Study of the Variations of Archaeological Marks at Neolithic Site of Lucera, Italy Using High-Resolution Multispectral Datasets

Abstract

Satellite images have been systematically explored by archaeologists to detect crop marks, which are considered as a proxy for the identification of buried archaeological remains. Even though several existing algorithms are frequently applied, such as histogram enhancements and vegetation indices, the detection of crop marks still remains a difficult task, while the final interpretation results can be very poor. This paper aims to present some of the current difficulties of "remote sensing archaeology" in terms of detection and interpretation of crop marks due to the crops' phenological variations. At the same time, the presented work seeks to evaluate the recently proposed linear equations for the enhancement of crop marks, initially developed for the eastern Mediterranean region. These linear equations re-project the initial n-space spectral into a new 3D orthogonal space determined by three components: a crop mark component, a vegetation component, and a soil component. For the aims of this study, the Lucera archaeological site (southern Italy), where several Neolithic trenches have been identified, was selected. QuickBird and GeoEye high-resolution satellite images were analysed, indicating that vegetation indices may mismatch some crop marks depending on the phenological stage of the vegetation cultivated in the area of the archaeological site. On the contrary, ratios from linear equations were able to spot these crop marks even in shadow areas, indicating that improvements and developments of novel methodologies and equations based on remote sensing datasets can further assist archaeological research.

Deploying multispectral remote sensing for multi‐temporal analysis of archaeological crop stress at Ravenshall, Fife, Scotland

“…From the VIs tested, variants of the Simple Ratio (SR, MSR, GRVI), NDVI (NDVI, GNDVI) and ARI2 produced the highest degree of separability between the two sampled regions. This supports results from other studies, which have found the SR to be one of the most effective band combinations for highlighting crop marks (Agapiou et al, , ; Bennett et al, ). When assessed in isolation, the red and green bands were more effective for detecting change across the sample region than the individual red‐edge and NIR bands.…”

“…Note: Agapiou et al, , ; Agapiou, Hadjimitsis, et al, ; Bennett et al, ; Birth & McVey, ; Challis et al, ; Deering, ; Jordan, ; Pearson & Miller, ; Verhoeven & Doneus, . ρ GREEN , ρ RED , ρ RED‐EDGE , ρ NIR : reflectance across broad green, red, red‐edge and near infrared wavelength bands.…”

“…Over 150 VIs have been utilised across different studies (Bennett et al, ) although their comparative effectiveness for crop mark detection has been found to vary depending on observing conditions (Agapiou et al, ; Aqdus et al, ; Aqdus, Drummond, & Hanson, ; Bennett et al, ; Cerra et al, ; Challis et al, ). Whilst there has been criticism that VIs are often applied haphazardly, without appropriate consideration of the vegetation properties they have been formulated to highlight, they are still commonly used for their simplicity and effectiveness in highlighting contrasts in vegetation health (Agapiou, Lysandrou, Lasaponara, Masini, & Hadjimitsis, ; De Guio, ). Indeed, recent work (Cerra et al, ) has evaluated a range of VIs to provide objective means of comparing information return through the application of information theory.…”

“…Alternative approaches have looked at information extraction techniques beyond the use of simple band algebra, such as Principal Component Analysis (PCA) or Tasselled Cap Transformations (TCTs) to transform spectral data into new information components that minimise redundancy, or the work by Doneus et al () on visualising the red‐edge inflection point and distribution of a crop's reflectance curve (Agapiou et al, ; Agapiou, Alexakis, Sarris, & Hadjimitsis, ; Aqdus et al, ; Bennett et al, ; Cavalli et al, ; Challis et al, ; Doneus et al, ; Lasaponara & Masini, ; Traviglia, ). There is an ongoing challenge to developing universally effective crop mark enhancement techniques as the most appropriate methodology will vary depending on the sensor spectral resolution, the type of landscape and stage of the growing season (Agapiou et al, ; Verhoeven & Sevara, ), and relatively few multi‐temporal studies have investigated how analytical techniques perform over an extended observing period (Agapiou et al, ; Agapiou, Hadjimitsis, et al, ; Bennett et al, ). However, extensive studies by Agapiou on a test site in Cyprus have explored how the performance of VIs changes over a phenological cycle, although these results have yet to be tested outside of the Mediterranean, or in comparison to RGB photography (Agapiou et al, ).…”

Multitemporal 2016-2018 Sentinel-2 Data Enhancement for Landscape Archaeology: The Case Study of the Foggia Province, Southern Italy

“…For the purposes of our investigations, the "Tavoliere delle Puglie", a plain in northern Apulia (Southern Italy), was selected as a test area due to its long human frequentation and, consequently, its high archaeological potential and variety of cultural heritage features, along with a huge amount of ancillary data and information also available in a GIS environment. In the past the area herein investigated has been widely studied also using very high resolution satellite data as QuickBird, World View [13], geophysical prospection [14,15], along with manned [16][17][18][19][20] and unmanned aerial archaeology prospection [21]. Archaeological recorders along with previous knowledge and information already available in GIS environment are herein adopted as ground truth [22][23][24][25][26][27] to assess the potentiality of Sentinel-2 data in known archaeological sites related to different periods and civilizations, from Neolithic era to Greek-Roman and Middle Ages [28][29][30][31][32][33].…”