PEER.REVIETYED ARTICTE

Photolineament Factor: A New Computer-Aided Method for Remotely Sensing theDegree to which Bedrock is Ftactured

Kenneth C. Hardcastle

AbstractPhotolineaments are often utilized during exploration forgroun dwater te sources in fracture d be drock; photolineamentsare thought to denote arcas wherc the bedrock may be rela-tively more fractured and, therefore, capable of storing andtransporting significant volumes of groundwater. It is sug-gested that three key charactefistics can be used to rank anarea's potential to store and transmit large volumes ofgroundwater: (1) the number of photolineaments, (2) thenumber of directional photolineament families, and (s) thetotal length of photolineaments which occur within or trav-erse an area of defined radius. The normalized sum of thesethree photolineament porameters is referred to here os aphotolineament factor value. A new computerized approachfor processing the thousands of photolineaments typicallycollected for large study sites (>1.0 km' or 2500 acres) yieldsa contourable grid of photolineament factor values. Such acontour map facilitates rapid quantitative ranking and selec-tion of discrete areas for further evaluation. Resulfs from a900 kmz (220,000 acre) study area in the Georgia Piedmontillustrate this new approach.

IntroductionMany subsurface features, such as fracture zones, faults, geo-logic contacts, and other bedrock discontinuities, have a sur-face expression which can be detected through analysis ofaerial photographs and satellite images. These subsurface fea-tures can greatly influence groundwater systems and are,therefore, important to determine in any program of ground-water resource exploration. The surface expressions of thesesubsurface features are often linear to curvilinear topographicdepressions or tonal discontinuities. "Fracture trace" or pho-tolineament analysis, in conjunction with detailed geologicmapping, is often applied as a practical means of delineatingpossible water-bearing bedrock features because of the de-monstrable relationship between some photolineament andbedrock features (Rich, tgza; Blanchett, 1957; Lattman, 1958;Lattman and Parizek, 1964; Wobber, 1967; Alpay, 1973; Ca-swell ef o.1., 1986; Mabee, 1992; Yin and Brook, 1992). Forexample, existing high-yielding wells generally occur proxi-mal to geologic contacts in the Greater Atlanta Region (Cres-sler ef o1., 1983), proximal to fracture-correlatedphotolineaments in crystalline metamorphic rocks in Maine(Mabee ef o1., 1990), and near lineaments in carbonate rocksin central Pennsylvania (Siddiqui, 1969).

At least one key assumption underlies groundwater stud-

Emery & Garrett Groundwater, Inc., 170 Waukewan Street,Meredith, NH 03253.

PE&RS

ies which utilize nhotolineaments: the photolineaments iden-tified are "real," that is, they are not mLrely an artifact ofeither the observers bias(es) and/or cultural activities. Oneapproach for gaining confidence on the reality of photolinea-ments is to collect and analyze data sets comprised of photo-lineaments observed by multiple analysts, on multiple scalesof imagery, and/or during multiple observational trials. Suchan approach is demanded by the inherent subjectivity ofphotolineament data collection (Wise, 1982); approximately30 percent or less of photolineaments observed on an imageare seen in the same place with the same orientation by dif-ferent analysts (Mabee et ol., 1.ggoi Podwysocki, 1974). Datasets developed with this approach, however, typically con-tain thousands of photolineaments for all but small study ar-eas (<500 acres). Such large data sets require computerizedanalytic methods for practical process times.

Two analytic methods known by the author are consid-ered viable for processesing large lineament data sets: (1) adouble filter process resulting in relatively few photolinea-ments (Mabee et aL, in press), and (z) the photolineamentfactor method described in this paper. Both methods can beused simultaneously for two different and complimentaryproducts: (1) discrete "fracture-correlated coincident photoli-neaments" (Mabee et aL, in press), and (z) a contour map ofohotolineament factor values which enables identificationind ranking of sub-areas of interest (this paper).

The new computerized method described in this paperis appropriate for analysis of large photolineament data sets.Results of this method facilitate ranking of discrete areaswith regard to their potential ability to store and transmitgroundwater. It is assumed that three characteristics allowsuch ranking: (1) the number of photolineaments, (2) thenumber of directional photolineament families, and (3) thetotal length of photolineaments which occur within or trav-erse an area of defined radius. The normalized sum of thesethree parameters is the photolineament factor value. The oc-currence of numerous photolineaments suggests the Iinearfeatures have a "strong expression," that is, they are repro-ducible or "real." Many directional families suggest there arenumerous intersections of photolineaments. A large totallength suggests the features are aerially extensive and, there-fore, likely to be intercepted by other linear features.

The computer program derives photolineament factor

Photogrammetric Engineering & Remote Sensing,Vol . 61, No. 6, fune 1995, pp. 73s-747.

