integration of geological data sets for gold exploration in nova scotia

8/10/2019 Integration of Geological data sets for Gold Exploration in nova Scotia

http://slidepdf.com/reader/full/integration-of-geological-data-sets-for-gold-exploration-in-nova-scotia 1/8

0099-1112.88 /5411–1585$ 02.25/0

©1988 American Society for Photogrammetry

and Remote Sensing

Integration of Geological Datasets for GoldExploration in Nova Scotia

G. F. Bonham-Carter, F. P. Agterberg, and D. F. Wright

Mineral Resources Division, Geological Survey of Canada, 601 Booth Street, Ottawa, Ontario KIA OE8, Canada

ABSTRACT: A variety of regional geoscience datasets from Nova Scotia have been co-registered and analyzed using ageographic information system (GIS). The datasets include bedrock and surficial geological maps, airborne geophysicalsurvey data, geochemistry of lake-sediment samples, and mineral occurrence data. A number of line features, includingstructural lineaments, fold axes and formation contacts, have also been digitized. The GIS uses a quadtree structure, ideallysuited to a mixture of polygonal-thematic (e.g., geological maps) and continuous "grey-scale" (e.g., remote sensing, airbornegeophysics) raster images. The goal of the study was to create a map showing areas favorable for gold mineralization, basedon the distribution of 70 known gold occurrences. Initially, a multi-element geochemical signature was generated using aregression analysis to find the linear combination of geochemical elements that best predict lake catchment basinscontaining a gold occurrence. A predicted gold occurrence map, based on the geochemistry alone, was produced. A methodusing Bayes' rule was applied to combine other factors important for gold prediction with the geochemical signature. Aunique conditions map shows all those areas where a unique set of overlap between the predictor maps occurs. For eachunique condition, an a posteriors probability was calculated, resulting in a map depicting probability of gold mineralization.This map confirms that the major known gold districts coincide with areas of high probability. Several new areas of high potential are indicated by the model, although exploration follow-up has not yet been carried out.

INTRODUCTION

A MAJOR ACTIVITY of government geological surveysconsists of mapping the composition and structure of theEarth's crust using b oth traditional field methods and advancedgeochem ical and geoph ysical techniques. The integration of such surveys, stored as paper maps and digital datasets for the purposes of mineral resou rce estimation and explo rat ion , is atask tailor-made for a geo graphic information system (GIS).

Despite the previous development of excellent software for spatial and statistical analysis of regional geological datasets.e.g., SIMSAG (Chung, 1983), CHARAN (Botbol, 1971), andGIAPP (Fabbri, 1985), mathematical tools for carrying out

mineral resource assessments have not been widely adopted.The re asons for this are many, but some important factors have been the dif ficulty of importing diverse data types intogeographically co-registered databases, the lack of goodcomputer graphics, and slow user interaction inherent in manysoftware packages. We believe that with GIS these factors can be ove rcome to a great ex tent.

In this paper we describe procedures for integratinggeological map data (polygonal, thematic) with structuralinformation (lines), lake-sediment geochemical data (pointdata associated with multiple attributes), airborne geophysics(raster images), and mineral occurrence data (points). We use both mult iple regression analys is and a new method of combining binary map pattern"sing Bayesian statistics tocreate a derived map showing areas favorable for gold

exploration in part of east mainland Nova Scotia.

SOFTWARE

We employed a quadtree*-based GIS (SPANS) for analysingregional geological datasets (TYDAC, 1987). SPANS uses a raster data structure with a variable pixel size. Raster images up to amaximum resolution of 215 by 215 pixels can be handled, althoughnormally most SPANS universes' use maps with a quad level* of 10to 12, i.e., with a size between 2 10 and 212 (1024 and 4096) pixels.The work described here was carried out on an 80386 PC with70 mb hard drive, a Number Nine color graphics card, and acolor monitor. SPANS will accept a variety of vector andraster data inputs, allows forward and backwardtransformations from about 20 p rojections to geograph ic(lat/long) coordinates, and provides a powerful set of

analytical tools for analysing multiple maps. Because SPANS permits the use r to move readily to DOS, other DOS-compatible software (e.g., editors, statistical packages, locally-developed programs) can be executed on mutually shared datafiles.

