AN EXAMINATION OF THE RELATIONSHIP BETWEEN 
RAPA NUI AHU AND TOPOGRAPHY USING STRUCTURE 
FROM MOTION AND GIS 
 
 
 
 
 
 
 
 
by 
ELIZA PEARCE 
 
 
 
 
 
 
 
 
 
 
A THESIS 
 
Presented to the Departments of Geography and Environmental Science 
and the Robert D. Clark Honors College 
in partial fulfillment of the requirements for the degree of 
Bachelor of Arts or Science 
 
Fall 2015 
An Abstract of the Thesis of 
Eliza Pearce for the degree of Bachelor of Arts 
in the Departments of Geography and Environmental Science to be taken Fall 2015 
Title: An Examination of the Relationship between Rapa Nui Ahu and Topography 
using Structure from Motion and GIS 
Associate Professor Mark F onstad 
Recent studies about Rapa Nui (Easter Island) have shed a new light upon the 
movement of the island's giant statues ( moai) and platforms ( ahu) showing that they 
were purposeful projects created by small communities around the island. What remains 
unclear, however, is the full meaning and purpose of the giant structures created by the 
ancient Rapanui. A current working theory is that the structures were costly signals to 
other clans on the island as a way to mark rare fresh water resources. For this study, 
imagery was collected of the south coast of the island and then used in conjunction with 
Structure from Motion to create topographic data for the area. Various GIS analyses 
were then run on different aspects of the landscape around the different types of ahu 
(those with moai and those without). None of the statistics run on the data acquired 
from running these analyses indicate any significant difference between the topographic 
placement of the two types of ahu. This lack of significance shows that topography does 
was not an influential factor in the placement of these features and whatever factors 
were involved, they did not have a topographical relationship. This means that other 
relationships like water sources were perhaps more important for ahu placement. 
ii 
  
iii  
Acknowledgements 
I would like to thank my advisors Dr. Mark Fonstad, Dr. Terry Hunt and Dr. 
Nick Kohler for taking the time to be on my thesis board and advise me during this 
process. I would especially like to thank Dr. Fonstad for his consultation and help with 
all the data. An additional thanks needs to be made to Dr. Hunt and Drs. Lipo and Lee 
from CSULB for accepting me as an undergraduate research assistant on their trip to 
Rapa Nui in early January 2015. This thesis is meant to tie into their continuing research 
about Rapa Nui and is based off of data and research conducted on the island. The work 
I have done would not have been possible without the amazing opportunity they 
provided me. 
I would also like to thank my amazing parents, Josie and Ken, who have stood 
by and supported me throughout my time here at the University of Oregon and in the 
Clark Honors College. Their help, encouragement, and engagement with all I do has 
been invaluable.  
 
A special thanks to CHC advisor Miriam Rigby for helping me through the 
logistics of the thesis defense and thank you to thank Reed College for providing their 
Thesis Template for the inspiration of many elements of this thesis template. It has 
saved me a lot of grief. 
  
  
iv  
Table of Contents 
Part 1: Introduction 1 Purpose 2 
Part 2: Background 4 Rapa Nui 4 Geography 4 History and the Conventional Narrative 6 Ahu 12 Costly Signaling and Water 18 Topography 22 Effect of Slope on Walking 23 Visibility in Archeology 25 
Part 3: Field Work and Data Analysis Methods 28 Data Collection 28 Structure from Motion and Topography 32 Topography 34 Ground Calibration 37 Processing 38 Errors 45 ArcGIS Processing 47 Ahu and Coastline Identification 49 Batch and Model Data Processing 53 Removing Extraneous Data 57 Elevation and Viewshed Analysis 60 Extracting Slope Data 64 
Part 4: Data Analysis and Discussion 67 Normalizing Data 69 The T-Test 73 Discussion 79 Errors 81 Future Research 82 
Part 5: Conclusion 84 
  
v  
Appendix A: Terms 86 
Appendix B: Tables 89 
Bibliography 110 
 
 
  
  
vi  
  
Accompanying Materials 
I have made all of my data and work files available to Drs. Terry Hunt and Carl 
Lipo in case any work I have done for this paper may be of use in their future research 
on Rapa Nui. Copies of my data and files will be kept with them in their Network 
Attached Storage. 
  
