PREVIOUS   NEXT   CONTENTS   ISSUE   HOME 

4. Identifying Shell Middens Using Remote Sensing Methods

In developing an understanding of shell midden site locations we used the following methods.

Digital elevation data can give both an indication of palaeoshoreline location, and the type of intertidal and offshore subtidal environments. For example cliffs, wave-cut notches and raised beaches along palaeoshorelines can be identified from DSMs (digital surface models) as breaks in slope. Following the identification of palaeoshorelines, the gradient of the littoral and sub littoral zones can also be assessed — extensive shallow bays would offer better access to marine resources and higher probability of the presence of shell beds, which often grow in large numbers in shallow areas. Therefore areas with large shallow bays or shallow offshore areas (whether still extant or palaeo-features that have since been exposed) were prioritised, since we have found (on Farasan) that shell middens are more likely to be concentrated around these features.

False colour composite satellite images can be used to identify changes in sediment; for example shallow palaeobays and palaeoshorelines are often evidenced by in-filled marine sediments. The different sediments and rocks show up on the images as different colours, allowing them to be distinguished from one another.

High-resolution satellite images can aid interpretation of landscape features such as palaeoshorelines (raised beaches, cliffs, wave-cut notches), and can also be used to identify shell midden sites. Google Earth is often the best source of such data, displaying imagery from a range of sources such as Quickbird and World View 1 and 2. In arid regions such as the southern Red Sea, the lack of vegetation means that shell mounds are often easily visible because the lighter colour of the shells stands out against the darker colour of the surrounding land surfaces. The unique spectral signature of shell middens can also be detected on high-resolution false colour images; however these datasets were not available for this research owing to cost constraints. High-resolution satellite images can also record the shadow cast by some larger shell middens, in this region reaching heights of up to 6m during the mid-Holocene.

The data are then brought together in a Geographic Information System (GIS) (we used ArcMap10) which allows us to reconstruct the palaeoterrain and model shell midden site location. Reconstructing the palaeogeography focuses on the physical landscape, and re-creating what the coastlines would have looked like during the period of shell midden accumulation. This includes both identifying the palaeoshorelines, and the nature of the subtidal environment (palaeo-offshore topography). Palaeoshoreline features were identified through analysis of changes in sediment (false colour composite images), change in slope or the terrain (DSM), and analysis of high-resolution satellite images. Reconstructing the morphology of the subtidal zone was accomplished using the same methods, but looking for wide flat areas rather than changes in slope/sediment.

Once the palaeolandscape was reconstructed, the shell midden site location model could be developed. This model was initially developed and tested on the Farasan Islands, using the location and distribution of known sites to predict the likely location of further sites. The large sample of sites on the Farasan Islands combined with satellite imagery has enabled us to develop this predictive model for shell midden location (Meredith-Williams et al. 2014). The model uses the palaeolandscape reconstruction to identify areas most likely to contain shell midden sites, based on where the known sites are located. This is predominantly in shallow bays, inlets, headlands and small offshore islands; the most common factor is the presence of accessible shallow sub-tidal environments, which would have been suitable for productive shell beds.

The final stage in the development of the predictive model is to ground-truth sites located from model prediction and satellite survey, to check for false positive results and to use these to refine the model. During field-testing, many sites were also measured, with dimensions and surface composition recorded. Selected sites were test pitted to obtain samples for dating and analysis. Application of this model using satellite imagery has enabled us to identify, remotely, large numbers of sites elsewhere, particularly on the Dahlak Islands.

Satellite imagery also enables us to address in more detail the likelihood and extent of sea crossings, raising a possibility/probability of sea crossings across the Red Sea from the African side to the Arabian, as well as to offshore archipelagos and within them. Finally a viewshed analysis was carried out using the GIS to calculate if it was possible to see the islands from their adjacent mainland from a position on the highest terrain.