GEOLOGY AND MINERALIZATION

The study area (Figure 1) is underlain by three major rock units. The Goldenville and Halifax Formations are Lower Paleozoic quartz wackes and shales, respectively. They areintruded by Middle Devonian granites (Keppie, 1984). Goldoccurs in quartz veins, usually confined to the Goldenville

Formation. Mining of gold has been carried out intermittentlysince the mid-19th century, to the present day. About 70 goldoccurrences are officially recorded in the study area(McMullin et al., 1986). About 30 of them have known produ ction.

The mechanism of gold mineralization is not well under-stood. Most of the gold-bearing veins are concordant andoccur at or near the crests of folds. The gold occurs withinquartz-carbonate veins with associated arsenopyrite and/or pyrrhotite and minor but valuab le amounts of galena,chalcopyrite, sphalerite, pyrite, and sometimes scheelite andstibnite. The veins are commonly confined to pyrite- or arsenopyrite-rich black shale horizons, and occur throughoutthe Goldenville Formation (Kontak and Smith, 1987). In someareas. the gold appears to be related to faults orientated NW-SE (e.g., Bonham-Carter et al., 1985a) . Some writers have

suggested that mineralization may be related to theGoldenville-Halifax contact (Graves and Zentilli, 1982).Discussion relating to the origins of the deposits are complex,and no consensus has been achieved. Proposals include (a)synsedimentary depositio n on the seafloor (b) deposition earlyin the geological history of the area from metamorphic fluidsand multicyclic remobilization of components duringdeformation and (c) deposition late in the orogenic historyfrom fluids derived either from granitic magmas or other sources deep in the crust.



PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING

Fig. 1. Location ma p, showing area in S.E . mainl an d of Nov a Scotia (inset), an d the principal geological units and gold occu rrences, from Wrigh t et a/. (1 988). Thirty-two

of the largest occurrences are sh own as open c ircles, flagged by number, and listed to show map sheet number , production and name. Solid squares show minor

occurren ces. Large open rec tangles indicate majo r gold-p roducing districts.

In this paper, GIS is used to examine empirically the spatialrelationship of the following factors to known goldoccurrences: multi-element lake sediment geochemistry,lithology, distance to formation contacts, and distance toanticlinal fold axes. A prob abilistic model is then developedfor predicting gold min eralization using these empiricalrelationships.

DATA INPUTS

The data input to the GIS were very diverse (Table 1). The

bedrock and surficial geology m aps were ras ter-scanned using

an Optronics 4040 at Canada Lands Data Systems (CLDS)

(Bonham-Carter et al., 1985b). Manuscript preparationinvolved tracing closed polygon bo undaries on to a stable

base, using a 0.006- inch b lack line. Identi fy ing numb ers wereassigned to each polygon, tagged by hand-digitizing, and used

as pointers to an associated attribute file. Output from theCLDS system con sisted of an arc-node vector file,subsequently converted to a raster format (Steneker andBonham-Carter, 1988). Geochemical data on lake sedimentsamples from about 550 sites were obtained as a samples ×varia bles ASCII file, containing analyses for 16 che micalelements for each sample. The catchment area or basinsurrounding each sampled lake was taken as the zone of influence of the sample. A map of catchment basins, one per sample, was drawn using a topographic map and raster-scanned as above. Again, each polygon number was used as a pointer to the associated sample reco rd in the att ribute file(Wright et al., 1988).

Fold axes and structural lineaments were table-digitiz ed. Thesame method was used to enter the locations of roads andtowns for reference purposes.



TABLE 1. SOURCES AND TYPE S OF VECTOR AND RASTER INPUT DAT A.

Name of M ap Type Digi tal Captu re Attributes

Bedrock geology Polygonal, themat ic Ras te r- scanning of polygon boundaries2

Map units

Surficial geology1 Polygonal, thematic Raster-scanning of polygon boundaries2

Map units

Lake catchment bas ins Polygonal , thematic Ras te r- scanning of polygon boundaries2

Lake se diment samp les, 16geochemical elem ents

Fold axes Lines Table digitizing3 Anticlines, synclines, age.