  
vii  
List of Figures  
Figure 1: Map of Rapa Nui showing basic topography and the major landmarks 4 
Figure 2: Rapa Nui Isolation 6 
Figure 2: Classic photo of just the moai heads 11 
Figure 3: Moai in Pop Culture 11 
Figure 4: Moai standing on an Ahu 12 
Figure 5. Image ahu 16 
Figure 6. Ahu 17 
Figure 7. Image Ahu with Various Iterations 17 
Figure 8. The UX5 Drone 31 
Figure 9. UX5 Flight Areas 31 
Figure 10. Final Checks of the Drone before a Launch 32 
Figure 11: Structure from Motion Diagram 33 
Figure 12: Camera Angles and Overlap in Photoscan 37 
Figure 13. Photoscan Processing: Sparse Point Cloud 41 
Figure 14. Photoscan Processing: Dense Point Cloud 41 
Figure 15. Photoscan Processing: Wireframe 42 
Figure 16. Photoscan Processing: Solid 42 
Figure 17. Photoscan Processing: Shaded 43 
Figure 18. Photoscan Processing: Texture 43 
Figure 19. Photoscan Processing: Orthophoto Output 44 
Figure 20. Photoscan Processing: DEM Output 44 
Figure 21. South Coast Orthophotos and Ahu Points 51 
Figure 22. Identifying “Image Ahu” 52 
Figure 23. Identifying “Ahu” 52 
Figure 24. Identifying “Ahu?” 53 
Figure 25. ModelBuilder Example 57 
Figure 26. Removing Extraneous Data Before 59 
Figure 27. Removing Extraneous Data After 59 
Figure 28. Elevation within 5m of Ahu 63 
Figure 29. Visual Raster Output of Observation Points 63 
Figure 30. 50m Coast Buffer Isolating Slope Data 65 
  
viii  
Figure 31. 50m Closest Coast Points Buffer Isolating Slope Data 66 
Figure 32. 50m Ahu Buffer Isolating Slope Data 66 
Figure 33. Elevation t-Test Results 76 
Figure 34. Area Visible by Ahu t-Test Results 76 
Figure 35. Distance to Closest Coast Point t-Test Results 77 
Figure 36. Slope Surrounding Closest Coast Points t-Test Results 77 
Figure 37. Slope Comparison between Image Ahu and Coastline t-Test Results 78 
Figure 38. Slope Comparison between Ahu and Coastline t-Test Results 78 
Figure 39. Slope around Ahu t-Test Results 79 
   
  
ix  
List of Tables  
Table: Skew Indicators 73 
Table 1: Basic Flight Data 89 
Table 2: Basic Ahu Data 91 
Table 3: Ahu Elevation Data 96 
Table 5: Closest Coast Points to Ahu 101 
Table 6: Data for Slope within 50m of Closest Coast Points of Ahu 103 
Table 7: Slope data for the Whole Coast Within 50m 106 
Table 8: Data for Slope within 50m of Ahu 107 
 
 
  
  
Part 1: Introduction 
In November of 2014, I was selected for an amazing opportunity to travel to the 
remote island Rapa Nui (Easter Island) as an undergraduate research assistant for 
Professors Carl Lipo and Christopher Lee from CSU Long Beach (CSULB) and Dean 
Terry Hunt of the Clark Honors College at the University of Oregon. Thus, I spent the 
first twenty days of January 2015 on an island in the Pacific helping to conduct 
research. Carl Lipo and Terry Hunt have been working on Easter Island for several 
years now and have made some ground-breaking discoveries regarding the movement 
of the giant stone statues on the island (moai) and the cultural dynamics of the island’s 
pre-historic people.1 In their continuing research on the Rapa Nui, they have been able 
to bring out students through Research Experiences for Undergraduate (REU) funding 
through the National Science Foundation (NSF) to help with various projects. 
The project I helped the professors work on while on Rapa Nui was collecting 
areal imagery of the island’s South Coast. This was done via the use of a drone and the 
result was incredibly high-resolution photos. From these images and structure-from-data 
technology, three-dimensional reconstructions of the landscape can be created along 
with composites of all the images that allow information and data to be gleaned 
regarding the topography of the ground and features upon it.  
Using this collected data, I was able to search out and identify pre-historic man-
made features such as the moai and ahu. Moai are the famous stone statues placed 
around the island and ahu are long, low platforms made of stone that were built by the 
                                                        1 Hunt, Lipo. The Statues that Walked. 2011 
 
