Guåhan (Guam) is an Unincorporated United States (U.S.) territory and the southernmost of the Mariana Islands located in the western Pacific (Figure 1). The northern part of the island is an uplifted limestone plateau with mostly limestone forest, while the southern part has rolling hills of volcanic origin with ravine forests and savanna grasslands. Typical for tropical areas, the vegetation is lush with a dense canopy and understory.
The CHamoru are the Indigenous people of the Mariana Islands and are thought to have inhabited this region for at least 3500 years (Carson 2014a). The archaeology of Guåhan is divided into the time periods of Pre-Latte (1500 BC–800 AD), Latte (800–1521 AD), Spanish (1521–1898), and Modern (1898–Present), which mark the cultural shifts of the CHamoru.
During the Latte Period, CHamoru people lived in villages all over the island and the most recognized remnant of this lifestyle are the latte, megalithic stone habitation structures (Figure 2). Latte are the unifying symbol the CHamoru people consider as a fundamental part of their present-day cultural identity. The presence of latte on Guåhan and the Northern Marianas is ‘proclaiming resilience in the face of centuries of colonialization, the massive destruction of WWII, and on-going globalization’ (Dixon, Bulgrin & Kottermair 2021: 39). Unfortunately, most latte sites have been disturbed, destroyed or left to dereliction. The cultural importance of the latte was the primary motivation to use lidar to find and preserve more unrecorded latte and other archaeological sites and features on Guåhan.
Initial Spanish interest in the Mariana Islands after 1521 waned when they realized that they did not contain the spices and riches they sought in Southeast Asia. Fray Diego Luis de San Vitores only established a small Jesuit mission on Guåhan in 1668. Within a short time, ‘…Spanish demands to abandon prominent elements of traditional culture and adopt imposed beliefs, institutions, and behavior’ (Montón-Subías & Dixon 2021:3) met with increasing CHamoru resistance. The Spanish Reducción of traditional CHamoru communities in the late 1600s and early 1700s then saw many former habitation sites and villages forcibly abandoned and their inhabitants consolidated into a few southern Guåhan villages under Spanish observation and indoctrination (Jalandoni 2011a).
In 1898 at the end of the Spanish-American War, the American Naval administration of the island began economic development of Guåhan and acculturation of the CHamoru people until World War II (WWII) and Japanese occupation. Inland villages fell into disuse or were only sporadically visited during early twentieth century agricultural activities, and later destroyed during post-WWII military construction and commercial economic development.
Today, there remains much disagreement on Guåhan between the military and Indigenous community because of the condemnation of CHamoru land and the destruction of prehistoric habitation and burial sites for military infrastructure (Dixon, Jalandoni & Kottermair 2022). In this context, there was a call in 2020 from the Guåhan Legislature to investigate non-destructive techniques to discover and preserve cultural sites (Kaur 2020). Therefore, in this research we explore a remote sensing technique that could be used to help better facilitate archaeological research and historic preservation on Guåhan.
Lidar (light detection and ranging) is a non-invasive remote sensing technique that works by sending out laser pulses and calculating the return time from surface reflections. This results in a point cloud with each point having a X, Y, Z location that represents the surface. The laser can penetrate a tree canopy the same way sunlight does, thereby also able to capture features on the ground below. However, the laser cannot penetrate areas where light is impeded from reaching the ground, such as buildings or extremely thick ground cover. The differences in collection methods change the perspective of the scanner, therefore ‘airborne lidar’ refers to lidar collected from an airplane, ‘terrestrial lidar’ is collected from the ground, ‘drone lidar’ is collected from a drone, and ‘satellite lidar’ is collected from a satellite. The individual points are post-processed and classified using algorithms to determine ground points and other classes like vegetation and water.
The classified point cloud can then be used to generate a digital terrain model (DTM) and a digital surface model (DSM), which are both a type of digital elevation model (DEM). These models are displayed in grid pattern with each grid or cell representing a single elevation value. A DTM only uses the ground points and shows the bare earth without any features, such as buildings or vegetation. In contrast, a DSM includes all features and shows the highest elevation value within a grid cell. In airborne lidar, the government agency or contractor usually supplies a post-processed point cloud and derivative products, like a DTM. Unfortunately, the resulting lidar-derived products are sometimes not suited for archaeological investigation and will require reprocessing (Lozić & Štular 2021).
In addition to being non-destructive, another benefit of airborne lidar is cost-effectiveness. Once the data is collected, usually by a government agency, the analysis can be done from an office. Fieldwork is one of the most expensive components of archaeological research, so when more analysis can be done from the office and reduce time in the field, it can reduce overall project costs (Jalandoni & May 2020; Jalandoni, Zhang & Zaidi 2022). Airborne lidar can also encourage citizen science participation as in the U.K. where volunteers and amateur archaeologists confined at home during the Covid-19 pandemic found dozens of archaeological sites using lidar-generated images (Fox 2020). Lidar is sometimes featured in mainstream media for highlighting exciting discoveries and fostering public interest in previously unknown aspects of their archaeological cultural heritage. Its potential use in locating latte stones, ‘a significant symbol and treasured birthright to contemporary CHamoru’ (Taitano & Liston 2021: 1), is especially appreciated when perceived as under threat by military construction and commercial development today.
