Data:A Toolfor

AirborneRadiometric
Reconnaissance MappingUsing
Geological
A GIS*
Abstract field verification of published field maps for the purposes of
Airbornegamma ray spectrometerdc'tq collectedby the Geo- minercI exploration.
logical Suwey of Canada, gridded to a pixel resolution of Despile the relatively low spatial resolution of the
725 mefres,r's used to create digital images that show the gamma ray spectuometer data, the areas studiedshow_quite
spatial distributions of gamma ray spectrometer(radiomet- itrong spatial associationsbetweenthe radioelement clusters
ric) data: equivalent Uranium (eU), equivalent Thorium and be&ock unik- The overlay technique was helpful in iso'
(eTh), and Potassium(o/oK).Theseradioelement images are Iating inconsistenciesbetween the two classifiedmapg sug'
currently used by the GeologicalSuwey to assistin geologi- gesting sitesfor further localized field mapping.
cal mapping and mineral exploration. Data interpretation is
traditionally made visually from hardcopy of pseudo-colored Objectives
single-channelimages,or from three-channelcolor compos- The purpose of the present study is to apply a cl-ustering
ite images. Visual interpretation can be augmentedby using technique to radioelement data and to comparethe resulting
an unsupewised classificationprocedure on the radioele- cluster map with a digitized geological map within a GIs soft-
ment data. The resulting clusters arc displayed as an image, ware package.This procedure should lead to a better under-
and interpreted by digitally overlaying the digitized geologi standing of the relationship between geological map units
cal map. This brings out the similarities qnd differencesbe- and radioelement patterns and help to identify locations
tween units determined from field mapping and units based where more detailed field work is needed. This work will
on radioelement response. serve to demonstrate the utility of image analysis and Gts
Imagesof the three radioelement channels were input to within the Geological Survey of Canada.
a migrating-means cluster analysis on an image processing
system.The rcsulting classified image was imported into a Background
cts. Other data setsin the GISincluded table-digitized bed- Migraiing-means clustering has been found to be an effective
rock and surficial geologr maps and a binary map showing technique for grouping pixels with similar radioelement sig-
the presenceof water bodies,derived from a density sliced natureJ into classes.Harris {1989) showed that radioelement
Landsat Thematic Mapper band 5 image, Lakesand bogs, as clusters were strongly correlated with mapped geology in the
weII as regions covered by particular surficial units, were Meguma Terrane of eastern Nova Scotia.
combined into a new binary image used to mask out regions Three-channel color composite plots prepared by the
where the radioelement rcsponsewas unrelated to outcrop. GeologicalSurvey of Canada(GSC,1990) and Intensity, Hue,
The classified radioelementimage was then compared with and Saturation (rrs) color composites prepared by the Can-
the geological map using two-map overlay and area analysis ada Centre for RemoteSensing(Grahamand Ford, 1991; Har-
cros s-tabul ation te chniques. ris ef o1.,1990)reveal striking relationshipsand also
The cross-tabulationcleaily identifies thosegeological apparentdiscrepancieswith geologicalmapping at a regional
units that have a distinctiveradioelement rcsponse.By rc- sCale.the three-channelimageis createdby the display of
classifying the map overlay, by imposing a color coding eU as red, eTh as green,and K as blue to createa color com-
schemethat enhancesbedrock geologt closses,the relation- posite.For the GeologicalSurvey enhancement(GSC,1990),
ship betweenthe bedrock geologr and radioelement rcsponse ipecific lookup tablesare used for each channel for an opti
is enhanced, The degreeof conelation betweenthe two car- mum product with the color strengthmodulated by total
tographic images is site dependent, rqther than global, as count.
might be expected.The conelation is not simply on the basis In the caseof the ms transform enhancements(Graham
of the average radioelementvalues, but also on the shape, and Ford, 1991),the radioelementimagesare assignedto red,
texture, and extent of the radioelement clusterc.Areas where green,and blue channels,followed by a transformationto the
the two maps dffir indicate zones of possible interestfor IHS coordinate system. The resulting Hue reflects variations
in all three radioelements.The Intensity information is re-
*Presentedat the Ninth Thematic Conferenceon Geologic Remote placed by Canada Centre for Remote Sensing airborne radar
Sensing,Pasadenma,California,8-11 February1993. data and the Saturation variable is held constant at a maxi-
mum B-bit value of 255. In the reversetransform of the IHS
back to ncg color space,the radar data are effectively merged