 
2  
prehistoric Rapanui. Some ahu have moai placed on top of them (image ahu) and there 
are also others that do not (ahu). While most of the moai are mapped and documented, 
ahu without associated statues are more likely to be accidentally overlooked from a 
ground perspective since they can simply look like a pile of stones. It is much easier, 
however, to identify their unique, long form from aerial photographs. In addition to 
locating these features, my study focuses on their locational relationship to the 
topography of the area.  
Purpose 
Almost all ahu and finished moai are found close to the coast. While it is 
sometimes thought that the moai face outwards towards the ocean (as a means of 
protection and look out), in reality they all face inwards towards the land. This is 
because rather than warriors or guardians, the statues represent ancestors meant to 
watch over and be respected by the people.2 For this study, I analyzed the relationship 
between the locations of ahu with their surrounding topography in order to see if there 
is a significant relationship between the two; to see if topography was a deciding factor 
in terms of where the ancient Rapanui decided to place and build these giant 
architectural feats, whether building locations were chosen randomly along the coastline 
or not.  
This is an important and relevant study because not much is known about the 
spatial reasoning behind why various features are placed where they are. A significant 
relationship between low topography and anthropogenic features would help support 
                                                        2 Hunt, Lipo. 2011 
 
 
3  
ideas that the ahu and moai serve additional functionality rather than simply being 
ancestors to respect. Drs. Hunt and Lipo have a working hypothesis that the statues and 
platforms relate to freshwater and other resources. Proving coastal access is near these 
sites, would thus tie my research into further data being collected and studied about 
coastal freshwater. Lower coastal elevation and slope generally indicate better ocean 
access, features placed in such locations could have been demarcations or claims to the 
area and the resources such an area provided.  
In order to fully analyze and understand this relationship, I asked a few broad 
questions: How are ahu structures positioned on the landscape? Is there a locational 
difference between ahu and image ahu structures that denotes a difference in purpose? 
Are these features near where water resources are more accessible? Are these features in 
places that make them more visible? These broad questions I later broke down into 
seven more specific questions related to the tests I was running: 
1. Are image ahu at higher elevations than ahu? 
2. Do image ahu have a larger viewshed than ahu? 
3. Are image ahu closer to the coast than ahu? 
4. Is the coastal slope near image ahu less steep than the coastal slope near ahu? 
5. Is the coastal slope near image ahu less steep than the general slope of the 
coast? 
6. Is the coastal slope near ahu less steep than the general slope of the coast? 
7. Is the ground slope around image ahu steeper than the ground slope around 
ahu? 
I used all of these questions to guide my research and methods in examining the 
relationship between the ahu and topography on Rapa Nui.   
 
 
4  
Part 2: Background 
Rapa Nui 
Geography 
 
Figure 1: Map of Rapa Nui showing basic topography and the major landmarks 
Source: http://www.crystalinks.com/easterisland.html 
Halfway between Chile and Tahiti Located lies the island of Rapa Nui, or as 
Europeans called when first landing there, Easter Island. Rapa Nui is small, only about 
63 square miles, and one of the farthest, most remote reaches of the world; Tahiti is 
about 2,500mi to the west, Chile about 2,300mi to the east and the closest inhabited 
land is Pitcairn Island, approximately 1,300 miles away (fig 2).3 Like most islands in                                                         3 "Easter Island Travel Guide." Travel Guide to Easter Island, South Pacific 
 