009s-1 1 12l95/6106-739$3.00/0O 1995 American Society for Photogrammetry

and Remote Sensing

PEER.REVIEWED ARTICTE

values for each node in a square grid comprised of partiallyoverlapping collection circles (the grid covers the entirestudv area). Gaussian curves are fit to normalized azimuthfrequency histograms of the photolineaments to derive thenumber, trend, and relative prominence of directional fami-l ies at each node (based on an algorithm of Wise ef o1.(1985)). The method described is designed to rapidly andsemi-quantitatively isolate areas ol interest and is most appli-cable to large study areas such as towns and counties. A 900km' study area in the Georgia Piedmont illustrates this newmethod. It is important to note that areas with 1ow photoli-neament factor values are also revealed. These areas may berelatively less fractured and less likely to transmit groundwa-ter and, therefore, cor-rld be potential targets for land usessuch as waste d isposal .

MethodPhotolineaments are drawn on different scale aerial ohoto-graphs and/or satellite images, digitized into a compirter. andthen the computer program is used to generate a contourmap of photolineament factor values.

Photolineament Data CollectionPhotol ineaments observed through obl ique and stereoscopicviewing are drawn on clear acetate overlays placed on aerialphotographs and/or topographic maps. Coverage extends atleast one average photolineament length beyond the studyarea boundaries to insure complete coverage. Numerousscales of photographs are typical ly used so as not to missfeatures which may be scale dependent (Table 1), but two orthree scales of imagery analyzed two or more t imes each isanother way of developing a large database of photolinea-ments (Mabee et al. ,1990). Once drawn, register points areidentified on each image, and the end points of each photoli-neament are digit ized into a computer. Each photol ineamentis described by i ts length, azimuth f in degrees east of north),and the lat i tude and longitude of each of i ts two end points.To draw and digit ize photol ineaments is the most t ime con-suming port ion of the photol ineament analysis described inthis paper. It is common to spend 1 to 2 hours drawing andanother hour digitizing photolineaments lor each photo-graph. There may be many photographs of a particular scaleto cover large study areas. At this time, only straight photoli-neaments a re cons idered.

Limitations of Photolineament DataThe accuracy of the geographic location of each photol inea-ment is proport ional ly lower on proport ional ly smaller scalephotographs because the photographs cannot be registeredexactly to geographic coordinates and there are sl ight photo-graphic aberrat ions due to lens curvature, airplane pitch, andphotographic edge effects (Table 2). Trend accuracy is about+Lo assuming a location accuracy of to.s mm on the photo-graph register points and photolineament end points fTable2). Zoom-Transfer Scopes can be used to correct for some ofthese inaccuracies, but the large time involved to transfereach photolineament is prohibitive for large data sets. Tomit isate the inaccuracies mentioned. the photol ineament fac-tor method is designed as a "f irst-cut" tool to isolate areas,not specif ic (dri l l ing) targets.

Analyses of photolineament density (number per area)are biased by photolineaments observed on larger scale im-ages because more photolineaments aru generally observedon larger scale images within a given area. However, there isnot a simple relat ionship between the number of photol inea-

740

TABLE 1. Tvprcqr Scmrs or ltnncrnv Usro tn PnoToUNEAMENT ANALysls

Scale Image Feature

1.:24,OOO1:58,0001:80 ,0001:130,0001:250,O00

1:1 ,000,000

Topographic mapCIR* aerial photographB&W** aerial photographCIR aerial photographB&W photomosiac ofSLAR**n imagesB&W photograph of sideilluminated, raised plasticre l ie f map*** *

Topographic lineamentsPhotolineamentPhotol ineamentPhotol ineamentPhotolineament

PseudoLandsatphotol ineament

*CIR : color infrared n* B&W - black and white *** SLAR : SideLooking Airborne Radar **** Wise and Grady, 1985

ments observed and the scale of the image, because of differ-ences in image type, image clari ty, land use, etc. (Table 3).This bias towards larger scale photolineaments, however,translates into a constructive bias towards photolineamentswhich are more accurate in geographic position when ana-lyzing a photolineament data set collected from differentscales of imagery (Tables 2 and 3).

Of course, the results of any analysis are dependentupon the nature of input data. In the case of photolinea-ments, the photolineaments must be correlated with featuresin the bedrock if they are to be used as indicators of such. Inthe photolineament factor method, input data can be pre-processed or input untouched. Pre-processing should focuson filtering out those photolineaments which cannot be cor-related with bedrock features, such as those which reflectcultural features: i .e., powerl ines, roads, human-made drain-age ditches, etc. If comprehensive data on the bedrock's fab-r ic elements exist, which wil l only be avai lable inenvironments where bedrock is nearlv evervwhere exposed.then the pho lo l ineanren l da ta can be h l te red to inc lude on lythose photolineaments with trends that match the trends ofbedrock fabric features.