Airborne lidar has proved useful in various contexts worldwide for archaeological research (for Mesoamerica: Chase et al. 2012; Europe: Bewley, Crutchley & Shell 2005; Asia: Evans & Fletcher 2015; and Australia: Davies, Turnbull & Lawrence 2016). However, there has been limited application of airborne lidar in the Small Island Developing States (SIDS) of the Pacific due to their remote geography and associated challenges. Airborne lidar has only been used in the archaeological contexts of the Pacific SIDS in American Samoa (Quintus, Day & Smith 2017), Tonga (Burley & Freeland 2022; Freeland et al. 2016; Parton et al. 2018), Vanuatu (Bedford, Siméoni & Lebot 2018), and Pohnpei (Comer et al. 2019). In American Samoa, Vanuatu, and Pohnpei the lidar data revealed land modifications indicative of a complex irrigated cultivation system and agriculture production, while in Tonga defensive structures and pigeon trapping mounds were identified.
In the Mariana Islands, lidar has been previously used in unpublished reports. On Guåhan, lidar was found to be less effective in discerning small-scale latte habitations and pre-WWII water cisterns in disturbed and vegetated contexts but has had some utility in identifying larger concentrations of concrete features and bomb craters in a post-WWII Seabee encampment (Dixon et al. 2017). On Tinian, lidar was effective in identifying large-scale WWII and post-war era military infrastructure and landscape modifications such as the historic American West Field (Gilmore et al. 2018). A paleo-terrain model was created to show an ancient shoreline of Guåhan during higher sea levels (Carson 2014b), but it was not mentioned whether lidar was used.
The novelty of this research, besides being the first academic archaeological investigation of using lidar on Guåhan, is that we explored multiple site types from multiple time periods unlike previous published research in Pacific SIDS that investigated a single site type. While not comprehensive, it is a thorough investigation of the use of lidar to explore the variety of Guåhan’s rich archaeological landscape.
The aim of this research was to determine whether the available airborne lidar was suitable for the identification of archaeological sites and features on Guåhan. Primarily we looked for latte (stone pillar structures), but we were also interested in finding agricultural terraces, gigao (fish weirs), and water wells; Spanish Period roads, bridges, structures, and forts; and Modern Period examples of WWII gun positions, bomb craters, developments, and fuel tank berms (Figure 1). The results can lead to better informed field surveys both in planning stages by targeting potential resources and upon discovery by enhancing documentation of previously unrecognised sites or poorly documented prehistoric habitation and land use sites such as those found in the southern Guåhan uplands. Lidar can also provide more accurate metrics of a site, especially for larger sites where field measurements might be difficult due to its size, obstructing vegetation, or unfavourable terrain.
Testing whether known archaeological sites on Guåhan are visible on the most commonly used lidar derivates, DSM and DTM, has helped to determine what site types can potentially be located using lidar. All the Modern Period sites were identified, as well as most of the Latte and Spanish Period sites. In the discussion we segregated which sites were best detected on which derivative, elaborated on challenges and limitations, and expounded on the research potential to find more archaeological sites. As more archaeologists and the community, both CHamoru and Guamanian, become familiar with lidar and what it can offer, this study encourages further exploration to potentially identify additional historic and prehistoric sites.
Topobathymetric airborne lidar data was acquired for the island of Guåhan by the National Geodetic Survey between January and July 2020 as part of a national shoreline survey. The data meets the United States Geological Survey Quality Level 1 standard which requires at least eight points per square meter or two points per square meter for bathymetry and a vertical accuracy of less than 0.15 meters. Island-wide lidar data was also acquired in 2007 and 2012, however, at a lower point density. We downloaded the classified point cloud in LAZ format and the DTM from the National Oceanic and Atmospheric Administration’s Digital Coast Data Access Viewer (NOAA 2021). It should be noted that the point cloud was classified only at a basic level meaning that many points remain unclassified including those representing vegetation. Also, the DTM was not detailed enough for this archaeological study.
Therefore, we further processed the LAZ files using Blue Marble Geographics Global Mapper® Pro v.24.0. A 1-meter-resolution topobathymetric digital terrain model (DTM), which is the most used lidar-derived product, was created using only ground [ASPRS class 2] and submerged topography  points. Similarly, a 1-meter resolution digital surface model (DSM) was created using all points except points pre-classified as noise  and water surface [41, 42]. Binning, using the minimum and maximum values, was the interpolation grid method used to generate the DTM and DSM, respectively. All other settings were default settings. Since a single latte feature is often under one meter in diameter, we generated a 0.25-meter resolution DSM and DTM to capture them. We then created a modified DSM (mDSM) that excluded points over three meters so that the latte features under tree canopy would be visible.
For facilitating visualization and interpretation, we created a hillshade of each lidar-derived product. The original DEM provides more detailed information but it is harder to visualise than the hillshade. While there are numerous visualization techniques, such as sky view factor and openness to name a few (Štular et al. 2012; Yokoyama, Shirasawa & Pike 2002), we used hillshade because it is the most common technique readily available in a GIS software.
High-resolution aerial imagery (U.S. Navy 2017) was also used to visualise the archaeological features and compare results with the 2020 lidar. However, conditions such as vegetation, water level (tide), and construction may have changed between collection of 2017 aerial imagery and 2020 airborne lidar. Since we were working on known archaeological sites under varying states of vegetation and preservation, we attempted to find the best visualisation of the site on available imagery. If the 2017 aerial imagery was insufficient, we cross-referenced archival Google Earth satellite imagery and historical photographs, maps, and artwork from our collections.
