Title: Journal of Caribbean archaeology
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00091746/00009
 Material Information
Title: Journal of Caribbean archaeology
Series Title: Journal of Caribbean archaeology
Alternate Title: JCA
Abbreviated Title: J. Caribb. archaelo.
Physical Description: Serial
Language: English
Publisher: Christopher Ohm Clement ;
Christopher Ohm Clement
William F. Keegan
Place of Publication: Gainesville FL
Publication Date: 2008
Frequency: annual
Subject: Archaeology -- Periodicals -- Caribbean Area   ( lcsh )
Antiquities -- Periodicals -- Caribbean Area   ( lcsh )
Genre: periodical   ( marcgt )
System Details: Mode of access: World Wide Web.
Dates or Sequential Designation: Vol. 1 (2000)-
General Note: Title from title screen (publisher's Web site, viewed Dec. 2, 2002).
General Note: Latest issue consulted: Vol. 5 (2004).
 Record Information
Bibliographic ID: UF00091746
Volume ID: VID00009
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 003345617
oclc - 41077527
lccn - sn 99003684
issn - 1524-4776


This item has the following downloads:

ejstrud2008 ( PDF )

Full Text

Journal of Caribbean Archaeology
Copyright 2008
ISSN 1524-4776


Bo Ejstrud
Institute of History and Civilization
Niels Bohrs Vej 9; DK-6700
Esbjerg; Denmark.

Maroons, or runaway slaves, form a conspicuous group in the narrative and study of slave
societies, being important exponents of the continuous rebellion against slavery. From a point
of view of cultural theory they also form an especially interesting group in that they both iso-
lated themselves from the slave society they escaped, but also constantly interacted with it. As
such, Maroon sites can inform us on the cultural conditions of people who are at the same time
both 'inside' and 'outside' established society, and their study will potentially provide insight of
theoretical significance far beyond local history and environment in the Caribbean. A main ar-
chaeological problem is to identify such sites, as they are inherently located in the most remote
and inaccessible places, difficult to access for the surveying archaeologist. In this paper a GIS-
based predictive model is developed for the island of St. Croix to identify the most likely places
for maroon settlements, thus making a more focused field work effort possible. The presented
model was made in preparation of a field trip in March 2007 as part of the Danish scientific Ga-
lathea 3 expedition, taking Danish scientists around the globe.

Any study on the period of Danish colonial-
ism on the U.S. Virgin Islands will necessarily
focus on plantations; their economy, organisa-
tion and populations. In a broader context, we
have to consider the position of the plantation
in systems of local and regional trade, the mer-
cantilist thinking and organisation behind the
entire colonial system, and the contacts with
both Europe and Africa. The triangular trade
of African slaves, American cash-crops and
European manufacturers and markets was what
made this world go around, and is in general
the mindset from which we have to view these
societies. But throughout the Americas, one
group of people broke out of the system, more
or less. Runaway slaves would take to the for-
ests and mountains to establish an independent
life away from plantation slavery. They were
known as maroons, a name derived from the

Spanish Cimarron, originally referring to cattle
gone wild.

In the dominant planter society these people
were a source of constant concern, because
they destabilized the slave system, not only by
running away, and thereby encouraging others
to follow their example, but also and not least
because the maroon societies were seen as cen-
tres for rebellion against slavery, and were in
effect in a state of constant warfare against the
slave owners. Maroons were fought vigor-
ously, and they often demonstrated a remark-
able ability to fight back. This happened eve-
rywhere where slaves were kept, and did so
from the very start of the colonial period. In-
deed one of the first Old World settlers in the
Americas was an African slave who escaped
his Spanish master in 1502 (Price 1979: 1), and
thus marronage is as old as colonization itself.

Journal of Caribbean Archaeology 8, 2008

Maroons and Landscapes

Maroons form a conspicuous group in both
the narrative and the study of slave societies.
Maroons were rebels and freedom fighters in a
cultural and economic system whose very
structure was based on inequality of human
beings. As such maroons are important sym-
bols of resistance and rebellion against slavery,
their significance in contemporary society ex-
tending far beyond a purely scientific interest
(cf Price (ed.) 1979; Agorsah (ed.) 1984;
Orser 2001).

