Application of Geochemical Datasets for Geological MappingTREK Mapping Project, Central British Columbia

M.Rahimi, C.J.R. Hart, J.J. AngenMineral Deposit Research Unit, The University of British Columbia

INTRODUCTIONRegions covered by glacial till and vegetation are always challenging areas for geological mapping. Enormous datasets and innovative methods and techniques are utilized to increase con�dence and extract more information for geological maps and exploration purposes. This project aims to clarify and to add more value to lithological information and bedrock geology of the Targeting Resources through Exploration and Knowledge (TREK) project. A lithological map was developed from geochemical basal till data by creating a geological classi�cation scheme from existing regional geochemical datasets. After evaluation and quality control of datasets, the uni-variant data were considered to best de�ne the elements and analytical methods of value from the existing datasets. Bi-variant and tri-variant diagrams were used to display and interpret improved geological information about the area of study. Classic data methods were utilized to provide a better understanding of statistical characteristics of major oxides, trace and rare earth elements for speci�c geological units.To extract the lithological information, data-driven and knowledge-driven methods were applied to raster and vector data.

Figure 1: Location of the area of interest (AOI) on the map of British Columbia

OVERVIEW OF DATASETSThe datasets are till analysis results from the TREK geochemistry project. Two till survey results were processed include: • Geochemical and Mineralogical Data, TREK Project, Interior Plateau, British Columbia; Geoscience BC, Report 2014-10 (Jackaman and Sacco, 2014). • Geochemical Reanalysis of Archived Till Samples, TREK Project, Interior Plateau, central BC (parts of NTS 093C, 093B, 093F & 093K); Geoscience BC, Report 2015-09 (Jackaman and Sacco, 2015). A total of 677 new samples, 1456 archived samples plus quality control samples were analyzed using di�erent methods for minor, trace and oxide of elements. The samples used in the process are basal till which can provide primary information related to local bedrock (Figure 2).METHODOLOGYThe main goal of this study is to extract as much information as possible from geochemical datasets to add con�dence to lithological interpretations of bedrock beneath till.The visualisation and data interpretation is divided to two major groups:• The statistics-based interpretation and visualisation of geochemical DBs • The spatial-based comparison and interpretation of information based on DBs processing

Final lithology lines

Lithology map from geochemical datasets

Dangles / spa al correc on

Lithology map / Primary unit name

Reclassi ed raster / Min-Max analy cal results

Mosaic raster / Blend algorithm

Integra on / weighted SUM

Evalua on / interpreta on

Till geochemistry dataset, Geoscience BC 2014-10, TREK area

Developed Lithology Map

Weighted Sum Integra on Model (WSIM)

Major element oxides / Selec ve

minor & trace elements

Evidence layer / Normalized raster

Raster informa on layer / Kriging approach

Validity / Evalua on

Till geochemistry dataset, Geoscience BC 2015-09, TREK area

Raster informa on layer / Kriging approach

Major element oxides / Selec ve

minor & trace elements

Rock name

A ribute table

Rock type classi ca on

Litho-line extrac on • Weighted Sum Integration Model (WSIM). The knowledge-based lithology lines were extracted and spatialy evaluated to make a �nal lithology map. Then the map was interpreted to determine possible rock types and add information to the attribute table. The brief steps of the model are shown in Flowchart 1.

Flowchart 2: Knowledge Driven/Supervised Weighted Sum Integration Model (WSIM), applied on Till geochemistry dataset, Geoscience BC, 2014-10 and 2015-09

SUMMARY OF RESULTS AND DISCUSSION

Correlation SiO2_pct Al2O3_pct

Fe2O3_pct MgO_pct CaO_pct Na2O_pct K2O_pct TiO2_pct P2O5_pct MnO_pct Cr2O3_pct

SiO2_pct 1.0 -0.3847 -0.9297 -0.7896 -0.6923 0.003563 0.607 -0.7585 -0.4765 -0.3486 -0.5015Al2O3_pct -0.3847 1.0 0.1627 -0.02415 0.1374 -0.05074 0.1406 0.08349 -0.1045 -0.1698 -0.1193

