image/svg+xml
79
VIII/1/2017
INTERDISCIPLINARIA ARCHAEOLOGICA
NATURAL SCIENCES IN ARCHAEOLOGY
homepage: http://www.iansa.eu
Thematic Review
Interpretive and Analytical Approaches to Aerial Survey in Archaeology
Ladislav Šmejda
a*
a
Czech University of Life Sciences Prague, Department of Ecology,
Kamýcká 129, 16521 Praha 6 – Suchdol, Czech Republic
1. Introduction
The concept of an analytical approach to archaeological
surface collection has been associated with processual
archaeology and its emphasis on sampling and the
quantitative aspects of the archaeological record (Redman
1987; Schifer
et al.
1978). These research strategies have
been systematically rethought, enriched with a number of
new observations and improvements and, most importantly,
brought into practice in central European archaeology by
M. Kuna (
e.g.
1994; 1998; 2000; 2004). This has occurred
in such a convincing manner that within one or two decades
they have become an integral part of the archaeological
methodology. Given the statistical evaluation of data and the
study of their spatial properties in Geographical Information
Systems (GIS), the discipline has gained a highly efective
tool which has signifcantly advanced our understanding
of the past (Gojda 2004a; Neustupný 1998; Neustupný,
Venclová 2000; Smrž
et al.
2011; Šmejda 2003).
The core of this article, which entirely subscribes to the
inspiration mentioned above, considers the idea that aerial
survey in archaeology can be understood in terms of both
an analytical and synthesizing (interpretive) methodology,
similar to that of surface survey by feldwalking (Šmejda
2009). In an analogous way to the development of the
techniques of surface collection of artefacts, in the feld of
aerial survey, we can also observe a movement from the
efort to identify individual spots of interest in the landscape
to a systematic study of entire landscape transects. In this
more recent approach, space is understood as a continuum
that is sampled in a certain controlled routine, the results
and interpretations being gained later, independently of
the process of data collection. The former approach, the
discovery of new “sites” through data collection, is a
synthesizing method because the interpretation of empirical
observations is conducted immediately during feld-walking,
while the latter is an analytical approach because only
the analysis (analytical decomposition) of the area being
investigated is conducted in the feld.
In order to discuss these strategies in the context of aerial
reconnaissance, it is frst necessary to compare the properties
of the two elementary categories of aerial photographs,
i.e.
so-called “oblique” and “vertical” photographs (Doneus
2000). They have traditionally been perceived as standing
Volume VIII ● Issue 1/2017 ● Pages 79–92
*Corresponding author. E-mail: smejda@fzp.czu.cz
ARTICLE INFO
Article history:
Received: 13
th
October 2016
Accepted: 19
th
June 2017
DOI: http://dx.doi.org/ 10.24916/iansa.2017.1.6
Key words:
aerial prospecting
sampling strategy
verticals and obliques
methodology
ABSTRACT
This article discusses two contrasting approaches to archaeological survey using aerial reconnaissance.
A more traditional strategy is to look for interesting spots in the landscape with a highly concentrated
archaeological record. These are usually called “sites”. This concept is still used in everyday practice,
despite its long-standing problematic character. The opposing approach divides the studied region into
analytical units, which are sampled for evidence in a standardized manner and only then is the collected
information subsequently interpreted. Varying densities of recorded facts across space are now studied
rather than the binary categories of “on-site” and “of-site”. In Czech archaeology, this operational
diference has often been classifed as the “synthesizing” vs. “analytical” research methodology. This
debate has been ongoing for quite some time in the context of feld-walking and surface collection of
archaeological fnds. This text examines an analogous problem in the feld of aerial survey, where it
seems to be closely connected to another long-standing methodological and terminological discussion:
the comparative usefulness of “oblique vs. vertical” aerial photography.
image/svg+xml
IANSA 2017 ● VIII/1 ● 79–92
Ladislav Šmejda: Interpretive and Analytical Approaches to Aerial Survey in Archaeology
80
in mutual opposition to each other as regards their technical
parameters and practical utility. The aim of this paper is to
evaluate oblique and vertical aerial photographs in terms of
the two above-mentioned survey strategies: synthesizing and
analytical approach.
2. Oblique and vertical aerial photographs
As their names suggest, the main criteria for distinguishing
between vertical and oblique photographs is the orientation
of the camera at the moment when the photograph is taken.
Verticals are produced when the camera’s optical axis
is oriented downwards, perpendicular to the horizontal
plane. For practical reasons, a small deviation (usually less
than 3 degrees) of the optical axis from the plumb line is
generally tolerated. Obliques are captured by cameras that
are tilted signifcantly from the vertical. We speak about
“low obliques” when the optical axis is tilted no more
than 30 degrees from the vertical, and “high obliques” that
typically point around 60 degrees away from the vertical.
In vertical photographs, the nadir (
i.e.
point on the ground
directly below the camera at the time of exposure) is located
approximately in their geometrical centre (principal point);
while in the case of high obliques the position of the nadir
is typically positioned outside the photo frame (Figure 1).
Another signifcant diference is that verticals are often taken
in so-called stereo pairs (subsequent frames have signifcant
overlap of their ground coverage), enabling a “three-
dimensional” perception during visual analysis and ofering
advanced possibilities of precision mapping (Risbøl
et al.
