**1. Introduction**

Independently, applying topographic imagery derived from LiDAR (light detection and ranging) or geophysical remote sensing methods in archaeological research is well-established in archaeology [1–21]. However, they are increasingly being applied together to create more robust understandings of social landscapes—including the emergence and long-term modification of built environments in the archaeological past (cf. [22–24]). This integration of aerial and terrestrial remote sensing methods has the potential to help tease apart the complexity of ever-evolving landscape palimpsests [25–27]. Archaeologists interrogate these large units of archaeological analysis (i.e., landscapes) at a given point in time, but they form over many millennia as a result of diverse human and natural processes that can build up, cut away, and rearrange the earth in ways that no singular remote sensing method can adequately elucidate. Moreover, from an anthropological perspective of remote sensing [28], the

integration of LiDAR-derived imagery and near-surface geophysical applications only enhances the ability of archaeologists to propose and explore new research questions and hypotheses apart from, or in conjunction with, excavations [29].

Our recent work at the Johnston Site in western Tennessee, USA (Figure 1) illustrates the efficacy of integrating these multi-scalar remote sensing tools to explore anthropological questions pertaining to Middle Woodland era (200 BC–AD 500) hunter-gatherer-gardener societies of the North American Midsouth, and to formulate new questions based on the results of such multi-scalar work. The Johnston Site, a satellite property of the better-known Pinson Mounds State Archaeological Park (PMSAP), is a large multi-mound center that has received very little attention by professional archaeologists since the site and the terrace on which it is located was first mapped in 1917 by E. G. Buck, a local civil engineer, hired by William Myer, a research associate of the Smithsonian Institution [30]. The integrated remote sensing approach and limited test excavations we used at Johnston resulted in a thorough evaluation of this 1917 map [31]. Our results afford us the ability to identify proxies for human-landscape interactions and environmental change in this area. This includes identifying areas where erosion has impacted the site and its monuments, as well as discovering shifts in monumentality at Johnston. Beginning to trace these changes allows us to lay the foundations for a landscape biography [32] of the Pinson Mounds vicinity that can be further developed with future research in this region.

**Figure 1.** Location of the Pinson Mounds Landscape in western Tennessee, USA (inset) and locations of Middle Woodland mound centers situated along the South Fork of the Forked Deer River discussed herein (primary map).

#### **2. The Johnston Site within the Middle Woodland Era Pinson Mounds Landscape**

The Middle Woodland period in eastern North America is characterized by the florescence of a near continent-wide social movement evidenced by novel religious rituals, elaborate craft production and exchange, and the rise of monumental ceremonial centers [33–42]. Alongside these changes came an increase in the importance of domesticated plant crops, some of which were associated with mortuary and other rituals, while a reliance on foraging wild plant foods and hunting was maintained; archaeologists refer to this subsistence pattern as the Eastern Agricultural Complex [43–48]. The Johnston site is part of the larger PMSAP, the largest Middle Woodland period ceremonial center in the southeastern United States (Figure 1). Spanning roughly 160 ha., Pinson Mounds exhibit a wide range of earthen monuments including Sauls Mound, the second tallest earthen monument ever constructed in North America at 22 m tall [49]. Other monuments include a large rounded geometric enclosure with a diameter of almost 340 m at its widest point, and at least 13 mounds comprised of low (ca. 1 m) and tall (ca. 10 m) rectilinear platforms, as well as small (ca. <1 m) and large (ca. 6.5 m) conical burial mounds. Aside from the impressive organization of labor and engineering required to construct the earthen monuments at Pinson, evidence for complex mortuary practices and the recovery of elaborate artifacts resembling those found in the Ohio Hopewell core area has positioned Pinson Mounds among the most important Middle Woodland centers for religious ceremonies, exchange, and pilgrimage in the eastern United States [49–51]. Even Hopewellian scholars working in Ohio have commented that Pinson is the, "premier Hopewellian center in the Southeast" because it was such an important destination for Middle Woodland societies [52].