Fe2O3_pct -0.9297 0.1627 1.0 0.7986 0.654 0.05312 -0.6977 0.8365 0.5539 0.4269 0.5567

MgO_pct -0.7896 -0.02415 0.7986 1.0 0.8021 0.2003 -0.7486 0.7228 0.426 0.4341 0.7726

CaO_pct -0.6923 0.1374 0.654 0.8021 1.0 0.5436 -0.6155 0.8408 0.341 0.151 0.7104

Na2O_pct 0.003563 -0.05074 0.05312 0.2003 0.5436 1.0 -0.1026 0.4295 0.2194 -0.06707 0.2527

K2O_pct 0.607 0.1406 -0.6977 -0.7486 -0.6155 -0.1026 1.0 -0.6415 -0.1531 -0.2264 -0.7349

TiO2_pct -0.7585 0.08349 0.8365 0.7228 0.8408 0.4295 -0.6415 1.0 0.5331 0.2243 0.6723

P2O5_pct -0.4765 -0.1045 0.5539 0.426 0.341 0.2194 -0.1531 0.5331 1.0 0.514 0.1907

MnO_pct -0.3486 -0.1698 0.4269 0.4341 0.151 -0.06707 -0.2264 0.2243 0.514 1.0 0.1451

Cr2O3_pct -0.5015 -0.1193 0.5567 0.7726 0.7104 0.2527 -0.7349 0.6723 0.1907 0.1451 1.0

The transformation in PCA is shown by scores and PCs derived from the classi�ed score-based PC layers. PCA was applied on the dataset, but on its own cannot produce predictions of speci�c geology units. It can produce a visual estimate layer of calculated values that is essential to determine di�erent rock types. Fuzzy-weights approach was used to cover the blind spots of Boolean operators for geological dataset. The lithological units were extracted in GIS fundamental format to predict knowledge-based geological layer.

Figure 5: Spatial distribution of 7 PCs based on minor and trace elements analytical results, eastern TREK area, central BC., The element loadings on the PCs from the target dataset are weighted using fuzzy function

124°0'0"W

124°0'0"W

53°0'0"N 53°0'0"N

10 0 105 km

±

Unknown

AA-ABA-DBB-BABADGDGneiss?RD

Evalua on / valida on

Geosta s cal analysis / Kriging

Fuzzy-weighted approach

• Eigen value – one criterion• The scree test• Propor on of variance

accountedNumber of PCs

Major element oxides

Lithology map from geochemical dataset

Extrac ng spa al informa on

Minor & trace element

Basal ll geochemistry dataset/ TREK area

Integra on / Interpreta on of spa al informa on

Principal Component Analysis (PCA)

PCs / Evalua on

Apply on raster dataset

No

Yes

Geos

paal

Info

rma

on S

yste

m (G

IS)

Correlation matrix as the foundation of geochemical studies (Table 1) provides positive correlation for TiO2 – CaO – MgO – Fe2O3 – Cr2O3 and a strong correlation between SiO2 and K2O where both have considerable negative correlation with other major element oxides.

Figure 5: Spatial distribution of 7 PCs based on minor and trace elements analytical results, eastern TREK area, central BC., The element loadings on the PCs from the target dataset are weighted using fuzzy function

PC1 PC2 PC3 PC4 PC5

PC6 PC7

Figure 6: Lithological layer using KFPCA (a) and Preliminary bedrock geology map (b) (Angen et. al., 2015) eastern TREK area. B: Basalt, BA: Basaltic andesite, A: Andesite, AB: Andesite basalt, D: Dacite, GD: Granodiorite, RD: Rhyodacite

CONCLUSION • Knowledge- based integration methods were applied to geoscience datasets to produce new scienti�c information that gives the possibility of advanced understanding of geological setting of the area. • The method strongly gives the possibility of sub-unit separation in bedrock units. • As the nature of geochemical datasets are more qualitative than quantitative, fuzzy approach gives the �exibility to determine the most appropriate lithology boundaries.

