You are on page 1of 14

This article was downloaded by: [University of Illinois Chicago]

On: 10 February 2015, At: 15:25


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office:
Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Cartography and Geographic Information


Systems
Publication details, including instructions for authors and subscription
information:
http://www.tandfonline.com/loi/tcag19

The Chamberlin Trimetric Projection


Albert H.J. Christensen
Published online: 14 Mar 2013.

To cite this article: Albert H.J. Christensen (1992) The Chamberlin Trimetric Projection, Cartography and
Geographic Information Systems, 19:2, 88-100

To link to this article: http://dx.doi.org/10.1559/152304092783786609

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy, completeness, or suitability
for any purpose of the Content. Any opinions and views expressed in this publication are the opinions
and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of
the Content should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising
directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or
distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can
be found at http://www.tandfonline.com/page/terms-and-conditions
The Chamberlin Trimetric Projection
Albert H.]. Christensen

ABSTRACT. A computer solution to the Chamberlin Trimetric projection is presented with a numerical method
that circumvents the need for closed formulas to analyze distortions. The existence and consequence of singu-
larities in the computer algorithms are discussed in detail, with suggestions to minimize their adverse effect.
Linear, area, and angular distortions are estimated for the Chamberlin projection, and compared with values
analytically computed for the Transverse Mercator and Albers Equal-Area Conic projections. In addition,
distance distortions for the three projections are computed, listed, and compared. The author concludes that the
Chamberlin projection is an excellent compromise between the other two, provided the discontinuities are
resolved.

KEYWORDS: cartography, projection, Chamberlin Trimetric, distortions, singularities, discontinuities.


Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

History of the Projection (1950). Additionally, the publishers of the Times Atlas must
agree, since their last editions have included North Amer-
n1947 the National Geographic Society (NGS) pub-

I lished The Round Earth on Flat Paper. The introduction


to this richly illustrated special issue (Chamberlin 1947,
39) contained a paragraph announcing the design of the
ican maps using the Chamberlin projection (Times Books
1990, plates 97 and 102). Certainly, the Chamberlin projec-
tion requires more manual labor than other well-known
projections (Bretterbauer 1989). But the overall low distor-
projection known as the Chamberlin Trimetric. The author tions quoted in Chamberlin (undated) must have justified
of the article and the projection, Wellman Chamberlin (soon its use. Furthermore, the Chamberlin projection can be ap-
after appointed NGS chief cartographer), explained in the plied anywhere on the globe. It is not limited-as is the
second edition of this article (Chamberlin 1950, 104) the Albers Equal-Area - to certain latitude ranges.
principle and execution of the projection. This terse, graph- Numerical methods, practical for several decades thanks
ical explanation is reprinted here (Figure 1), and is the only to computers, could have supplied the missing proofs. Sur-
one by Chamberlin that can be found in the literature (Sny- prisingly, it was only recently that the situation changed,
der and Steward 1988). with Bretterbauer (1989). In this paper, Bretterbauer prec
Later literature has relatively few references to the pro- sented a table of distortions for a North American map on
jection. In the area of theoretical cartography, there seem the Chamberlin projection, constructed with a numerical
to be no major works that include studies and analysis of method over a uniform grid. The grid was computed with
the Chamberlin projection, an omission that at first sight formulas derived for the plotting of a graticule of parallels
might look odd in light of the preeminent place NGS has and meridians. This paper does not treat the determination
among map publishers. However, the construction and na- of the graticule in particular, but the more general forward
ture of the projection proposed by Chamberlin was purely (latitude/longitude to rectangular coordinates) and inverse
geometric. The formulas required to convert longitude and cases for the projection. Computer solutions for those two
latitude to polar or rectangular map coordinates were not cases are the bases for the numerical procedures described
derived for many years. Consequently, there was little in- in this paper.
centive to derive the distoption formulas for studying the The objective of this paper is to present a complete de-
behavior of the projection and its properties, nor was it scription and analysis of the Chamberlin projection. The
considered feasible to objectively evaluate the merits of the paper contains the following discussions:
Chamberlin Trimetric against those analytical projections
in use at that time.
Although there was a lack of theoretical proof, the con- 1. Description of the manual method
struction method seemed simple and the results good 2. Outline of the automated solution to the forward and
enough to make the case for the Chamberlin projection. In inverse transformations
fact, since the original announcement, NGS has published 3. Study of the singularities present in the automated
10 wall and atlas maps in the Chamberlin projection (Table solution (the forward transformation)
1), all using the manual procedure described in Chamberlin 4. Analysis of the distortions (the linear singularity, the
point singularity, and the plotting of the distortions)
5. Comparison of the Chamberlin distortions with dis-
Albert H.I. Christensen is a member of the GIS Project Office, tortions in conformal and equivalent projections used
IBM Corporation, 10401 Fernwood Road, Bethesda, MD 20817. for continental maps

Cartography and Geographic Information Systems, Vol. 19, No.2, 1992, pp. 88-100
In the Qumberlin Trirnetric Projection,
:Na1:ionaUjeographirSoci£ty cart1Jgraphers
sek.ct thru pOUlts determined by the size
an.dshape oEthe map l1J be drawn. For t:he
March, 1950,.N,ationaUjeographic map of
..Afrira, the poutl:s,on the globe, are arA,S,C.
Thethree light aLlved Lines connecting these
points form a spherical triangle. rliese lines
are the 9 reat circles or shortest distances
between points A, 13, C.
rfu. dark straisht lines als~ J1tU5ure the
true distances between POUl.ts A, S, C.
Tfz.eseare the sides of the spherical triangle
flattened out. TfI£Y form the basis for
plotting the flat map.
Tfz.edifference between these triangles
i11.diratesthe approximate distortion
front projE'c~ 3. parr of the curved
surface of the globe UpOJ"l. flat paper.
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

HAVINt; }r1AJU<.ED ON THE FLAT SHEET OFP.APER


mE POINTS A, B, C, THE MAP MAKER. THEN-
I

\;... ,
\1
\ \
v !
J
J
/

I /
/

" /
/
.5------

-- Draws Lines COl1.ll£ctinS- tlu- centers of the


small triangks f":"0d.uced.. These lines fOrfn;
the meridians aiuJ..parall.£ls of the fla.t nwp .