In all studies, correlation must be sought to justify use ofany photolineament analysis results, Correlation is generallya two-step process of comparison between trends of photoli-neaments and bedrock structural fabric data elements: (1)rose diagrams are generated for photolineament data basedon dif ferent sub-groupings of the data set ( i .e., cumulative,within various radius sub-areas around mapped outcrops,those photolineaments that are underlain by specific rocktypes and/or geologic sett ings, etc.), and (Z) specif ic photol i-neaments which traverse exposed bedrock are comparedwith mapped structures.

Photolineament Factor AlgorithmTo quantify the degree to which bedrock in a given locationmay be fractured or contain other discontinuities such asfault zones or pronounced differential weathering along com-positional layering, three photolineament parameters are as-sessed: the number of photolineaments, the number ofdirectional photolineament families, and the total length ofphotolineaments. The normalized sum of these three parame-ters is defined as the photolineament factor value. The occur-rence of numerous phbtolineaments suggests the linearfeatures have a "strong expression," that is, they are repro-ducible or "real." Many directional families suggests thereare numerous intersections of photolineaments. A laree totallength suggests the features are aerial ly extensive and] there-fore. likelv to be intercepted bv other linear features.

PE&RS

PEER.REVIEWED A R I I C T E

Scale

TneLe 2. AccuRncy or PHoroLrnEnvENTS FRoM DrrrERErr ScnLe lvncrs photol ineament famil ies with these trends could be posi-tively weighted to help reveal areas containing large numbersof photol ineaments with such trends. Another considerationpossibly important for weighting photolineament data are theregional, resolved normal stresses on the vert ical planes as-sumed to be represented by the photol ineaments. Those withthe least amount of resolved normal stress are most likely tobe "open" and therefore potential ly capable of storing andtransmitting appreciable amounts of groundwater. For exam-ple, i f the regional stress f ield is dominated by west-directed,horizontal compression, then those planar elements in thebedrock detected as photol ineaments which trend east-westhave the least amount of resolved normal stress (zero, in fact,i f they are col l inear with the compression direct ion) as com-pared to features with different trends. Qualification of theielevance of any of these weighting scenarios can, in part, beachieved through comparison of contour maps of photol inea-ment factor values with both existing well data and by actualdri l l ing of test wells. Such quali f icat ion is currently under-way; prel iminary results are presented below.

All software used runs on a PC under the oos operating envi-ronment. A modified version of Ltsc;s csMaP software is used todigitize the photolineaments. Photolineament analysis softwareis based in part on the work of Wise el o1., (1985) and incorpo-rates a Gaussian curve-fitting algorithm written by F. Salvini.Earlier versions of this software for domain analysis v'ithout Ihephotol ineament factor algori thm were developed by D. Wise, S'Mabee, and the author. Processing a 100 by 100 grid and a f i leof 10,000 l ineaments requires about 10 minutes on a 486 ma-chine. Contouring is done by Schrieber's QUIcKSURF and finalmap products are in Autodesk's autoc.lo.

ExampleA recent study of groundwater resources in Cobb County,Georgia, (about 900 km'z area) uti l ized the described photoli-neament factor algorithm as one component of its explora-tion program (Emery & Garrett Groundwater, Inc., report,1991). The study showed that there is very good correlationbetween the trend of photolineaments and both foliation andtectonic fracture families on both local and county-widescales which iustif ied use of photolineaments and the photo-Iineament factor algorithm to identify areas of interest. Theseareas were then evaluated in detail for their groundwater po-tential.

SettingCobb County is underlain by unglaciated, multiply deformed,crystall ine rocks (Hurst, 1956; Higgins and Mcconnell, rgzs).Regional layering and foliation trends northeast but varies lo-cally. Key fracture family trends identified are, in order ofprevalence: 92" , 116' , 145", 68" , 43o,24", and 2 ' (a l l

TeeLe 3. Nuvsen or PuotoLtnenN,tEt'tts OesEnveo oru DtrreRrnt ScnLeIMAGES FoR ,q 900-xw2 Sruov AnEn

Approx. AverageLength of

Photolineaments *

Approx. EndApprox. PointTrend Location

Accuracy Accuracy

1:58 ,000

l :80 ,0001 : 1 3 0 , 0 0 01:250,000

+ 2 9 m

+ 4 0 m+ 6 5 m

+ 1 2 5 m

* based on 95 percent level on a probability versus length plot of thephotolineaments from each scale of image.