Lineaments, faults Lines Table digitizing3 Length, orientation

Airborne radiometrics1 Raster, grey scale Gridded from digitalflight line data4

K, eTh, eU, plus ratios

Airborne magnetics1 Raster, grey scale Gridded from digitalflight line data4

Total field, vertical gradient

Landsat MSS Raster, grey scale Comp uter compatible

tapes5

4 spectral bands

Mineral occurrences Points Digital database6 Elements, status

Roads1 Lines Table digitized 3 Major, m inor

Towns, cities1 Points Table digitized3 Size

1 not directly used for gol d prediction in this study2 by Canada Lands Data Systems, Environment Canada3Centian digitizing table, using TYDIG (part of SPANS system)4Gridding by geophysical personnel, Geological Survey of Canada5Purchased from Ca nada Cen tre for Remot e Sensing6CANMINDEX, National mineral o ccurrence database, Geological Survey of C anada

Airborne geophysical images were imported in an 8-bitraster format by downloading from a VAX mainframe. Eachimage (UTM projection) was geo-referenced using the

southwest corner in UTM coordinates and image dimensionsin metres. A Landsat MSS image was geometrically correctedon a micro-based image analysis system (EASIPACE softwaredeveloped by Perceptron Incorporated) to a UTM base andimported to SPANS by a similar route.

Finally, point data de fining the locations of gold occ urrenceswere downloaded from CANMINDEX, a mineral occurrencedatabase maintained at the Geological Survey of Canada(Picklyk et al., 1978), updated with data from McMullin et al.(1986 ).

MAP C REATION

Maps in the quadtree structure were created for each of the polyg onal-thematic inputs (e .g., geology) and grey-scale raste r inputs (e.g., Landsat). A title and legend was created for eachmap by using a text-editor to add entries to ASCII dictionary

files. Screen images were saved in a browse* file for futurereference. Re-display of browse file images is virtuallyinstantaneous, and files could be re-ordered for demonstrations.

It was possible to superimpose any point or line file as vectors onto any image for display purposes. Vector files could also beconverted into quadtree maps by creating corridors or buffer zonesaround lines. This is particularly important for geological problems,where "distance to" linear features is often significant in studies of mineralization. For example, 20 corridors were spaced at 0.25-kmintervals (0 to 5 km) around anticlinal axes, thereby creating a mapshowing distance to these structures. As will be shown below, asignificant proportion of gold occurrences lie close to fold axes, andthis corridor map is important for modeling gold mineralization. Inaddition, corridor maps showing distance to northwest trendinglineaments, distance to the granite contact, and dist ance to the

Halifax-Goldenville contact were prepared, using the same corridor-generating routine, permitting an analysis of gold occurrences inrelation to these linear features.

DATA ANALYSIS

INTRODUCTION

Although gold exploration has been carried out in NovaScotia for over 100 years, the 70 gold "occurrences" representonly the discovered gold resources of the study area. The purpose of spat ial data integration was to make a map whichwould predict the location of new deposits. The new map was based on those factors that are associated wi th the loca tion of known gold occurrences.

The predictive strategy for mapping areas favorable for goldmineralization involved two stages. In the first stage, the multi-element geochemical data (16 elements) were combined into asingle new variable. As we sh ow below, th is variablerepresents the prediction that gold mineralization occurs usingknown gold occurrences as a dependent variable. Each lakesediment sample is assumed to exhibit a geochemical responserepresentative of the rocks and mineralized zones occurring inthe catchment area of the sampled lake. Regression was usedto combine the ge ochemical var iables into a weighted sum that best pred icts whether a basin contains a known occurrence.The resulting map of predicted gold occurrences may be usefulfor locating new deposits from the geochemistry alo ne, but itcan only help in those areas covered by the sampled catchment bas ins. There are several other factors observab le throughoutthe region that may be useful guides to gold mineralization.These are combined with the geochemical evidence in thesecond stage.