 
5  
the Pacific, Rapa Nui is volcanic and the three points of its roughly triangular shape 
consist of the extinct volcanoes whose activity formed the island: Rano Kau, Poike, and 
Terevaka (fig. 1).  The general landscape is that of rolling grassy hills interspersed with 
rough lava fields and old volcanic craters and parasitic cones. Given the island’s 
volcanic origins, the island is mainly basalt and andesitic rock as well as some scoria 
and obsidian. The soil on the island consists mainly of loams and clays.4 
Just south of the Tropic of Capricorn, Rapa Nui’s climate is subtropical with an 
annual average temperature of 68.5°F and some rain falls 140 days out of the year. 
March to June are the rainiest months and August to December the driest but heavy 
rainfall, often in squalls, can occur any time during the year.5 While the island sees a 
good amount of rainfall throughout the year, it is quickly absorbed by the porous 
volcanic bedrock so the land rarely remains wet. There are also no permanent natural 
sources of freshwater on the island; no perennial rivers and only two lakes, both 
standing water and located in the craters of volcanoes, though stagnant water can also 
collect in lava tubes. 
Unlike many tropical islands, Rapa Nui has no barrier reef. This leaves the 
island exposed to the ocean whose rough seas have, over time, created cliffs along much 
of the shore line. There are three sandy beaches on the island but all are on the two 
more northern sides of the island. The southern coast has spots where coastal access is 
easier but these remain rocky. 
                                                        4 "About Rapa Nui." Easter Island Statue Project Official Website RSS. 5 "Easter Island Travel Guide." 
 
 
6  
 
Figure 2: Rapa Nui Isolation 
This image shows Rapa Nui (Easter Island) and its distance to other inhabited lands. 
The island labelled “Rapa” refers to Rapa Iti not Rapa Nui.  
Source: http://www.bibliotecapleyades.net/arqueologia/eastern_island/easter01.htm 
 
History and the Conventional Narrative 
Despite the island’s isolation, Rapa Nui is home to some of the most fantastic 
prehistoric architectural feats known, the moai. There are hundreds of these large 
statues placed all around the island, with each one carved from a quarry on the side of a 
volcano called Rano Raraku and then transported, sometimes all the way across the 
island, until they reached their final permanent destination (fig. 1).  
 The earliest radio carbon dating places initial settlement of Rapa Nui by 
Polynesians around 1200 CE and Europeans first came in contact with the island in 
(RapanNui) 
 
 
7  
1722 when the Dutch explorer Jacob Roggeveen stayed for a few days on the island.6 
Roggeveen and subsequent explorers and scientists observed the large statues and 
relatively small number of indigenous people (a few thousand) and wondered how such 
monumental structures could have been created.7 As the chief pilot of Cpt. Don Felipe 
González’s 1770 voyage wrote: 
“That a people lacking machinery and materials for constructing any 
should be able to raise the crown or headpiece on to a statue of such 
height causes wonder, and I even think that the stone of which the statues 
are made is not a product of the island, in which iron, hemp, and stout 
timber are absolutely unknown. Much remains to be worked out on this 
subject.”8   
The base assumption underlying this and following ideas about the statues, is that 
ingenious tools and a large population must have been needed to create them. Thus, 
there must have been must have once been a great, ancient Rapanui civilization which, 
after having made hundreds of statues, for some reason collapsed consequently when 
Europeans arrived they encountered the “depleted” population. This in turn has given 
rise to speculation as to what caused this population collapse. Some look to various 
aspects of native oral traditions or use pieced bits or evidence while others like go to the 
extreme to explain moai. 
The idea that no one knows how the moai were constructed and moved or by 
whom has resulted in the majority of those uneducated in the subject believing that 
Easter Island is a bare and desolate place with no people, just statues. The truth, 
however, is that there has been continuous occupation of the island throughout its 
                                                        6 Corney; Roggeveen. The Voyage of Captain Don Felipe González. 1908 7 Corney; Roggeveen. The Voyage of Captain Don Felipe González. 1908 8 Corney; Felipe González De Haedo. The Voyage of Captain Don Felipe González. 1908 
 