Figure 1. Manual construction of the Chamberlin Trimetric projection. (Original illustration on page 104 of Chamberlin [1950J is
reprinted by permission from the National Geographic Society)

6. Criteria for selecting the optimal configuration of pro- reduces the overall distortions to remarkably low figures.
jection parameters However, these results are achieved at a cost: linear dis-
continuities inside the map area.

It is well known in theoretical cartography that condi-


tions of conformality and equivalence are mutually exclu- Manual Method
sive. This paper is intended to show that the Chamberlin According to Chamberlin (undated), one selects three points,
projection, without completely meeting both conditions, "determined by the size and shape of the map to be drawn."

Cartography and Geographic Information Systems 89


Table 1. Chamberlin Trimetric basic triangles. (Reprinted by permission from the National Geographic Society)

Map Point A Point B Point C


Australia 134°E,08°5 1l00E,32°5 158°E,32°5
Canada Atlas 98°13'OO"W 135°W,40 N0
55°W,400N
61°39'00"N
Eastern South America 63°33'00''W 35°13'00''W 58°33/00''W
8°08'00"N 5°47'00"5 34°35'00"5
Northwest South America 69°W,25°S 85°W, looN 55°W, lOON
Southern South America 43°W,18°S now, 18°5 now, 56°5
Europe Wall Map 15°E, nON 8°W,33°N 38°E,33°N
Africa Wall Map 19°03'00''W 59°03'OO"E 20000'OO"E
24°25'00''N 24°25'OO''N 35°00'00"5
Canada Wall Map 150000'OO''W 45°00'00''W 97°30'OO"W
60000'OO''N 60°00'OO"N 50000'OO"N
North America Wall Map 150000'00''W 35°00'OO"W 92°30'00"W
55°00'OO"N 55°00'OO''N 10000'OO"N
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

South America Wall Map 800W,09°N 35°W,06°S 71oW, 53°5

-- .•.....•.•.

---------.---- /

/-
_----~t-
I:
//".... / !
I' :
f '
/ I
/ '
/
/ :
I
/ I

"

Figure 2. (a) (left) Distances and reduced azimuths on the sphere. (b) (right) The forward-case solution. The centroid i of the triangle
of intersections is the transform of point P on the sphere. Shown are the pairs of plane-reduced azimuths at vertices A' and B'.

These three points, A, B, and C, taken two at a time, de- the maps, the triangle is exactly or very approximately isos-
termine three arcs of great circles on the surface of the celes, with the base on a parallel. The selection criterion is
globe. The points and arcs form a spherical triangle. When discussed in detail later. Once the vertices were selected,
one examines a list of parameters for wall and atlas maps the construction of the projection required the computation
in the Chamberlin projection provided by NGS (Table 1), of the spherical distances among the three vertices. Because
it appears that Chamberlin selected the three vertices of the objective of the projection is to map a large country or
each spherical triangle so that a large part of the map area a continent, all operations were executed on the sphere.
would fall inside the triangle itself. Among these maps, the The work described in this paper was also executed on the
Canada atlas map is the one that seems to have been con- sphere.
structed with some other guidelines in mind, because the By applying the principal scale selected for the map, the
area within the triangle is not much larger than the outside cartographer then used the lengths of the spherical arcs to
area. The apex of the basic triangle is at a rather low latitude determine the shape and size of a plane triangle NB'C'.
(61° 39' N), leaving almost all the Northwest Territories, a This triangle and its vertices are hereafter called "basic"
good part of British Columbia, Labrador, Newfoundland, and its edges "bases," after Chamberlin (undated).
and northern Quebec outside of the triangle. In seven of As is usual in manual procedures, the map was con-

90 Cartography and Geographic Information Systems


Table 2a. Parameters for the four Chamberlin projections and their plots.

Radius of the sphere 1.0


Radius of the limiting circle = 22.243°
Area of the spherical zone (theoretical) = 0.468 radians
(numerical) 0.475
Coordinates of the centroid
of the four basic triangles: Longitude 90°
Latitude + 38°29/ 53~/5848
Grid origin Longitude -125°
Latitude 15°
Grid end Longitude 55°
Latitude 75°
Grid interval 2.471435529°
All graticules and uniform grids for plots of the four triangles were created in the same system, Albers Equal-Area with
first standard parallel at + 27°, second standard parallel at + 65°, longitude of origin = - 90°, and latitude of origin =
+46°.
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

Table 2b. Effect of the size of the basic triangle over the deformations (coordinates rounded to the nearest second).

Basic Triangles
T1 T2 T3 T4
Vertex A longitude 75°00'00" 68°48/55" 62°56'26" 57°30/50"
latitude + 300000 + 252036 + 202630 + 152138
Vertex B longitude -1050000 -1111605 -1170334 -1222910
latitude + 300000 + 252036 + 202630 + 152138
Vertex C longitude 900000 900000 - 900000 900000
latitude + 532954 + 604428 + 675903 + 751338
Triangle area (radians) 0.091 0.205 0.372 0.604
Triangle area as part of 0.194 0.438 0.795 1.291
the limiting circle
Deforma tions:
Minimum Linear (%) - 0.807 1.105 1.711 2.019
Maximum Linear (%) + 2.452 + 2.077 + 1.573 + 0.946
Maximum Angular n + 1.6 + 1.6 + 1.6 + 1.6
Minimum Area (%) 0.606 1.279 2.197 3.335
Maximum Area (%) + 2.081 + 1.335 + 0.318 0.951
S = Sum of squares of linear
deformations over circular area,
after Maling (1973, 71) 0.562 0.671 0.487 0.838
Areas between linear deformations
as parts of the limiting circle
-1.0 and + 1.0 0.493 0.629 0.000 0.000
-1.5 +1.5 0.688 0.827 1.000 0.000
-2.0 +2.0 0.882 0.991 1.000 0.997
-2.5 +2.5 1.000 1.000 1.000 1.000
-3.5 +3.5 1.000 1.000 1.000 1.000

structed by splining the features to be represented inside troid of the relatively small triangle, herein called the triangle
a graticule of parallels and meridians. The cartographer de- of intersections, formed by iABi, BCa' nd iCA'Refer to Figure
termined the position of each graticule intersection in the 1, reproduced from Chamberlin (1950), for a more detailed
following manner: Let P be a point on the sphere, and SPAI explanation.
SPB' and Spc the spherical distances from P to the three At this point, the reader may notice that a fact has been
vertices A, B, and C (Figures 2a and 2b). Selecting one of omitted: two nontangent circles intersect at two points, not
the bases first - for example, A'B' - the cartographer traced one. In describing the manual procedure, one may usually
an arc of a circle, with center in A' and radius SPAIand a ignore the second intersection, because the cartographer
second arc, with center in B' and radius SPB'The two arcs nearly always knew on which side of the base the two arcs
intersect in point iAB.The same procedure was followed should have intersected.
with the other two bases, B'e' and e' A', to obtain inter- The descriptions also ignore that in the limit, when the
sections iBCand iCA'The graticule point sought is the cen- sum of the two distances SPAand SPBis very nearly the