Recent work by Agterberg (in press) has provided a newmethod for combining map p atterns using Bayesian statistics.The simplest kind of map for this exercise is one which showsonly the presence or absence of a single theme – a binary

pa ttern. Although the metho d is not confined to binary maps,

most geologists tend to think of predictor variables that



INTEGRATION OF GEOLOGICAL DATASETS FOR GOLD EXPLORATION

are either "anomalou s" or "background," so the thresholding of maps into binary form is appealing. For example,"bac kground" levels of a geochemical element cover aconcentration range believed to be associated with the particular rocks an d so ils o f the area; "anomal ous" levelswould be above this range, and might be due to mineralizationor other process es. Each binary map is associated with positive

and negative weights, depending upon whether or not the pattern is present. Such weights are more easily interpretedthan regression coefficients. The weights are determined usingthe locations of known deposits, so it is assumed that sufficientexploration has been carried out to make reliable estimates of the coefficients. The final product from the second stage is anew predicted gold map that should reflect the locations of known mineralization, as well as provide new target areas.

MULTI-ELEMENT GEOC HEMICAL SIGNATURE

Determining the multi-element geochemical signature that best predicts those lake ca tchment basins containing knowngold occurrences (Wright et al., 1988) involved adding a newattribute column to the lake sediment file indicating whether each lake basin contains a gold occurrence (score = 1) or not(score = 0). In practice this was achieved using the SPANS point result* option, thereby attaching the lake sediment bas innumber to the gold point file, and using this information toupdate the geochemical attribute file. This modified file wasthen entered in to SYSTAT, a DOS-co mpatible statistical package t o carry out regressio n analysis as il lustrated below .

Let Y be a binary variable denoting presence/absence of agold occurrence. Let X j , j = 1, 2,..., 16, be the concentrationvalues of the 16 geochemical elements, log transformed tostabilize the variance. Then let

be the predicted gold occurrence at the ith catchment basin,where the coefficients, b are determined by ordinary least-squares regression. In practice, this was carried out by astepwise method, reducing the number of variables andcoefficien ts requiring interpretation.

These regression coefficients represent a multi-elementgeochemical signature for predicting gold mineralization. Anew column for predicted gold occurrence, Y, was added tothe lake sediment attribute table. This new attribute columnwas converted to a map based on the catchment basins,subdivid ing the range of predicted values into discrete classes.This step was carried out using the SPANS modelinglanguage, written in a special ASCII file ("equation.inp")*

reserved for this purpose.BINARY MAP ANALYS IS

In order to combin e other fa ctors with the geochem icalsignature, the second stage of the analysis employs the newmethod described by Agterberg (in press) for modelingconditional probabilities. This method is more con venient touse than multiple regression fo r several rea sons. First, it avoidsthe requirement to subdivide the region into cells, each cellassociated with an attribute list (e.g., geochemical elements,"distance, to" measures, presence/absence of mineralization,and rock type). In order to capture the geometrical informati onabout "distance to" linear features adequately, a very largenumber of small sampling cells must be created, and this isundesirable because of the resulting large attribute file and

degree of spatial autocorrelation present in such a dataset.Secondly, the binary map method is better able to cope withthe problem of missing data. For example, the lake catchment basins do not c over the wh ole study area , whereas the other maps (rock types, "distance to" maps) occur ubiquitously.

Using regression, one must either assume mean values for those missing observations, or simply omit those regions withincom plete d ata.

The e quations for the map pattern analysis are as follows. LetP pri or be the a priori prob abili ty of a go ld depos it occurringwithin a small area of arbitrary but known size (e.g., 1 km 2).The a priori odds are th en defined by

O pri or = P pri or / (1 - P pri or ).

The a posteriori odds can be exp ressed as

and the a posterior probabili ty of a gold deposit occu rring is

P post = O post / (1 + O post).

The weights for the jth pa ttern are determ ined from

The c onditiona l probability terms a re ca lculated from

where

Adt = number of 1 km2 units containing a deposit in the totalstudy area, Adj = number of 1 km 2. units containing a depo sit in pattern j, A j = area of pattern j, km2, and A t = total study area, km2.