Next, archaeological data was compiled from the records of the authors and imported into ArcGIS Pro 2.9.0 by Esri along with aerial imagery from 2017. Initially, we only selected features of interest where the exact location was known, but many of the records had only approximate locations. Each selected feature was investigated in the aerial imagery and the lidar-derived layers to determine which features could be detected in which reference layer.
A feature confidence level (adapted from Lozić & Štular 2021) was then assigned to categorize the suitability of lidar at identifying archaeological sites on Guåhan (Table 1). However, since we used ideal examples, the confidence level we determined is the highest that we expect for a particular site type using default parameters for the various DEMs.
|0||None||No feature is visible.|
|1||Low||A feature is visible, but the shape is not clearly defined or possibly an anomaly of lidar data.|
|2||Medium||A feature is visibly detected but incomplete or indeterminate whether natural or anthropogenic.|
|3||High||A feature is clearly visible and is verified either by expert knowledge or ground-truthing.|
The archaeology of Guåhan is known from local knowledge and grey literature more than from academic publications. Therefore, this project required intimate archaeological knowledge and local technical expertise for site selection. The sites selected were based on the authors combined archaeological and GIS experience on the island of over 100 years. We aimed at sampling a variety of sites from all time periods across the island. Considering all the sites included in this study, only the archaeology of Fort Soledad (Driver & Brunal-Perry 1994) has been academically published, while some of the Litekyan (Ritidian) water wells are mentioned in passing (Carson et al. 2014).
Pre-Latte Period archaeological sites are only identifiable by subsurface investigation or small artifacts like a cache of pre-latte ceramics or marine shell ornaments. Unfortunately, there are no associated surface features or land modifications of appropriate size for lidar detection that are definitively pre-latte. As a result, this study covered only a sampling of Latte, Spanish, and Modern period archaeological sites that have potential for being visible on the lidar.
For Latte Period sites we sampled latte, gigao, wells, and agricultural terraces. The latte stone forms are unique to the Mariana Islands (Morgan 1988). Latte are stone columns with a cup-shaped capstone in two parallel rows of pillars (Figure 2). The number of paired sets vary between six and fourteen. These megaliths come in various sizes, the tallest known is House of Taga on Tinian where the remaining column stands 4.6 m tall (Figure 2), while the As Nieves quarry on Rota reveals other attempts at erecting large latte (Russell 1998). The majority appear to have been around 1 m tall. They are found mostly on the presently inhabited islands of Guåhan, Saipan, Luta (Rota), and Tinian, but are also recorded on Aguiguan and Pagan plus smaller mostly uninhabited northern islands today (Athens 2011). The locations of some of these structures were recorded in historical Spanish accounts but were not always accurate due to name changes or the foreigner’s poor understanding of the complex geography (Jalandoni 2011b). Even where previous archaeological surveys recorded Latte Period structures, the positional data is often inaccurate due to the use of low-accuracy GPS receivers and thick canopy cover resulting in GPS signal error or steep terrain masking the location on a small-scale topographic map or aerial photograph.
For this study we investigated two repositioned latte sets, Senator Angel Leon Guerrero Santos Latte Stone Memorial Park in Hagåtña often referred to as the Latte Stone Park and a latte monument near the Nikko hotel in Tumon, because these sets were reconstructed from previous archaeological settings and located in public areas. They are the ideal examples of what a latte set would have looked like without its a-frame structure resting on top during the Latte Period and their location is known and easily verifiable.
Other Latte Period archaeological sites potentially recognizable on lidar include gigao or coral stone fish weirs, traps, and pens in coastal bays and estuaries (Dixon, Gilda & Mangieri 2012); agricultural terraces with stone walls on inland slopes (Highness & Leap 1992; Dixon, Schaefer & McCurdy 2010); and traditional hand-dug freshwater wells in coastal settings (Carson et al. 2014). For gigao or fish weirs, we chose Abo Cove because of their visibility from the shoreline (Gosser et al. 2004). Then we examined the northern coastal lowlands of Litekyan, Haputo, Jinapsan, and Tarague sites for reported wells. Finally, we inspected the Sigua drainage for reported agricultural terraces (pers. comm. David Defant, July 2022).
As a result of Reducción, most of the inhabitants on Guåhan were clustered into towns that the Spaniards developed. Archaeological sites from this period include rural forts, bridges, roads or bull-cart paths, ovens or kilns, rock-lined wells or water courses, wetland ponds with ditches or dikes, small churches or chapels, and public buildings such as schools (Tomonari-Tuggle et al. 2018). The largest structures or ruins remain visible and include the Plaza de España in Hagåtña, Fort Nuestra Señora de la Soledåd (Fort Soledad) in Humåtak, and Talaifak Bridge in rural Hågat. Many have been restored in the last 25 years by the Guam Preservation Trust, a non-profit, public corporation.
Some less widely known are rural colonial features including wooden plank roads of the Camino Real near Atantano’, rural domed “Spanish” ovens, later water cisterns, and hilltop bonfire lookouts or Vigía for incoming ships (Madrid 2014). Due to their size and poor preservation, these archaeological features are more challenging for lidar identification and not included in this study.
When the Manila Galleons from Acapulco ceased their yearly visits to Guåhan in 1815, so too did the yearly influx of imported products and livestock, hence many rural church and clergy farms and properties were often abandoned. By WWII, Japanese and American bombing and then reconstruction of the colonial capitol of Hagåtña and the port in Apra Harbor destroyed most Spanish colonial remnants of urban Guåhan, often placing the footprint of the historic past beyond the reach of lidar.
