Documenting the evolution of tourmaline-bearing localities in the United States

This dataset contains geological interpretations for tourmaline-bearing localities that are listed in the Mindat.org database. This Excel spreadsheet is separated into three sheets containing variables describing environmental and mineralogical characteristics of tourmaline-bearing sites in the United States. Localities listed under the "Training sites" sheet have descriptions that are based upon mineralogical and geological reports made by contributors to the Mindat.org database. More information regarding the retrieval of the different variables listed can be found at zenodo.org/record/4110642#.X5bxc0eSnYU

The locality descriptions were then used to make predictions regarding how these localities likely evolved through geologic time. The sites were given labels that contain a series of geological processes listed under the columns "Simple_Evolution" and "Complex_Evolution" of the "Site evolution" tab. Geologic processes included in the Simple and Complex evolutionary sequences have been described below.

Simple_Evolution: HE: hypogene (hydrothermal activity with fluids originating from deep within the Earth) hydrothermal activity. I: igneous rocks or some magmatic process. M: metamorphic process. S: sedimentary rocks or active sedimentary processes. SE: supergene (hydrothermal activity with fluids originating from at or near the Earth's surface) hydrothermal alteration. 

Complex_Evolution: CaMgSe: calcium-magnesium-rich carbonate sediments. CaSk: calcic skarns. CP: complex pegmatites. FC: felsic composition igneous rocks. G: greisenization. HE: hypogene hydrothermal activity. IC: intermediate composition igneous rocks. M: metamorphic process. MC: mafic composition igneous rocks. PSe: clay-rich sediments. S: sedimentary rocks or active sedimentary processes. SE: supergene hydrothermal activity. Serp: serpentinization. SiSe: siliciclastic sediments. SP: pegmatites with simpler mineralogies dominated by quartz, microcline, and muscovite. VSe: volcanoclastic sediments. WSk: tungsten-rich skarn. WSnSk: tungsten-tin-rich skarn. 

The labels assigned to the training sites were then used to train random forests and decision tree algorithms to assign evolutionary sequences to unseen mineralogical and environmental data. The "Test sites" tab contains 389 sites located within the conterminous United States and Alaska. These sites were randomly selected and mineralogical and environmental surface models were used to extract the associated variables (more information regarding the methods used at zenodo.org/record/4110642#.X5bxc0eSnYU). Results from the decision tree (DT) and random forests (RF) classifications of the Simple and Complex evolutionary sequences are listed under the "Site evolution" tab for each test location.