 
8  
known history, and while its population was diminished due to disease and slaving to a 
mere 110 people at one point, there are still decedents of the indigenous people living 
on the island today.9 There is a common misconception among the general public that 
the Moai are simply large stone heads due to years of their depiction in popular culture 
as such. This idea has been propagated by the classic stock image from the quarry Rano 
Raraku (fig. 3), where the statues were carved, as well as their portrayal as simply heads 
in various media forms, from comics to cartoons to movies (fig. 4). In reality though, 
the statues are full-bodied and when properly placed, stand upright and erect (fig. 5).  
While there have been many different narratives regarding the island’s pre-
history over the years, most have since been dismissed by the general public, that the 
statues were made by aliens10 or that they were the work of white Egyptians and 
Incans11. In their place is the current conventional narrative of the island, an idea 
presented by Jared Diamond in his book Collapse. By his account, when the 
Polynesians first arrived on it Rapa Nui was a lush paradise and a prosperous 
civilization was established. As time went on, a ruling class forced their people to create 
moai in order to honor the ancestors and encouraged a so-called moai cult of intense 
building competition. In order to build and move these moai wood and plant materials 
were required and so the Rapanui cut down all of the trees that once existed on the 
island. With no more trees, the topsoil washed away and food became scarce. The 
Rapanui were then unable to sustain themselves on the island and they turned to 
                                                        9 Hunt, Lipo. 2011 10 Däniken. Chariots of the Gods?: Unsolved Mysteries of the Past. 1970 11 Holton "Heyerdahl’s Kon Tiki Theory and the Denial of the Indigenous Past.” 2004 
 
 
9  
cannibalism as their civilization collapsed. They caused their own demise through 
ecocide.12 
Diamond’s theory, like the ones before, though, still labors under the assumption 
that there must have once been a great civilization under which the moai were 
constructed that disappeared before the Europeans arrived. There is a new explanation 
as to how the moai were moved, however, that goes against this idea. Using the physical 
characteristics of the moai, Drs. Carl Lipo and Terry Hunt have constructed a new fact-
based and clarified narrative for the island, one where the pre-historic population 
remained stable and that the populace encountered by the Europeans was, in fact, a 
healthy and thriving community, one which only began to collapse post-contact due to 
disease and slavers. Lipo and Hunt were able to show that moai were not moved in a 
completely finished state. Rather, when the moai were initially carved, they had a much 
further forward center of gravity. The resulting forward lean meant that through the use 
of three ropes tied around the head area, if the moai was rocked back and forth while 
twisting it forward, it could be moved by relatively few people.13 If the moai could be 
moved this easily, there would have been no need for a big population to supply the 
laborers needed to move the statues and thus there would be no need to try to explain 
some sort of disappeared society. With this revealed, the changes in landscapes and land 
cover become less important to the overall story, with palm tree loss being explained by 
the introduction of rats that ate the palm seeds and saplings instead of clearing for 
                                                        12 Diamond. "Twilight at Easter." Collapse: How Societies Choose to Fail or Succeed. 2005 13 Hunt, Lipo. 2011 
 
 
10  
agriculture by the natives.14 This is supported by the observations of Jacob Roggeveen 
in 1772: 
“Nor can the aforementioned land be termed sandy, because we found it 
not only not sandy but on the contrary exceedingly fruitful, producing 
bananas, potatoes, sugar-cane of remarkable thickness, and many other 
kinds of the fruits of the earth; although destitute of large trees and 
domestic animals, except poultry. This place, as far as its rich soil and 
good climate are concerned, is such that it might be made into an earthly 
Paradise, if it were properly worked and cultivated; which is now only 
done in so far as the Inhabitants are obliged to for the maintenance of 
life.”15  
Hunt and Lipo’s explanation shows that the indigenous Rapanui were smart and well 
aware of their surrounding environment, unlike their depiction in previous narratives. 
Rather than labor being forced upon people, Hunt and Lipo show that moai making and 
moving was a group activity of choice and there was deliberation behind the making 
and moving of each one.   
                                                        14 Hunt, Lipo. 2011 15 Corney; Roggeveen. 1908 
 
 
11  
 
Figure 2: Classic photo of just the moai heads 
Buried moai in Rano Raraku.  Photo taken by author 
 
Figure 3: Moai in Pop Culture 
From left to right: depictions of moai in the movie Night at the Museum, The Simpsons, 
and a Batman comic. Source: http://www.moaiculture.com/popculture.html 
 
 
12  
 
Figure 4: Moai standing on an Ahu 
Ahu Nau Nau. Once scattered, this ahu has been reconstructed and these maoi with their 
pukau re-erected to stand on top as they once did.  Photo taken by author 
 