For Spanish Period sites we sampled a road, bridge, structural complex, and fort. The road-cut next to Asan Point ridge was sampled because it is hypothesized to have been a part of the Camino Real used during the Spanish Period, before modification by the U.S. military to construct Route 1 Marine Corps Drive in 1944–45. This segment of the Camino Real from Hagåtña to Piti and Apra Harbor passing between two cliff faces is portrayed in painted images from visiting European artists in the early 1800s (Tomonari-Tuggle et al. 2018). For Spanish colonial architecture, we sampled reconstructed Talaifak Bridge, Plaza de España, and Fort Soledad because they are prominent landmarks in guided tours and local history.
Archaeological remains from historic American to modern times included in this lidar study reflect a conscious shift from European to American and Asian colonial institutions and values, both socio-economic and increasingly geo-political. After 1898, Apra Harbor and Sumay hosted not only the port of Guåhan with coal and later fuel oil storage, but a Pan-American clipper hotel for visitors in route from Hawaii to Asia, telegraph cables that tied Guåhan to world events, and barracks for the US Marines. Foundations of these abandoned communication and trade sites in Apra Harbor are often recognizable to lidar but overgrown. In contrast, foundations of Hagåtña that grew from a sleepy colonial outpost to a growing commercial hub providing goods and services to CHamorus and the first American territorial military dependents are largely undetectable today due to the destruction during WWII and subsequent development in modern times.
Sites and features related to WWII include those of the Japanese military from 1941–1944, such as anti-aircraft positions and concrete fortified bunker entrances or tunnels. However, they were often buried or destroyed by 1944 to post-war American construction of expanded naval and air force bases and the rebuilding of war-devastated areas and new urban communities. Beginning in the late 1960s, infrastructure like new roads, power, public water, and telecommunication utilities were built to improve the lives and health of the local community outside of Hagåtña.
For Modern Period archaeological features, we sampled gun positions, bomb craters, developments, and fuel tanks. We first looked for WWII Japanese anti-aircraft positions within the USDA Brown Tree Snake Pen in Anderson Air Force Base (AAFB). Past field surveys have recorded that three features were C-shaped facing opposing directions, constructed of stone and soil-filled 50-gallon drums, and covered in soil excavated from a shallow depression (Dixon et al. 2010). Next, we looked for bomb craters in shallow waters off the coastal areas below Orote Point at the entrance of Apra Harbor. There was a Japanese airfield on Orote Point and the area is known to have been under attack by the American Army Air Corps in 1944 (Dixon, Jones & Nelson 2017). Third, we explored abandoned American military hospitals and later housing developments in Harmon. In 1945, the 137th Army Hospital was constructed in the area for the invasion of mainland Japan, as shown on the Fifth Naval Construction Brigade Area Allocations map (Pacific National Historic Park Guam 1992). By the 1950s, American military housing development of tents, wooden houses, or metal Quonset huts on concrete pads in the same general location of Harmon (Craft & Byerly 2019) had been destroyed by typhoons or later removed. Finally, we looked for abandoned fuel tank berms on the ridgetops above Sasa Bay. The earliest WWII Sasa Bay fuel tanks are no longer present although some berms remain (Dixon et al. 1999).
The results of evaluating the visibility of archaeological sites included in this study from aerial imagery and lidar-derived products are summarised in Table 2. For every site we indicated whether it is visible in the 2017 aerial imagery, its confidence level and resolution in the DTM, DSM, and mDSM derived from the 2020 airborne lidar. It should be stressed that the confidence level refers to the hillshade of the DEMs and not the DEMs themselves.
|TIME PERIOD||ARCHAEOLOGICAL SITE||2017 IMAGERY VISIBILITY||2020 AIRBORNE LIDAR|
|Latte||Hagåtña Latte Stone Park||no||0||0||3||0.25 m|
|Tumon Latte Memorial||partial||0||2||3||0.25 m|
|Abo Cove gigao||yes||0||0||nd||1 m|
|Northern coastal wells||no||2/3||0||nd||1 m|
|Sigua agricultural terraces||no||2||1||nd||1 m|
|Spanish||Asan road||no||1||0||nd||1 m|
|Talaifak Bridge||yes||1||0||nd||1 m|
|Plaza de España||yes||2||3||nd||1 m|
|Fort Soledad||yes||2||2||nd||1 m|
|Modern||AAFB anti-aircraft gun positions||no||2||0||nd||1 m|
|Orote bomb craters||no||3||0||nd||1 m|
|Harmon housing development||no||3||0||nd||1 m|
|Sasa Bay fuel tank||no||3||0||nd||1 m|
The latte at Hagåtña Latte Stone Park were not visible on the imagery because they were completely obfuscated by large trees. The latte set at Tumon Latte Memorial was partially visible on the imagery and DSM, where vegetation did not obscure it. After filtering points over three meters in the mDSM, the foliage of the large trees was removed and both latte sets were visible with a feature confidence of 3 (Figure 3). No features were detected in the DTM because the points representing the latte were filtered out during the DTM creation. It should be noted that this ideal condition of upright latte in perfect alignment is unlikely in the natural settings of latte today because they are usually overgrown and fallen over.
The gigao in the Abo Cove wetlands were visible on imagery but not visible on any of the lidar-derived products resulting in a feature confidence of 0 (Figure 4). The gigao remains are not in good condition to begin with and are increasingly being covered by mangroves off the shoreline. Their piled rather than stacked coral walls are too low, eroded, and partially submerged at high tide to be distinguished from the water and noise on the former reef flat.