Cartography and Geographic Information Systems 91


40'

+20'
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

Figure 4. Ellipses of distortion for the Chamberlin projection and


graticule of parallels and meridians in the same projection. The
three vertices A, B, and C, with longitudes and latitudes, re-
spectively, equal to - 40°, 35°; + 80°, 35°; and + 20°,
- 36.0592897", define a relatively large equilateral basic triangle.

-120' -100'

Figure 3. The effect of singularities on lines, greatly enlarged.


(a) (top) A witness test: sections of great circles far away from the
loci of singularities. (b) (middle) Obtained with double-precision
arithmetic. Framed area measures 10 by 4 m on the ground. (c)
(bottom) Obtained with single-precision calculations. Framed area
is 30 by 12 km.

length SAD of the base AB, one of the three contributing


intersections is undefined. Furthermore, when the grati-
cule point approaches one of the basic vertices, only one
intersection is well defined. In the manual construction and
prior to the determination of the graticule, these singular-
ities could have been avoided by an adequate selection of
basic vertices. If the singular situations had been unavoid-
able, the cartographer would have realized, after a quick -110' -100' -90' -80' -70'

visual inspection, which intersections were well behaved


and should have been retained and which were unstable Figure S. Basic triangle T2 (Table 2b) and isogram of area dis-
and should have been rejected. tortion for the Chamberlin projection. Plotted in the same projection.

of singularities. Unlike its manual counterpart, the singu-


Automated Solution to the Forward
larities in an automated procedure are unavoidable. The
and Inverse Transformations automated solution to the Chamberlin forward transfor-
Following the manual method explained above, the design mation computes the vertices of the intersection triangle.
of a general computer procedure is not an overly difficult Each vertex is the result of an operation in which the data
task (Bretterbauer 1989). Therefore, a detailed account of consist of the length of a base and the spherical distances
the computer routines is not included in this work. How- from point P to the endpoints of that base.
ever, the task becomes more difficult with the consideration One readily apparent approach for executing this oper-

92 Cartography and Geographic Information Systems


Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

Figure 6. Basic triangle T2 (Table 2b) and isogram of maximum Figure 7. Basic triangle T2 (Table 2b) and isogram of linear
angular distortion for the Chamberlin projection, plotted in the distortion along parallels and meridians for the Chamberlin pro-
same projection. Asymmetries in this isogram are due to chance jection, plotted in the same projection. The graticule was omitted
coincidences of the uniform grid and bases of the projection. The for the sake of clarity. The geographic area is exactly the same as
grid is not registered to the centroid of the triangle. Hence, the in Figures 5 and 6. The lack of symmetry in the plot is explained
plot does not show a zero distortion value. The distortion value in the caption for Figure 6.
plotted near the center of the figure is the minimum found in the
grid of computed values.

ation is trigonometric: the solution of a plane triangle with


the three sides known. Let the endpoints of the base be A'
,
AREA OF TRIANGLE
and B'. iAB, the vertex triangle corresponding to AB in the OF INTERSECTIONS
triangle of intersections, is determined by solving the plane
triangle formed by three known distances SPN SPB, and SAS-
and the positions of A' and B' (Figures 2a and 2b).
A second approach is geometric, and replicates the man-
ual method: the intersection of the two circles centered in
the endpoints A' and B', with radii respectively equal to
the spherical distances SPA and SPB'
Both approaches lead to double solutions, with singular-
ities occurring when the two solutions tend to be coinci-
dental. In the first approach, the double solution appears,
because the triangle PAB may be located equally well at
either side of base AB. In the second approach, the two
solutions are due to the double sign of a square root. There Figure 8. A cross section of the distortions for the Chamberlin
is no discernible advantage in using one approach over the projection, for the extense basic triangle in Figure 4: area, an-
other. This author, after a few tests, chose the trigonome- gular, and minimum and maximum linear distortion on the line
tric approach, which is slightly faster. defined by vertex C and the centroid T of the basic triangle.
Whichever approach is used, the two intersections are
located symmetrically on both sides of the base used to intersection out of each pair, as was done in the manual
compute them (Figure 2). For each base, two intersections operation.
are determined, resulting in six for the three bases. These In the manual construction, the choice between the two
six intersections, taken three at the time, originate eight intersections is obvious. In fact, the cartographer would
possible intersection triangles. The centroid of one of these create only one intersection for each base by tracing the
eight triangles is the solution. It might be possible to select arcs toward the righf (left) of, for example, A'B', if the point
the correct triangle out of the eight. It is simpler and more on the sphere was to the right (left) of the arc AB. The
elegant to reduce the alternatives by first choosing the proper ambiguity is solved in the computer routine by first com-

Cartography and Geographic Information Systems 93


I
I
I
I T4
, I
" I
i"-
I', I

/ 'If
I I
I I
I I
/... I
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

" I
, I
'!---
I " , Figure 11. Isograms of area distortions for the four basic triangles
in Table 2b, plotted in Albers Equal-Area projection, with the
parameters listed in Table 2a. The four isograms must be inter-
preted with Figure 10.