Ahu 
While the moai are generally the first thing that comes to mind when “Easter 
Island” is mentioned, there are also other large archeological structures that can be 
regularly seen throughout the landscape, ahu. These are long low platforms upon which 
the moai were placed. These places are considered to have once been ceremonial 
locations and gathering places for the prehistoric Rapanui, community locations for 
local “clans” or family groups. Rather than simply being ceremonial sites, people would 
live and farm in the areas around the ahu with the large stone statues placed with their 
backs to the sea and looking down on them. Even with this understanding, the ahu and 
their purpose could have meant many things to the ancient Rapanui and whatever the 
 
 
13  
original intent, it is well documented that burials took place there at least during 
historical times. As Captain Cook, leader of the third European party to disembark on 
Rapa Nui, noted in 1774: 
 “The gigantic statues, so often mentioned, are not, in my opinion, 
looked upon as idols by the present inhabitants, whatever they might 
have been in the days of the Dutch; at least I saw nothing that could 
induce me to think so. On the contrary, I rather suppose that they are 
burying-places for certain tribes or families. I, as well as some others, 
saw a human skeleton lying in one of the platforms, just covered with 
stones.” 16 
Visitors to the island continued to note the remains found near ahu and a century later in 
1886, William Thompson, the paymaster of the U.S.S. Mohican recorded this 
description of ahu: 
“The platforms differ greatly in dimensions, but the general plan and 
characteristics are inevitably the same. Many of them are in a fair state of 
preservation, except that the images have been thrown down and the 
terraces in the rear obliterated or strewn with rubbish, while others have 
been reduced to a state of complete ruin. The platforms are usually 
located near the beach, and on a high bluff some of them are quite near 
the edge, overlooking the sea. The general plan consists of a front 
elevation composed of blocks of stone fairly squared and neatly fitted 
together without cement, a parallel wall forming the inside boundary, 
built of uncut stone, inclosing small chambers or tombs placed at 
irregular intervals. [Thompson’s reference’s to “front” and “rear” are 
opposite as, when confronted with all of the maoi fallen over, he 
assumed that the moai faced outwards towards the sea when they’d 
originally stood rather than inland] Loose bowlders [sic] fill the spaces 
between the tombs and form the horizontal plane of the platform, into 
which are let the rectangular stones which constituted the base upon 
which the statues stood. The façade stones are large and heavy, and in 
some cases the smooth surface presented could not well be attributed to 
the ride implements at the command of the builders and must have been 
produced by friction or grinding. Long wings composed of uncut stone 
extend from the platform proper, built up to the summit at the ends. In 
the rear of the platform a few steps descend to a gently sloping terrace, 
                                                        16 Cook and Furneaux. A Voyage towards the South Pole and round the World. 1777 
 
 
14  
which terminates in a low wall and is bounded by a squarely built wall 
raised above the ground so as to join the top of the platform.”17 
William Thompson is considered to have conducted the first archeological investigation 
on the island during his two week stay and many others would follow, noting the ahu, 
describing their features and categorizing them. There remains a general similarity 
throughout these descriptions, all noting the same general rectangular form and winged 
ends with the large wall in back and sloping ramp in front (fig. 5).18  
While ahu that supported moai (image ahu) were the main subject of 
investigation and note for those initial explorers and archeologists to the island, there 
are structures on the Rapa Nui that have a largely similar basic shape to the image ahu, 
being low and long, but are generally simpler in their construction and have no 
associated statues. These ahu are also found near the coast but they tend to have a much 
greater variation in size, and, rather than the wings and squared ends of the image ahu, 
their ends tend to taper (fig. 6). Sometimes there is a front slope and strong back wall 
associated with them but this is generally not the case. Captain Cook described some of 
these features in his journal: 
“Besides the monuments of antiquity, which were pretty numerous, and 
nowhere but on or near the sea-coast, there were many little heaps of 
stones, piled up in different places along the coast. Two or three of the 
uppermost stones in each pile were generally white, perhaps always so, 
when the pile is complete. It will hardly be doubted that these piles of 
stone had a meaning; probably they might mark the place where people 
had been buried, and serve instead of the large statues.”19 
                                                        17 Thompson. Te Pitot e Henua, or Easter Island. 1886 18 Beardsley. Spatial Analysis of Platform Ahu on Easter Island. Dissertation. 1990 19 Cook and Furneaux. 1777 
 