The traditional water wells in the northern coastal areas of Litekyan, Haputo, Jinapsan, and Tarague Were not visible on the imagery because they are obscured by vegetation. However, they Were clearly visible on the DTM with a feature confidence of 2 and 3 since not all wells have been field-verified (Figure 4). Several circular depressions are in linear alignments within or behind back-dune depressions, which is the ideal location for a well to access the shallow freshwater aquifer. In addition, they are in coastal areas such as Litekyan in the northern plateau that were not targeted for invasion by the Americans during WWII. And we know from early historical records and archaeological evidence that coastal villages on the northern plateau were densely inhabited during the Latte Period. We were able to identify ten wells at Litekyan, three at Haputo, at least five at Jinapsan, and six at Tarague. However, the number of traditional versus later historic wells is not conclusive until it is verified through ground-truthing and archival research.
The walls of the agricultural terraces in the Sigua drainage were not visible on the 2017 aerial imagery but visible on the DTM with a feature confidence of 2 (Figure 4). The walls Were also partially visible on the 2020 Google Earth imagery. They are in an area on broad slopes above the Sigua River. The terraces appear as four wide and roughly parallel alignments, described as made of stone. Each appear perpendicular to the slope in an orientation that might facilitate drainage into each other and then the river during seasonal rains.
The Asan Camino Real road-cut, not to be mistaken for the adjacent modern Route 1, was not visible on the imagery (Figure 5). A feature was visible on the DTM but not clearly defined resulting in a confidence level of 1. Talaifak Bridge was restored in 2013 and was visible on imagery but the DTM had a feature confidence of 1 and DSM confidence level of 0 because it cannot be differentiated from its surrounding. The Plaza de España was easily visible on imagery and DSM with a feature confidence of 3. Features, such as buildings, walls, and pathways restored after WWII, can be detected on the DTM with a feature confidence of 2. As expected, Fort Soledad was large enough and in open space to be visible on imagery. Both the DTM and DSM have a feature confidence of 2 because of the clearly visible land modification (levelling the hilltop). However, the structures are filtered out in the DTM and inconclusive in the DSM.
The three Japanese anti-aircraft defensive berms were not visible in the imagery but were recognized on the DTM on the AAFB plateau with a feature confidence of 2 (Figure 6). WWII American bomb craters on the coast below Orote Point were not visible on the imagery but noted in the DTM with a feature confidence of 3. The Harmon housing development was not visible on the imagery but the roads in a grid pattern were recognizable on the DTM with a feature confidence of 3. The WWII Sasa Bay fuel tank circular berms were not visible on the imagery, but easily detectable on the DTM with a feature confidence of 3.
The motivation for this paper was to determine which archaeological sites on Guåhan can be detected using lidar. Apart from the Abo Cove gigao, Talaifak Bridge, and Asan road-cut, we were able to detect all archaeological sites sampled using lidar-derived products with medium or high confidence.
The DTM was beneficial for detecting land modifications to the ground surface where the feature was classified as ground points and not filtered out. Generally, the algorithms applied to classify ground versus non-ground points filter out structures and vegetation. As a result, archaeological built-up features, like latte, cannot be detected in the DTM.
For Latte Period sites, the northern coastal water wells offer one of the clearest examples of the advantage of using a DTM because they were distinctly visible. Water wells are significant as an indication of Latte Period habitation and a substantial labour investment, using only shell and stone tools, normally associated with village level occupation. A field survey of the water wells may incorrectly record their location and dimensions (depth and width) or fail to see them entirely. A lidar survey can likely identify all wells, accurately map their locations, and systematically record their dimensions. Of course, the results of the lidar survey will need to be ground-truthed.
For Spanish Period sites, the DTM was the preferred method for identifying large-scale land modifications that were undertaken in preparation for building major structures such as forts. For example, Fort Soledad was only identifiable on the DTM because of the levelled hilltop and anthropogenic shape. Similarly, only the structural footprint left by the walls and land modifications of the Plaza España helped in their detection.
For Modern Period sites, the DTM proved particularly useful for detecting the geometrically shaped remnants of the AAFB anti-aircraft gun positions, Orote bomb craters, Harmon housing development, and Sasa Bay fuel tank berms that were not visible in the imagery.
The DSM was beneficial for detecting some archaeological structures; however, these were also visible on the imagery. Since the DSM was created using the maximum elevation within a grid cell it showed whatever was at a higher elevation, which could either be the archaeological feature or vegetation if it was covering the archaeological feature. Similarly, the visibility of an archaeological site on imagery is often also limited by vegetation. For example, the dense tree canopy at Tumon Latte Memorial obscured some latte features on both the DSM and imagery.
The mDSM solved the issue of vegetation obscuring the feature by excluding points in the DEM creation that exceed a certain height above ground. Our mDSM excluded all points above three meters, filtering out the parts of the tree that exceeded that height. This approach worked well in our two sampled latte sites since the features were below the foliage of the tree.
One benefit of using DSM or mDSM over imagery that should be noted is for measuring heights of features. In contrast to imagery, which is only visual, the DSM also provides elevation information. The ability to measure height from a desktop can be especially useful in small developing island states where access to sites can be difficult.
Other factors that affect the detection of archaeological features in the lidar are whether they are intact, have a distinct pattern, distinctive vegetation, and a measurable degree of land alteration. For example, if an archaeological wall is intact, then it is easy to identify during pedestrian surveys but also in lidar surveys. If a feature has a distinct pattern such as a circular fuel tank berm or rectangular foundation, that can also help its detection.