Init ial lv. a studv area is divided into a number of nodesin a square grid (Fi[ure 1). Each node is the center of a circleof specif ied radius for col lect ion of photol ineament data. (Acircle is used instead of a square, because square col lect ionareas are inaccurate due to the uneoual distances from thecenter of the square to dif ferent points along i ts edges.) Thespacing of the nodes and the radius of the circles varies witheach study area and photolineament data set. Generally,these are lhor".r so that the circles overlap neighboring cir-cles by 30 percent to 50 percent in area to smooth the databy area averaging (similar to many other contouring algo-rithms). In addition, to yield meaningful statistics for thenumber of direct ional famil ies (Wise, pers. comm., 1985), theradius of the collection circle is selected through trial and er-ror so that an average number of :15 or more photolinea-ments pass through the collection circles.

To calculate the number of directional families, a rosediagram is constructed at every node in the grid. Each familyis defined by separate "rose petals" with heights > 50 per-cent of the tallest oetal (normalization makes the tallestpeaks of each rose diagram : 100 percent). The "cut-off"height of 50 percent is arbitrarily selected, but ensures use ofonly those directional families which are prominent. Rose di-agrams are constructed using an algorithm which fits Gaus-sian curves to normalized azimuth freouencv histoerams ofthe photol ineament data (Figure 2: Wise et i t . . tsss).

After values have been derived for the three parametersat all nodes, each parameters values are normalized usingthe average value for that parameter. The average value isbased on that parameter's range within the study area. Nor-malization enables independent weighting of each parameter.The normalized values for each of the three parameters arethen added together for a single sum value referred to as thephotolineament factor value (Figure 1). Each of the normal-ized values are weighted independently during the summa-tion step to facilitate enhancing one or more parameterswhich may be thought to be more or less important in termsof a study's obiectives. For example, i f long photol ineamentsare thought to be the most significant parameter, then thenormalized value for photolineament length can be multi-ol ied bv some constant. such as three. then summed withihe other two parameter values. each mult ipl ied by only one.Photolineament factor values are plotted at each node in thegrid and contoured to aid in visualization and selection ofdiscrete areas for further evaluation (Figure r).

Other weighting scenarios and variations on the photoli-neament factor algorithm are currently being explored. Forexample, specific trends may be suspected to be more impor-tant to the groundwater system than others based on fieldwork, such as identification of brecciated brittle fault zoneswhich consistently trend north and northeast. Nodes with

PE&RS

1.:Z4,OOO1;58,0001:80 ,0001 : 1 3 0 , 0 0 0 * "1 :250,000 - *

1 :1 ,000,000* *

* . a l l photol ineanents wi th in 20 km of the center of the study areawhich, therefore, resuits in slightly larger numbers because the col-lection area includes some area not covered by the larger scales.

2.0 km

z . 5 K m

4.0 km9.0 km

+ 0 .8 "

+ 0.85"+ 0 .9 ': t t t . / 5 -

Scale n>300 m

g2"lo

4529288521.46

868

J 5 Z /

4506288s2146

86875

PEER.REVIEWED ARI ICTE

Ph,otolinea,n'uen tsCobotr,d, tnn awqm&t of @eq0'1d aotlbh?,,d &rtoorr dclo 6csel\lploal m&s @tcl.grrr,d,qr l:21K. l:58K,l:80X, l:18OK, sdt:250KUee of nultfitb tc&sfidDbs co4,.cl,btt ote dottsttaally mhst/frpr?d.14fbb datc bcsrol plotolilvarrcnt*

PlratoHerlorlocrnts 6r /---n-oaalyscd bqsed oat c \ ll l-sgtwr gril of 7d__-)(*-co|J, ,ct t ,nc{ , rcbt ( i . { l .ol sct t& ocld. \t-11-ffi of ovedzp. /l-7 I:Vahnt tor tir tlrre \]-\FItorllrtct rt cssesed ar ll--fi*colanJated ct rcclr aodr \ ll /

'

bose,d orl tlw photo- )LjLl'Itrr,onrctrls tslt:bh oc'' { | . l.throv4h tl?4 co,J[,ai{4,'^ Xlcirctr crntrr.rd o* th,ot i lY t.

Bff&" of S/lrtdy Arco y'

DatuE*:oltt7tb ol uohn oeddton@t ort ,p& br ttr. gdd,AU pltrrtol&rcowttJs tttltbhpcsr ttirougrr tlu cobotbrta{m!a o't rrsad?lr. total nturb.t, M.tol &f.ot{,r?4l lttt|,ll{,,c. edto,r'l let?4th of plotolktrur''trtsor!3 oalculdLdIa tfui;r eest tltttt ot l0glratol&t@rrelrb. 3 dnot{oarcftbrzdl{l,t, od c 26bm totdlrlltetLTlt ,twrr|l,

" ol d{tl,ct{,rrcl

lod)bs Lc Mn d tbr,oltghe ganarabn1otsto llt:tbqqo"Ul?'n (*c Ftgnst 2).

