WP4 Geochemical and other exploration data preprocessing

CRS Laboratories Oy, Sokli Oy, Geopool Oy, GTK, UTU
The mappable criteria corresponding to mineral system’s critical elements defined WP1 are manually derived from the datasets selected in WP1. Some of the initial input data require only basic processing, like interpolation or proximity analysis, whilst others must be treated with more complex, yet standard, preprocessing such as denoising, outlier detection, spatial filtering, geophysical interpretation.
The geochemical data requires compositional data preprocessing (CoDa), which does not yet have standard procedures in prospectivity modelling. CoDa techniques consider elements as a part of the whole elemental composition and thus avoid spurious interpretation of the geochemical datasets such as soil geochemistry, often used in MPM as input data layers.

CRS Laboratories Oy, Sokli Oy, Geopool Oy, GTK, UTU
In the AIMEX project, we will deal with pairwise log ratios and produce them as inputs into the following data processing steps. The output from this task provides the input for the machine learning algorithms used in the project, both directly for prospectivity modeling and for generating further derivatives using feature engineering.
We will prepare evidence datasets with different levels of preprocessing to find out which of the manual preprocessing steps can be replaced by machine learning. CRS laboratories will be involved in this task by bringing in knowledge on the quality control of laboratory assays.
Sokli Oy needs pre-processing of the exploration data so that it can be used as input data in the mineral prospectivity mapping. The same applies to the data provided by the two in-kind partners, Boliden Kevitsa Mining Oy and Mawson Oy. Geopool Oy will be working on hyperspectral data correlation with geochemical datasets.