Photogrammetric Engineering & Remote Sensing, D. F. Graham

Vol. 59, No. B, August 1993, pp.7243-7249. Intera Information Technologies (Canada)Ltd., 2 Gurdwara
Road,Suite 200, Nepean,Ontario KzE 7A2, Canada'
0099-1112/93/590B-1243$03.00/0 G. F. Bonham-Carter
01993 American Societyfor Photogrammetry GeologicalSurvey of Canada,Room 694, 601 Booth Street,
and RemoteSensing Ottawa, Ontario K1A 0EB,Canada.

PE&RS
 east. The study area is on 1:50,000-scalel.lTSmap sheets
42D19,15, and 16. A Canada Centre for Remote Sgasing air-
borne radar image enhancesthe area'srugged topography
which, along with dense vegetation, has greatly impeded
field mapping (Figure 1),
The general area hosts several precious metal (Herrlo -
Au) and base metal (Wiuston Lake - Zn, Cu, Ag, and Au) de-
posits and occurrences associatedwith supracrustal metavol-
canic/metasedimentarysequencesof the Abitibi-Wawa
Subprovince of the Superior Province. In addition, numerous
mineral occurrences (Ce, Nb, Cu, Fe, Ti) are hosted by the
highly radioactive Coldwell Alkaline Complex, the largest al-
kaline intrusion in North America {Walker et al.,tgg7).
The area is underlain by generally east-westtrending
metavolcanic/metasedimentaryrocks occurring in supracrus-
tal sequencesof the Abitibi-Wawa Subprovince which are
surrounded by poorly subdivided granitic/syenitic and gneis-
sic lithologies. The supracrustal sequenceshave undergone
up to four periods of deformation with large scale faulting
occurring in several zones. This sequenceis intruded by the
Coldwell Alkaline Complex (ODM, 1972a;7972bt Mitchell et
aI., 7982; Schniedersand Smyk, 1989; Walker et o/., 1991).
In general, each lithologic unit can be distinguishod based on
its gross radiometric signature: the Coldwell Complex is
highly radioactive, the metavolcanic and sedimentary units
have low to moderate radioactivity, and the granitic/gneissic
units are moderate to highly radioactive. The Coldwell Com-
'
plex has been interpreted as being subdivided into three cen-
/l Y- .':.'
LAKEsuPERroR
/
\::: ::'ii' ters of intrusion. To the east, ferroaugite syenite was
|
intruded into a border gabbro. West of the first intrusion,
,' g7ow nepheline syenite was emplaced. Between these two inhu-
I sives, a third intrusive center occurs, composed of a quartz-
\ amphibole syenite.
t
Methodology
Table 1 and Figure 2 summarize the input data used and im-
age analysis and cIs manipulation steps. The final goal is to
produce a confusion matrix showing the degree of overlap
between radioelement clusters and bedrock units.