Normalization of Vahtes

'%, % t%o

oahns ora tht ttrrlctnl3rrcton la tirc ftoc,tlorla sttoruoa.

Su:rn:m,ationAU tlalc trorarrltllzad, scluas qila&d togot'Itrl to? orw labrl;&t PHAIOIJNEAIIENT FAercR

tor.h vahn k tntupcabilqtaldglde,d. ht collctartls q b, & c

uldch qr dltrrtrtlrattl U11 o,r&rdy'r o0lbcfloeo

ot /zz* b%+ a2/ro= 2.0/ooro

(Tlu o'rc Lt tlut tlu colbctbtt citrlz.)

for,,_Bach, Node Contour MPlptolltrr,onurat Faatotoctrrcs cn contorlptdto gcapttbalt!

"tt d

ct ar sr{th ,tlgtruotrA SUotr ar.rs@r trlos! ll}.Vwnrbr:rdrr bU h{giltt,",aofrnd D.droctcqo6b ol trstcdll/rngWh nb&I u.Ut ortd,oil.. tLrttor., tart.tttor ftif'ttttr atah/[lfort

Figure 1. Schematic flow chart for the computer-aided method of deriving photolineament factor values forcontour maos.

tlrr tto,Itln tor ouh poraalutcr,d tqch rr&" b trornu,IkedUy *tt&tg L W tlr ont.tvryooalile of that porwncttt &tfiu lrhr{,1 o*ro- Ttw evoqo

Valueson pbllod ono octa ryr.

Processing

4 .1

s.8

8.5

8.9

2.9

2,8

2,7

z.o4.2

8.7

8.9

PEER.REVIEWED ARI ICTE

( 4 )

W

( 3 )

Double shoothed

30 40 50 60 30 40 50 80

N o r m a l i z e d H i s t o g r a m S m o o t h e d D a t aData

'o l;. il"

60

Gaussian-Curve Info. Rose Dragram Ploi

Figure 2. Description of Gaussian curve-fitt ing algo-rithm which determines the trend, width, and rela-tive height of directional families. (1) Normalizeddata from 30o-60o portion of a histogram (normali-zation makes the maximum peak on all histogramshave the same height). (2) Same data smoothed bytwice passing a 10o running average over the datawith Gaussian curves for comparison. (3) Gaussiancurves fit to the smoothed data with centerpoint/azimuth (A7, A2), peak height (HL, H2), and widthat half-height (W1,W2) for each gaussian noted. (4)Same curues as in (3) but plotted in rose diagramform. Modified after Wise et al.,1985.

directions in degrees east of north). These trends are defined

by the synoptic, Gaussian-curve-based rose diagram of '1.432

fracturei measured in detail at about 100 outcrops scatteredthroughout the County. The 43" and 68o trends include bothfractuies and layering/foliation of the host rocks. The lattercan represent discontinuities related to locally well devel-oped differential weathering along compositional layeringwhich can have significant impacts on groundwater systems(Crawford, pers. comm., 1991). These seven trends are

matched or-overlapped by the trends of photolineament fam-

i l ies (EGGI report, issr). Rs expected, this correlat ion is best

at the local sCale where comparison is made between frac-

tures recorded in outcrops and the photolineaments that oc-

cur within 2 km of the outcrops studied (EGGI report, 1991;

Hardcastle, 1992). The local scale of Z km is used for com-parison because fracture families can differ in outcrops sepa-iated by as little as 2 km in the County based on theavailabie data (EGGI report, 1991)' Detai led discussion of the

fracture fabric, bedrock geology, and correlation of thesewith photolineaments is offered elsewhere (EGGI report,1 9 9 1 t .

Photolineament Factor AnalysisThe correlation documented between photolineaments andbedrock fractures/discontinuities supported using the photo-