But even from the most dispassionate scien-
tific view, maroons form significant cultural
groupings to investigate and understand, their
study potentially contributing insight of theo-
retical significance beyond their local history
and environment in the Caribbean. On one
hand these people did rebel against slave con-
ditions, seeking the most remote and inaccessi-
ble places in order to avoid contact with the
established planter society and hence recap-
ture. On the other hand they did not stand in
complete isolation from the societies towards
which they rebelled (Orser & Funari 2001).
Throughout the Americas these people were
dependent on regular interaction with sur-
rounding communities, not only in the form of
the much dreaded raids on isolated plantations
and settlements, but also as trade and exchange
with people both inside and outside the estab-
lished plantocracy (Vibaek 1966: 154f; Orser &
Funari 2001). There are even contemporary
accounts of maroons visiting the local markets
to procure necessities, obviously doing this at
great personal risk (Oldendorp 1777). It is this
exact mix of isolation and interaction which
makes Maroon societies and settlements so
interesting. These people were in many ways
both outside and inside the plantation economy
system, part of it and secluded from it, and in-
vestigating the types of cultural response to
this situation, and how such a system could
work, is culturally significant by any scale.

The presence of maroons on St. Croix is
widely documented in the historical sources, as
has been discussed and presented in a recent
paper in this journal (Norton & Espenshade
2007), but the memory of them also lives in
the landscape itself, where the place names
Maroon Ridge and Maroon Hole in the North-
western quarter can be found on any modern
map, marking the location of important sanctu-
aries for escaped slaves, and both supporting
an interaction with a local oral tradition on ma-
roon hideouts. According to Vibaek (1953:
153), a more organized existence for the ma-
roons even developed at Maroon Ridge, then
named Maroon Hill or "Maron Bierg"; at least
the Moravian historian Oldendorp (1777) de-
scribed how "a large number of Negroes" had
established there, the approaches of the area
protected not only by the dense vegetation, but
even by small pointed stakes of poisoned
wood. Oldendorp wrote his account on the
basis of having lived on the Danish Virgin is-
lands for 18 months in 1768-69, and must be
considered a well informed writer.

That an island as small as St. Croix could
sustain an organized and independent maroon
population seems surprising, and the condi-
tions here can hardly be compared to the large,
independent and well known maroon commu-
nities in for instance Jamaica (Campbell 1988),
Surinam (Price 1990) or Palmares in Brazil
(Anderson 1996; Orser & Funari 2001). As
shown by Hall (1985), this type of "interior"
grandmarronage became increasingly diffi-
cult on St. Croix during the second half of the
18th century, when the general intensification
of the plantation economy meant that even the
marginal areas to the North were settled by
planters, reclaiming large tracts of the forested
hills. Instead, Hall claimed that grand marron-
age had either to be maritime in nature, seek-
ing means to get off the island, or benefit from
the relative anonymity of the towns. Such
types of marronage would not leave many rec-
ognizable archaeological remains. This inter-

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

pretation leaves the question what became of
the maroons in the Northern Quarters after
1750. Even today tracts of forest are preserved
on St. Croix, where individuals and small
groups could hide.

The early history of the maroons in the hills
of St. Croix forms an intriguing story, not in
spite of the relative small size of the island, but
exactly because of it. While it may be imme-
diately understandable, that relatively large
communities of runaway slaves could escape
and establish into the forested hills and moun-
tains of Jamaica or Brazil, the confined space
of St. Croix makes the question of their spatial
and cultural organization even more pressing.

The history of maroons is mostly told
through the written accounts of their oppres-
sors, or through the traditions of oral history
and ethnography. Archaeology plays an im-
portant role in voicing out the history of the
Maroons, as the material culture of these peo-
ple is the only contemporary reflection of cen-
tral aspects of their society, made by them-
selves. Maroon settlements can be expected to
answer questions of subsistence, structure and
external contact, and are thus fundamental in
understanding Maroon societies. An obvious
and very practical problem to archaeology is to
locate Maroon settlements. This reflects the
double problem that not only were these settle-
ments inherently placed in the most remote and
difficult areas, but also that, at least in the case
of St. Croix, the settlements would reflect rela-
tively small bands of people, possibly moving
between places, and thus leaving few traces at
each individual site, and such settlements are
always difficult to find in the field. But al-
though difficult, other periods and cultural set-
tings have proven it possible to find even such
elusive settlements. One telling example -
among many similar in archaeology- is Late
Palaeolithic settlements in Denmark. Having
been known in Northern Germany since the
1930s, only stray finds of the so-called Ham-

burgean Culture were known in Denmark, al-
though settlements had been found almost
right up to the Danish-German border. The
lack of settlements in Denmark was for many
years explained in various ways as being a re-
sult of solifluction or of the proximity of the
ice sheet during the last glaciation. But in
1981 the first settlement was found (Holm &
Rieck 1983). As is so often the case in archae-
ology, when the first discovery was made, sev-
eral followed. The settlements turned out to be
small camp sites, reflecting a few days of stay
by family-sized groups. The finds are simply
small scatters of flint and occasionally a tent-
ring of large stone slabs. Such sites are inher-
ently difficult to locate, and seen as archaeo-
logical entities, the maroon sites in the Carib-
bean must be rather parallel to the equally
small and elusive Danish Palaeolithic sites, the
latter demonstrating that with a concerted ef-
fort -and some luck- it is possible to locate
them in the field.

