APPLICATION OF CATCHMENT ANALYSIS TO REGIONAL STREAM ...
Transcript of APPLICATION OF CATCHMENT ANALYSIS TO REGIONAL STREAM ...
APPLICATION OF CATCHMENT ANALYSIS TO
REGIONAL STREAM SEDIMENT DATA SETS
Dr. Dennis Arne
Consulting Geochemist
PGeo (BC), MAIG (RPGeo), MAAG (Fellow), MCIM
Telemark Geosciences/CSA Global
AIG Exploration Geochemistry Symposium – December 3, 20181
AIG Exploration Geochemistry Symposium – December 3, 20182
Statement of problem
Background to catchment analysis
Examples of interpretive approaches
Using catchment analysis to plan surveys
Conclusions
OUTLINE
ACKNOWLEDGEMENTS
Former colleagues at CSA Global & ioGlobal
Yao Cui, British Columbia Geological Survey
Funding from Geoscience BC
Support from the AIG WA branch to attend
AIG Exploration Geochemistry Symposium – December 3, 20183
STATEMENT OF PROBLEM
How do we correct stream sediment data for the effects of variable
geochemical background due to different lithological units?
How do we take into account surficial processes such as scavenging
of metals onto secondary Fe and Mn oxides, clay minerals or organics?
How do we correct for the differential effects of dilution in catchment
basins of differing area?
What is the optimal catchment area for sampling and has the region
been effectively sampled? (i.e. are there opportunities for further
exploration? Are we over- or under-sampling).
AIG Exploration Geochemistry Symposium – December 3, 20184
AIG Exploration Geochemistry Symposium – December 3, 20185
THEMATIC MAP VS TARGETING PRODUCTS
How do we get from this …. To this?(From Arne et al., 2018)
AIG Exploration Geochemistry Symposium – December 3, 20186
GRIDDED PERCENTILE IMAGE OVERLAIN WITH STRUCTURAL TRENDS
(From Arne & Blumel, 2011)Or this?
AIG Exploration Geochemistry Symposium – December 3, 20187
Collate sample information and analytical data
Verify location datums and analytical units
Geochemical data conditioning
Determine optimal analytical data set and data quality
Treat values below lower limits of detection
Obtain DEM at suitable resolution
Pit-fill to remove topographical sinks
Derive or import hydrology layer
Validate sample locations against hydrology
Locations may need adjustment
Generate individual catchments for each sample
Each sample to be attributed with entire catchment
Attribute samples/catchments with lithological information
Geological legends must be simplified for statistical analysis
CATCHMENT ANALYSIS METHODOLOGY
AIG Exploration Geochemistry Symposium – December 3, 20188
DERIVATION OF CATCHMENTS - 1
Derived from digital elevation model (DEM).
Suitable data sources include Shuttle Radar
Topography Mission (SRTM – 30m resolution
at equator) or Advanced Spaceborne
Thermal Emission and Reflection Radiometer
(ASTER – 15 to 90m resolution), plus other
commercial satellite, government and
geophysical survey elevation data.
Note minor locational issues with some of
these data sets.
Project area is the White Gold District, Yukon
Territory, Canada.
SRTM DEM for the White Gold District, Yukon, Canada.Squares are 25 km2. Black dots are mineral occurrences.
White Gold District
AIG Exploration Geochemistry Symposium – December 3, 20189
DERIVATION OF CATCHMENTS - 2
Regional Geochemistry Survey samples
from British Columbia, Canada were mainly
collected prior to the widespread use of
GPS.
i.e. physical transcription of data from paper maps
and manual entry into GSC database (plenty of scope
for error!)
Original topographical maps used to locate
samples may have been inaccurate
compared to present day terrain data.
Sample locations were adjusted to lie on
correct drainage, based on original
topographic maps.
Sample location from Vancouver Islandre-located onto TRIM hydrology layer with scan
of original paper map sheet as background
(From Arne & Brown, 2015)
AIG Exploration Geochemistry Symposium – December 3, 201810
DERIVATION OF CATCHMENTS - 3
Previously done by tracing out the
heights of land surrounding the
catchment basin by hand.
