Study of the chemical and mineralogical correlations

The correlation coefficient measure the strength of the linear relationship between two variables, here between two elements concentration. We will use the GeoLime dictionary to analyse the correlations between the differents elements present in the Rocklea Dome datas.

import pandas as pd
import geolime as geo
from pyproj import CRS
import numpy as np

geo.Project().set_crs(CRS("EPSG:20350"))

First we need to import our data from Geolime .geo files.

dh = geo.read_file("../data/dh_hyper.geo")

dh.user_properties()

['X_COLLAR',
 'Y_COLLAR',
 'Z_COLLAR',
 'X_M',
 'Y_M',
 'Z_M',
 'X_B',
 'Y_B',
 'Z_B',
 'X_E',
 'Y_E',
 'Z_E',
 'Fe_pct',
 'Al2O3',
 'SiO2_pct',
 'K2O_pct',
 'CaO_pct',
 'MgO_pct',
 'TiO2_pct',
 'P_pct',
 'S_pct',
 'Mn_pct',
 'Fe_ox_ai',
 'hem_over_goe',
 'kaolin_abundance',
 'kaolin_composition',
 'wmAlsmai',
 'wmAlsmci',
 'carbai3pfit',
 'carbci3pfit',
 'Sample_ID',
 'Fe',
 'Fe2o3',
 'P',
 'S',
 'SiO2',
 'MnO',
 'Mn',
 'CaO',
 'K2O',
 'MgO',
 'Na2O',
 'TiO2',
 'LOI_100',
 'Depth']

Chemical Analysis

Correlation Analysis

Correlation Matrix

Now we can create the correlation matrix for the hyperspectral data:

geo.correlation_heatmap(
    geo_object=dh, 
    properties=["Mn_pct","P_pct","Fe_pct","CaO_pct","MgO_pct","S_pct","Al2O3","TiO2_pct","SiO2_pct","K2O_pct"]
)

A correlation coefficient (r) close to 0 indicates a weak linear relation between the two variable. Positive values of r indicate a positive correlation (values of both variables increase together) while negative values of r indicate a negative correlation (values of one variable increase while values of the other variable decrease). When r=1 or r=-1 it”s a perfect correlation which means that both variables evolves together at a fixed speed.

Magnesium oxide MgO and Calcium oxide CaO are highly correlated (r = 0,9). Same for Aluminium Al2O3 and Titanium TiO2 who have a correlation coefficient of 0,7. We can also notice the Silicium SiO2 and Aluminium and the Silicium and Titanium which correlation coefficients are respectively 0,6 and 0,5. We also have a strong negative correlation between iron Fe and Silicium (r = -0,9). Finally iron also have a noticeable negative correlation with Aluminium (r = -0,7) and Titanium (r = -0,6).

Graph for the positive correlation relations are the following. We can see that the first one which is the one with the higher correlation coefficient is the most linear.

2D Scatter Plots

geo.scatter(geo_object=dh, property_x="MgO_pct", property_y="CaO_pct")

This positive and very strong correlation between these two oxide indicate the presence of dolomite. This is a anhydrous carbonate mineral composed of calcium magnesium carbonate (CaMg(CO3)2).

geo.scatter(geo_object=dh,property_x="Al2O3", property_y="TiO2_pct")

geo.scatter(geo_object=dh, property_x="Al2O3", property_y="Fe_pct")

geo.scatter(geo_object=dh, property_x="SiO2_pct", property_y="Al2O3")

This correlation can be the formation of kaolinite (Al2Si2O5(OH)4). Kaolinite is a clay mineral present in the Rocklea dome site.

geo.scatter(geo_object=dh, property_x="SiO2_pct", property_y="TiO2_pct")

Graph for the negative correlation relations are the following. We can see that the first one which is the one with the lowest correlation coefficient is the most linear.

geo.scatter(geo_object=dh, property_x="SiO2_pct", property_y="Fe_pct")

geo.scatter(geo_object=dh, property_x="Al2O3", property_y="Fe_pct")

geo.scatter(geo_object=dh, property_x="TiO2_pct", property_y="Fe_pct")

Negative correlation relation show that the concentration in iron increase when the concentration in Aluminium, Titanium or Silicium decrease.

We can do the same for the assay data and compare the result to see if we have the same correlations :

This correlation indicates the formation of kaolinite, which we already have have thanks to the hyperspectra data.

We don’t find the same results as in the hyperspectral file because some datas are missing with the assay file. Indeed we don”t have a lot of informations on the Phosphore, Sulfure, Ferric oxide, Titanium or sodium oxide. This is less precise than the hyperspectral datas.

Now that we have study the different correlations between the chemical elements we can use the hyperspectral data to study the correlation with the mineralogical information.

Mineralogical Analysis

Correlation Analysis

We can make the correlation matrix including all chemical elements and mineralogical informations.

geo.correlation_heatmap(
    geo_object=dh, 
    properties=["Fe_pct","Al2O3","SiO2_pct","K2O_pct","CaO_pct","MgO_pct","TiO2_pct","P_pct","S_pct","Mn_pct","LOI_100","Fe_ox_ai","hem_over_goe","kaolin_abundance","kaolin_composition","wmAlsmai","wmAlsmci","carbai3pfit","carbci3pfit"]
)

First we can notice a high correlation coefficient between the Ferric oxide abundance (Fe ox ai) and the concentration in iron (Fe %). There is also a high negative correlation coefficient between the iron concentration and the kaolin abundance which means that when iron concentration increases, kaolin abundance decrease or the opposite.

When we look at the aluminium concentration (Al3O3), we notice correlation coefficients of 0,8 with the kaolin abundance and 0,6 with the white micas and Al-smectite abundance. Aluminium is linked to the formation of kaolin and white micas.

Calcium oxide (CaO) and magnesium oxide (MgO) have ahigh correlation coefficient (0,8) with the loss-on-ignition measure. This means that when the LOI percentage increases, the concentration in CaO and MgO increases too.

Finally kaolin abundance is linked to the carbonate abundance (r = 0,8).

According to the article called The Rocklea Dome 3D Mineral Mapping Test Data Set there are four main geochemical variables (FeO, Al2 O3 , SiO2 , and TiO2 ) and two main mineralogical variables found thanks to hyperspectral data (ferric oxide abundance and kaolinite abundance). What we found with the correlation analysis is similar to this, the four geochemical elements that were the most correlated are FeO, Al2O3, SiO2 and TiO2.For the mineralogical informations, the kaolin abundance is linked to three of these variables and the ferric oxide abundance is linked to the fourth one which is iron.

The other article from Fouedjio explains these correlations. The strong positive correlation between iron and the hyperspectrally derived ferric oxide abundance confirms the richness in iron of this site. Then the strong positive correlation between alumina and titanium may be caused by heavy minerals associated with detrital material in clay-rich layers of the channel. The northern part of the Rocklea dome channel iron deposit is full of extensive calcretee deposits which may explain the high correlation between calcium and magnesium oxides. They also talk of the negative correlations found in the data set. They explain the strong negative correlation between iron and silica with the intercalation of Al-clay-rich sedimentary layers (rich in silica and low in iron) with iron-oxide-rich horizons which are relatively low in silica. Finally, this negative correlation is coherent with the positive correlation beteween silica, alumina and potassium.