Maroon settlements have yet to be found on
St. Croix. Therefore the first important step in
an archaeological study of the maroons on St.
Croix is to locate where they lived. This can
only be done in the field, but due to the practi-
cal problems of surveying for such settlements,
it would be useful to guide surveys by produc-
ing an informed guess about where in the land-
scape surveying might be most profitable. The
purpose of this paper is therefore to develop
and present a preliminary spatial model for
guiding future field work.

Methods and Data

Informed and uninformed guessing has
always been a part of archaeological field
work. But formal methods have developed
since the 1970s in the form of predictive mod-
elling. A recent and comprehensive overview
of predictive modelling is given by van Leusen
et al. (2005), but in brief a predictive model is
a computer generated map showing areas of

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

"high" and "low" probabilities of archaeologi-
cal sites being present. They can be based on
statistical analysis of the spatial properties of
already known archaeological sites (inductive
or data-driven models), or on hypothetical as-
sumptions on the "best" places to live
(deductive or theory-driven models). Some
methods, such as the one used here, allow for
the combination of empirical analysis with hy-
pothetical modelling, but due to the lack of
known Maroon settlements on St. Croix, the
model will have to be deductive and hypotheti-

A vast array of methods has been used in
producing such models, and the map outcome
of "highs" and "lows" given in an equally di-
verse number of scales. In this case a tech-
nique called Dempster-Shafer theory is used.
Dempster-Shafer based modelling has increas-
ingly been used for this purpose by archaeolo-
gists around the world, replacing older -mostly
regression based- methods (Ejstrud 2003;
2005b; Canning 2006; Hernandez 2006; Ducke
2007). The method is an extension of Bayes-
ian probability theory, and is based on statisti-
cal rules of combination for handling an array
of individual parameters, which can support
one or more hypotheses. There are several dif-
ferent versions of the mathematical 'belief
function' by which the actual calculation is
done (Smets 1994), but the implementation of
Dempster-Shafer theory in the GIS system
IDRISI (Eastman 2003) seems to be based on
what Smets labelled the Transferable Belief
Model (1994). Although the actual -and fairly
complicated- mathematics behind the IDRSI
module is something of a black-box, this has
the advantage that the interface is very easy to
use in an analysis, even for the non-statistically
minded archaeologist. What is important to
know are the more general workings, assump-
tions and conditions of the method, rather than
the actual equations.

Dempster-Shafer theory provides a flexible
and theoretically valid framework for model-
ling past human behaviour. Unlike many other
methods in probability theory it allows for the
assumption that ignorance is built into of our
knowledge of a field. To archaeology this
seems an important assumption to make, as
some degree of ignorance about a specific field
is inherent in most archaeological work. The
method is based on determining a decision
space of possible outcomes. In this case, that
would be [presence] or [absence] of a maroon
settlement. The degree to which we cannot
decide between the two is our ignorance, and
can be described mathematically through infer-
ence using the Dempster-Shafer statistical
rules of combination on the variables. In dis-
tinguishing between variables that support
presence, variables that support absence, and
recognizing the fact that our body of knowl-
edge is not complete, Dempster-Shafer model-
ling is a strong tool for archaeological predic-
tive modelling (Ejstrud 2003; 2005b).

The method works within the [0; 1] range of
probabilities, although the result should not be
interpreted as a formal probability of finding a
site. At best it can be seen as reflecting the
suitability of a certain location for settlement,
given a chosen set of variables and on a gradi-
ent from 0 to 1: whether people actually set-
tled there is a different story. For this reason
the resulting models are probably better re-
garded as indicative rather than predictive: in
a strict sense they do not predict sites, but do
give a general indication of the most likely
spots to find them.

Two main datasets are used in the analysis.
The first is a Digital Elevation Model (DEM)
derived from contours on the standard topog-
raphical maps in 1/24,000, which has been
digitized by USGS, and is freely available for
download for instance through http://
www.geocomm.com. The contours are inter-
polated into a raster model, and the unit con-

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

averted to meter. This gives a highly detailed
model, and probably the best available for the
area (Figure 1). The DEM is a basis for devel-
oping many other important variables in the
landscape, as shall be shown below.

