Geophysical correlation: global versus local perspectives

Abstract

Robust anomalies that point to the same buried features frequently occur in diverse data sets from multi‐method surveys [e.g. ground penetrating radar (GPR), magnetic gradiometry, electrical resistivity]. Relationships between such corresponding anomalies traditionally have been noted subjectively, through visual comparisons of mappings or overlays of geophysical results. These circumstances create a paradox because theory and correlational studies generally suggest largely independent dimensions, while mappings of geophysical results often illustrate parallel anomalies which suggest that robust correlations should exist. Through application of local Pearson's r in small neighbourhoods, with radii ranging from 0.71 to 1.6 m, a means is offered, and demonstrated, to quantify and map local correlations between two geophysical data sets. A case study examines relationships between apparent electrical resistivity, a GPR depth‐slice, and magnetic susceptibility data acquired at Army City, a Great War troop support town in Kansas, now obliterated except for subsurface remains. Intricate spatial patterns of positive and negative correlations between these modalities are illustrated that vary in complexity with neighbourhood size. Neighbourhoods of moderate size seem preferable because the spatial patterning of correlations is enhanced locally and region‐wide. With focus given to positive correlations that point to parallel anomalies, overlays of high correlation frequently correspond with robust anomalies observed in each data set and offer objective criteria for assessments of correlation. However, it is also demonstrated that mutually robust anomalies do not necessarily exhibit high correlation and that high correlations frequently exist in regions devoid of apparent anomalies. The issue of assessing jointly parallel correlations between three or more geophysical data sets is also confronted using a form of local principal components analysis. It is shown that high eigenvalues associated with first principal components that exhibit loadings in the same direction point to zones of mutually high parallel correlation.

Reconstruction of a Segment of the UNESCO World Heritage Hadrian’s Villa Tunnel Network by Integrated GPR, Magnetic–Paleomagnetic, and Electric Resistivity Prospections

“…A major problem with these two approaches is given by the fact that the different physical quantities used to map the same archaeological features generate, in most cases, structural lineaments that are slightly displaced, with respect to each other, due to measurement errors, instrument sensitivity, etc. Consequently, these maps may display very low correlation, even when a buried object is characterized by significant contrasts of such different physical quantities [23]. To overcome this problem, it is possible to apply statistical methods of data integration, for example local Pearson [23] and principal component analysis [24].…”

“…Consequently, these maps may display very low correlation, even when a buried object is characterized by significant contrasts of such different physical quantities [23]. To overcome this problem, it is possible to apply statistical methods of data integration, for example local Pearson [23] and principal component analysis [24]. In any case, it is important to note that none of these techniques should be used to combine magnetic anomalies or vertical gradients of the total magnetic field intensity with grids that represent the distribution of a physical quantity in the ground.…”

Exploring the Consistency of Data Collected in Archaeological Geophysics: A Case Study from the Iron Age Hillfort of Villasviejas del Tamuja (Extremadura, Spain)

“…We also calculated the corresponding values of the Pearson correlation coefficient (Table 3), which measures the statistical relationship between two variables. In particular, this coefficient is appropriate for geophysical studies applied in archaeology to identify any correlations existing between datasets collected using different geophysical methods [15,33,34]. The obtained values of this coefficient show that the best correlation is between the ERT and GPR results and that the worst correlation corresponds to the GPR and magnetic results.…”

“…We detailed the geological knowledge of the substrate in this zone to support the interpretation of geophysical methods. Currently, it is very common in archaeological studies to apply several geophysical methods to investigate anomalies that correspond to the same structure, because different methods may reveal different aspects of the subsurface [14][15][16][17]. In this study, we analysed the main anomalies revealed by each method and evaluated the graphical and numerical spatial correlation.…”

Potential of Virtual Earth Observation Constellations in Archaeological Research

“…The proposed concept is built upon the original work carried out by the Committee on Earth Observation Satellites (CEOS), which formulated the concept of virtual constellations, which was then extended by Wulder et al [16]. Although fusion and integration analysis of various remote sensing datasets have been reported in the past for archaeological research (see for instance [2,18,19,20,21,22,23], the concept of virtual constellations, proposed by CEOS is more than complementary observations. It is a framework for synergistic and coordinated use of earth observation sensors to increase the data availability, minimizing “unnecessary redundancy and costs” [16].…”