However, Pinson represents only one, albeit the largest, collection of earthen mounds in this section of the South Fork of the Forked Deer River (SFFDR) in western Tennessee. It is centrally positioned amongs<sup>t</sup> a landscape of three Middle Woodland ceremonial centers in the region that encompasses nearly 100 km2, an unusual collection of sites for this region of the U.S. Using Sauls Mound as the center of Pinson, the Elijah Bray mound site is situated roughly 8 km upstream from Pinson on a terrace overlooking the confluence of Clarks Creek and the SFFDR. This comparatively small site is comprised of at least two conical burial mounds measuring 5.5 and 3 m respectively, with an associated artifact scatter spanning roughly 4 ha [44] (p. 15). The focus of this research, the Johnston site, is much larger than Elijah Bray. Johnston covers roughly 48 ha of a terrace overlooking the SFFFDR 6 km northwest of Pinson (Figures 2 and 3). The site is characterized by a collection of 10 rectilinear platform and conical mounds. This concentration of mound centers in a condensed stretch of a major tributary river to the Lower Mississippi River Valley is exceptional relative to the absence of mound centers in neighboring drainages across west Tennessee. This calls into question the historical development of this landscape during the Middle Woodland period, as well as the possible situational nature of pre-Contact American Indian use of these separate mound centers. Questions revolving around the unknown social, historical, and environmental contingencies that certainly influenced the indigenous use of this landscape motivated our archaeological research in this area of West Tennessee.

#### *Previous Research and Cartography at the Johnston Site*

Eastern North America has a long history of naturalists, antiquarians, and early professional archaeologists mapping indigenous earthen monuments [53–57]. Sometimes these people had professional backgrounds in surveying, sometimes they did not. It was not uncommon for some well-funded researchers to hire local surveyors to conduct mapping projects. Today, archaeologists using modern technologies like GIS software and aerial and terrestrial remote sensing methods are documenting the mixed successes of these early site surveys [2,22,58,59]. LiDAR, and geophysical surveys across eastern North America have shown that sites mapped more than 150 years ago were sometimes quite accurate. However, sometimes features were drawn di fferently than we might be able to discern today. These di fferences may relate to simple mistakes, generalizations, or overactive imaginations.

Like Pinson, the Johnston site was initially investigated by William Myer, an associate researcher of the Smithsonian Institution, who hired local civil engineer E. G. Buck to produce maps of both sites in 1917 [30] (p. 32), [49] (p. 52). However, unlike Pinson, the early map of Johnston arranged by Myer [60] has not been su fficiently reexamined using modern methods (e.g., [49,59]) to determine what this landform looked like at the time of early European expansion into West Tennessee. This is important to understand the broader Middle Woodland landscape along the SFFDR because the work of Mainfort and colleagues [59] identified numerous discrepancies in Myer's 1922 map of Pinson. For instance, they argue the elaborate enclosure walls that span the exterior boundaries of Pinson, as well as the "Inner Citadel" and their associated intersecting mounds, might not have existed, and thus may have been embellished or severely impacted by plowing. Identifying such discrepancies led to the title of their article "*Mapping Never-Never Land*", from which we derive the title of our article.

**Figure 2.** The 1917 map of the Johnston Site by E.G. Buck presented in Myer [31] and first published by Kwas and Mainfort [30]. Shown here courtesy of the National Anthropological Archives, National Museum of Natural History.

The work of Mainfort et al. [59] provides a cautionary lesson on cartographic 'artifacts' and the potential for embellished earthen architecture at Johnston. Nevertheless, to begin understanding earthen monuments at Johnston, and identifying how they might have changed since their initial mapping, we had to begin by assessing the original map of the site as first published by Kwas and Mainfort [30]. We use the 1917 map as a comparative documentation of the site prior to more than 100 years of agricultural impact. The full sketch map of Johnston depicts 10 mounds situated on a north-south oriented terrace (Figure 2). The details of this map are described in an unpublished manuscript by Myer [31] and discussed in detail in an article by Kwas and Mainfort [30].