The a priori prob abili ty P prior can be estimated as A dt/A t.

Bayes' rule assumes that the patterns are conditionallyindependent. This will not always be the case, and a generaltest for conditional independence can be made by comparingthe predicted versus observed number of deposits, as described by Agterberg et al. (in press).

In order to determine the optimum cutoffs for classifying pa tterns in to binary presence/absence (absence reflec ts "not present" as opposed to unknown) , the weights W+ and W - can be ca lculat ed for a succession of cutoffs and, und er normalconditions, the maximum value of (W+ - W -) gives the cutoff atwhich the predictive power of the resulting pattern ismaximized.

The numerical area calculations were made in SPANS using,,area analysis"' of the map in question and a "point result"' of the map by the gold deposi t point file. Weights for each patternwere computed using an external program. The final mapshowing a posteriors probab ilit ies was ca lculated using theSPANS modeling language, after creating a "uniquecon dition s"' ma p. This map consists of the set of unique polygons, each one defined as that area with a unique overlap




(a)

(b)

(c) (d)

of the binary patterns being modeled. Finally, the a posteriors probab ili ty map can be displayed. A "poin t select "* is usefulto show which patterns are actually present for each goldoccurrenc e, and the associated a posteriors probab ility value.

RESULTS

Figure 2a shows a map of the geochemical signatureobtained using the regression coefficients in Table 2 andthresholded to a binary pattern. Several cutoff thresholds weretried, as discussed in Agterberg et al. (in press), to ma ximize

(W+ - W -). The coastline, fault contact, and limit of lakecatchment areas were displayed as vector overlays. Figures 2b,2c, and 2d show the mapped areas of the three bedrock units.These three patterns are mutual ly exclusive, and together cover the whole study area. As a consequence, no W- weights wereused, as shown in Table 3, although it is to be noted that theW+ weights can actually be negative.

FIG. 2. M ap patterns used to predict gold (Au) occurrences. Black dots show locations of known gold occurrences. (a) Geochem ical signature.

Note that, outside the c atchment basins, the signature is unknown. (b) Goldenville Form ation. (c) Halifax Formation. (d) Devonian Grani tes

Figure 3 shows four different types of corridor map patterns.In each case, successive corr idors were crea ted aro und a vector feature, at intervals of 0.25 km out to 5 km. Optimal cutoffs

(Table 3) were calculated by finding the distance at which (W+

- W -) was maximized, as shown in Agterberg et al. (In press).Finally, the map of P po st was generated (Figure 4). The

program was set up with in teractive prom pts so that seve ral




(a) (b)

(c) (d)

alternatives could be tried experimentally, omitting one or more maps to evaluate the robustness of the results to changesin the assumption s of the model.

From the weights in Table 3, it is clear that the presence of grani te stron gly downweigh ts the pro babi lity of go ldmineralization, whereas the presence o f the favorablegeochemical s ignature and proximity to an ticlinal axes are,

stron g positive factors. The presence o f the GoldenvilleFormation, particularly where in close proximity to the Hal ifaxcontact, is moderately favorable. The proximity to granite and proximity to northwest lineaments have little effect on the probability map, at least in the study area.

FIG. 3. Map patterns used to p redict gold occurrences, based on corridor neigh borhoods rou nd a linear feature. Black dotsshow loca tions of k nown gold occ urrenc es. (a) Anticline axes with cor ridor s. ( B) N.W. lineaments with corridors. ( C)Goldenville-Halifax contact with corrido rs. (D) Granite contact with corridors.

TABLE 2. REGRESSION COEFFICIENTS1 AND THEIRSTANDARD ERRORS GIVING THE MULTI-ELEMENTGEOCHEMICAL SIGNATURE THAT BEST PREDICTS GOLDDEPOSITS.

Element b S.E.b

Au 0.196 0.021

As 0.009 0.009

W 0.037 0.029

Sb 0.005 0.022

Constant 0.1280 0.024

1Stepwise regression was u sed resu lting i n the selection of four out of the original 16 elements, from Wright et al.(1988).