 
15  
As Cook noted, it is impossible to understand exactly what these different forms of ahu 
meant to the Rapanui but they undoubtedly served some purpose.  
Some archeologists have tried to classify Rapa Nui prehistory into various 
periods and fit various types of ahu construction within these time frames but in their 
study of the inland image ahu complex of A Kivi-Vai Teka, Mulloy and Figueroa found 
that: 
“From the point of view of image ahu architecture, a single, coherent, 
continuously developing pattern of ideas is represented. In terms of 
general conception and apparent cultural function as well as detailed 
architectural characteristics this sequence of structures demonstrates a 
clear and detailed, unbroken chronological progression such as might be 
expected from the architectural reflection of the activities of a single 
continuously developing society. No evidence is interpreted as revealing 
a chronological break in the sequence such as might suggest a population 
replacement, the intrusion of a new cultural pattern or even a period of 
sudden cultural renaissance. The evidence indicates that, from the point 
of view of image ahu architecture, this part of the local history can most 
meaningfully be seen as a single period of uninterrupted development 
characterized by gradual introduction of new ideas, the expansion of 
themes and improvement of capacities.”20 
Indeed, several ahu, particularly image ahu, show signs of reworking, rebuilding and/or 
maintenance taking place since their initial creation. This can be noticed when there are 
square-ended outlines indicating where a previous ahu was once built or if an ahu 
angles in the middle rather than being straight (fig. 7). In addition, some of the large 
blocks used as the outside walls of the ahu can be identified as the heads and bodies of 
moai that presumably fell, broke and were then repurposed. These maoi generally have 
the rounder, less long head shape that is characteristic of older moai styles. It should be 
                                                        20 Mulloy and Fueroa. The A Kivi-Vai Teka Complex and its Relationship to Easter Island Architectural Prehistory. 1978 
 
 
16  
noted, however, that the simple lack of discernable difference between features does not 
indicate that no timeline or chronological difference exists. 
 
 
 
 
 
 
Figure 5. Image ahu 
The central part of the image ahu is thicker, more solid, and rectangular and is where 
the moai (now fallen) once stood. To either side stretches out the squared “wings” 
while in front is the sloping ramp where evenly spaced smooth, rounded rocks were 
placed. A historic wall has been made using/intersecting the image ahu.  
 
 
 
17  
 
Figure 6. Ahu 
A well preserved ahu with no associated moai. The construction is still the same with a 
retaining outer-wall of larger stones, but the ends of the structure are much pointier and 
there are less associated defining factors.  
 
Figure 7. Image Ahu with Various Iterations 
This image ahu has had several stages to it. The front rectangular outline and wings are 
clear as are another set behind and at a slightly different angle.  
 
 
18  
Costly Signaling and Water 
The construction and moving of a moai would have been a long term and 
significant event given their size and the distance they were moved to locations around 
the island. While no one can know exactly the rationale behind the Rapanui’s 
construction of the enormous moai and ahu, one concept that could have been an 
underlying driver to their construction is costly-signaling. 
Costly-signaling is the notion that large structures and other “frivolous” visible 
possessions indicate a level of wealth, health, or status because those in possession of 
such signals must be able to handle the “cost” placed upon the individual. This cost is 
often the diversion of energy to create the feature, like with peacocks putting energy 
into growing their elaborate tails rather than getting bigger themselves (it does not have 
to be a conscious decision). In modern human societies, the cost is often monetary as 
people buy objects that have little functional value except to display the wealth of the 
owner.21 These costly signals are generally visual and moai and ahu would fall into this 
category, their size and scale indicating to others that those involved in their 
construction had enough resources and “wealth” to deal with the cost that came from 
erecting such features.  
The moai could also be involved in costly-signaling by delineating members of a 
group. As Smith and Bliege say, “another type of collective good that may be a form of 
costly signaling involves punishing those who free-ride on the group’s cooperative 
activities or otherwise violate group-beneficial norms.”22 There were once several 
                                                        21 Hunt, Lipo. 2011 22 Smith and Bliege Bird. "Costly Signaling and Cooperative Behavior.". 2005 
 