The second dataset is the Cronenberg/
Jagersberg map from 1750 (Figure 2). This
map, which was drawn in c. 1/30,000, is an
excellent and highly detailed source for the

land use and land cover of the island, based on
an original survey. The map was largely pre-
pared by the competent Lieutenant Johann
Cronenberg, who started his work in 1747, and
had completed main parts of the work by
March 1750 when he was arrested and ex-
pelled from St. Croix due to an amorous affair
with a married woman. A Lieutenant Johann
Jagersberg, who had previously been dis-
missed from his job as surveyor on St. Croix


0 5 10

Figure 1. The DEM used in this analysis. Overlain are the island's two towns, the administrative division in
"quarters", and Maroon Ridge.

Figure 2. The Cronenberg/Jagersberg map. Here scanned in two sections.
Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

due to laziness, was called back to complete
the survey and the map, which was shipped
back to Copenhagen in 1750.

The result was marvellous. The map shows
the settlement in much detail, with every field
surveyed and their crop type given. Plantation
main buildings, locally known as
"greathouses", are marked precisely on the
map, while the slave villages are shown in sig-
nature format. In all 201 houses are marked on
the map. In some instances, especially on the
sugar plantations, more than one building is
marked on a single plantation, so the number
of buildings reflects somewhat less than 200
plantations. The completeness and quality of
this map is outstanding for it's time, and gives
us a very complete picture of the colonial set-
tlement (Hopkins 1989).

The original map has been scanned, geo-
referenced and digitized for use in a Geo-
graphical Information System (GIS), and has
proven to be very precise, except in the north-
ern quarters, where the maroons would have
been located. But in this case the map is used
to locate the plantations and infrastructure of
the European settlers, so the imprecision in the
northern quarters is not be a problem for the
analysis. Using this map, the model will show
potential areas of Maroon settlements in the
period around the mid 18th century.

The data are analysed in a 50 m resolution
giving the final map a nominal precision of

The Geography of Grand Marronage

As most any writer on the subject of ma-
roons has noted, the success of grand marron-
age is dependent on topography and environ-
ment. But unlike many other situations where
one might use predictive modelling, the model
for maroon settlements will describe a balance
between two partly opposite considerations:

while finding suitable places to live, the ma-
roons also had to keep out of sight of the exist-
ing settlement of towns and plantations.

Visibility from plantations

Modelling areas out of sight from the plan-
tation is a relatively straightforward task using
cumulative viewshed analysis (van Leusen
1999). Viewshed analyses show which parts
of a landscape can be seen from a given loca-
tion. The analysis uses the DEM (Figure 1) to
model visible and invisible areas from that lo-
cation (Figure 3). Visible areas are normally
coded with the value 1, invisible with a 0. In a
cumulative viewshed analysis one simply adds
up the result from several analyses.

The basis is the plantation main buildings
on the 1750 map, and using the DEM it is pos-
sible to calculate how many houses can be
seen from any point on the map or how many
houses can see each part of the map (Figure 4).
The variable is not considered to reflect di-
rectly where runaway slaves could hide. Al-
though the calculation of a viewshed is rela-
tively straightforward, the reality behind the
model is slightly more complicated. First of
all, using the DEM, vegetation is not taken into
consideration. Secondly the algorithms behind
viewsheds do not take into account the increas-
ing difficulties of seeing an object or a person
at increasing distances. Rather than modelling
direct visibility of a runaway slave, this vari-
able is thought to reflect a cognitive variable.
Areas of the island that are visible from the
plantations would also be the ones that a slave
planning to escape would try to get past. The
visible land is that controlled by the planters,
and intuitively it seems a valid goal for an es-
capee to take to the areas beyond that.


Another way to describe a good hiding
place is to maximize the distance to the planta-

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

Viewshed analysis

Visible -
Not visible -,,, i

&~ Jr

Figure 3. Left: The principle of viewshed analysis in a GIS, where visibility is determined from a given location
across the DEM. Right: As an example, a viewshed (black areas) is calculated from the a location on waterfront at
Christiansted, St. Croix, and overlain the DEM. For this example viewing distance is set to 3500 m, and town
buildings are not taken into account.



Figure 4. Viewshed analysis showing the number of main houses in 1750 that could be seen from any part of the
island. The unit is number of houses visible.

tions, again represented by the main buildings
in 1750. But Euclidian distance is not neces-
sarily describing the actual effort of going
across the landscape, especially in a hilly area,
so in stead a Cost Surface Analysis (CSA) was
used. CSA is a type of analysis which can de-
scribe the cumulative difficulties of moving
through the landscape by using the DEM (van
Leusen 1998; Ejstrud 2005a). The analysis
works from two maps. The first map gives the

'cost' of moving through any given cell, using
time or energy as the unit. The second holds
one or more starting points from which the
cost of movement is calculated cumulatively.
Cost is typically a function of the topography
of the landscape, as a flat surface would be
faster to move through than a sloping one.
Walking speed through each cell is here calcu-
lated using the formula derived by Garenflo &
Gale (1990):

