on a modern ecological survey, where information regarding vegetation cover are derived from digital ground photos taken at different times. The aim is to estimate the effect of different substrate typologies on vegetation cover (substrate suitability). As it is often the case in ground cover imaging, information on substrate typologies are available as compositional data, e.g., the area proportion occupied by a certain substrate. We develop a novel procedure for managing compositional covariates within a Bayesian hierarchical framework and illustrate it with data from a gypsum outcrop located in the Emilia Romagna region, Italy.