• Hybrid statistical methods and Geospatial information system (GIS) allows: o to model and interpret the data using the parameters of the real world. o to attribute a set of information relevant to the speci�ed geology domain. o to connect the results to geological characteristics and get the most substantive results from the DBs.

The method enables PCs to associate to geological classi�cation using geologic factors controlling the components. The results are interpreted using rock geochemical characteristics loading on each PC and geological information of the area. Correlation of results were evaluated using bedrock geology information of the area. An example is the extracted units which have spatial distribution in the north and has correlation with the Eocene Quanchus Suite granodiorite. The geochemical combination of this unit most likely con�rms the granodioritic rocks.

10 0 105 km

±GF

XW

'4

XW

XW

#*

#*

%,

#*

")

#*

GF

#*#*

XW

XW

#*

#0

XW

#*#*

%,#*

%,")

#*

%,

%,

&-

XW

%,

")#*

#*

#*")")

")")

")

#*")")

")

")")#*

")

XWXW

")

")")GFGF")

XW

")

XW

124°0'0"W

124°0'0"W

125°0'0"W

125°0'0"W126°0'0"W

54°0

'0"N

53°0

'0"N

52°0

'0"N

Andesite#* Basalt#* Basaltic Andesite#* Chilcotin Basalt'4 ConglomerateGF Crystal Tuff") Dacite") Dacite Aphanitic") Dacite Horn/Px") Dacite Plag") Dacite Plag-Bt&- Diorite

Eocene Undivided%, Granite%, GranodioriteGF Lapilli tuffGF PyroclasticTuff

XW Rhyolite

XW Rhyolite Bt-Plag

XW Rhyolite Bt-Qz

XW Rhyolite Horn-Plag

XW Rhyolite Undivided#0 Tonalite

Vitreous black dacite

A

B

BA

Calcalk_B

D

Di

Flood_B_HPT

Flood_B_LPT

GD

MDi

MG

Meta

QD

QDi

QM

R

RD

Sed

The following methods were applied on the datasets: • Principal Component Analysis (PCA) . The results of extracted information from each PC corresponds to the objects of interest in the AOI. Contribution of the data as vector points or pixels in this multivariate method is the basis of extracting the similarity from the dataset. The contribution can be as a crisp number or a fuzzy membership based on the data characteristics.

Table 1: Classical (Pearson's) correlation matrix values Figure 3: Bi-variant diagrams, SiO2 vs. element oxides

Figure 4: Weighted- Raster layers using kriging approach Flowchart 1: Steps of Principal Component Analysis (PCA), applied on till geochemistry dataset, Geoscience BC, 2014-10

Lithology polygons vector layer, extracted from raster layers using Weighted Sum Integration Model (WSIM) is shown in �gure 7 and attributed by chemical analyses information for di�erent rock types from felsic to basic.Attribute table re-processed and classi�ed for volcanic/ plutonic/ sedimentary/ metamorphic rocks using database information and the latest bedrock geology map of TREK area. To evaluate the results, rock samples are plotted on the lithology layer.

Figure 7: Lithological layer using WSIM, attributed by chemical analyses information for di�erent rock types. B: Basalt, BA: Basaltic andesite, A: Andesite, DI: Diorite, D: Dacite, GD: Granodiorite, MDI: Monzodiorite, MG: Monzogranite, RD: Rhyodacite, QDi: Quartz diorite, QM: Quartz monzonite, R: Rhyolite, Sed: Sedimentary rock, Meta: metamorphic rock

B

R to B

R to A

GD

124°0'0"W

124°0'0"W

53°0'0"N 53°0'0"N

10 0 105 km

±

Vanderhoof Metamorphic Complex

Chilcotin Group

Coast Plutonic Complex Undifferentiated

Skeena Group

Telkwa Formation

Ootsa Lake Group

Permian-Triassic(?) sediments

Ashman Formation

Cache Creek UltramaficsCache Creek Complex

AL

BE

RT

A

U . S . A .

