Fig. 9
Allanite classification based on their Th/U ratio and the fraction of initial 206Pb (f206) values. Typical uncertainties for f206 values are between 10 and 20% (2σ, see Additional file 1). a Database of allanite data. Black squares indicate magmatic allanite (344 analyses). Central-Alpine Bergell intrusion: granodiorite BONA—tonalite SISS (Gregory et al. 2007, 2012; Airaghi et al. 2019; Santamaría-López et al. 2019); Atesina Volcanic Complex (AVC): rhyolite (Gregory et al. 2007); Cima D'Asta Pluton (CAP) granodiorite (Gregory et al. 2007; unpublished data); Ecstall Pluton: granodiorite BC (Gregory et al. 2007, 2012); Berridale Batholith, southeast Australia: granodiorite TARA (Gregory et al. 2007, 2012); Sesia Zone: orthogneiss (Giuntoli et al. 2018b); Central Alps: metatonalite—metagranodiorite—metagranite—metadiorite (Gregory et al. 2012). Grey circles indicate metamorphic allanite (including 525 analyses of low-pressure allanite and 686 high-pressure). Sesia Zone: mica schist—metaquartzite—blueschist—eclogite—Glaucophane-epidote schist—orthogneiss—leucogneiss (Regis et al. 2014; Giuntoli et al. 2018b; Vho et al. 2020); Austroalpine basement: pegmatites (Hentschel et al. 2020); Nevado-Filábride metamorphic complex: mica schist (Santamaría-López et al. 2019); Western Alps: metagranite, Mont Blanc (Cenki-Tok et al. 2014) and micaschist, Acceglio Zone (unpublished data); Longmen Shan: metapelites (Airaghi et al. 2018, 2019); Himalaya: metapelites (Goswami-Banerjee and Robyr 2015; unpublished data); Central Alps: metapelites (Janots et al. 2009; Janots and Rubatto 2014); metatonalite—metagranodiorite—metagranite—metadiorite (Gregory et al. 2012). Red circles indicates Aln analyzed in this study. b Visualization of the predictions of a machine learning classification model produced using the Random Forest algorithm including bootstrap aggregating (See Additional file 3: Decision tree for the classification model). A classification three containing 107 nodes was refined and used as a predictive model. The allanite data produced in this study (red dots) are 100% classified as “metamorphic” by the model. The classification accuracy is above 95% based on the out-of-bagging error calculation