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

V = 6 e-3'5 TAN(s)+0.05 |

Where V is the velocity in km/h, s is the slope
of the landscape in degrees, and e is the base
for natural logarithms. Having calculated the
speed by which one can move through every
50x50 m cell in the GIS system, it is easy to
calculate the time it would take to pass through
the cells, and this is the 'cost' used in the

The model could be refined by combining
landform with information of vegetation. A
dense forest is more difficult to pass than an
open plain. But even without using this refine-
ment, the model seems to give a reasonable
description: it is noteworthy that Maroon
Ridge is part of a narrow area which is furthest
away from any plantations, measured in walk-
ing distance (Figure 5). This is a good indica-
tion that maroons actually did seek to maxi-
mize distance away from their former owners,
and by inference that formal modelling of this
kind can actually give valid results.

Settlement density

But the visibility of and distance to individ-
ual houses cannot in itself describe the state of
being "away" from the planter society. One
may be hiding in a local hollow between two

houses, and still be in the middle of the settled
area. To supplement the variable above, the
number of main houses on the 1750 map was
added within a radius of 2000 m across the en-
tire island. This map gives the density of the
settlement, and thus the general areas to avoid
(Figure 6).


Settled areas are not only defined by
houses. The roads on the Cronenberg/
Jaegersberg map were digitized and a simple
distance function was fitted, to mark the imme-
diate corridors (500 m) around the roads.
There are only few roads on the 1750 map, and
probably only the main road network has been
mapped. This variable may therefore be un-
derestimated in this model. It is assumed that
proximity to roads will discourage maroon set-


The maroons lived in the hills. Or at least
the maroons at Maroon Ridge did, and there is
historical evidence of maroons hiding every-
where possible on the island (George Tyson,
personal communication March 2007). As this
model will focus on the communities in the
mountains mentioned by Oldendorp (1777),
elevation has been considered, higher values

73 94
8B 26

Figure 5. Cost surface analysis using the main houses as basis and moving over the DEM. The unit is minutes.
Journal of Caribbean Archaeology 8, 2008 8


Maroons and Landscapes

Figure 6. Density of main buildings on the 1750 map. The unit is number of houses.

being better for settlement. The relation was
expressed as an S-shaped increase between 75
and 200 m, and a similar decrease between 300
and 350 m (Figure 7). The highest point on the
Island is Mount Eagle at 355 m. The values
are decreased here, as it is assumed that no one
would settle on the highest peaks of the moun-


Living in the hills, the Maroons also settled
in the most sloping areas. But to establish a
settlement proper, there has to be at least some
level ground to settle on. This is also noted in
the historical sources. Quoting a source from
1803 Jamaica, Price describes settlement

Modelled relation between elevation and Maroon sites
( 0.50
a 0.30
0.00 ..
0 50 100 150 200 250 300 350 400
Elevation (m)

Figure 7. The modelled relation between elevation and maroon sites.
To S-shaped functions are fitted; One increasing between 75 and 200
m and another decreasing between 300 and 350 m elevation. Areas
between 200 and 300 m are considered equally good for hiding.

Journal of Caribbean Archaeology 8, 2008

placed "on a ground which compared to those
precipices, might be called a plain" (Price
1979a: 6), noting similar conditions for Ma-
roon settlements in Surinam.

Calculating slope from a DEM is a standard
component of any raster GIS system, and this
variable has been used in almost every ar-
chaeological predictive modelling effort. A
seemingly universal trend is that settlements
can be found on slopes up to 6-7, while they
are rare in more sloping areas. In the current
area and situation, settlements may be ex-
pected at slightly higher slopes. To model this,
an S-shaped function is used to convert the
map of slopes with probabilities decreasing in
the range between 5 and 100 slope.

This variable can be mis-
leading in karst areas, where caves
can be used for settlement. In such
cases the relation between settle-
ment and slope may be reversed,
letting the most sloping areas be the
ones with the most cave entrances
at the surface. In all likelihood
such a use would render the relation
between surface slope and human
settlement very, if not too, compli-
cated to model. Coastal caves were
used for Maroon settlement on St.
Thomas (Hall 1986), and the place


7 49
23 98

Maroons and Landscapes

name Maroon Hole on St. Croix may indicate a
similar use here.

Gill et al. (2002) mentions some tendency
for karsting in the northern area, but the phe-
nomenon does not seem to extend beyond the
immediate coastline. The historical record in-
dicates that these caves, on the slopes just
North of Maroon Ridge, were inhabited by ma-
roons, and this means that for this specific area
the predictive model does not show the correct
locations of maroon settlement. On the other
hand, that would mean that caves would be a
focal point for surveys, being geographically
as precise as any predictive model.