TABLE 3. WEIGHTS FOR MODELING POSTERIORPROB ABILITY OF A GOLD DEPO SIT OCCURRING IN A 1 KM 2 AREA

Map Pattern W+ W -

N.W. Lineamen ts -0.0185 -0.0062

Anticline axes 0.5452 -0.7735

Geoche mical Signatu re 1.0047 -0.1037

Goldenville-Halifax Contact 0.3683 -0.2685

Granite Contact 0.3419 -0.0562

Bedrock geolo gy1

: Halifax Formation -0.2406 –

Goldenville Formation 0.3085 –

Granite -1.7360 –




FIG. 4. Map of a posteriori probability of a gold deposit

occurring in a 1 km 2 area. Black dots show locations of known

gold occurrence s.

1A ternary pattern where units are mutu ally exclusive, andno negative weights are used.

DISCUSSION

This study could have been carried out using a mix of existing computer programs for image analysis and statisticalanalysis. However, the advantages of using a GIS were

• relat ive ease of importing diverse map inputs, and creating aco-registered database;

• ability to move betwe en GIS and other DOS-compatiblesoftware packages;

• interactive graphics capability, with windowing, mapoverlays, and vector overlays permitting experimentation not previ ousl y practic al;

• integration of corridor generation, unique conditionsmapping, area analysis, and modeling; and

• the browse feature, which is very useful for keeping track of both the development and final stages in a data integration projec t.

CONCLUSIONS

Spatial data integration for mineral resource assessment andexploration using digital databases is greatly facilitated using aGIS in association with other software. Advanced GIS packages may provide breakthroughs which will bridge thegap between the traditional manual overlay approach and

mathem atical methods usin g multivariate statistics and imageanalysis. The method of combining map patterns usingBayesian statistics is practical and intuitively appealing because it is closer to the “sea t-of-the-pants” approach of theexpl oration geol ogist than are statistical re gress ion methods. Inthe Nova Scotia example, the map showing probability of goldmineralization indicates several areas of favorable mineral potent ial , wi th no known occurrences. Although the predictedgold map is useful itself, the real benefit of this study for anassessment of gold p otential in Nov a Scotia would derive fromgeologists performing their own integration experim ents, giventhe datab ase and the GIS with which to manipulate the data.

A forthcoming paper (Agterberg et al., in press) discussesthe problem of estimating un certainty of the probabilityestimates, and using a goodness-of-fit test for the assumption

of conditional independence. Uncertainty is due to manyfactors, but two important sources of error are associated withthe estimates of the weigh ting factors, and with the incomp letecoverage of one o r more data layers.

ACKNOWLEDGMENTS

This work was supported by the Geological Survey of Canada under the Canada-Nova Scotia Mineral DevelopmentAgreement, 1984-1989, a subsidiary to the Economic RegionalDevelopment Agreement. We acknowledge the contributionsof seve ral No va Scotia M ines and Energy geologists, par ticularly Peter Rogers and Dun can Kappi e. Al Sangster,Andy Rencz, M ike Steneker (Geological Survey of Canada),and Jeff Harris (Intera Technologies) made important

contributions to the work. TYDAC personnel providedvaluable technical support. We thank Andy Rencz and JimMerchant for the ir comm ents.

GLOSSARY OF SPANS TERMS

Area Analysis - An operation which produces a table of areas for each map class. Two-map area analysis produces a two-way table of areas of class overlaps.

Browse File - A directory of screen images saved in compact formdirectly from the graphics board. Can be re-ordered and re-displayedquickly.

Equation.inp - Text file containing statements to control modelingand classification of maps . Created by the operator using a texteditor.

Potential Mapping - A series of functions used for, interpolation of point data.

Point Result - Used in conjunction with a statement in Equation.inp,this operation adds one or more attribute columns to a point file,indicating the attribute value of one or more maps at point locations.

Point Sample - Operation to generate new set of points on a gridwith a pre-set spacing. Points may be confined to selected themes.Used in conjunction with point result to "resample" a series of mapson a regular grid and produce an attribute file.