lineament factor method. Results of the photolineament fac-

tor analysis helped locate areas presumably underlain by

PE&RS

highly fractured bedrock and/or bedrock containing,numer--on"r oi pronounced discontinuities which might be favorablefor grorindwater development. All 1:24,OOO-, 1:58,000-' and1:80",000-scale photolineaments greater than 300 m in length(n-13,000) *"i" p.o""tted for photolineament factor valuesusing a 300-m radius collection circle overlapping 30-per-cent]These photolineament data were used because theyrepresented ihe three largest of six scales of imagery ana-lyied and, therefore. have the highest geographic accuracy(Letter than about +40 m). No pie-processing was conductedio .u*ou" the few photolineaments which represent culturalfeatures. In the unglaciated crystalline terrain of the centraland southern Appilachian Mountains, east coast of NorthAmerica, photoiineament data are used by the author with-out pre-processing. This simplification is thought lustified be-cause: (i) the relatively few photolineaments which reflectcultural features "fall out" statistically, unaffecting results(Hardcastle, unpublished data); and (2) studies of correlationshow that mosf photolineaments do match fabric elementsmeasured in the-bedrock (Hardcastle, '1992). Average valuesin the collection circles for the number of photolineaments,number of directional families, and total length of photolineament were 19, 4, and 26 km, respectively. The three pho-tolineament parameters were evenly weighted. Derivedohotolineament factor values were contoured and plotted onmylar for superposition on the USGS 7'l '-minute-topographicmips of the quadrangles which covered the study area. Ap-proximatelv 70 areas with relatively high photolineamentiactor values were identified in this 900-km2 study area(EGGI report, 1991).

One of the eight quadrangles which covered part of thestudv area is used forlllustration of the photolineament dataand iesults of analysis with the photolineament factor algo-rithm. Photolineaments in this quadrangle reveal a promi-nent northeast-trending fabric which occurs in anortheast-trending swath (Figure 3a)' This swath is underlainby a layered package of variably erosionally resistant quartz-it"es, gneisses, and Jchists (Higgins and McConnell, roza).Simp"le visual evaluation of these photoline,ament data wouldgenerally result in focus on the pronounced fabric in the area6f the northeast-trending swath. The photolineament factormethod, however, reveals areas of potential interest through-out the quadrangle (Figure 3b). The numeric values of thecontourJenable quantitative ranking of these areas.

As expected,^areas with high photolineament factor val-r.r"s o"c,ttltt regions of intersections and along second and-higher order streams. The non-glaciated, crystalline bedrockinlhe region is characterized by a rectilinear drainage-pat-tern (Creisler et a1.,1983) which is thought to reflect the ero-sional control imparted by fracture systems and differentialweathering of layering in the bedrock.

A few contour highs occur along drainage divides be-cause of the greater tlian average number of photolineamentsand number of directional families of photolineaments whichbegin or terminate in these areas (Figure -+). Weighting thepaiameter for photolineament length in- the.photolineamentiactor algorithm removes these "anomalous" high value-areas(Hardcastle, unpublished data). The areas identified withhigh photolineament factor values are sometimes obvious

".r"d ulro not so obvious on the topographic map' but are

clearly shown and therefore can be ranked obiectively.basedon the photolineament factor value contours (Figure 4J'

It must be emphasized that areas with high photolinea-ment factor valueJ are not necessarily areas with largegroundwater resources nor are they specific drilling targets'

PEER.REVIEWED

' - - , 1 - \fr.t::-- r s : t - - :-,-+J"&

h*\--1 \ . r t l _\.:l\N\-- \. {-+i_<

N / \ \ \ - ' - \ 1 \ ,l ' (--- . . \ - - - \ ' \ i \

*X'\=)\*$- \ ' 1 , , V - - ) . |:i--V?-ti \- \-::-\\ -\/ \--- \ \:

- ) L - \ 2 . - \

'\i<Jfi\-'N.)z<

:i'zi}&?

ffi.$+

:s*\-l]'--.:("-\Y.

(nk/?'!,1'\ \'x'27r-

r(:i;ffi--x{-:.t,

\iP+- i r \ , \G\- \. I' . h - ' r

;

, '

i,\:\ '- \ .

F- / t / ^ \ i / \ l

:\ i- ' -\- l-,rf

,-Jr

. i "Y'x

vr\\=-'\'i k-'-7' ./ .AL:BL--:--=.<- X\ //l-54../ - \ r i i \ ' I / . J /

I / ^ \ . - \\ /

' !( . ' t.-\4 z. l 'q er-1,tffi

a4 ' J7 J fJ J 5 2 J d -

GN

lNr 30'

MNc 00'

$ € ' o d

KEY

Ixl l',n{. l-X ,,ru* .,*l \ | ropouneamenls

l / , Ipho to l ineaments

1:80K B&WPhotolineaments

Er $' 0d

Figure 3. (a) Raw photolineament data for one of eight 7'lrminute quadrangle map areas which cover the example studyarea.

ARI I ctE

PE&RS744

PEER.REVIEWED ARI ICTE

UO

G N ^

,tU.'fMN 1+]' * \ l l]}r

UTM I*n

,t\*---'-*\\t)1 D

O!"oo

i

L- s s ' o dE{ $', 0d

Pholol ln.omonl loclor volu€s r .prcs€nl lho KEsum of normol lzod volucs tor lhal

1) totol numbar ol phofol ln€om.nts;2) totol lenoth of pholol inood.nls,

a 51 numbc. -ot

dlr . i i lonol tomll l .sof photol ln.om€nls found wlthln clrculor or.osSOOm In .odlus thich ov.r lop 3OZ. Ar.ot wlthhigh voluo! oro mosl l lk. ly und.r loln by l roclur.dbcdrock copoblo o{ l ronsmli l lng signl f lcontvolumrs of groundwolcr.