Slope in itself does not describe much on
the properties of the landscape. Again using
the DEM, the landscape can be classified into
eleven different features: peak, ridge, saddle,
flat, ravine, pit, convex hillside, saddle hillside,
slope hillside, concave hillside, and inflection
hillside. Any pixels not assigned to these
classes are "unclassified". Estimating that
relatively flat areas are preferred, the classifi-
cation has been deductively assigned values as
shown in Table 1.

Using this classification the model will
point out the ridges, saddles and flat areas. It
must be added that the use of this variable is
entirely speculative at this stage. For Danish
prehistory, where the variable has been tested
against actual sites, it cannot explain settle-
ment location Class Value
(Ejstrud 2003). Peak 0.4
But for hills such
as the ones on St. Ridge, saddle, flat 0.6
Croix, a classifi- Ravine, pit 0.2
cation of landform Hillsides (all) 0.3
seems more Unclassified 0.5
meaningful than
meaningful than Table 1. Values assigned to the
in the flat Danish topographic classification.

Water is an indispensable part of human
life, and must be considered essential for site
location in any part of the world. A drainage
network was generated from the DEM, as the
modem streams shown on the topographical
maps are visibly affected by modem activities.
By using the DEM the original network of
streams can be reconstructed. The next step
was to analyze distance from the streams.
Again cost surface analysis was used rather
than Euclidian distance. This analysis revealed
that only very few areas on St. Croix are actu-
ally more than 15 minute walk from a stream.
Given that the choice of settlement locations
for maroons must have been a balanced trade
off between convenience and security, this
cannot be considered a long distance. Impor-
tant as this variable generally is, it carries little
significance on this island. An S-shaped func-
tion was fitted for areas between 15 and 30
minutes walking distance from a stream, the
final map showing places that were avoided.

The Model

The resulting model is made using the
Dempster-Shafer based module build into
IDRISI (Eastman 2003), using the variables
listed above. The resulting model expresses
the minimum conditional probability that any
given area is suitable for maroon settlement.
What is seen on the model is that a larger area
than Maroon Ridge itself is covered by the
model. Large tracts of land could have been
used for hiding places, at least around 1750.
Nonetheless the model does point to certain
locations where the suitability is particularly
high for maroon settlements, and given the
variables chosen here (Figure 8).

For Maroon Ridge itself, the areas with the
highest probabilities are distributed in three
locations. In the westernmost part at an area
around 6451'56.2"W / 17045'58.8"N, at the

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

centre of the ridge at 64052'05.8"W / 170
45'52.5"N, and to the east in a area around 640
51'33.2 W / 17045'46. IN, all positions given
in WGS84. Altogether these areas cover 2-21/2
hectares (5-6 acres).

For ease of use, the model has been con-
verted from a cell-based to a vector based GIS
format, so that it can be layered on top of other
maps. This map could be brought into the
field on any portable computer to guide a con-
certed effort to locate actual sites (Figure 9).

0 34


Figure 8. The predictive model, showing the relative suitability of the landscape for Maroon settlements.

, -- .dd3- '. r -KE' -j;^ ^"- J '- *" :'--. ," '-
. . .^ ', ,*..i . . .
r- "- '_ .\ V OA
",.As $,,.. .S A, i
-- -'-- ^r -.'. *-" -- "-_- .;
A : 'r. ~': .% -

^ . .. ^^ ^ ^ <^ .',. ,. ........ .^ ; -. "
'- "" ",!^ ?-'" '%7. ji,]":-" .___ -' -. , ,, *_ ;. %

*, r.-.'- "' ,,"-, '" --ti-'- "a". ~" .

- '.. 9 -W A . :-. ... ,

Figure 9. The model converted to vector format and drawn on top of a modem topographical map. From USGS
(1958, with corrections 1978).

Journal of Caribbean Archaeology 8, 2008



Maroons and Landscapes

To validate the model itself, it is important also high "relative suitability".

to look in areas where lower degrees of suit-
ability are indicated. Otherwise these models
could potentially end up as self-fulfilling
prophesies. Still, to locate the first sites it
would not bias the data to focus on the most
suitable areas.


The model should present the potential of
using predictive modelling for directing the
archaeological field work necessary to under-
stand the maroons of St. Croix. A map show-
ing only areas with the highest values demon-
strates this (Figure 10). Selected here are areas
with a model value of more than 0.85. The
value of 0.85 is somewhat arbitrarily chosen,
but as an experience from earlier studies
(Ejstrud 2003; 2005b) would reflect a very

When comparing to the topographical map,
not all of these areas seems likely to actually
have been settled. This is an expected out-
come of a purely statistical modelling method:
archaeology is still a discipline involving hu-
man assessment. But there are areas which do
look promising, both on Maroon Ridge itself,
but even in other parts of the area.

