Data to Support Predictive Models for Detrital Titanite Provenance with application to the Nanga Parbat syntaxial massif, western Himalaya."

The files published here are metadata that are being used to support a manuscript currently (Jul, 2023) submitted to Journal of Geophysical Research: Earth Surface. The intention of these data and code is to support a publication that is about generating a predictive categorisation scheme for the mineral titanite. The code to generate the titanite classification schemes (Titanite_Random_Forest_Model1.ipynb; Titanite_Random_Forest_Model2.ipynb; and Titanite_Random_Forest_Model3.ipynb) was created in Python3, using Jupyter Notebook. Each of those files also provides more motivation for why a predictive categorisation scheme for the mineral titanite is desirable, and other similar context. Chiefly, the dataset and random forest models published here will allow us to trace titanite in detritus. In general, Models 2 or 3 are probably the most useful models. Model 1 is mostly just provided to enable comparison to similar previously published works (e.g. https://doi.org/10.1111/ter.12574). Briefly, the supplementary file “Table_S4_Titanite_Machine_Learning_Dataset.csv” is required to generate the models and must be saved in the appropriate directory for the code to run. Your unknowns must contain the same elements and identical labelling for the code to successfully run, unless the files and code are modified. Broader descriptions of these matters are contained in each file. Any new published data are titanite compositional or isotopic data collected by LA-ICP-MS. Description of how those data were collected is given in "OSullivan_et_al_Supp..." file. Some of the data, information or code in this submission may be subject to change after journal review, in which case a new version may be published. References for the dataset compilation are provided in File S3. If you have any queries contact:<br> Gary O’Sullivan, Trinity College Dublin