A general challenge we identified in using lidar for detecting archaeological sites is the vegetation on Guåhan. Archaeological features are often intermingled with vegetation, especially in the dense tropical forests or savanna grass areas, making it difficult to use height above ground as a filtering technique. The mDSM we created was effective at revealing the latte sites but only because they were obscured by dense canopy with no understory. However, many latte sets, particular in the middle of the island, are surrounded by swordgrass (Miscanthus floridulus). If vegetation was classified in the point cloud, it could be used as a filter. Other methods may need to be tested and some of these are explored in the next section.
Three archaeological sites in our investigation had specific challenges: Abo Cove gigao, Talaifak bridge, and Asan road-cut. The Abo Cove gigao and Talaifak bridge were visible only in the aerial imagery and not the lidar-derived products. A possible explanation for the gigao is that the walls were submerged during data capture and high sediment content in the water created noise that obfuscated the feature. Therefore, in some circumstances low relief archaeological features may be easier to detect on imagery than lidar if they are not covered by vegetation and have contrasting colours. A possible solution would be using the intensity values of the lidar to potentially reveal the feature.
The Talaifak Bridge was not visible on the DTM because the points that represent the bridge were non-ground points. It was also not visible on the DSM because it was indistinguishable from the features around it. However, using the actual elevation grid rather than the hillshade or both layers combined might help in differentiating the feature. We could not detect the historic Asan road, just parts of the road-cut at the top of the ridge, because of major modifications to the landscape when the northern end of the limestone ridge was quarried to build Route 1. However, the land modification is clearly visible in the DTM.
As noted earlier, one major limitation of lidar is that the features of the site need to be of a certain size and arrangement, therefore precluding any Pre-Latte Period sites because they are mostly subsurface or too small. This limit applies to smaller sites from other time periods as well, such as ceramic scatters that are considered sites and prehistoric burials on Guåhan. Also, some of the detected site types sampled were in near ideal condition in terms of structural integrity, arrangement, and surrounding vegetation. Many others might not be as easily detected in their more natural condition.
Finally, there is also the ambiguity in differentiating features with a similar appearance in Guåhan’s archaeology. For example, there are visual parallels between double-alignment trails and agricultural terraces or some bedrock outcrops, hand-dug wells and bomb craters, and fish weir walls with post-war coastal modifications. Our rationale for distinguishing bomb craters from hand-dug wells demonstrated how essential it is to know the context of the site and stressed the importance of local knowledge. Lidar can be useful, but it is just one tool among other archaeological techniques. If a site detected on lidar cannot be corroborated using existing archaeological documentation or local knowledge, it needs to be verified with field visits.
Using the latte sites, we identified that changing the resolution, visualisation, and algorithm parameters of the DEM can improve detection of this site type. These strategies can be applied to other site types to also improve their specific detection.
A potential next step is to reclassify the point cloud to remove vegetation and other surface features to create a digital feature model (DFM). A DFM is a DTM that includes archaeological features and other structures that are not necessarily ground points (Pingel, Clarke & Ford 2015; Lozić & Štular 2021). Additional sites, both known and unknown, should be detected more easily on the DFM leading to a higher feature confidence level. Machine learning can also be used to facilitate classification. For example, knowing that latte sets are on average in pairs of six to ten can help in their classification. Then further modelling, such as Bayesian statistics, can be employed to predict new sites.
The visibility of the Asan road-cut, the paths at Plaza de España, and the remaining roads of the post-WWII military housing and hospital development in Harmon offer some potential for using lidar to detect more Spanish roads. It could be especially useful for identifying the Camino Real from Hagåtña to Humåtak in the hill above Asan Park to the port of Piti and former elevated passages around Atantano Marsh (pers. comm. Anthony Alvarez, Sept 2022).
Finally, this work can be expanded to finding archaeological sites in Saipan, Rota, Tinian, and other Mariana Islands that have available lidar data and in some cases very different historical trajectories involving German and Japanese presence for decades before WWII.
The purpose of this research was to ascertain the effectiveness of using lidar for finding archaeological sites on Guåhan from different time periods. Using either DTM or DSM, the two most widely used lidar-derived products, we were able to detect archaeological features from the Latte Period, Spanish Period, and Modern Period. Modern Period sites and many of the larger Spanish Period sites are known from the military and historical records. However, finding Latte Period sites are where lidar can provide a unique source of data. While there are historical data on these archaeological sites, they can be unreliable and the verifiable accuracy of site locations in the field is low.
While we focused here on the default lidar-derived products most archaeologist will have access to, future research can extract more information from the lidar data using more complex settings. Still, this work serves as the foundation to eventually build an integrated multi-scale interpretation of archaeological features in the prehistoric and historic landscape by combining environmental factors, structured expert knowledge, and the automation of archaeological feature detection and extraction that can lead to more archaeological sites being discovered.
Development for the global tourism industry and military expansion have led to archaeological sites disappearing under new or renovated structures, underscoring the importance of lidar in planning for the future now. By finding more effective ways to identify archaeological sites and their environmental settings, this project has demonstrated the potential use of lidar to improve archaeological research before construction to benefit CHamoru cultural heritage. Furthermore, this work has been encouraged by the CHamoru people interested in non-invasive methods for identifying and preserving their cultural heritage.