Figure 9. Isogram of maximum linear distortion for the Cham- Table 3. Comparison of linear and angular distortions.
berlin projection. The relatively small area has vertex A in its
center. The basic triangle is the same as in Figure 4. Sections of Linear Area Angular
edges of the basic triangle are also illustrated. The contoured grid [%] [%] [0]
was uniform in latitude and longitude, with an interval slightly Transverse Mercator 10.0 20.8 0.0
less than 3 minutes of arc. Albers Equal-Area 6.4 0.0 7.1
Chamberlin Trimetric 2.1 1.3 1.6

'>0.
Parameters for the projections: Transverse Mercator-central
meridian at - 90°, scale factor = 1.0; Albers Equal-Area-
standard parallels at +25° and +53°, central meridian at
-90°; Chamberlin-as listed for T2 in Table 2b.

intersection selected is that which causes the same sign in


its reduced azimuth (aA and a8 in Figure 2b) as its spherical
counterpart (aA and a8 in Figure 2a). The same operations
"0. are repeated for the two other bases, Be and CA. The re-
sulting three points configure the triangle of intersections.
The centroid i is the projection of P (Figures 2a and 2b).
The inverse transformation is solved rather easily by a
rough estimate of the position on the sphere, followed by
an iteration of the forward solution. With a well-configured
basic triangle, such as the one indicated as T2 in Table 2b,
the inverse solution for a point interior to T2 requires five
iterations of the forward solution to ensure an accuracy of
/0· -'3-'-0_ \~.
10 arc seconds.

I ~f. c -c.:.-_II~· -,rio:----=9~· -;0' -70' -60'


Singularities in the Forward Transformation
Figure 10. The four basic triangles and the circular map area, The computer procedure must consider not only the gen-
plotted in Albers Equal-Area projection, with the parameters listed eral case of forward transformation, but two special cases
in Table 2a. where the plane coordinates are not continuous functions
of longitude and latitude: for points on or very near one of
puting the spherical and plane azimuths to the point and the great arcs, and for points nearly coincidental with one
the two possible projections, and then comparing the signs of the three basic vertices.
of the azimuths reduced to the d~ection of the base. The In subsequent derivations, the following should be noted:

94 Cartography and Geographic Information Systems


1. The programs prepared for the transformation invoke routines, of an order that depends on whether single-
mathematical routines that generally are supplied to- or double-precision arithmetic is used for all arguments
gether with the compiler for the language in use 3. The results returned by these mathematical routines
2. There is an error associated with the results of these generally cannot be trusted if the function is discon-
tinuous for the value of the argument
4. To avoid inconsistencies (catastrophes is a better word),
the program should test the arguments for borderline
conditions before passing them to the mathematical
routine in question

Regardless of the technique used to detect singularities


in functions, the calling program must then assign values
that will bring the results exactly to the limiting case.
In the particular case of the trigonometric approach de-
scribed previously, the detection of a singularity occurs in
the routine that solves the plane triangle by Heron's for-
mula. The same routine forces results to fit the logic of the
singular case. Although elementary in nature, the proce-
dure is given here: Let 5j {i = 1,2,3} be the known edges,
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

and CXi the unknown, opposite angles. In the expression for


the angles, three differences (P - 5i) are computed, where

P = semiperimeter = (51 + 52 + 53)/2

If one of these differences (P - Si) is zero or negative, a


condition that could only be created by the accumulation
of round-off errors, the subroutine forces the following re-
sults on the angles:

exi = 180°; i = {1,2,3}


Figure 12. Isogram of finite distance distortions for the Cham- CXj = 0°; j = mod(i,3) + 1
berlin projection, using triangle T2 (Table 2b) and its centroid as CXk = 0°; k = mod(j,3) + 1
center.

Table 4. A comparison of distance distortions.

Transverse Albers
Center of Mercator Equal-Area Chamberlin
Distances Max. Min. Max. Min. Max. Min.
Centroid (See Table 2a) + 2.61 - 0.00 + 3.04 - 2.91 -0.62 -0.66
Basic Vertex (') + 6.93 + 1.43 + 6.06 - 5.89 +1.87 -0.69
Long. = -70°; Lat. = +45° +5.33 +0.76 +2.97 -2.64 +0.69 -0.64
The distortions are expressed in percents of the spherical distances.
(') from the basic vertices listed for T2 in Table 2b: Vertex A for computing Transverse Mercator distortions; Vertex C
for the Albers Equal-Area distortions. For the distortions in the Chamberlin projection, the three vertices yield the same
results.

Table 5. Precision in the distortions of the Albers projection and the radius c of the small circle.

Distortions
Linear Area Angular
Radius Maximum Minimum
[radians] [%] [%] [%] [0]
Analytical values from infinitely 1.00855 0.99152 1.00000 0.97612
Snyder (1987, 291) small
Numerically computed 10-3 1.00858 0.99152 0.99995 0.97749
10-4 1.00856 0.99152 0.99995 0.97626
10-5 1.00855 0.99152 0.99997 0.97613
10-6 1.00855 0.99152 0.99999 0.97613
10-7 1.00855 0.99152 0.99999 0.97612

Cartography and Geographic Information Systems 95


The exception in the geometric approach is treated in a that transforms lines rather than points, and that incor-
similar manner. Even if the singularities are anticipated, porates a mechanism for smoothing the spikes. The routine
they still have adverse effects, as the following section will would be considerably more complex than the point-by-
show. point approach used in the most popular cartographic
libraries.

Linear Singularity
Point Singularity
An excellent test for any transformation routine is to plot The second class of singularity, where P coincides with or
a dense family of curves, such as arcs of great circles, in is very near one of the basic vertices, is easier to deal with
the projection in question with a large, accurate pen plotter. than the first class (points on or near one of the bases).
It might be possible to distinguish any small irregularities First a distance D must be estimated, considering the kind
caused by the type of singularities discussed here by ex- of arithmetic used and the range of the coordinates in the
amining the result obliquely, at a very acute angle. The map area. If the distance from a point P to one of the basic
effect is illustrated, greatly enlarged, in Figures 3a, 3b, and vertices is less than D, then P is assigned the coordinates
3c. A set of great-circle arcs was created at an angle with of the basic vertex.
another arc C of a great circle, traced in bold lines in the Both classes of singularities have a prominent place in an
three figures, and projected using T2 in Table 2b as a basic analysis of distortions of the Chamberlin projection, as the
triangle. For an arc C located arbitrarily in the interior of author will prove in the following sections.
the basic triangle and far from the singularities, the pro-
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