Pa

c i f i c oc c e a

n

!

!

!

!

!

\

VICTORIA

Vancouver

Prince George

Nazko

Vanderhoof

Anahim Lake

110°0'0"W

115°0'0"W

115°0'0"W

120°0'0"W

120°0'0"W

125°0'0"W

125°0'0"W

130°0'0"W

130°0'0"W

135°0'0"W

135°0'0"W

60°0

'0"N

55°0

'0"N

55°0

'0"N

50°0

'0"N

50°0

'0"N

B R I T I S H C O L U M B I A

Y U K O N T E R R I T O R Y

U. S

. A.

Figure 1: Location of the area of interest (AOI) on the map of British Columbia

OVERVIEW OF DATASETSThe datasets are till analysis results from the TREK geochemistry project. Two till survey results were processed include: • Geochemical and Mineralogical Data, TREK Project, Interior Plateau, British Columbia; Geoscience BC, Report 2014-10 (Jackaman and Sacco, 2014). • Geochemical Reanalysis of Archived Till Samples, TREK Project, Interior Plateau, central BC (parts of NTS 093C, 093B, 093F & 093K); Geoscience BC, Report 2015-09 (Jackaman and Sacco, 2015). A total of 677 new samples, 1456 archived samples plus quality control samples were analyzed using di�erent

U . S . A .!!!!!

\\\\\

occ c ee a

n

!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!

!!!!!!!!

!

PPaa

cc ii ff i c

AL

BE

RT

A

VVICTORIA

VaVaV ncouvvveeeer

Princeee George

NNaaazko

VaVaV nderhoof

Anahim LLLLaaakkke

120°0'0"W125°0'0"W130°0'0"W135°0'0"W

55°0

'0"N

50°0

'0"N

B RR I TT I S H CC OO LL UU MM B I AA

U. S

. A.

!

!

!

!

Prince George

Nazko

Vanderhoof

Anahim Lake

123°0'0"W

123°0'0"W

124°0'0"W

124°0'0"W

125°0'0"W

125°0'0"W

126°0'0"W

126°0'0"W

54°0

'0"N

54°0

'0"N

53°0

'0"N

53°0

'0"N

52°0

'0"N

52°0

'0"N

TREK mapping outlineTREK geochemistry project

124°0'0"W

124°0'0"W

125°0'0"W

125°0'0"W

126°0'0"W

126°0'0"W54

°0'0

"N

54°0

'0"N

53°0

'0"N

53°0

'0"N

52°0

'0"N

52°0

'0"N

Intrusive rockMetamorphic rockSedimentary rockUltramafic rockVolcanic rock

20 0 2010 kmO

Figure 2 : Distribution map of basal till in AOI, subdivided based on attributed geology unit from database.

5

REFERENCES. Angen, J.J., Westberg, E., Hart, C.J.R., and Rahimi, M. (2015) Preliminary bedrock geology of the TREK area, Central British Columbia. Geoscience BC map 2015-10-01, Scale 1:250000.. Cheng, Q., Bohnam Carter, G., Wang. W., Zhang, Sh., Li, W. and Qinglin X., (2011) A spatially weighted principal component analysis for multi-element geochemical data for mapping locations of felsic intrusions in the Gejiu mineral district of Yunnan, China, Computers and Geosciences, V. 37, P 662-669.. Grunsky, E.C., (2007)Advances in Regional-Scale Geochemical Methods, The Interpretation of Regional Geochemical Survey Data, Geological Survey of Canada, Natural Resources Canada.. Jackaman, W. and Sacco, D. (2014, 2015), Geochemical Datasets, TREK Project, Interior Plateau, British Columbia; Geoscience BC data.

For more information, contact:Mana Rahimi, PhDmrahimi@eos.ubc.ca