The next step in developing a more thor-
ough study would be to go through existing
field reports on other Maroons sites. Unfortu-
nately such reports would mostly be unpub-
lished or published as grey literature in local
reports, which are very difficult to access. In-
ternationally available literature is mostly in
the form of outlines of field work, or interpre-
tations of field work, but rarely presents the

..... ...... .. . . . .
-, ,
.,-.::-,-- :-: ,_,, -, 1, ,. _iv-" *: l."1 -", '-. .., ,
f^ ^ '-r ^' "-* -* ". ''. ^ ^- ^ --"- (' '-_- .. .. r,-, J) r" ^-- 1." '

Figure 10-. ThBe 1'' h "ghe-st vauesfthemode "( "e, d 'o t of a m t o.pgahc map.
-... .. 3 ... ,.

N 1.-' :- .. ....-z 1 .
Fig"u:r .--,e_ ,; T,
"., -4 -. C f --
Figure 10. The highest values of the model (red areas) drawn on top of a modem topographical map.

Journal of Caribbean Archaeology 8, 2008

Maroons and Landscapes

field work itself, so it is difficult to get an im-
pression of what to expect, and hence what to
look for in the field.

Further work would include coupling the
presence and disappearance of maroons on St.
Croix with the development of land use during
the 18th and 19th centuries, trying to understand
the conditions from a dynamic perspective,
rather than a single static picture. Hall's hy-
pothesis that maroons disappeared from the
area during the second half of the 18th century
is worth investigating further.

During fieldwork done in March 2007, the
focus was on the plantation remains, and only
limited time was available to go into the field
of Maroon sites. The area around Maroon
Ridge was visited to get an overall impression
of the landscape and vegetation, but no sys-
tematic survey was possible. The most strik-
ing result was the importance of the very dense
vegetation, which should have bearings not
only on the modelling effort, but will certainly
effect field work to an extent that could not
have been foreseen from a desk-based study
done in Denmark. Nonetheless, the strong re-
lation between marronage and geography
makes GIS a strong tool in getting closer to the
structure, organization and lives of these peo-
ple of the hills, and hopefully this paper has
demonstrated some of the potential.

Acknowledgements. The present work was carried out
as part of the Galathea 3 expedition under the auspices
of the Danish Expedition Foundation. This is Galathea
3 contribution no. P7. Thanks are due to the many re-
searchers we met at St. Croix. Mr. George Tyson pro-
vided insightful and invaluable guidance into the soci-
ety, history and landscape of St. Croix. Dr. Olasee Da-
vies, Mr. Myron Jackson, Mr. Joel Tutein and Mr. Paul
Chackroff have kindly commented on an earlier version
of this paper, while Dr. Brian Daley generously pro-
vided GIS assistance while working on the island. The
paper improved through the comments made by two
anonymous referees. All errors remain my own.

References Cited

Agorsah, E.K. (ed.)
1994 Maroon Heritage. Archaeological Ethno-
graphical and Historical Perspectives. Canoe
Press, Kingston.
Anderson, R.N.
1996 The Quilombo of Palmares: A New Overwiew
of a Maroon State in Seventeenth-Century Brazil.
Journal ofLatin American Studies 28(3):545-566.
Campbell, M.C.
1988 The maroons of Jamaica 1655-1796: A history
ofresistance, collaboration, and betrayal. Bergin
& Garvey, Granby, MA.
Canning, S.
2006 'BELIEF' in the Past: Dempster-Shafer theory,
GIS, and archaeological predictive modeling.
Australian Archaeology 60:6-15.
Ducke, B.
2007 Ein archaologisches Pradiktionsmodell ffr das
Bundesland Brandenburg: Methoden,
datengrundlagen und Interpretation. In
Archaeoprognose Brandenburg II, edited by J.
Kunow, J. Miller & F. Schopper, pp. 235-257.
Forschungen zur Archaologie im Land Brandenburg
10, Wunsdorf.
Eastman, J. R.
2003 Idrisi Kilimanjaro. Guide to GIS and Image
Processing. Clark University, Worcester.
Ejstrud, B.
2003 Indicative models in Archaeology: Testing the
methods. In Landschaftsarchdologie und
Geographische Informationssysteme: Prog-
nosekarten, Besiedlungsdynamik und prdhistorische
Raumordnungen./The Archaeology ofLandscapes
and Geographic Information Systems: Predictive
Maps, Settlement Dynamics and Space and Terri-
tory in Prehistory. Archdoprognose Brandenburg I,
edited by J. Kunow & J. Miller, pp.119-134.
Forschungen zur Archaologie im Land Brandenburg
8, Wuinsdorf.
2005a Cost Surface Analysis and ancient roads: A
comparison. In Temps et spaces de l'homme en
soci&td. Analyses et modules spatiaux en archeolo-
gie, edited by J.-F. Berger, F. Bertoncello, F.
Breamer, G. Davtian & M. Gazenbeck, pp. 135-
140. APDCA, Antibes.
2005b Taphonomic models. Using Dempster-Shafer
theory to assess the quality of archaeological data
and indicative models. In Predictive Modelling for
Archaeological Heritage Management: A research
agenda, edited by P.M. van Leusen & H. Kamer-
mans, pp. 189-198. Nederlandse Archeologische
Rapporten 29. Amersfoort.