jected lines appear as smooth curves (Figure 3a). Distortions in the Chamberlin
By using one of the bases as arc C, the plots acquire quite
Trimetric Projection
a different aspect (Figures 3b and 3c). The transversal am-
plitude of the spikes near arc C depends primarily on the
angle formed by the secant line with the base; and, in a
Analytical Expressions and
minor way, on the configuration of the basic triangle; as a Numerical Method
well as on the distance from the intersection of the line Linear, angular, and area distortions on map projections
with the base to the nearest vertex of the basic triangle. usually are derived with general formulas from functional
Figure 3b is the result of projecting with double-precision relationships of the type:
routines. The area shown measures 10 m by 4 m on the
x = /'. (<p,A)
ground, with spikes measuring approximately 6 cm across
Y = fy (<p,A)
the length of the line. This amplitude varies much more,
however, with the type of real arithmetic used. Figure 3c An analysis of distortions in this traditional manner for the
is the result of an execution with single-precision arith- Chamberlin projection is ruled out by the lack of such ex-
metic. In this case, the plot covers a ground area of 12 km plicit functions. Therefore, in order to study its distortions,
by 30 km, and the spikes illustrated measure approximately a numerical method was devised that permits drawing and
1 km. At 20 km from C, the transversal amplitude of the presenting conclusions in an objective manner. The results
spikes is reduced to 200 m. of this numerical analysis are presented in Figures 4 to II.
These figures attest to the need for double-precision To estimate the distortions at a point P, with longitude
arithmetic. At a ten-millionth scale, if single-precision arith- A and latitude <p, a circle is placed on the sphere, with
metic is used, the spikes would be 0.1 mm long, seven radius c and centered at P. A number of points are selected
times larger than the resolution of cartographic-quality laser ("stroked," in computer graphics parlance) at regular in-
plotters. tervals along its circumference. The points are then pro-
Again, notice that the above results were obtained with jected with the Chamberlin forward-case routine. The shape
the previously described mechanism, with the solution of they form is the projection of the circle. Figure 4 shows
the formulas always bound to their logical values. Had that several such projected shapes, which vary from almost-
mechanism been absent, the mathematical routines would perfect circles to rather elongated ellipses.
have failed, acknowledging an exception and possibly re- The Tissot Indicatrix, a well-known concept in mathe-
turning incorrect values to the calling program. The con- matical cartography, is the transform of an infinitely small
sequences would have been catastrophic for the entire plot, circle defined on the surface of the sphere. The Indicatrix,
not just for the minute part illustrated here. known also as the Ellipse of Distortion (Maling 1973, 64),
Many interrupted projections show linear discontinui- can be used to calculate the diverse distortions in a projection.
ties, but these occur only at the map borders. The Cham- The distinction between a Tissot Indicatrix and the shape
berlin projection is exceptional in that the linear loci of determined by numerical means lies in the approximation
singularities are inside the map area. to the infinitely small radius used to generate the Indicatrix.
Can anything be done to minimize the effect of these This radius is a limit, not usable with numerical means. To
singularities? One can make the basic triangle large enough apply the numerical method, a small finite value c must be
to encompass the whole map area. Unfortunately, increas- used. Guidelines were searched in the literature as to how
ing the size of the basic triangle also increases the distor- precise any computation of distortions should be, in an
tions. The size of the basic triangle and its effect on the attempt to substantiate the selection of a value for c. Work-
distortions are discussed later. ing back from this precision, it was then possible to esti-
A better alternative is to prepare a forward-case routine mate in a pragmatic manner a proper value for c. In addition,

96 Cartography and Geographic Information Systems


the application of the numerical method ought to be pre- The symmetry of the basic triangle also explains the null
ceded by a proof of the programs, and a decision on the angular distortion at the centroid. There is no predominant
real arithmetic to be used. The Appendix includes an out- effect at the center of symmetry. Hence, the transform of
line of how these issues were resolved, as well as an ex- the circle is a circle, and the angular distortion, which is
planation of the determination of distortions by the numerical approximately proportional to the difference between the
method. semiaxes of the ellipse, becomes zero.
All the basic triangles in this work are equilateral. This
Plotting the Distortions is partly because their symmetry seems to reduce the influ-
Once established, the numerical method can be used to ence of the triangle configuration in the analysis of errors
generate arrangements of distortion values at uniform in- and distortions, and partly because the projection and its
tervals of latitude and longitude. To distinguish them from distortions are invariant with the orientation of an equilat-
the graticules of parallels and meridians provided for ref- eral basic triangle on the sphere.
erence in the figures, those arrangements are herein called Because of the symmetry of the equilateral triangle, the
"grids." These grids were then contoured into the various diagrams of distortions tend to exhibit a circular shape in-
isograms that illustrate this article. In all figures and tables side the basic triangle. In particular, the isolines at the cen-
presented here, the linear and area distortions are ex- ter of the triangle should be almost exactly circular (Figure
pressed as percents of the unit value. The angular distor- 5), if they are created with a very fine grid and plotted in
tions are in decimal degrees. a Chamberlin projection. With triangles of increasing sizes,
Some of the lines in the isograms are more angular than the isograms near the centroid take a more triangular shape,
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

