Assembling a site of acquisition : knowledge production and drone survey at Dunbeg

1 University of Oxford, Oxford, Oxfordshire, England, UK ABSTRACT: Geo-spatial visualising technologies are finding dynamic articulation within contemporary archaeology. With increasing regularity, archaeologists are using methods like drone-based photogrammetry to construct immersive spaces for research, analysis, and public-facing historical reconstructions. The rate at which they have been folded into the discipline, however, has outpaced efforts to critically theorise them. Too often these “new” forms of archaeological media are handled unreflexively. Often they are presented as easily knowable or self-evident. This paper attends to what it identifies as the contingencies inherent to the production of such media. Using theorists like Donna Haraway and Karen Barad, it specifically attends to notions of “partial objectivity”, “situated knowledges” and “embodiment in contemporary archaeological practice. Centred around a series of observations conducted as part of an ethnography of the Discovery Programme’s involvement in the Cherish Project (a collaborative EU-funded research initiative designed to monitor the impacts of climate change on coastal heritage sites in Ireland and Wales), it targets processes of data acquisition for photogrammetric modelling at the site of Dunbeg Fort in Co. Kerry, Ireland.


INTRODUCTION
It had been getting down to the wire. After four days of rainouts and windstorms we were finally going to be able to get the drone up and One of the core deliverables for Cherish is the production of a baseline dataset comprised of digital visual media products ranging from highly accurate point clouds to hachure plans and thousands of "raw" digital images. In making digital visualisation-and 3D modelling in particularso central to its cause, the Cherish Project is emblematic of the "digital turn" being sustained within archaeology more generally. Whether it is in the construction of immersive spaces for collaborative research or public facing historical reconstruction, archaeologists-particularly in Europe-have embraced advancements in photogrammetry, laser scanning, and VR/AR with palpable enthusiasm (BENDICHO, 2013;FORTE, 2014;2012;FORTE;PIETRONI, 2009;FORTE;KURILLOU, 2012;MORGAN, 2009;CORNS et al, 2017;KENNEDY, 2015;SHAW, 2009;DRAP et al, 2017). The discipline, however, is turning towards the digital with a concerning lack of critical reflexivity, particularly in its relationship to the digital visual media that it finds itself now capable of producing. Theoretical sectors of the discipline, certainly, have readily embraced poststructural, extra-lingual, and phenomenological discourse since the 1980s. From the advent of Ian Hodder's "post-processualism", to the work of scholars like Yanis Hamilakis, Chris Fowler, or Gavin Lucas, contemporary discourses on new materialism, assemblage theory, or relational realisms have found innovative articulations in archaeological circles (HODDER, 1982;HODDER, 1997;HAMILAKIS, 2017;HAMILAKIS;JONES, 2017;FOWLER, 2013;FOWLER;HARRIS, 2015;LUCAS, 2012). Yet the development of such concepts has been dogged by the persistent schism or gap between the theories proposed and their respective potential for methodological application. Often it is the case that such theoretical advances either cannot find convincing articulation or evidence within given datasets, or that a given theoretical framework is simply so broad and enveloping that it superficially can apply to all data (LUCAS, 2012). Furthermore, the majority of the literatures attempting to resolve these schisms have primarily focused on their application in traditional excavation contexts (FOWLER, 2013). As archaeology-in-practice becomes increasingly centred around methods of digital visualisation, however, new problems emerge. For one, archaeology has struggled to acknowledge the reflexivity inherent to the media it traditionally produces. Whether 2D, 3D, or otherwise, archaeologists too often produce and circulate media where "landscapes, lifeways, human and non-human entities are portrayed as static, timeless, uncomplicated or easily knowable" (PERRY, 2009;PERRY;BEALE, 2015).
Research initiatives like the Cherish Project are designed to incorporate suites of technological resources and instruments that update and evolve rapidly. As such, our capacities to empirically understand both the methods pursued and media produced as either socially constructed-or partially subjective-are frequently outpaced by the rate at which new methods are adopted and employed. The practice of digital survey, monitoring, 3D modelling, or analysis, then, warrants further and consistent critical theorising.
On an earlier trip to Ireland, I had been given a crash course on the content the Discovery Programme had already generated for Cherish, including models of Dunbeg Fort both before and after its collapse. These demonstrations evidenced the clear efficacy this media has to perform a range of analytical tasks necessary to make functional recommendations about future monitoring concerns. However, given the conceit of the project-namely its attendance to cultural resources that will likely either be disappeared or seriously compromised in the not-to-distant future-it's entirely possible that much of this data/ media will be the only remaining record of some of these sites. As such, they are far more than just self-evident analytical tools. They are loaded with a distinct potential to produce nuanced material necessarily iterative and multi-sited by design; the data derived in these initial stages is ultimately commuted to desktops and various softwares for processing and visualisation at later stages.
As such, the configuration and circumstances dictating both the protocol and enactment of on-site data acquisition will likewise hold considerable bearing and influence over the data as its translated across interfaces for visualisation and rendering. On the one hand, the relationship between the performed material enactments with instruments or landscapes in order produce either images or spatial coordinate data, say, is relatively evident. Furthermore, the relationship between such data and their application in analysis contexts is likewise often straightforward. I came to Dunbeg, however, hoping to tease out a slightly thornier relationship; namely that between the performed material engagements inherent to the acquisition of Cherish data and the knowledges such data could carry. As a result, I entered the field with particular concerns in mind; the types of embodiment produced through these procedures, for example, or how conditions necessary to conduct drone-based landscape surveys were predicated on the construction of contingent frameworks of reference. In attending specifically to the stages of acquisition observed at Dunbeg, what I will illustrate through the latter portions of this paper is how the dense interrelation of the components and interventions assembled in acquisition become co-constitutive of what such data can both do or mean as it coheres into tools for teaching, analysis, or reconstruction. Attending to the aspects of contingency present in the derivation of archaeological data is not necessarily novel. Doing so, however, is necessary in order to more meaningfully theorise the digital objects and media products such practices eventually come to comprise.