Figure 3. (b) Contour map of photolineament factor values for the same quadrangle area, graphically revealing areas of inter-

est.

PEER.REVIEWED ARI ICTE

A

frFigure 4. Portion of the photolineament factor contour map (Figure 3b) superim-posed on a portion of the uscs 7'lz-minule topographic map of the quadrangle.Contours facil i tate rapid, quantitative separation of discrete areas of interest.Note the high values along pafts of the Chattahoochee River and NickajackCreek, especially where the Creek makes abrupt changes in direction. These ar-eas are targets for detailed investigation of their groundwater potential.

5 .25 0 .5 mitc

.!-?9.----9-----j: kllom.tor

Many of the 7O areas with the relatively high photolineamentfactor values identified were later deemed unsuitable forgroundwater resource test well drilling because of inappro-priate land use, low recharge characteristics, poor yieldingbedrock type, and/or other hydrogeologic concerns (EGGI re-port, 1991). Specific drilling targets are not identified withthe photolineament factor method because of the nature ofthe algorithm, the location inaccuracies of the photolinea-ment data, and, in this study, the use of col lect ion circleswith 300-m radi i .

Thirty-three of the areas identified, however, weredeemed "favorable" based on detailed hydrogeologic evalua-tion of their potential to yield groundwater and were recom-mended for testing to the Cobb County-Marietta WaterAuthority (EGGI report, 1991). Geophysical analyses and testwell drilling are currently underway in a number of these"favorable areas." Preliminary results in one such "favorablezone" is a combined air- l i f t yield in excess of r,ooo,o0o gpd(gallons per day) from three of three wells drilled. This vol-ume of groundwater greatly exceeds that produced by mostwells in the region.

746

SummaryThree photolineament parameters are thought to indicate thedegree to which bedrock may be fractured and, therefore,provide a guide for selecting discrete areas where groundwa-ter resources may exist: (1) the number of photolineaments,(Z) ttre number of directional photolineament families, ond(31 the total length of photolineaments. The normalized sumof these three parameters, calculated for photolineamentswhich occur within or traverse an area of defined radius, istermed here as the photolineament factor value. A new com-puterized algorithm facilitates rapid processing of the largephotolineament data sets (n>>1.000) typically collected forstudies of regional groundwater resources (>10 km, or 2500acres). The resulting contour map enables quantitative rank-ing and selection of discrete areas for further evaluation.

AcknowledgmentsNumerous people have contributed to the ideas and com-puter programs discussed in this paper. I wish to thank S.Mabee, D. Wise, F. Salvini, and L. Hills for their sisnificant

PE&RS

PEER.REVIEWED ARI ICTE

inputs in software programming. S. Mabee, J. Emery, D.Tinkham, and ]. Brooks are also appreciated for illuminatingand thought provoking discussions and for reviews of earlyversions of this manuscript. Thanks are also extended to twoanonymous reviewers. Use of the sample data from CobbCounty, authorized by P. Karr of the Cobb County-MariettaWater Authority, is greatly appreciated.

RefelencesAlpay, A.O., 1s73. Application of Aerial Photographic Interpretation

to the Study of Reservoir Natural Fracture Systems, /our. of Pet.Technology, 25(1.):37 -45 .

Blanchet, P.H., 1957. Development of Fracture Analysis as an Explo-ration Method, Am. Assoc. Pet. Geol. Bull., 4^1i1748-1'759.

Caswel l , W.B., D.B. Hi l l , and M.F. Eichler , L986. L ineoments, High-Yield Bedrock Wells, and Potential Bedrock Discharge Areas in

the Maine Portion of the Portland and Bath 2-Degree Sheets,Open-Fi le Rep. No. 86-67, Maine Geol . Survey.

Cressler , C.W., C.J. Thurmond, and W.G. Hester , 1983. Ground Waterin the Greater Atlanta Region, Georgia, Georgia Geological Sur-vey Information Circular 63, Georgia Geologic Survey, Depart-ment of Natural Resources, Environmental Protection Division,Atlanta, Georgia.

Emery and Garrett Groundwater Inc (EGGI), 199'1,. Groundwater Ex'ploration and Development Program, Phase I, Vo1. 1, submittedto the Cobb County Marietta Water Authority, Cobb County,Georgia, 98 p.