The authors would like to acknowledge the advice of several colleagues, including David DeFant, Blaž Miklavič, Mike Carson, Jolie Liston, Wakako Higuchi, Lon Bulgrin, Ross Winans, Sama Low-Choy, Olympia Terral, James Oelke Farley, and Anthony Alvarez. The anonymous reviewers are thanked for their constructive criticism that greatly improved the manuscript. This research was funded by a 2020 Griffith University Postdoctoral Fellowship and a 2020 Australian Academy of the Humanities Travelling Fellowship.
The authors have no competing interests to declare.
Athens, S. 2011. Latte Period Occupation of Pagan and Sarigan, Northern Mariana Islands. Journal of Coastal and Island Archaeology, 6: 314–330. DOI: https://doi.org/10.1080/15564894.2011.555806
Bedford, S, Siméoni, P and Lebot, V. 2018. The anthropogenic transformation of an island landscape: Evidence for agricultural development revealed by LiDAR on the island of Efate, Central Vanuatu, South-West Pacific. Archaeology in Oceania, 53(1): 1–14. DOI: https://doi.org/10.1002/arco.5137
Bewley, RH, Crutchley, SP and Shell, CA. 2005. New light on an ancient landscape: lidar survey in the Stonehenge World Heritage Site. Antiquity, 79(305): 636–647. DOI: https://doi.org/10.1017/S0003598X00114577
Carson, M. 2014a. First Settlement of Remote Oceania Earliest Sites in the Mariana Islands. Springer, NY. DOI: https://doi.org/10.1007/978-3-319-01047-2
Carson, M. 2014b. Paleo-Terrain Research: Finding the First Settlement Sites of Remote Oceania. Geoarchaeology, 29: 268–275. DOI: https://doi.org/10.1002/gea.21457
Carson, M, Dixon, B, Peterson, J and Jalandoni, A. 2014. Site Definitions in a Complex Archaeological Landscape: An Example at Ritidian, Guam. In: Carson, M (ed.), Guam’s Hidden Gem: Archaeological and historical studies at Ritidian, 2663: 105–111. British Archaeological Review International Series. DOI: https://doi.org/10.30861/9781407313054
Chase, AF, Chase, DZ, Fisher, CT, Leisz, SJ and Weishampel, JF. 2012. Geospatial revolution and remote sensing LiDAR in Mesoamerican archaeology. Proceedings of the National Academy of Sciences, 109(32): 12916–12921. DOI: https://doi.org/10.1073/pnas.1205198109
Comer, DC, Comer, JA, Dumitru, IA, Ayres, WS, Levin, MJ, Seikel, KA and Harrower, MJ. 2019. Airborne LiDAR Reveals a Vast Archaeological Landscape at the Nan Madol World Heritage Site. Remote Sensing, 11(18): 2152. DOI: https://doi.org/10.3390/rs11182152
Davies, P, Turnbull, J and Lawrence, S. 2016. Remote sensing landscapes of water management on the Victorian goldfields, Australia. Journal of Archaeological Science, 76: 48–55. DOI: https://doi.org/10.1016/j.jas.2016.10.009
Dixon, B, Athens, S, Welch, D, Ward, J, Mangieri, T and Rieth, T. 1999. Archaeological Inventory Survey of the Sasa Valley and Tenjo Vista Fuel Tank Farms, Piti District, Territory of Guam, Mariana Islands. Honolulu: IARII.
Dixon, B, Bulgrin, L and Kottermair, M. 2021. Latte Period Cultural Heritage in the Commonwealth of the Northern Mariana Islands. In: I estoria-ta Guam, the Mariana Islands and Chamorro Culture An exhibition of “Let’s turn around the world” Official Program of the Fifth Centenary of the First Round the World, pp. 37–46. Madrid: Museo Nacional de Antropologia.
Dixon, B, Gilda, L and Mangieri, T. 2012. Archaeological Identification of Stone Fish Weirs Mentioned to Freycinet in 1819 on the Island of Guam. Journal of Pacific History, 48(4): 349–368. DOI: https://doi.org/10.1080/00223344.2013.856666
Dixon, B, Jalandoni, A and Kottermair, M. 2022. Traditional places in conflict and their historic context: Ritidian, Guam. Archaeological Perspectives on Conflict and Warfare in Australia and the Pacific, 54: 89–105. DOI: https://doi.org/10.22459/TA54.2021.05
Dixon, B, Jones, R and Nelson, I. 2017. Archaeological Monitoring in Support of Remedial Investigation at Orote Point (Spanish Steps) Trap and Skeet Range, Naval Base Guam. Prepared for: AECOM Technical Services, Inc. Honolulu, Hawaii, Prepared by TEC, Guam.
Dixon, B, Rudolph, T, Jones, R, Welch, D and Nelson, I. 2017. Archaeological Data Recovery in Support of the J-011B Utilities and Site Improvements at Naval Base Guam Telecommunications Site, Guam. Cardno, Guam.
Dixon, B, Schaefer, R and McCurdy, T. 2010. Traditional Farming Innovations during the Spanish and Philippine Contact Period on Northern Guam. Philippine Quarterly of Culture & Society, 38(4): 291–321.
Dixon, B, Walker, S, Schaefer, R, Whippy, J and Walker, C. 2010. Archaeological Investigations Conducted in the Territory of Guam Supporting the Joint Guam Build-up Environmental Impact Statement: Threshold Report No. 3. Prepared for Department of the Navy Naval Facilities Engineering Command, Pacific Pearl Harbor, Hawai‘i. Prepared by TEC, Guam.