TnsLE
Teeu1. Surr,ttunnv or Dnn Sounces.
Map/Imagery OriginalData/Legend DigitizedClasses
Bedrock Geologyt Metasediments,
Metavolcanics, igp.eous rocks 1) Volcanics
Granite, Quartz Monzonite, 2) Granite
Figure1. Location Alkaline
of the Coldwell Complex, Mara- Granodiorite
thon,Ontario,Canada. the ex-
Theshadedareaidentifies Nepheline Syenite, Granite- 3) Quartz Syenite
Anairborne
tentof the Complex. sARimageenhances the Quartz Syenite
ruggedtopographywhichhasimpededlield mapping. Augite Syenite a) Augite Syenite
Quartz-Amphibole Syenite, 5) Syenodiorite
Syenodiorite Hornblende,
Syenite
with the radiometric data. Radiometric data are a measure of
Surficial Bedrock plateaus, plains, 1) Outcrop
the concentration of radioelements,providing information on ridges and knobs
Geology*
surface cover and in some caseslithology. Radar, becauseof
Other (Drift types) 2) Surficial Cover
its side viewing geometry, enhancesterrain morphology and
surface roughness.Radar data, in the IHS transform, provide Iandsat Thematic Density slice lookup table of 1) Water
Mapper Data TM band s. 2) Lake
a good high resolution cartographic base for the presentation
of radiometric data, which is useful for reconnaissancemap- Gamma Ray Migrating-means clustering Radioelement
Spectrometer of U, Th, and K. clusters
ping.
DataS
andGeology
Location t ODM, 7972a;1972b.
Marathon is situated along the north shore of Lake Superior * Gartner, 1980.
between the towns of Schreiber in the west and Hemlo in the s Gsc,1990.

t244 PE&RS
 Bedr*k gooloty mp
(F.om ODii 1972o,b)

gurllcal polyoons grouped

s outcroD and aurllclal
d$9r

Aress ot aurfidal covsr and

wrt6r bodlos arc rcm0v6d flom
Bedr@k rod Clustcr mapl

Ggn?,ato raster bllmsPs

lo sample radloelomonts

Pro$nt si.tl9tl6
a5 box plois
loa uch ol€mant

figure 2. Methodolo$t digital raster and vector processing.

DataPreparation image showing land and water (Table 1 and Figure 2). As
water absorption is high in the infrared wavelengths, the dig-
Digitization ital values are near zero and water can be readily isolated.
A bedrock geological map (ODM, 7972a; 7972b) and sur- Ternary radiometric data based on a 1000-metreline
ficial geological map (Gartner, 1980) were table digitized in a spacing and resampled to a pixel size of 125 metres were ob-
hybrid raster-vector cIs to form the initial data layers in the tained through the Mineral ResourcesDivision of the Geolog-
database.Table 1 shows the geological units that were com- ical Survey of Canada (GSC 1990). This three-channel data
bined to create the five general unit classes.The older rocks set was already LrM georeferenced.
into which the Coldwell Complex was intruded were The radiometric data were grouped into 16 classes,using
grouped into (1) volcanics and (2) granite.Within the Cold- a migrating-means clustering algorithm (PCI, 19S1).This al-
well Complex, the classescorrespond to the three intrusive gorithm performs an unsupervised clustering using the K-
phases: (3) quartz syenite, (4) augite syenite, and (5) syeno- meens minimum-distance method. The cluster image was
diorite. The surficial geological map was simplified for digi- imported into the GISfor further processing.Within the GIS,
tization to distinguish outcrop and surficial cover as two the 16 radiometric clusters were reclassified into five classes
map classes. to eliminate small polygons. Collectively, these small poly-
gons comprised only three percent of the total databasearea
and are related to radiometric spikes within much larger
clusters.
RasterProcessing
Landsat Thematic Mapper {rv) band 5, having a resolu-
Datalntegration
tion of 30 metres, was registered to a Universal Transverse
Mercator (uru) base by extracting ground control points Create Overlav Mask
from 1:50,000-scale topographicmap sheets.Thesedata were In order io remorreregions covered by water and surfi-
density sliced using a lookup table (Lur) to create a binary cial sediments, a binary image was created as a mask. The