"SITUATION AND ENTANGLEMENT"
Scientific media is often used to derive, illustrate, or support knowledge claims. The difficulty, however, is that "it is all too easy to assume that scientific images show exactly 'the things themselves as they appear' without paying attention to the considerable work it takes for scientists to produce such pictures" (VERTISI, 2015, p. 8). As a consequence, any discussion of scientific media-particularly those used in Gathered under term of "situated knowledges", Haraway foregrounds circumstance and orientation in order to contend that "objectivity turns out to be about particular and specific embodiment", that "only partial perspective promises objective vision" (HARAWAY, 1988, p. 582-583).
Photogrammetry proves a particularly interesting test-case through which to consider Haraway's notion of partial objectivity. Long before the advent of structure from motion (IE, the most contemporary iteration of photogrammetry now widely used to generate 3D models), the Discovery Programme employed photogrammetric methods to aerial photography in order to generate georectified "ortho-images". Every photograph possesses a degree to radial distortion emanating from the focalisers within it. In the case of aerial photography, this distortion becomes particularly exacerbated at or around points of elevation. As such, overlapping images were pixel-matched in order to evenly distribute focus across a collection of photographs, eventually cohering into 2D images that could then be accurately geo-rectified to the pixel. As such, the empirical value of a geo-rectified photogrammetric image is produced through the systematic amalgamation of partial perspectives. Ironically, however, it is this very same process that effectively obscures the partiality inherent to the resulting product once reified as ortho-image.
Yet perspective, situation, or partiality, in this case, pertain to much more than the distribution of focus or distortion across a photograph. Any data, be it a photograph from a drone or a sparse point cloud from a laser scanner, is the result of a performed intervention between a given practitioner and the phenomena being observed.
As such, another issue with the already established problem of assuming scientific images simply show "the things themselves as they appear" lies in the correspondence between representation/ abstraction and notions of realism it suggests. In an effort to "disentangle realism from its traditional representationalist formulation", Karen Barad has suggested that "images or representations are not snapshots or depictions of what awaits us but rather condensations or traces of multiple practices of engagement" (BARAD, 2007, p. 51, p.53). As a theoretical physicist, Barad (via the philosophy-physics of Neils Bohr) contends that even a simple abstracted concept like the position of a particle in motion: …cannot be presumed to be a well--defined abstract concept; nor can it be presumed to be an individually determinate attribute of independently existing objects. Rather, position has meaning only when an apparatus with an appropriate set of fixed parts is used.
And furthermore, any measurement of position using this apparatus cannot be attributed to some abstract, independently existing object but rather is a property of the phenomenon-the inseparability of the object and the measuring agencies (BARAD, 2007, p. 139).
Barad, in so squarely foregrounding entanglement, dramatically reconfigures the perceived distance between an observer and the phenomena observed. Such a suggestion is also foundational to one of the more striking and instructive philosophical notions she develops.
Deferring again to Bohr, Barad elaborates the notion of "complementarity" to construct a radical theory of knowledge production. As one of the central notions within Bohr's larger theory of quantum physics, the law of complementarity contends that any particle in motion cannot have simultaneously determinate values. The very act of measuring a hypothetical particle's position, then, means the same particle's momentum becomes necessarily unmeasurable. As a result, the material entanglement required to either perform scientific measurement or produce reliable descriptions of observed phenomena negates any supposed representational distance between the production of knowledge and the knowledge produced.
Bohr argues that scientific practices must therefore be understood as interactions among component parts of nature and that our ability to understand the world hinges on our taking account of the fact that our knowledge-making practices are social-material enactments that contribute to, and are a part of the, the phenomena we describe (BARAD, 2007, p. 26).
If we accept that "we are part of the phenomena we describe", and that the content we produce In its rendering of threatened archaeological landscapes, the Cherish Project protocol for photogrammetric modelling roughly mandates the following: 1. After a preliminary site assessment, a series of control targets-clearly visible from the air-are distributed across the area of interest.
2. These targets are then registered within a given geographic coordinate system using either an on-site Real Time Kinetic network or a virtual base-station (both configured to available satellite networks).

3.
A series of automated overhead drone-based photographs programmed using the DJI Groundstation application.

5.
All the locative and photographic data is then commuted over to Agitsoft Metashape, a modelling software where the entire collection of photographsafter the geo-referenced control targets have been located and populated across the image cache-are processed into sparse point clouds, dense point clouds, meshes, and textures.  is the cohesion of these datasets that ultimately resolves in exhaustive and accurate 3D models, the respective value of the resulting models is also predicated upon the systematic fitting together of partial or mediated perspectives.
Because both vision and position are mediated and configured in accordance with these apparatuses, they are also necessarily partial.
The objectivity they afford, then, should be framed with these conditions of partiality firmly in mind.
As a consequence, the resulting photographs (the likes of which will ultimately be geo-rectified stitched together as a three-dimensional point cloud) are much more than a self-evident representation of features on a landscape; they are a condensation of intervening measurements, performed movements, and dispersed vision inherent to this process. performance. Furthermore, as digital objects that will be perpetually re-visualised and re-produced as they move across interfaces and platforms for any number of known or yet-to-bedetermined reasons, it is vital to remember that these conditions of partial objectivity, situation, or embodiment remain inscribed into the data as it travels and changes. As such, given their unique propensity to produce further knowledge claimseither through analysis or illustration-these factors will ultimately comprise new generative configurations in new contexts; crucially shaping how we might come to know or understand threatened heritage spaces like Dunbeg Fort.