Evans, D and Fletcher, R. 2015. The landscape of Angkor Wat redefined. Antiquity, 89(348): 1402–1419. DOI: https://doi.org/10.15184/aqy.2015.157
Fox, A. 2020. Amateur Archaeologists Studying Aerial Maps of the U.K. Spot Dozens of Hidden Historical Structures. Smithsonian Magazine. Available at https://www.smithsonianmag.com/smart-news/amateur-archaeologists-make-discoveries-during-lockdown-180974886/ Last accessed 25 May 2020
Freeland, T, Brandon, H, Burley, DV, Clark, G and Knudby, A. 2016. Automated Feature Extraction for Prospection and Analysis of Monumental Earthworks from Aerial LiDAR in the Kingdom of Tonga. Journal of Archaeological Science, 69: 64–74. DOI: https://doi.org/10.1016/j.jas.2016.04.011
Freycinet, LC. 1825. Voyage autor du monde par ordre du Roi, sur les corvettes de sa Majesté l’Uranie et La Physicienne, pendant les années 1817–1820. Paris: Chez Pillet Aîné, Imprimeur-libraire, 1. DOI: https://doi.org/10.5962/bhl.title.15862
Gilmore, K, Leclerc, E, Hille, P, Kurashina, H and Carucci, J. 2018. “Illuminating the Obscure: Using Legacy LiDAR Data to Define and Interpret a WWII Airfield on the Island of Tinian, Commonwealth of the Northern Mariana Islands (CNMI).” Society for American Archaeology Conference Presentation in Washington DC, 11–15 April 2018. DOI: https://doi.org/10.13140/RG.2.2.13407.10406
Gosser, D, Dixon, B, Gilda, L and Torres, T. 2004. Cultural Resources Survey and Limited Subsurface Excavations at Naval Ordnance Annex and Waterfront Annex, Territory of Guam. AMEC/Ogden Environmental and Energy Services Co., Inc., Honolulu.
Jalandoni, A. 2011a. Casa real or not real? A Jesuit mission house in Guam. Unpublished Master Thesis from the University of the Philippines. Manuscript on file, Micronesian Area Research Center, University of Guam, Mangilao.
Jalandoni, A and May, SK. 2020. How 3D models (photogrammetry) of rock art can improve recording veracity: a case study from Kakadu National Park, Australia. Australian Archaeology, 86(2): 137–146. DOI: https://doi.org/10.1080/03122417.2020.1769005
Jalandoni, A, Zhang, Y and Zaidi, NA. 2022. On the use of Machine Learning methods in rock art research with application to automatic painted rock art identification. Journal of Archaeological Science, 144: 105629. DOI: https://doi.org/10.1016/j.jas.2022.105629
Kaur, A. 2020. Senator calls for halt to buildup construction after three more burials disturbed. Pacific Daily News. 22 July. Available at https://www.Guampdn.com/story/news/2020/07/22/senator-calls-stop-buildup-construction-after-3-burials-disturbed/5484267002/ [Last accessed 22 July 2020].
Lozić, E and Štular, B. 2021. Documentation of archaeology-specific workflow for airborne LiDAR data processing. Geosciences, 11(1): 26. DOI: https://doi.org/10.3390/geosciences11010026
Montón-Subías, S and Dixon, B. 2021. Margins are central: identity and indigenous resistance to colonial globalization in Guam. World Archaeology, 53(3): 419–434. DOI: https://doi.org/10.1080/00438243.2021.1999851
Morgan, W. 1988. Prehistoric Architecture in Micronesia. Austin, Texas: University of Texas Press. DOI: https://doi.org/10.7560/765061
National Oceanic and Atmospheric Administration (NOAA). 2021. 2020 NOAA NGS Topobathy Lidar: Guam. Digital Coast Data Access Viewer. Charleston, SC: NOAA Office for Coastal Management. Available at https://coast.noaa.gov/dataviewer. [Last accessed Jul 19, 2021].
Pingel, TJ, Clarke, K and Ford, A. 2015. Bonemapping: A LiDAR processing and visualization technique in support of archaeology under the canopy. Cartography and Geographic Information Science, 42(sup1): 18–26. DOI: https://doi.org/10.1080/15230406.2015.1059171
Quintus, S, Day, SS and Smith, NJ. 2017. The efficacy and analytical importance of manual feature extraction using lidar datasets. Advances in Archaeological Practice, 5(4): 351–364. DOI: https://doi.org/10.1017/aap.2017.13
Russell, S. 1998. Tiempon I Manmofo`ona: Ancient Chamorro Culture and History of the Northern Mariana Islands. Micronesian Archaeological Survey Report 32, Saipan: CNMI Division of Historic Preservation.
Štular, B, Kokalj, Ž, Oštir, K and Nuninger, L. 2012. Visualization of lidar-derived relief models for detection of archaeological features. Journal of archaeological science, 39(11): 3354–3360. DOI: https://doi.org/10.1016/j.jas.2012.05.029
Tomonari-Tuggle, M, Rieth, T, Tuggle, D, Bell, M and Knecht, D. 2018. A synthesis of archaeological inventory and evaluation efforts on the island of Guam. Volume II: AD 1521–1950. Prepared by International Archaeology LLC, Honolulu.
Yokoyama, R, Shirasawa, M and Pike, RJ. 2002. Visualizing topography by openness: A new application of image processing to digital elevation models. Photogrammetric engineering and remote sensing, 68(3): 257–266.