Journal of Caribbean Archaeology 8, 2008


Maroons and Landscapes

Gorenflo, L.J. and N. Gale
1990 Mapping Regional Settlement in Information
Space. Journal ofl iimh,. ,. "t.... 1Archaeology
Gill, I.P, D.K. Hubbard, P.P. McLaughlin & H.H.
2002 Distribution and types of porosity within the
Kingshill aquifier. In Geology and Hydrogeology
of the Caribbean Islands Aquifer System of the
Commonwealth of Puerto Rico and the U.S. Virgin
Islands, edited by R.A. Renken et al, pp. 123-125.
U.S. Geological Survey professional paper, 1419.
Hall, N.A.T.
1986 Maritime Maroons: "Grand Marronage" from
the Danish West Indies. The William and Mary
Quarterly, 3rd ser., 42(4):476-498.
Hernandez, A.A.
2006 Strategic Location and Territorial Integrity: The
Role of Subsidiary Sites in the Classic Maya King-
doms of the Upper Usumacinta Region. Internet
Archaeology 19.
Holm, J. & F. Rieck
1983 Jels I The First danish Site of the Hamburgian
Culture. A preliminary Report. Journal ofDanish
Archaeology 2:7-11.
Hopkins, D
1989 An extraordinary eighteenth-century map of the
Danish sugar-plantation island St. Croix. Imago
Mundi 41:44-58.
van Leusen, M.
1999 Viewshed and cost surface analysis using GIS.
In New Techniques for Old Times. CAA 98. edited
by J.A. Barcel6 et al., B.A.R. Intl. Ser. 75:215-
van Leusen, M., J. Deeben, D. Hallewas, H. Kamer-
mans, Ph. Verhagen & P. Zoetbrood
2005 A Baseline for Predictive Modelling in the
Netherlands. In Predictive Modelling for Archaeo-
logical Heritage Management: A research agenda,
edited by H. van Leusen & H. Kamermans, pp. 25-
92. Nederlandse Arkeologische Rapporten 29, Am-
Norton, H.K. & C. Espenshade
2007 The challenge of locating Maroon refuge sites at
Maroon Ridge, St. Croix. Journal of Caribbean
Archaeology 7:1-17.
Oldendorp, C.G.A.
1777 Historie der caribischen Inseln Sanct Thomas,
Sanct Crux und Sanct Jan, insbesondere der
dasigen Neger und der Mission der evangelischen
Brfider unter denselben. Barby. (Reprint 2000,
Verlag fur Wissenschaft und Bildung, Berlin)
Orser, C.E., Jr.
1998 The Archaeology of the African Diaspora.
Annual review of i i, d. .. 27:63-82.

Orser, C.E., Jr. & P.P.A. Funari
2001 Archaeology and slave resistance and rebellion.
World Archaeology 33(1):61-72.
Price, R
1979a Introduction. In: Maroon Societies: Rebel
Slave Communities in the Americas. 2nd edition,
edited by R. Price, pp. 1-30. John Hopkins Univ.
Press, Baltimore and London.
1990 Alabi's world. The John Hopkins University
Press, Baltimore.
Price, R. (ed.)
1979 Maroon Societies: Rebel Slave Communities in
the Americas, 2nd edition. John Hopkins Univ.
Press, Baltimore.
Smets, Ph.
1994What is Dempster-Shafer's model? InAdvances
in Dempster-Shafer Theory of Evidence, edited by
R.R. Yager, J. Kacprzyk & M. Fedrizzi, pp. 5-
34. Wiley, New York
Vibak, J.
1966 Dansk Vestindien 1755-1848. Vestindiens stor-
hedstid. Vore Gamle Tropekolonier, 2. udg., bd.
2. Fremad, Kobenhavn.

Journal of Caribbean Archaeology 8, 2008


University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs