Professional Documents
Culture Documents
How Mercator Constructed His Projection
How Mercator Constructed His Projection
[2] A solution for the problem was found by Mercator himself 23 years
later, in 1569, with the construction of his world map Nova et aucta orbis
terrae descriptio ad usum navigantium emendate accomodata, that is,
‘’New and improved description of the Earth corrected for use in
navigation’’. [*] In this map meridians and parallels form a regular mesh of
rectangles where the spacing between parallels increases with latitude in
such a way that rhumb-lines [3] (that is, those lines having a constant
course on the surface of the Earth) are represented by straight segments
making true angles with the meridians. What Mercator didn’t know (could
not have known) is that his invention was well ahead of its time, as
1
2
navigation was still not prepared for using a chart where the places were
located by latitude, longitude and true geographic directions. Two
hundred years had still to pass, until the longitude problem was solved,
the spatial distribution of magnetic declination was known and the pilot’s
reluctances removed, so that Mercator’s projection could be fully adopted
in navigation and the old ‘plane chart’ definitely dismissed.
[4] According to the testimony of Walter Ghim, the mayor of the town of
Duisburg and Mercator’s first biographer, he said more than once that his
invention corresponded to the squaring of the circle in a way that nothing
seemed to be lacking save a proof. Mercator did not explain how (and
when) the invention was conceived and his method has been the object of
a long and inconclusive debate among historians for more than a century.
However this cryptic statement contains some clarifying elements which
combine perfectly with the conclusions that we have reached in our
research.
The goal of our research was to identify the method used by Mercator to
calculate the projection adopted in his world map of 1569. [5] Here is the
summary of the presentation. [*] First a quick overview of Mercator’s
world map is made focused on its most relevant geometric features; [*]
second a cartometric analysis of the chart’s geographic graticule is
presented, with the purpose of estimating the errors associated with the
construction method, which we have called the ‘error signature’ of the
chart; [*] third, a few of the methods described in the literature are
reviewed, taking into account the error distribution determined before. A
new method, based on the use of a table of rhumb is proposed and
tested; [*] and finally, the results obtained before are re-assessed in the
light of Mercator’s comments and explanations about his projection, and
the best solution is chosen.
2
3
previously kept in the municipal library of Breslau, [9] Poland, was lost
during World War II; only this facsimile remains.
[10] The map contains two scales graduated in one degree intervals: [*]
one scale of longitudes along the equator, from 0˚ to 360˚; [*] and one
scale of latitudes, from 66.5˚S, at the bottom, to a little less than 80˚N, at
the top. [*] Near the lower right corner there is an abacus, entitled
Organum Directorium, [11] composed of a graticule of meridians and
parallels for the northern hemisphere, matching the one of the map. [*]
From its lower and top left corners radiate seven straight lines
representing the seven classical rhumbs. [*] Centred on these same two
points two quarter circles graduated in one-degree intervals are shown.
According to the instructions given by Mercator in one of the legends of
the map, the Organum was intended to be used to solve simple problems
of navigation, like finding the coordinates of the point of arrival given the
course and the distance from some point of departure.
3
4
the circle. With this quantity and the radius of the circle (which is a little
less than 90 equatorial degrees) we can easily determine the original
ordinate of the parallel, a value not necessarily equal to the vertical
distance measured on the paper, due to distortion. [16] Notice that this
result does not depend on the physical distortion of the abacus because
we don’t make any direct measurements on it.
[17] Thus the next step was to determine the ordinates of all parallels of
the Organum, compare them with the theoretical values of the Mercator
projection (the so-called meridional parts, which are calculated by a
formula) and derive an error distribution. [18] The results are shown in the
table at left, were the last column is the difference between the ordinates
we have determined from the Organum (the Ys in the table) and the
theoretical values of the projection (the MPs in the table). These
differences are illustrated on the graph at right as yellow circles, together
with a smoothed line, which was interpolated through the individual
values. To this line, which we consider to reproduce very closely the
original errors associated with Mercator’s method, we have called the
error signature of the chart. [19] The next phase was to compare this error
signature with the errors produced by the various methods described in
the literature. Here, we have assumed that the scale of latitudes in the
Organum is identical to the one of the chart. This is a more than
reasonable assumption given the fact that the Organum was supposed to
replace the main chart in the resolution of navigational problems.
This is not surprising if we recall that most of those proposals don’t take
into account the errors affecting Mercator’s graticule and that none of
4
5
them considers the distortion of the paper. I will refer to only two of the
methods reviewed by D’Hollander, which illustrate those facts. [21] The
first was proposed by the Finnish geographer Nils Albert Nordenskjöld, in
1889; the second, by the Portuguese Captain Abel Fontoura da Costa, in
1933. [*] Both adopt a concept equivalent to the one introduced by Pedro
Nunes in 1537, in which a series of rectangular grids with growing
latitudes are piled up, where the proportion between the length of the
meridian and parallel arcs are locally conserved in each of them. In the
solution of Nordesnkjold, the range of each partial grid is 10 degrees of
latitude (as in the figure). The method of Fontoura da Costa is based on
the same principle but uses variable latitude intervals. These were chosen
in such a way as to reproduce the errors affecting the copy of the chart
kept in the French national Library. What the author didn’t realize with
this ad hoc solution was that about a half of those errors were caused by
physical distortion of the paper and had nothing to do with Mercator’s
method. [22] Here we have the comparison between the corresponding
error curves and the error signature of the chart. It is clear that none of
the solutions is good enough: one because the errors are too small; and
the other because they are too large.
5
6
theory, as we can see in the graph, showing the errors associated with his
method, together with the error signature of the chart.
When Mercator engraved his world map, in 1569, at least two different
tables had been calculated in central Europe: [34] the one by the English
mathematician John Dee, around 1558, in his Canon Gubernauticus; [35]
and the other described mathematically (although not calculated) by
Pedro Nunes, in his Opera, published in 1566. Other tables of rhumbs
were made after 1569, such as the one by Edward Wright, in 1599, and by
Simon Stevin, in 1605. All of them, except the one by Wright, were
calculated by solving iteratively small triangles along a rhumb-line on the
sphere.
The fact that I want to stress now, and this is an important point of our
presentation, is that constructing a Mercator’s graticule with a table of
rhumbs is very simple and intuitive. [36] On square paper, [*] draw a
horizontal line representing the equator, and graduate this line with a
scale of longitudes. [*] Draw a series of straight and equally-spaced
meridians, perpendicular to the equator. [*] Choose one of the traditional
rhumbs, for example the fourth rhumb (45º), and represent it as a straight
segment starting from the lower left corner and making an angle of 45
degrees with the meridians. All we have to do now is to trace the parallels
of the projection in such a way that this straight line will match the fourth
rhumb on the sphere. [*] For the purpose, we will be using this table of
rhumbs, the same as before. [*] The first point of the rhumb will be the
lower left corner of the graph, whose coordinates are 0 degrees of
latitude and longitude. [*] Now, take the second point on the list, with
latitude 10º and longitude 10º 3’. [*] Trace a vertical line representing its
meridian, using the scale on the equator, and find the intersection with
6
7
the red line. (and now comes the magic part) [*] The horizontal line
containing this point will represent the parallel of latitude 10º N in our
projection. [**] Repeat the procedure with the other points until all
parallels have been found. [37] The work is done. Very simple, is it not?
But how can we be sure that all rhumb lines, and not only the one used to
construct the projection, are represented correctly? We can prove it
mathematically today, but not at Mercator’s time. The only way available
to him would be to experiment with all the seven rhumbs and confirm
that they were indeed depicted as expected [*]. Under this light
Mercator’s comment that nothing seemed to be lacking in his invention
save a proof begins to make sense.
But can we conclude, just because the method is simple, intuitive and
historically plausible, that Mercator actually used it to construct his
projection? Certainly not. We would be making the same mistake as the
researchers of the past, who didn’t care to compare the output of their
methods with the real map. Thus the next step will be to identify, among
the various types of rhumb tables, the one replicating as closely as
possible the error signature of the chart. [38] Two obvious candidates
come immediately forward: John Dee’s Canon Gubernauticus, and the
table of Pedro Nunes. Surprisingly none of them comes even close to the
error signature. [*] The one by Dee because is too accurate and [*] the
one by Nunes because the errors have the wrong sign. So other types of
tables of rhumbs had to be considered. Before proceeding, let me explain
very briefly how a table of rhumbs can be constructed.
7
8
[41] One additional reason, other than their simplicity and performance,
makes these two solutions ideal candidates, which is the fact that they
produce rhumb lines that never reach the pole, a feature shared by all
other tables using equal intervals of longitude. Being a mathematician and
a cosmographer who was well aware of the properties of the rhumb-line,
Mercator was certainly sensitive to this fact.
[42] Having concluded that the errors produced by a certain kind of table
of rhumbs closely match the error signature of Mercator’s chart, a
relevant question remains: would it be possible to achieve a similar result
using an alternative method, based on a different requirement? In theory,
yes. Instead of having forced the rhumb lines to be represented by
straight segments, by spacing the parallels appropriately, Mercator could
have used the concept enunciated 30 years ago by Pedro Nunes, that is,
that the proportion between the lengths of the parallels and meridians
arcs on the sphere should be conserved on the chart. [43] That would lead
to using this simple formula, identical to the one proposed by
Nordenskjold and many others. Such possibility is implicit in one of the
maps’ legends, where Mercator states that he has increased progressively
8
9
9
10
[47] Whichever method Mercator actually used to solve the problem, his
genius did not consist in developing a new projection from the void (ex
nihilo) but in assuming that a solution existed and in being able to use
effectively the knowledge that was available to him at the time. [48] In
other words: on the shoulders of giants.
10