others. This difference is caused by the contouring pro- while tending to recover the circular shape toward the out-
gram, and by the intervals selected for the grids of distor- side of the triangle.
tions. The larger the intervals, the stronger the angularities However well known, the question of basing conclusions
in the lines. on numerical grids deserves comment here, because that
The isograms of angular and area distortions are easily is the focus of this paper. Provided that the functions rep-
conceived and plotted (Figures 5 and 6). Linear distortions resented as isograms of a grid are continuous, and that the
are more problematic, since they should be considered in sampling interval has been judiciously chosen, the iso-
pairs of values (maximum and minimum) or along parallels grams would offer a fair idea of the function's behavior.
and meridians. They are usually pictured as the semiaxes But the Chamberlin projection is not continuous, as ex-
of a small number of distortion ellipses, as in Figure 4. plained earlier. Because of the singularities, distortion is-
However graphic, this mode of illustration lacks the power ograms turn out to be - up to some point - deceptive. If the
of isograms and the objectivity of the labels. Therefore, the positions of the grid intersections are such that all singu-
linear distortions are presented in Figure 7 as two sets of larities are missed, the picture offered is too favorable. Con-
isolines, approximately vertical for the distortions along versely, if one or more grid intersections are very near the
meridians, and approximately horizontal for those along locus of singularities, the perturbations appear and are pic-
parallels. The reader will have to use both values to esti- tured with an amplitude equal to at least the grid interval.
mate the total distortion at a point. Consequently, the isograms offer a bleak view (Figure 9).
Notice the elongated shape of the ellipses away from the Since the two alternatives are equally unsatisfactory, this
basic triangle in Figure 4. Also remarkable are the strong author chose one or the other according to the purpose of
curvature of the projected edges of that triangle, and the the illustration. In Figure 9, the singularities were pro-
angle distortions. In Figures 4 and 8, these characteristics voked. In Figure 11, they were successfully avoided. In
of the Chamberlin projection were purposely exaggerated others, the outcome was left to chance, as was the case in
by using an unrealistically large basic triangle. Figure 4 shows the isograms selected to illustrate this section (Figures 5, 6,
only one quadrant of the whole picture. It is unnecessary and 7).
to show the entire picture, since the large triangle is equi- Figure 9 is a view of the area near one of the vertices of
lateral; the other quadrants would show the same set of a basic triangle. In this case, the uniform grid was comple-
ellipses arranged symmetrically with respect to the three mented by extra points located close to the bases. The grid
bases. interval selected, 3 minutes of are, was small relative to the
Figure 8 is a cross section of the distortions in the basic 2.47-degree interval used for the rest of the figures in this
triangle. The vertical scales are not shown, because of the paper. With extra points and that short interval, singular-
disparity of measures and units. It should be noted that ities are clearly seen. The significance of the isogram labels
the distortions do not reach a minimum at the centroid, in the figure is relative. Microscopic changes in the posi-
except for angular distortion. The behavior of the area of tions of the extra points could increment the corresponding
the triangle of intersections is also interesting. There are distortion values to any value within the range of the real
two points at which the area is nil-one at the vertex, which arithmetic in use (the forward transformation used to com-
is easy to anticipate, and the other near the straight line pute the distortions behaves like a function with an infinite
defined by the projections of the other two vertices. The limit for a finite value of the independent variable).
sign of the area is given by the formula used. Where the The distortion isograms in Figures 5 (area), 6 (angular),
intersection triangle results with its vertices listed in the and 7 (linear) correspond to a Chamberlin projection based
same direction as the vertices in the basic triangle, the sign on the triangle indicated as T2 in Table 2b. For brevity's
of the area is positive. Conversely, if the listing sequences sake, this projection is referred to as T2. The grid was con-
are reversed, the sign is negative. structed with an interval equal to 2.47 degrees, and the

Cartography and Geographic Information Systems 97


isolines were plotted in projection T2. Other details of the cedure outlined by Maling was not rigorously followed,
projection and map area can be found in Tables 2a and 2b. because a grid in spherical coordinates with its origin in
The projections of the bases that make triangle T2 are plot- the common centroid would have tended to be tangent to
ted with thick lines. The sharp indentations in the isograms the three bases, which are precisely the loci of the singu-
for linear and angular distortions are the results of chance larities. Instead, a grid uniform in longitude and latitude
coincidences between grid intersections and bases of the was used (Table 2a). Singularities were avoided by choos-
projection. ing an appropriate grid interval. Only those grid elements
Distortions were also computed for the two well-known within the circle were included in the summations. This is
Transverse Mercator and Albers Equal-Area Conic projec- also the cause of the broken isolines in the periphery of the
tions. The distortions were computed with the formulas in isograms in Figure 11, as well as in Figures 5 and 6.
Snyder (1987) for the parameters listed under Table 3, and If one looks at Table 2b, it appears that projections TI,
for the circular area specified in Table 2a. These results, T2, and T3 are nearly the same. Which is optimal for all
and the maxima and minima obtained by the numerical distortions? From the plots in Figure 11, one might con-
method for projection T2, were then assembled in Table 3. clude that TI is the best. But if one wants to maximize the
Table 3 shows the linear distortions of the Chamberlin band around the line of no area distortion, T2 seems the
projection as several times smaller than those of the other best. Column T3, Value 5, shows minimums of distortion
two more traditional and widely used projections. In ad- and range of variation for the area distortion. But T2 has
dition, its maximum angular distortion is several times slightly better linear distortions. T4 is clearly inadvisable.
smaller than that of the Albers projection, while its area The author selected T2 for the more detailed illustration of
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