Now automated using a number of
software packages – note that not all
allow determination of a catchment
from an arbitrarily located sample
point rather than a stream outlet.
Generally iterative as location
problems are recognised by
nonsensical catchments.
Sample points are from the Yukon
regional geochemical survey (RGS)
program.
Sample locations in green with derived catchment basins in orange outlines.
AIG Exploration Geochemistry Symposium – December 3, 201811
A low -80# Au from the Yukon open file
data in the stream leading from Golden
Saddle led to the assumption that stream
sediment data didn’t work in this area.
While subtle anomalies are present at
Golden Saddle and the Coffee deposits
using raw element data, an index using As
and Sb levelled for dominant catchment
lithology provides an enhanced response.
Note that the raw Au is only weakly
anomalous at Golden Saddle if an
average of the primary and field duplicate
sample are plotted.
WHITE GOLD DISTRICT, YUKON TERRITORY, CANADA
(From Arne & MacFarlane, 2014)
AIG Exploration Geochemistry Symposium – December 3, 201812
Raw data
Residuals following regression against regolith control elements
Data levelled by dominant catchment lithology
Data levelled by presence/absence of a particular unit
Data levelled by catchment weighted background value
Multiple regression analysis using catchment geology
Productivity analysis
Weighted sums model
Weighted sums model adjusted for catchment area size
Residuals following regression against principal components
Machine learning algorithms
DATA ANALYSIS APPROACHES
Inc
rea
sing
ma
the
ma
tica
l co
mp
lexity
NORTH VANCOUVER ISLAND CASE STUDY
AIG Exploration Geochemistry Symposium – December 3, 201813
• Data compiled for 1835 samples in total
• Catchment basins delineated for 1725 samples
(From Arne & Brown, 2015)
IMPORTANCE OF LITHOLOGY
• Geology is interpreted in terms of lithology, rather than map units.
• The distribution of Cu is clearly influenced by the Karmutsen basalt; difficult to
distinguish Cu anomalies associated with porphyry Cu deposits in those areas.
AIG Exploration Geochemistry Symposium – December 3, 201814
Raw Cu DataMt Hall Gabbro?
(From Arne & Brown, 2015)
LEVELLING Cu DATA FOR LITHOLOGY
• Levelling by dominant lithology or presence/absence of basalt both reduce
lithological effects.
AIG Exploration Geochemistry Symposium – December 3, 201815
Levelled by Dominant Lithology Levelled by presence/absence basalt
(From Arne & Brown, 2015)
WEIGHTED AVERAGE BACKGROUND VALUES
• Average background values have been estimated for most lithologies.
• These have been used to calculate weighted average background values, which
were then used to calculate robust Z-scores ((observed – median)/interquartile range).
AIG Exploration Geochemistry Symposium – December 3, 201816
Z-score Cu “Residuals”
(From Arne & Brown, 2015)
AIG Exploration Geochemistry Symposium – December 3, 201817
Methods based on levelling against known geology depend on the following
assumptions:
That the samples are located correctly
That the geology of the area is accurately known
That the lithological character of mapped units is known and consistent
That there are no minor but geochemically distinct lithological units in the area that
have not been mapped (i.e. black shales)
That erosion from all lithological units is similar so that the contribution from each unit
is consistent and represented by its areal extent within the catchment
Methods that make use of principal component analysis depend on the following
assumptions:
That the principal components represent recognisable geochemical processes
(lithology, scavenging, mineralisation)
ASSUMPTIONS USING CLASSICAL APPROACHES
PRINCIPAL COMPONENT ANALYSIS
AIG Exploration Geochemistry Symposium – December 3, 201818
• The most important principal components from
regional stream sediment geochemical data
usually reflect bedrock lithology.
• PC1 from northern Vancouver Island separates
felsic and mafic lithologies.
Sc
ale
d C
oo
rdin
ate
s
PC1
PC2
PC3
Felsic
Mafic
• PC1 provides a better discriminator for the Karmutsen basalt than raw Cu, which is
influenced by the Mt. Hall Gabbro.