The two platform mounds (Mounds 4 and 5) at the center of the site are the largest and most visible today (see also Figure 3). A pair of conical burial mounds (Mounds 1 and 2) are drawn at the northern edge of the terrace. Only Mound 1 is currently still visible. Three additional small conical mounds are situated along the bluff line that rises above the SFFDR floodplain (Mounds 3, 6, and 7), while three more small conical mounds are drawn south of Mound 4 (Mounds 8, 9, and 10). The dimensions of these monuments as they appeared in 1917 were recorded (Table 1) and provide baseline measurements that we can compare to the results of our research presented here. Low-lying parallel embankments (Walls K and L) are discussed as once being 3 m wide and 0.75 m tall and depicted on the Johnston map to have extended from Mound 4 north to Mound 1, leaving an open turn west near Mound 3 that led to a spring at the base of the bluff. Myer discussed the embankments being most visible in 1917 at either side of Mound 1, where they are drawn as solid constructions. Elsewhere on the map

of Johnston, the parallel embankments and other walls are represented by dotted lines, indicating they were ephemeral or proposed to have once existed in these locations. This includes embankments (listed as 'walls') on the 1917 map shown to have extended from Mound 4 to Mounds 8 and 9, and from Mound 9 to Mound 10. Associated with the mounds and embankments are small features labelled 'streets' that Myer discussed as leading from beneath Mound 4 outward to the cardinal directions, and from Mound 5 toward the cemetery (i.e., 'Graveyard' on the Johnston map). Very little is known of the cemetery and it was not surveyed during our work because it is located on private property. Within the parallel embankments between Mounds 1 and 4 are two roughly 13.5 × 13.5 m rectilinear features labelled 'A' and 'B'. These are described as 'structures' (i.e., buildings) because daub was visible in the soil at these locales when plowing occurred. Springs are also listed along the base of the western edge of the terrace.

**Figure 3.** The Johnston Site. (**a**) Aerial photograph of Johnston during 2015 field season. Numbers reference mounds easily visible upon ground inspection; (**b**) LiDAR-derived visualization of Johnston. Map produced in ArcGIS 10.6 using a 1-m resolution digital terrain model (DTM) and the Relief Visualization Toolbox [61,62] by overlaying a Sky-view visualization at 50% opacity onto the color-stretched DTM displayed at three standard deviations and set to refresh with the display extent. Numbers reference easily visible mounds.

Kwas and Mainfort's work on the Johnston site included site visits and examinations of surface-collected artifacts by residents who lived near the site. Their visits called into question whether any of the embankments ever existed; they commented that none of them were present during their site survey in the early 1980s. Moreover, they noted that only Mounds 1, 4, and 5 were visible and could be confirmed as indigenous earthen architecture. An examination of ceramics collected from the site suggested contemporaneity with the nearby Pinson Mounds. However, their analysis of projectile points and fragments from Johnston site differs significantly from such tools recovered from Pinson. While projectile points from Pinson can be characterized by Middle Woodland stemmed variants, the tools from Johnston are dominated by Late Archaic variants (ca. 4000–1000 BC). From the ceramic

and lithic evidence at Johnston, Kwas and Mainfort sugges<sup>t</sup> that the Johnston site was a potential predecessor to Pinson Mounds, with the mounds constructed around the first century B.C. [30] (p. 39).

**Table 1.** Shape and dimensions of mounds at Johnston as reported by Myer [31] in Kwas and Mainfort [30] predecessor to Pinson Mounds, with the mounds constructed around the first century BC [30] (p. 39).


1 Calculations have been converted from the 1917 measurements reported in feet to meters.

Applying a multi-staged [11,58,63] and multi-scalar remote sensing approach to Johnston provides an opportunity to thoroughly assess the accuracy of the 1917 historic map of the site, in addition to its descriptions of mapped earthen monuments. This approach integrating LiDAR-derived imagery and multi-instrument geophysical surveys allows us to harness the strengths of both techniques, and in doing so, fully realize the topographic and subsurface signatures of pre-Contact human manipulation of this terrace landform.