distortion is better than that of the Transverse Mercator the distortions already discussed as Figures 5, 6, and 7, and
projection by a factor of 16. This table substantiates what for the comparison presented in the next section.
intuition, or insight supported by the few manual tests pos- Regrettably, it is not evident from the TI, T2, and T3
sible at the time, must have told Chamberlin-namely, that tests what advice could be given that would be valid for
his creation was an excellent compromise between the con- the many practical cases possible. The poor differentiation
formal and equivalent alternatives. among the performances of those three configurations pre-
cludes any positive recommendation. Only if most of the
Size of the Basic Triangle and Its Effect area to be mapped would nicely fit into a circle, as would
be the case for Australia, would the analogy justify a rec-
on Distortions
ommendation such as T2. For more irregular areas, such
Once routines for computing the projection distortions are as Canada, the experiment described here would perhaps
available, it is possible to investigate the effect of the size show the cartographer the way for determining the optimal
of the basic triangle on the distortions, and, it is hoped, configuration.
determine an optimal size of this triangle for a given tar-
geted map area.
Table 2b was compiled with those two purposes in mind. Equidistant Property of the Chamberlin
For the sake of simplicity, and also because the deforma- Trimetric Projection
tions in the Chamberlin projection exhibit a strong central In Snyder (1987, 192) the Chamberlin projection is qualified
symmetry, a circle was selected as the map area under study. as an "approximately 'three-point equidistant' projection."
Distortions were computed for four basic triangles, indi- As Table 4 shows, the choice of words could not have been
cated in Figures 10 and 11 as TI, T2, T3, and T4. To ease more suitable. The Chamberlin projection is exactly equi-
the plotting and interpretation of the test results, these four distant for the lines measured between the three vertices
triangles were defined with the same centroid and with of the basic triangle. In addition, it is possible to show that
linear dimensions differing by approximately equal amounts. between the projection of a base and the straight line de-
To save space, only the area distortion is illustrated for fined by the projection of its endpoints, there are points
the four triangle configurations, while the size of the four whose distances to those endpoints are true.
plots is quite reduced. Contouring was limited to a few If the picture must provide a general idea of the property
integer values, to ease the interpretation of such small-size on the map area, the illustration of distance distortions is
figures (Figures lIa to lId). These figures, as well as Figure not an easy matter. These distortions are not scalar func-
10, were plotted using an Albers Equal-Area projection. tions, which easily can be pictured by means of isolines.
Table 2b was not affected by the choice of the projection Rather, they could be equated to tensor magnitudes, since
to be used in the plots. All listed areas were computed on at each point there exists a vector for each angle of mea-
the sphere. The line indicated as "Value 5" is the result of surement. The illustration of such functions exceeds the
an attempt to follow the procedure outlined in Maling (1973, scope of this paper. Instead, a number of experiments were
71): carried out, summarized in Table 4. The first column con-
S = I (1 - a) (1 - b) d.AlIdA tains the "Center," the point from which spherical dis-
tances were determined to the intersections of a uniform
where a and b are, respectively, the major and minor sem- grid. The grid intersections were projected with the Cham-
iaxes of the ellipse of distortions, and dA is the area of a berlin routines, using triangle T2 (Table 2b) and the corre-
spherical quadrilateral at a point (<p,A) with sides equal to sponding planar distances computed and compared with the
.6.lp and .6.A. The domain of these summations was a cir- spherical ones. As verification, some of the results were plot-
cular area, with the radius indicated in Table 2a. The pro- ted as isolines. For the reasons mentioned above, only one

98 Cartography and Geographic Information Systems


of the plots, Figure 12, is included in this work. The maxima Appendix: Magnitude of c and Computation
and minima for each center were recorded in Table 4. of Distortions by the Numerical Method
The distortions obtained for the Chamberlin projection
To apply the numerical method for estimating distortions,
with the centroid of the basic triangle as center of distances,
the radius c of the small circle must first be determined.
shown in Figure 12, are remarkable for their uniformity
The principle involved is that, within a certain approxi-
over the whole of the map area. The extreme variation is
mation, the numerical method should yield the same re-
under 0.04%, which for a map of 1 m square, would rep-
sults as the analytical formulas, if these formulas were
resent a variation of 0.3 mm for the largest distance possible
available. The accuracy chosen for the approximation is based
from any point in the map to the center. In addition, the
on the number of decimal figures with which distortions
small overall value for the distortion should also be consid-
are ordinarily listed in textbooks.
ered when judging the equidistance property in the Cham-
A search in a classical book yielded three decimal places
berlin projection.
Gordan-Eggert 1939), and in a more recent book, seven
For reasons of space, Table 4 was shortened to three
decimal places (Snyder 1987). The mean, five decimal places,
entries. Two of them correspond to centers placed at char-
was chosen for this paper, with the certainty that the pre-
acteristic points of the Chamberlin projection, and the third
cision would be more than appropriate to support any con-
to an arbitrary position outside the basic triangle.
clusions made with small-scale mapping in mind.
The two projections already used in comparisons (Albers
Routines for expressions of distortions and transforma-
Equal-Area and Transverse Mercator), with the parameters
tions for two widely-used projections, Transverse Mercator
listed under Table 3, were used with the results shown in
and Albers Equal-Area, were next prepared and incorpo-
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

Table 4. From this table, it is possible to conclude that the


rated into a program that stroked and projected a small
Chamberlin projection behaves considerably better than the
circle with a modifiable radius.
other two, with respect to distance distortions.
The same program computed area, linear, and angular
distortions from the projected figure by the numerical
Conclusions method. The maximum and minimum linear distortions
Discontinuities exist in the solution to the Chamberlin pro- were evaluated from the maximum and minimum distance
jection that may cause wriggles in lines. Tests using single between the N projected stroked points Cj and the pro-
and double precision seem to indicate that wriggles in lines jected center of circle P.
are only noticeable if one uses single-precision arithmetic.
A complete computer solution to the effect of the discon- a = [ max{d;}i i = 1, N ] Ie
tinuities is a double-precision routine that transforms lin~s b = [ min{dJ; i = 1, N ] Ie
rather than points, and smoothes the spikes. With four where
examples, it was shown that a substantial increase in size
of the basic triangle causes remarkable increases in the dis- dj = distance from P' to Cj
tortions. Other than this conclusion, the four examples pro- c = radius of circle
vided no strong indication of how to select the optimal
basic triangle. Finite distance distortions are remarkably low The area distortion, or exaggeration of area (Maling 1973,
and uniform. All the distortions compared favorably with 67), was computed as the ratio between the area of the
those of the Transverse Mercator and Albers Equal-Area closed figure formed by the N points Cj, and the area of
the circle.
projections. Additionally, the Chamberlin projection is not
The maximum angular deformation w was estimated as
limited to narrow strips along a particular meridian or par-
allel, as are the other two. Therefore, provided that its dis- the maximum departure from a right angle displayed by a
continuities are smoothed out, the Chamberlin projection pair of directions at P', which correspond to two directions
should be considered an excellent candidate for small-scale on the sphere forming a right angle at point P. For each of
mapping with minimal values for all distortions. the N points Cj, the projection of a second point Qj at right
angles with Cj was also computed and projected. The angle
ACKNOWLEDGMENTS at P' between Ci and Qi was taken as the angular distortion
The author thanks IBM Corporation, for the encouragement re- for that pair of directions. The maximum distortion ob-
ceived in preparing this paper and for the use of the equipment tained through N determinations was accepted as the max-
needed to compile the tables and plot the figures, and, in partic- imum angular distortion w. The error in the determination
ular, Thomas Delaney, IBM, General Systems Division, for his of w is equal to 2"IT/N.The error in the other distortions
unfailing support. The author is grateful to the National Geo-
depends less on N. When computing a grid of values, N
graphic Society (NGS) for its permission to include Table 1 in this
paper and to reprint Figure 1 from Chamberlin (1950). To Charles cannot be too large, for reasons of processing time. N =
F. Case and David Miller, both from NGS, thanks for the patience 120, which means that the error in III is of the order of 0.05
with which they met and satisfied the author's many requests. The radians, was the value used for computing all the grids
author is particularly indebted to J.P. Snyder at the U.S. Geological needed for this article. Therefore, in Table 2b, w in degrees
Survey for reviewing the manuscript and offering thoughtful, and is listed with just one decimal place.
tactful, suggestions. Note that if the transform of the small circle was truly
plotted as an ellipse, once its two semiaxes a and b had

