Geoadditive Bayesian models for forestry defoliation data: a case study