AIG Exploration Geochemistry Symposium – December 3, 201819
Raw Cu Data Inverse PC1
PC1 VS RAW CU DATA
Raw Cu Inverse PC1
(From Arne & Brown, 2015)
REGRESSION AGAINST PC1
• Cu can be regressed against PC1 (lithological control).
• High positive residuals represent Cu above predicted background.
AIG Exploration Geochemistry Symposium – December 3, 201820
(From Arne & Brown, 2015)
METHOD VALIDATION
• Data processing methods were validated against known occurrences.
• Known occurrences in the upper 90th percentile plotted relative to the proportion
captured by raw Cu data.
21
Cu Occurrences No Cu Occurrences
>9
0th
Pe
rce
ntile
<9
0th
Pe
rce
ntile
True
Positives
False
Positives
False
Negatives
True
Negatives
Single Element Methods
Cu
Pro
du
ctiv
ity
We
igh
ted
Su
ms
Mo
de
l
WSM
* C
atc
hm
en
t A
rea
AIG Exploration Geochemistry Symposium – December 3, 2018
AIG Exploration Geochemistry Symposium – December 3, 201822
• Geochemistry in catchment areas >10 km2 mainly reflects regional background
• Catchments > 10 km2 have not been effectively sampled
Sampling Effectiveness
SAMPLING EFFECTIVENESS
(From Arne & Brown, 2015)
CONCLUSIONS
The effects of elevated background metal contents are easily filtered
using a variety of methods of varying sophistication.
Processing methods using several commodity and pathfinder elements
weighted for the effects of dilution work best.
Existing mineral occurrences & deposits are more readily apparent and
new targets evident in processed data.
An empirical assessment of sampling effectiveness indicates whether
there are still exploration opportunities in under-sampled areas and defines
the optimal density of follow-up sampling.
AIG Exploration Geochemistry Symposium – December 3, 201823
REFERENCES
AIG Exploration Geochemistry Symposium – December 3, 201824
Arne, D.C. and Bluemel, E.B., 2011, Catchment analysis and interpretation of stream sediment data from QUEST South, British Columbia, Geoscience BC, Report 2011-5, 25 p. URL < http://www.geosciencebc.com/s/2009-023.asp> [November 2018]
Arne, D. and MacFarlane, B., 2014, Reproducibility of gold analyses in stream sediment samples from the White Gold District and Dawson Range, Yukon Territory, Canada. Explore, No. 164, p. 1-10. URL <https://www.appliedgeochemists.org/explore-newsletter/explore-issues> [November 2018]
Arne, D.C. and Brown, O., 2015, Catchment analysis applied to the interpretation of new stream sediment data from northern Vancouver Island, Canada (NTS 102I and 92L). Geoscience BC Report 2015-4, 41 p., URL <http://www.geosciencebc.com/s/Report2015-04.asp> [October 2017]
Arne, D.C., Mackie, R., Pennimpede, C., Grunsky, E. and Bodnar, M., 2018, Integrated assessment of regional stream-sediment geochemistry for metallic deposits in northwestern British Columbia (parts of NTS 093, 094, 103, 104). Geoscience BC Report 2018-14, 96 p. URL < http://www.geosciencebc.com/s/2016-028.asp> [November 2018]
Arne, D., Mackie, R. and Pennimpede, C., 2018, Catchment Analysis of Re-analyzed Regional Stream Sediment Geochemical Data from the Yukon. Explore No. 179, pp. 1-13. URL <https://www.appliedgeochemists.org/explore-newsletter/explore-issues> [November 2018]
Mackie, R.A., Arne, D.C. and Brown, O., (2015): Enhanced interpretation of regional stream sediment geochemistry from Yukon: catchment basin analysis and weighted sums modelling; Yukon Geological Survey, Open File 2015-10. URL <http://data.geology.gov.yk.ca/Reference/69462> [November 2017]
Mackie, R.A., Arne, D.C. and Pennimpede, C. (2017): Assessment of Yukon regional stream sediment catchment basin and geochemical data quality; Yukon Geological Survey, Open File 2017-4, 36 p. URL <http://data.geology.gov.yk.ca/Reference/79430> [October 2017]