PE&RS
 classeswithin a single bedrock class).The resulting 25-class
map shows that each radioelementclassis dominantly asso-
ciated with a particular bedrock unit.
The large-numberof classeswithin the Coldwell Com-
plex indicatesgradationallithologiesmaking the separation
into discretelithologies difficult. It can be seenin Figure 4,
however,that the cluster map doesrevealthree distinct ra-
diometric proportions which show minor correlationwith
the three mapped intrusive centersfound in Figure 3. Two of
3l\- the Granite classesin this overlay map (Figure 5) enhance
two localized targetswhich appearto be isolated:A at San-
L)
toy Lake and B along the northwesternedgeof the Coldwell
Complex, warranting further study. Confirmation of these
truE ND
targetsby statisticalmeanswill be necessaryto confirm va-
lidity.
f l vor",,i"" sv"nir" sv"rit"
leu"* lnueit"

l--l c',,,r, sv"noaiorin furalysisol Confusion

Matrix
F]
I 10 km
An area cross-tabulationor confusion matrix is a two-dimen-
Figure3. Bedrockgeolos/map with overlaymask, sional table summarizingthe areal overlap of all possible
combinations of the two input maps, As there are the same
number of radioelementclassesas there are bedrock geology
map units, the confusion matrix can be arrangedso that the
largestelementsappearin the principal diagonal.The Kappa
coefficient of agreement(Rosenfeldand Fitzpatrick-Lins,
1986) then servesas a measureof correlationbetweenthe
two maps.
The confusion matrix, T (Table 2A), is organizedwith
the five geologicalmap classesas columns and the radiomet-
ric clustersas the rows. Values in the matrix are areasof
overlap in squarekilometres.The classesare orderedwithin
the confusion matrix such that the diagonalelementsof the
matrix (coincidentclassnumber) indicate the amount of
agreementand the off-diagonal elements indicate the amount
of disagreementor confusion.Perfectlycorrelatedmaps
would have zero values in the off-diagonalpositions.
LEGEND
To calculatethe value of Kappa (x), the observedarea
proportions,pt (Table 2B), and the expectedareapropor-
E Clustet 1
I Cluster 3
! ct
",".
s
tions, q,, (Table 2C), are calculated.Details relating to these
n Clustet 2
I Cluster 4 '
to k.
- computationscan be found in Appendix 1.
The overall Kappa value calculatedfor these data is
0.46, indicating a moderatelystrong agreementbetweenra-
Figure
4. Radiometric mask.
clustermapwithoverlay
diometric clustersand the generalizedbedrock geologyunits

mask was built by applying a threshold to the LandsatrM

image (for detectingwater) and the digitized map of surficial
cover (Table 1). The mask was applied to both the bedrock
map and the radioelementcluster map, removing the effects
of water and cover from the subsequentanalysis (Darnley,
1973),Figures3 and 4 show the application of the overlay
mask to the bedrock geologyand radiometric cluster maps.

Results
Evaluation
Visual
The two maps - bedrock geology (Figure 3) and radiometric
LEGEND
clusters (Figure 4) - were combined to form an overlay map, :.^' o.
shown in Figure 5, where each of the five cluster classes
NvcT fl ccr ffiocr mil|n
Nvc2 f,o"z mo., Iace
overlao each of the five bedrock classes. Based on the bed- N\ ve !eca VZ@' -lss4 [ ]n*
rock geology, the overlay map is coded to determine how the N vca IGca V- a" m^ca
bedrock classes are subdivided by the radiometric cluster NS vc5 lecs Qo"s llllllacs
classes. In Figure 5, the legend annotation reflects this poly- '
1oT;-
gon shade pattern (i.e., Vcl represents the intersection of the
volcanic bedrock geology class with radiometric cluster class Figure5. Overlaymap of geolos/ - clusteruniquecondi-
1..Vcl through VcS are assigned a unique map shade pattern. tions.
The density of the shade pattern varies between cluster

1246
 TneLE
2. CnLcuuttot{or lGppeCoernctrnror AoReeuerur.
(A) Corrustor.rMenx f wmt EleueursIa. (B) OesenvroPnopornorusp,y.SHnoeoVALUE5 tNTHE
PRINCIPAL
DteeorurL
RepnrserurAREAS (C) ExpecreoPRopoRnoNs
oF AcREEt',tErur. Qri,wrH PnrrucrprdracoNlLSxaoeo.(D) Colonror.rir-lGppl ano PERcENT
AGREEMENT,