Quad level - Defines the pixel resolution of a specific map layer. Theminimum pixel size in metres is determined by dividing the width of the universe in metres by 2 raised to a power equal to the quad level.Usually in the range of 9 to 12; must be <15.

Quadtree - A raster data structure that uses a variable pixel size,depending on the spatial homogeneity of the image. Efficient for datacompression of thematic maps, and allows for fast search of thedatabase with Morton coordinates - a referencing system that usesquad level and quad position.

Unique Conditions - An operation which produces a map where the polygons are defined by the overlap combinations of up to 15selected input maps. Used for modeling operations.

REFERENCES

Agterberg, F.P., in press. Systematic approach to dealing withuncertainty of geoscience information in mineral exploration,APCOM89, Las Vegas , March 1989.

Agterberg, F.P., G.F. Bonham-Carter, and D.F. Wright, in press. Stat istical pattern recognition for mineral exploration, Pr ocee di ngs CO -G EODATA Symp osium on Comp uter Applications in Reso urce E xplora tion, July 1988, Espoo,Finland.

Bonham-Carter, G.F., A.N. Rencz, and J.R. Harris, 1985a.Spatial relationship of gold occurrences with lineamentsderived from Landsat and Seasat imagery, Meguma Group, Nova Scotia, Proc. 4th Thematic Conference, Remot e Sensing for Exploration Geology, San Francisco, April 1985, pp. 755-768.

Bonham-C arter, G.F., D. J. E llwood, I.K. Crain, and J.L.Scantland, 1985b. Raster Scann ing Techniques for theCapture, Display and Analysis of Geological Maps, CanadaLands Da ta Systems, Report R003210, 12 p.

Botbol, J.M., 1971. An application of characteristic analysis tomineral exploration, Proc. 9th Int. Sym. on Techniques for Decision-Mak ing in the M ineral Indust ry, CIM M, SpecialVolum e 12, pp. 92-99.

Chung, C.F., 1983. SIMSAG: Integrated computer system r use in evaluation of mineral and energy resources, lour. Math.Geol., Vol. 15, pp. 47-58.




Fabbri, A., 1985. Image Process ing of Geolo gical D ata, Van Nostran d Rein hold C ompa ny, 244 p.

Graves, M.C., and M. Zentilli, 1982. A review of the geologyof gold in Nova Scotia, Geology of Canadian Cold Depo sits,CIMM , Special Paper, pp . 233-242.

Keppie, J.D., 1984. T he Appalach ian co llage, The CaledonideOrogen Scandinavian and Related Areas (edited by D.C. Geeand B. Stuart), Wiley and Sons.

Kontak, D.J., and P.K. Smith, 1987. Meguma gold: The bestkept secret in the Canadian mining industry, Prospectors andDevelo pers Association of Canad a Annual Meeting.

McMullin, J., G. Richardson, and T. Goodwin, 1986. Cold Compilation of the Meguma Terrane in N ova S cotia, NovaScotia D ept. of Mines and Energy, O pen Files 86-055, 05 6.

Picklyk, D.D., D.G. Rose, and R.M. Laramee, 1978. Canadian Miner al Occurrence Index (CANMINDEX) of the Geological Survey of Canada , Geological Survey of Canada Paper 78-8,27 p.

Steneker, M., and G. F. Bonham -Carter, 1988. Computer Program for Converting AreNode Vector Data to Raster Format, Geological Survey of Canada, Open File 103/88, 16 p.

TYDAC, 1987. Spatial Analysis System Reference Guide,Version 3.6, TYDAC Technologies Inc., 1 600 CarlingAvenu e, Ottawa, C anada, K1Z 8R 7, 300 p.

Wright, D.F., G.F. Bonham-Carter, and P.J. Rogers, 1988.Spatial data integration of lake-sediment geochemistry,geology ;And gold o ccurrences, Meguma Terrane, Eastern Nova Scotia, Prospec ting in Areas of Glaciated Terrain,CIMM Meet ing, Halifax, Sep t. 1988, p p. 501 -515.

integration of geological data sets for gold exploration in nova scotia

Documents