#### **3. Materials and Methods**

#### *3.1. LiDAR-Derived Imagery and Examination of the Johnston Site's Historic Map in GIS*

Aerial LiDAR integrates laser scanners, airborne vehicles (e.g., airplanes, helicopters, or drones), and highly accurate geospatial positioning instruments like real-time kinematic global positioning systems (RTK GPS) to record a diverse range of reflection information [64,65]. LiDAR-derived datasets are publicly available through most state governments in the U.S. This is because the method holds grea<sup>t</sup> potential for applications in a variety of civil and research sectors that span environmental hazard and ecological studies to heritage managemen<sup>t</sup> and archaeology. However, the quality of publicly available LiDAR datasets varies. We obtained LiDAR-derived digital terrain models (DTMs) from the State of Tennessee (https://www.tn.gov/finance/sts-gis/gis/gis-projects/gis-projects-elevation), which has high-resolution coverage for most of the state. LiDAR data across our project area along the SFFDR was collected in the winter of 2011 and 2012 in collaboration between the United States Army Corps of Engineers (USACE) and the Tennessee Division of the United States Department of Agriculture's Natural Resources Conservation Service (USDA NRCS). Data were collected with a nominal pulse spacing no more than one point every 70 cm. Downloadable bare-earth DTMs were produced from these data that were geospatially referenced using the NAD 1983 horizontal datum and the NAV 1988 vertical datum. The DTMs were tested to 1 m horizontal, and ≤18 cm vertical accuracy.

We created a LiDAR-derived hillshade visualization of the Johnston DTM in ArcGIS 10.6 by ESRI (Redlands, CA, USA) for preliminary analysis by altering the DTM symbology to display the topographic data in greyscale using the hillshade effect (Stretch Type: None; Z factor: 0.075). However, recent LiDAR-based archaeological research has highlighted numerous issues in identifying and interpreting landscape features from uni-directional hillshades alone [66–68]. Therefore, we also applied a suite of visualization methods available in the Relief Visualization Toolbox 2.2.1 (RVT; Institute of Anthropological and Spatial Studies, Ljubljana, Slovenia) [61,62] to the DTMs in order

to better assess the presence or absence of landscape features at the Johnston Site. These included multi-directional hillshade (16 directions, sun angle 35◦), principal components analysis (n = 3) of the multi-directional hillshade, simple local relief model, sky-view factor, positive and negative openness, and local dominance. Each visualization technique o ffers enhancement or deemphasis of topographic characteristics that allow archaeologists to elucidate the presence or absence of subtle landscape features (see Table 2). In-depth descriptions of these visualization methods can be found in [61,62,69]. When assessing the RVT imagery, we often created 'blended' images to enhance the RVT outputs with reference to the original LiDAR-derived DTM or other RVT imagery. An example of one blended image we used can be seen in Figure 3b, where we overlaid the Sky-view factor depicted in a greyscale color stretch at 50% opacity over the color stretched DTM. This helped us understand the correlation of subtle elevation changes to the range of values for height above mean sea level. Another blended image we used included overlaying the Sky-view factor depicted in a greyscale color stretch at 50% opacity over the Local Dominance imagery depicted in a color stretch. This blended imagery is presented and discussed in our results but o ffered better clarity for subtle elevation changes in both the Sky-view and Local Dominance imagery. To assess the validity of the 1917 Johnston map against the LiDAR-derived visualizations we created of the site, we georeferenced the 1917 map over an analytical hillshade visualization of the DTM in ArcGIS 10.6 using the 'Adjust' transform and adding control points on the 1917 map to the three most intact earthen mounds visible at the site today (i.e., Mounds 1, 4, and 5).


**Table 2.** Visualizations applied to the Johnston Site DTM in the RVT [61,62]. Information from [69,70].

#### *3.2. Magnetic Gradiometer Survey*

Magnetometry has become one of the most frequently applied near-surface geophysical survey methods to archaeological research. As such, it has received extensive discussion on the foundation of the technology [8,71,72]. We utilized a Foerster Ferex 4.032 DLG Karto 4-sensor fluxgate gradiometer (Institut Dr. Foerster GmbH & Co. KG, Reutlingen, Germany) to survey 14.25 ha of open field space at the Johnston Site (Figure 4a). Data were collected on a 40 × 40 m grid system that was laid out using a

Topcon RTK GPS (Livermore, CA, USA) and a Trimble robotic total station (Sunnyvale, CA, USA). Raw data were collected at a 0.5 × 0.1 m resolution.