A. Geologicll tep B. Geologicll l|rp

Cl uslel Class 2 4 Total Cluslcl Cls33 3 4 Total

llap lrap
99.4 0.4 0.9 364.6 .108 o 0 .001 396

86.7 212.9 2.9 3.8 15.9 322.2 094 ,:192 003 .004 017 .350

34.3 43.6 4.9 r6.5 106.5 008 .037 l{l{Z 005 .018 r16

4 10.7 9.0 24.4 27.S 11.9 83.9 4 .o12 010 .027 530 013 092

5 ll 1.4 8.8 0.0 30.7 42.O .001 @2 010 0 046

1:,W,,,';,,';:,
Tord 369.3 357.0 80.0 37.0 75.9 9't9.2 Total .402 389 .0s7 040 082 L000

c. Geological llap D. MPA TAGSEEIEM

Clultcr Clas Clas

4 Total 1 0.537 0.723
lrap
::::::::::::::
.154 .034 .016 033
iri:1Y,19 396 0.zll5 0.G61

2 140 .!36 030 .014 .030 350 0.409

.047 .045 ::rOlO .005 010 117 0.305 0.333

4 036 035 008 ':(161',,, .@8 091 5 o.707 0.73 1

or8 0t8 .004 002 :Oli,::::l 045

Tolal 400 388 086 041 085 |.000

in e-xposedbedrock areas.Conditional kappa values, off-diag-

onal elements of the confusion matrix, and visual interpreta--
Statistical
PlotEvaluation
A series of box plots are used to graphically show the distri-
tion of the overlay map indicate several kinds of confu;ion.
bution of radiometric values within each geological class.
The granite (class2) is most often confusedwith both the
Thesedisplay the minimum, maximum, 25th percentile,TSth
volcanics-andthe quartz syenite.Confusion can be expected
percentile,and mean (Figure6). In general,the low radio-
betweenthe volcanics and the granitesbecausecount levels
metric values of the granitic and volcanic country rock dis-
are much lower than those of the Coldwell intrusive units.
tinguish them from the Coldwell Alkaline Complex units.
Of greater interest are the areas which are intemreted as
The unique radiometric signatures which provide the
quartz syenite according to the radiqmetric data but are
separation of colors within the initial integrated IHSimages
mapped as granite.Thesevisually confirmed areasof dis-
originate mainly from the contrast in the potassium results.
agreement(Figure 5), again highlight targets A and B. Their
As the K count rate is influenced by the percentageof feld-
spatial -locationssuggestareas for further detailed mapping.
spars, acidic rock types such as granite and quartz syenite
Target B has a strong spatial association with the Coldwell
show elevatedvalues as comparedto the other lithologies.
Complex,where detailedmapping may support the revision
Although the box plots (Figure 6) reveal a similarity be-
of the western limit of the intrusive body. As the radiometric
tween the volcanicsand granites,as comparedwith the
classesare similar for these two targets,lithological similari-
syenite units, granite is distinguished by the greater potas-
ties or associationsmay be supportedby detailedfield map-
sium content. The three syenite units have relatively more
ping.
eU, eTh, and K. Low eTh values of augite syenite separateit
Conditional Kappa (Table 2D) gives a value for the areal
from the other syenites. The quartz syenite and syenodiorite
associationwithin individual cluster classes,Cluster class 5
are radiometrically quite similar, except that K is higher in
has greatestconditional Kappa value, indicating the best
the former and eU and eTh are higher in the latter.
agreementoccurs with the syenodiorite unit within the Cold-
well Complex. In order of decreasingagreement,the volcan-
ics are followed by the granites, euartz syenites, and augite Conclusions
syenite, Referring to Table 2A, the confusion matrix shows The IHs transform is a valuable algorithm for the presenta-
that the radiometric class4 has almost equal areapropor- tion of integrated data for the purposes of reconnaissance
tions associatedwith both the augite syenite and quaitz mapping. The count rate variations of radiometric potassium
syenite, making these two units difficult to separateradi- predominantly control the degreeof Hue separability in the
ometrically. IHStransform. It is proposed that the radioelement cluster