Hardcastle. K.C., 1992. Correlation of Lineaments and Fracture Fab-ric in the Piedmont: Implications for the Study of FracturedBedrock Aquifers, So. East- Geol. Soc. Amer., Abs. and Prog..24 (2 ) : 19 .

Higgins, M.W., and K. I . McConnel I , 197a. The Sandy Spr ings Groupand Related Rocks of the Georgia Piedmont: Nomenclature andStratigraphy, Short Contributions to the Geology of Georgia,Georgia Geol . Survey, BuI l . 93, pp. 50-55.

Hurst, V.J., -1956. Geologic Map of Kennesow Mountain-Sweat Moun'tain area, Cobb County, Georgia, Georgia Geol. Survey MapRM-2,1. :24,OO0 scale.

Lattman, L.H., 1958. Technique of Mapping Geologic Fracture Tracesand Lineaments on Aerial Photographs, Photogrammetric Engi-neering, 24:568-576.

Lattman, L.H., and R.R. Parizek, 1964. Relationship between FractureTraces and the Occurrence of Ground Water in Carbonate Rocks,

lour. of Hydrology, 2:73-91.

Mabee, S.8., 1,992. Lineaments: Their Value in Assessing Groundwa'ter Availability and Quality in Bedrock Aquifers of Glaciated

Metamorphic Terranes - A Cose Study, PhD dissertation, Univ.

of Massachusetts, 567 p.

Mabee, S.8. , K.C. Hardcast le, and D.U. Wise, 1990. Correlat ion of

Lineaments and Bedrock Fracture Fabric: Implications for Re-

gional Fractured-Bedrock Aquifer Studies, Preliminary R"sults

from Georgetown, Maine, Proc. FOCUS Conf , on Eostern Reg

Groundwater Issues, (3) :283-297.

in press. A Method of Collecting and Analyzing Lineaments

for Re[ional-Scale Fractured-Bedrock Aquifer Studies, Ground'

water.

Podwysocki, M.H., 1974. An Analysis of Fracture Trace Patterns in

Areas of Flat-Llting Sedimentary Rocks for the Detection of Bur-

ied Geologic Structure, NASA publication X-923-74-200, God-

dard Space Flight Center, Greenbelt, Maryland, 67 p.

Rich, f.L., 1928. Jointing in Limestones as Seen from the Air., Amer'

Assoc. Pet . Geol . Bul l . , 12[B):861-862.

Siddiqui, S.H., 1S69. Hydrogeologic Foctors Influencing Well Yields

aid Aquit'er Hydraulic Properties of Folded and Faulted Carbon-

ate Rocks in Central Pennsylvanio, Ph.D. Dissertation, The Penn-

sylvania State University, 502 P.

Wise, D.U., 1982. L inesmanship and the Pract ice of L inear Geo-Art ,

Geol . Soc. Amer. Bul l ,9:886-888.

Wise, D.U., and L.T. Grady, 1985. Topography of the Southern Ap'

palachians, Dept. of Geology and Geography, University of

Massachusetts.

Wise, D.U., R. Funic ie l lo, M. Parot to, and E. Salv in i , 1985. Topo-

graphic Lineament Swarms: Clues to their Origin from Domain

Analysis of l ta ly, Geol . Soc. Amer. Bul l . ,96:952-967.

Wobber, F.J.,1.967. Fracture Traces in Illinois, Photogrammetric En'

g i n eer i ng. 33:499-506.

Yin, Z.Y. , and G.A. Brook, 1992. The Topographic Approach to Lo-

cating High-Yield Wells in Crystalline Rocks: Does it Work,G roundwoter, 30:96-1O2.

(Received 15 Apr i l 1993; accepted 16 June 1993; revised 12 August

1993 )Kenneth C. HardcastleKen Hardcastle is a structural geologist withEmery and Garrett Groundwater, Inc., specializ'

i Sf ing in the development and protection of

) T g* if"o il?:il ffiliJ:? S111il:i,"i:TF Xii"" ::i:lected at bedrock exposures and from aerial and digital im-ages. Primary studies of tectonics and the occurrence andmovement of groundwater are in the eastern United States.He received his doctoral degree from the University of Mas-sachusetts at Amherst in 1989.

HydroGlS '9 -lnternational Conference on Application of GIS inHydrology and Water Resources Management

April 16-19, 1996, Vienna, Austria.

Sponsors: International Association of Hydrological Sciences' Ground Water Commission and InternationalCommittee on Remote Sensing and Data Transmission Division on GIS; Universitiit fur Bodenkultur;UNESCO; International Association of Hydrogeologists; and ASTM.

For further information contact: Ivan Johnson, Chairman, GIS Division, ICRSDT,7474 Upham Court,Arvada, CO 80003; tel: and fax:303-425-5610.

PE&RS