2015). Obliques are very rarely obtained in this way, their
analytical potential thus being, technically speaking, more
limited.
Verticals versus obliques can be compared based on
practical considerations of data collection and processing, but
not necessarily the most important one for a full appreciation
of the actual potential of aerial photographs. No image taken
by an optical sensor with a central projection of rays (all
conventional cameras) captures the surface of the Earth truly
vertically (orthogonally), thus making what we understand
as a plan or map. This radial distortion of an image due to the
vertical ruggedness of the terrain is explained in Figure 2.
There is no simple transformation relationship between the
central projection of any photo and the orthogonal map or
plan. Correction of this type of distortion can be computed
from a series of overlapping images, in which the apparent
dislocation of points on the individual photographs can be
explained by diferences in their elevation. If stereo pairs of
photographs are not available, a digital elevation model of
the terrain can help to re-project a photo onto a horizontal
plane (Hampton 1978).
Adjustments of the horizontal positions of captured data
must therefore always be computed for both verticals and
obliques. For this type of processing vertical photographs are
much less problematic, because the perspective distortion as
well as displacement due to elevation variances generally
increase with the distance from the nadir. In vertical photos,
these positional shifts as well as the distortions of shapes
and lengths are smaller and more regularly distributed across
the photo frame than is the case in high-angle obliques.
However, it is clear that all photographs require a geometric
correction before they are used for planimetry (measurements
of distances, angles and areas). Therefore it might seem
more suitable to link the diference between “oblique” and
“vertical” imaging more generally with the strategy of data
collecting (synthesising/interpretive vs. analytical), rather
than with the type and orientation of the camera.
3. Scale of photographs
Archaeologists, and especially those insufciently acquainted
with vertical aerial photos, sometimes highlight the issue
Figure 1.
Footprints of oblique (A) and vertical (B) aerial
photographs covering an archaeological site. The crosses mark the
nadirs of individual photographs,
i.e.
the points directly below the
camera positions. Note that they are located outside the covered area
in the case of obliques, while they coincide with the centres of vertical
photos (after Hampton1978, Figure 9).
image/svg+xml
IANSA 2017 ● VIII/1 ● 79–92
Ladislav Šmejda: Interpretive and Analytical Approaches to Aerial Survey in Archaeology
81
that the nominal scale of available vertical images is smaller
than that required for fne-grained studies of archaeological
heritage and that no details are visible. In many cases this is
true of imagery taken for purposes other than archaeology,
but in principle there should be no dramatic diferences in
this respect between vertical and oblique photographs, and
this can be easily exemplifed. To better understand this, we
can consider imaging on flm to illustrate the principle, even
though flm has largely been replaced by digital technology
nowadays (Verhoeven 2007). We know that the nominal
scale of an image on a flm depends on the ratio between
fight height (altitude above the terrain) and the focal
length of the camera. When photographing the landscape
using a common hand-held camera with a standard lens of
focal length f=50 mm from an altitude of 500 m, we get
an image on the negative at a scale of 1:10,000 (500/0.05).
For hand-held oblique photography, the use of a lens with a
signifcantly longer focal length (a so-called telephoto lens)
is mostly impractical in aerial prospection because such an
arrangement can capture only small views and the image is
too enlarged to be held steadily in the viewfnder because of
constant vibrations and turbulence afecting the aircraft and
its crew during the fight. In addition, the necessity to use
a fast shutter speed in order to avoid blurred images calls
for a wide aperture, which may in some cases decrease the
sharpness of certain parts of the picture. Hence in oblique
photography we can hardly obtain a signifcantly higher
nominal scale than the value stated above.
Obtaining vertical images at approximately this same
scale is not particularly a problem (for example, with the
once common wide-angle aerial camera with f=152 mm
from an altitude of 1,520 m above the ground). To give an
example from central Europe, a limited number of verticals
with this scale are available in the military archive of the
Czech Republic in Dobruška (Břoušek, Laža 2006), although
more frequently we can fnd photos there with a nominal
scale ranging from 1:20,000 to 1:30,000. Nevertheless, large
format negatives (18×18 cm or more recently 23×23 cm) can
be enlarged without any signifcant loss of detail. Thus, we
can conclude that in the end, we are working with enlarged
oblique and vertical photographs of comparable scales (see
also Doneus 1997; Palmer 2005, 103–104). Furthermore,
the scale of oblique photographs dramatically decreases
from the foreground to the background of the image, which,
together with the distortion of shapes due to perspective,
usually leaves parts of oblique photographs useless for
detailed analysis.
Oblique photography using medium or large format flm
still has the advantage that we can get a greater enlargement
of the details on the positive compared to vertical imaging
from a greater height, but today most oblique photographs
are probably taken on small format flm or, increasingly, by
a digital sensor, the resolution of which has only slowly been
improved to approach the standard common in analogue
photography. Past studies have concluded that the necessary
density of data was not present in the primary digital record
Figure 2.
The concept of radial distortion of an image due to vertical ruggedness of the terrain on an aerial photograph. There is no simple
transformation relationship between the central projection of the photo and the orthogonal map or plan. The correction of the distortion can be
derived from a series of overlapping images, in which the apparent dislocation of points a, b, c on the individual photographs can be explained
by diferences in their elevation. Using the method of intersecting radial lines, their correct locations A, B, C on the map can be derived (after
Hampton 1978, Figure 17).