PE&RS
 POTASSruM STATISTICS FOR EXPOSED BEDR.OCK TJNITS
(ccns) and J. Harris of Intera Information Technologies, for
their invaluable comments. M. Trinidade of ccns assistedin
T 2.97 tt
the preparation of graphics for the paper. Airborne radar data
were provided by CCRS,Data Acquisition Division, while the
s
T I*Ii airborne radiometric data were supplied by the Mineral Re-

| | trf Y
2-3r t<

= r.F e sources Division of the Geological Survey of Canada.

z '.* ?
F

25
tTtl
I I o.e
0.6
I
fi
t
References
Darnley, A.G., 1973. Airborne Gamma Ray Suwey Techniques - Pres-
ent and Future, Proceedingsof a Panel on Uranium Exploration
Methods, IAEA-PL-490/15,International Atomic Energy Agency,
Vienna, Austria, pp. 67-105.
Gartner, I.F., 1980, Norilrern Ontario Engineeilng Geologr Tenain
Study, Data Base Map, Heron Bay, Ontario Geological Suwey,
Map 5093, Scale1:100,000.
Geological Survey of Canada (GSC),1990. Seven-ChannelGamma
Ray SpectrometerData Suwey, Coldwell Complex, Mamthon,
%.,2 F
Ontario, Mineral ResourcesDivision, Project leader: K.L. Ford.
D.a' =
z
ta-q ><
Graham, D.F., and K.L. Ford, 1991. The Integration of Airborne Ra-
,,.,, & dar and Geophysical Data for ReconnaissanceGeological Map-
tr.<
-
a
ping in the Marathon-SchreiberArea, Northwestern Ontario,
F
Proceedingsof the 14th Canadian Symposium on Remote Sens-
".," 2 rng, pp. 95-98.
Harris, f.R., 1989. Clustering of Gamma Ray SpectrometerData Using
a Computer Image Analysis System, Statistical Applications in
the Earth Sciences(F.P.Agterberg and G.F. Bonham-Carter,edi-
URANruM STATISTICSFOR EXPOSED BEDROCK UNTTS
tors), Geological Suwey of Canada, Paper 89-9.
Hanis, J.R.,R. Murray, and T. Hirose, 1990. IHS Transform for the
Integration of Radar Imagery with other Remotely SensedData,
Photogrammetric E ngineeilng & Remote Sensing, 56(12) :1631-
t< g
g 1641.
T]-
rt+
J
Mitchell, R.H., and R.G. Platt, 1982. The Coldwell Alkaline Com-
z - plex, Protozoic Geologyof the Northern Lake Superior Area,
E
GAC-MAC Field Trip Guide Book, Trip 4, Winnipeg.
2
Ontario Department of Mines and Northern Affairs (ODM),1972a.

25
tY
II
I Map 222 0, Manitouwadge-Wawa Sheet,Geological Compilation
Series,Scalel:253,44O,
1972b.Map 2232, Nipigon-SchreiberSfteef,GeologicalCom-
pilation Series,Scale!i253,44O.
PCI, 1991. KCLUS - K-means Clustering, EASIPace UsersManual,
Pace Multispectral Analysis Package,V5.0a, pp 2-67 to 2-67.
Rosenfeld,G,H., and K. Fitzpatrick-Lins, 1986. A Coefficient of
Agreement as a Measure of Thematic Classification Accuracy,
Photogrammebic Engineering& RemoteSensing,52(2):223-227,
Schnieders,B.R., and M.C. Smyk, 1989.Annual Repoftfor the
Figure6. Boxplotsshowing
the distribution Schreiber-Hemlo Resident Geologist's District, 1989, Ontario
of uranium, andpotassium con- Geological Suwey, Miscellaneous Papet 147, pp. 139-159.
thorium,
centrations
for eachbedrock
unit. Walker, 8.C., R.H. Sutcliffe, C.S.f.Shaw, and G.T. Shore,1991. Geol-
ogy of the Coldwell Alkaline Complex, Ministry of Northern De-
velopment and Mines, Summary of Field Work and Other
Activities, Ontario Geological Survey Miscellaneous Paper 157,
pp. 107-116.
mapping be considered as an additional tool for presenting
gamma ray data to the geologist, in addition to the ternary
IHSwith radar or TM images.Cluster maps could also be Appendix
combined with radar by assigning the cluster class to Hue To calculate the kappa coefficient of agreement,an observed
and radar to intensity, and holding the Saturation constant. area proportions matrix and an expected area proportions
The advantageof the clustering method is that the prod- matrix are calculated and combined using the following
uct is a classifiedimage,like a geologicalmap. The disad- equations:
vantagesare (1) that classifications are area-dependentand T,
each region must be analyzed separately (i.e., clusters are not Pr: t (1)
universal) and (2) that water and surficial cover need to be
masked for the method to work effectively. where T.. is the total area of comparison. Then the expected
proportion, q;1,is determined from
Acknowledgments Qt : Pt.Pt {2)
The authors would like to thank the reviewers, V. Singhroy
and R. Brown of the Canada Centre for Remote Sensing where pr. is the marginal total for the i-th row, and p., is the

t24a
marginal total for the l+h column of the proportions matrix. calculated (Figure 6d) showing the breakdown of agreement
Then Kappa is by classas shown in Equation 4: i.e.,
n n
-
Pn Qii (4)
2p,,- 2qu
i-7 i- 1
K 't : -
Pt-Qn
n
(3)
The "percent agreement" of each cluster per geology class is
1-2qu simply the area proportion of a single cluster in each geology
class as compared to the total area of the cluster (4/X,.). This
where r ranges from - 1 (perfect disagreement)to + 1 (per- measurementincludes random correlations by chance which
fect agreement)with a value of zero indicating that the agree- consistently over estimatesthe value of agreement.Kappa
ment is no better than that expected due to chance. corrects for correlations due to chance and gives a more use-
A conditional Kappa value for each cluster class can be ful measure of areal association.

CALL FOR PAPERS

Speclal lssue
on
S orrcoPY PnoTOGRAMMETRT

Photogramrneffic Engineerlng and Remote Senslng

The American Society for Photogrammetry and Remote Sensing will publish a Softcopy Photogrammetry
special issue of PERSin August 1994. This issue will contain commentaries,invited and contributed articles.
Authors are especially encouragedto submit manuscripts on the following topics:

r Definition issues, functional requirements, and design considerations.

r Data issues, acquisition, volume, compression,visualization, and reduction.
I Digital orthophotography, production and use.
r Softcopy photogrammetry, image analysis, and GIS, the all-in-one system concept.
r Conversion of operational production systemsinto softcopy, requirements and implications.
I Standards, testing procedures, accuracy, performance, and system evaluation.
r Trends in technology, state-of-the-art,and future directions.

All manuscripts, including invited articles, will be peer reviewed in accordancewith establishedASPRSpolicy
for publication in PE&RS. Authors who wish to contribute papers for this special issue are invited to mail
five copies of their manuscript to:

Raad A. Saleh, ERSC,University of Wisconsin-Madison,1225 West Dayton St., Madison, lW 537O6;Phone;

(608) 263-6584,fax: (60S)262-5964,internet:[email protected].

I All papers should conform to the submission standards in "Instructions to Authors" that appearsmonthly
in PE&RS.
r Papers should be free from promoting a specific commercial product.
I Paperswithout funds for color printing may be subject to rejectionby ASPRSHeadquarters due to budget
constraints.
t Cover image of the special issue may be selected from one of the submitted articles. To qualify, photos
must be at least 9"Xg" in size, and can be prints or transparencies.
r Electronic submission will be accepted using the above e-mail address, if the manuscript contains no
graphics. However, final manuscripts must be in hardcopy, accompanied by DOS ASCII disk.
r AII manuscripts must be received by 5 ]anuary 1994 in order to be considered for publication in this
special issue.

PE&RS