Cartography and Geographic Information Systems 99


been determined, the other two distortions could have been helps prove that the computer routines prepared for the
found with the known expressions: numerical method are error-free. It would have been more
difficult to prove the routines correct with the Chamberlin
p = a'b
transformation routines, since there was nothing to com-
w = 2'arcsin((a - b)/(a + b)) (Maling 1973, 67)
pare the results with. Instead, they were tested and proved
However, the values arrived at by the numerical method correct against the theoretical values found for the Albers
can be accepted even if the small circle was transformed and Transverse Mercator projections.
into something very different from an ellipse, as will hap-
pen if the small circle contains one or more singular points. REFERENCES
The distortions were computed analytically for the Trans-
verse Mercator and Albers projections, with routines pre- Bretterbauer, K. 1989. "Die trimetrische Projektion von W. Cham-
pared with formulas extracted from Snyder (1987; 58, 100) berlin." Kartographische Nachrichten, vol. 39, no. 2, pp. 51-55.
Chamberlin, W. Undated. "A Description of the Chamberlin Pro-
for a number of points. With the same data, the distortions
jection." Internal memorandum. Washington: National Geo-
were computed by the numerical method with different graphic Society.
radii, starting with a relatively large value and decreasing Chamberlin, W. 1947, 1950. The Round Earth on Flat Paper. Wash-
it by powers of 10, and using very large values for N. ington: National Geographic Society.
With the resulting distortions, several short tables were Jordan-Eggert.1939.Handbook of Geodesy. Stuttgart, Germany: Army
prepared for both projections. Table 5 shows the Albers Map Service. Translated in 1962 by Martha W. Carta.
experiment, for a transformation established with the pa- Maling, D.H. 1973. Coordinate Systems and Map Projections. London:
rameters listed in Snyder (1987, 291), and for the data point George Philip & Son Ltd.
Downloaded by [University of Illinois Chicago] at 15:25 10 February 2015

used therein. Snyder, J.P. 1987. Map Projections, A Working Manual, Professional
In Table 5, it is easy to see that, with a decreasing radius Paper 1395. Reston, Virginia: U.S. Geological Survey.
Snyder, J.P., and H. Steward, editors. 1988. Bibliography of Map
for the circle, the distortions computed by the numerical
Projections, Bulletin1856. Reston, Virginia:U.S. GeologicalSurvey.
method converge to the theoretical values listed in the up- Snyder, J.P., and P.M. Voxland. 1989. An Album of Map Projections,
per row, and that the numerical method will warrant re- ProfessionalPaper 1453.Reston, Virginia:U.S. GeologicalSurvey.
sults similar to those obtained with rigorous formulas if a Times Books. 1990. The Times Atlas of the World, eighth edition.
radius c equal to 10-7 radians is used. In addition, Table 5 London: Times Books.

DON'T GET LEFT BEHIND!


if you can't make it to ACSM meetings, our proceedings are the next best thing.
Keep up with the latest research and technology!
ASPRS/ACSM 1992 ANNUALCONVENTION GIS/LIS '91 ACSM/ASPRS 1991
Proceedingsfromthe 1992 AnnualConventionin Albuquer- AtGIS/LIS'91 in Atlanta,topicsincluded FALLCONVENTION
que, NewMexico.TopiCSincludeoptimalsurveyingtech- resourceand environmentmanagement, Atthe 1991 ACSM/ASPRS Fall
niques,countyremonumentation,successfulproposal transportation,realestate,implementa- Conventionin Atlanta,topicsincluded
preparation,howto createa successfulsurveyingprogram, tion,trainingand education,determining CPS/LISlegalaspectsofsurveying,
localplanecoordinatesystems,GRIDsystemtax mapping needsand costs,planningand support, GPSand the vertical,standards for
withAutoCADand DCA,GPSand the U.S.ArmyCorpsof user interfaceissues, erroranalysis, digitaldata bases, digitalmap filesfor
Engineers,topographicmappingwithGPS,certificationand standards,data quality.raster and vector cartographicapplications,measuring
standards, spatialstructure and designprocess,usingGIS geoprocessing,and visualization and positioningtechniquesforthe land
to developa digitalproductgenerationsystem,rendering techniques.A valuablereferencethat will surveyor,automatedapplicationsof
hydrographicfeaturesusingthe MKS-Aspect, and GPS helpyoustay informedon the latest terraindata, spatialmodeling,remonu-
advancesin GIS/LIS, mentationofsurveycomersand GPS,
techniquesforGISusers. and high-accuracyregionalnetworks,
Volume l-ASPRS/Softcover/606 pp./Order #T713 Softcover/l045 pp./Two-Volume Set
Volume 2-ACSM/Softcover/294 pp./Order #T714 Order #T7ll Softcover/500 pp./Order #T712
Price per volume: $15 Member; $25 Nonmember Price: $45 Member; $75 Nonmember Price $15 Member; $25 Nonmember
American Congress on Surveying and Mapping

All orders must be prepaid. Send orders to ACSM,5410 Grosvenor Lane, Bethesda, MD 20814-2122. To order by telephone
using a credit card please call (301) 493-0200. Fax: (301) 493-8245. Allowthree to four weeks for delivery. Checks must be in U.S.dollars,
payable in the United States. CODorders not accepted. Prices are subject to change without notice.
QTY, ORDER# TITLE PRICE SENDTO: C-P492

Date Day Phone _ /__


Name

Shipping add 10% Organization


Maryland residents add 5% sales tax Address
Member # TOTAL
o Mycheck or money order is enclosed. Card # _
Clty State __ ZIP_
Please charge my: 0 MasterCard 0 VISA Exp. Date _ Signature _

100 Cartography and Geographic Information Systems

You might also like