**Figure 4.** Geophysical data collection underway at the Johnston Site. (**a**) ERH collecting gradiometer data with the Foerster Ferex DLG Karto; (**b**) APW collecting magnetic susceptibility data using the Bartington MS2 m and the MS2D field loop; (**c**) ERH collecting EMI data with the Geonics, Ltd. EM38-MK2.

Gradiometer data were processed in TerraSurveyor 3.0.36.0 (DW Consulting, Barneveld, The Netherlands) using typical destagger, destriping, and low pass filter processes prior to interpolating the data to a 0.1 × 0.1 m resolution for georeferencing in ArcGIS 10.6.

#### *3.3. Large-Area Surface Magnetic Susceptibility*

Magnetic susceptibility is a measure of a soil's ability to be magnetized when introduced to an artificial magnetic field [73,74]. Increased magnetic susceptibility has been identified as a proxy for cultural activity (e.g., low-intensity fires, midden formation, soil movement & manipulation) [6,75–77]. We collected large-area surface magnetic susceptibility data to complement the results of our gradiometer survey and aid in its interpretation, as well as provide a coarse resolution 'reconnaissance' survey of areas we could not survey with the cart-based gradiometer used in this study (cf. [78]). We collected volume magnetic susceptibility using the Bartington MS2 m and MS2D field loop (Bartington Instruments, Oxon, UK) across approximately 24 ha of the Johnston Site (Figure 4b). Data were collected in 20 m increments and spatially situated with a handheld GPS. Recorded data were an average of three readings spaced 10–20 cm apart. The instrument was zeroed prior to each reading. In some areas where we identified 'hot spots' of high magnetic susceptibility (mounds and non-mound open areas), we increased the resolution of our horizontal coverage from 20 m to 5 or 2 m. Data were visualized in ArcGIS as color-coded points, in addition to a raster interpolated using nearest neighbor gridding.

## *3.4. Electromagnetic Induction*

Electromagnetic induction meters (EMI) measure subsurface phenomena by emitting an electromagnetic field and measuring the response to that field as it is moved around the survey area [73,79–82]. Slingram EMI instruments can measure two responses to the electromagnetic field, the Quadrature-phase (QP), which represents apparent soil conductivity, and the In-phase (IP), which represents apparent volume magnetic susceptibility. Some EMI instruments can do this simultaneously. For instance, the Geonics EM38-MK2 (Geonics Limited, Mississauga, ON, Canada) we used in this study measures QP and IP at two different depths simultaneously (Figure 4c). This instrument has electromagnetic coil separations of 0.5 and 1 m, resulting in an approximate maximum depth penetration of 0.75 m and 1.5 m for conductivity data, measured in millisiemens per meter (mS/m), and 0.3 m and 0.6 m for magnetic susceptibility data, measured in parts per thousand (ppt) when operated, as we did, in the vertical dipole mode. Data were collected every 0.5 m along transects spaced 0.5 m apart. Data were downloaded and processed using the TerraSurveyor software package, with typical application of despike and either high-pass or low-pass filter operations applied before being interpolated to 0.25 m pixels and exported to ArcGIS.

#### *3.5. Test Excavations of Geophysical Anomalies*

After our LiDAR-derived imagery and geophysical datasets were processed and analyzed, we examined a non-random sample of the identified topographical and geophysical anomalies through test excavations. Excavations were conducted in a range of trench sizes, by hand, in arbitrary 10 cm levels until intact features were identified. We screened a 10–25 percent random sample of all excavated non-feature soil matrix (e.g., plowzone) using 6.35 mm (0.25") mesh. The entirety of all features we identified and excavated were processed using water flotation. This excavation methodology ensured a sufficient sample of displaced non-feature artifacts, and the total recovery of artifacts and ecofacts from feature contexts. The test excavations are presented in detail elsewhere [83]. The excavations we discuss here are intended to provide the reader examples of how our integrated remote sensing approach led to success in identifying surface and subsurface archaeological features that help us enhance our understanding of this important site, and better situate it within the Pinson ritual landscape [83].