 
19  
different family “clans” living in around Rapa Nui that were associated with specific 
areas of the island.23 In talking to Dr. Lipo, he is of the opinion that rather than a top-
down order to build moai involving specific carvers and transportation laborers (as 
some narratives say), these groups of people would carve, move the moai, and build the 
ahu themselves in addition to maintaining their other sustenance related activities.24 If 
this were indeed the case, such action and involvement in the endeavor would help to 
show who was involved, invested, and a part of that particular family group. Erecting a 
moai would make it clear who was to partake in the “clan’s” resources as opposed to 
outsiders and freeloaders.  
In regards to what the moai and ahu could be signaling, with Rapa Nui’s small 
size it makes sense that they could have been signaling the possession of some resource. 
Since lithic mulching was used as a farming method, food resources were spread out 
across the whole island.25 Thus, as a resource, land and food were not the limiting 
factor. Rather, it would make sense to place these statues near a scarcer resource, a 
resource such as water.  
Given the volcanic geology of the island, there is little to no standing water on 
Rapa Nui. This is a problem. There are basins carved into the rock called taheta26 to 
help hold rainwater, but this is not enough. For a society to survive, a more reliable 
source of water is needed. There is evidence, both historical and archeological that the 
                                                        23 Tilburg. Among Stone Giants. 2003 24 Talk with Carl Lipo January 2015 25 Ladefoged. “Soil Nutrient Analysis of Rapa Nui Gardening.” 2010 26 Tilburg. Among Stone Giants. 2003 
 
 
20  
wells were constructed by the ancient Rapanui. As Captain James Cook described in his 
journal during his 1777 visit to the island:  
“[My men] could find no water except what the natives twice or thrice 
brought them, which, though brackish and stinking, was rendered 
acceptable, by the extremity of their thirst…Towards the eastern end of 
the island, they met with a well whose water was perfectly fresh, being 
considerably above the level of the sea; but it was dirty, owing to the 
filthiness or cleanliness (call it which you will) of the natives, who never 
go to drink without washing themselves all over as soon as they have 
done; and if ever so many of them are together, the first leaps right into 
the middle of the hole, drinks, and washes himself without the least 
ceremony… What the natives brought them here was real salt water; but 
they observed that some of them drank pretty plentifully of it, so far will 
necessity and custom get the better of nature!... On the declivity of the 
mountain towards the west, they met with another well, but the water 
was a very strong mineral, had a thick green scum on the top, and stunk 
intolerably. Necessity, however, obliged some to drink of it; but it soon 
made them so sick, that they threw it up the same way that it went 
down.”27 
It is expected that these are puna, wells where the ground has been cut away into the 
side of the slope in order to access the water table. It is obvious, however, given the 
brackish nature of the proffered water that many of these wells were near the coast 
where the sea water mixes with the water table. 
 It makes sense geologically that it would be easier to access freshwater near the 
coast because as the land comes down to meet the ocean, the water table is relatively 
closer to the surface. At the shoreline where they meet, fresh water discharges into the 
ocean but the volume depends on the tide. At high tide there is little discharge as the 
increasing tide creates a “hydraulic dam” blocking the fresh water, but as the tide ebbs 
the position of maximum discharge moves towards the ocean.28 The result is less salty 
                                                        27 Cook and Furneaux. 1777 28 Urish and McKenna. “Tidal Effects on Ground Water Discharge Through a Sandy Marine Beach.” 2004 
 
 
21  
brackish water. Given the water shortage on the island, it makes sense that the Rapanui 
took advantage and made the most of this coastal freshwater resource. In his journal 
Captain Cook described one such well: 
“The little [water] we took on board, could not be made use of, it being 
only salt water which had filtered through a stony beach into a stone 
well; this the natives had made for the purpose, a little to the southward 
of the sandy beach so often mentioned, and the water ebbed and flowed 
into it with the tide.”29 
The Rapa Nui locals would have been well adapted to drinking brackish water as Cook 
observed and this geological phenomena can still be observed today when free roaming 
horses can be seen drinking ocean water at low tide. This does not happen everywhere 
though. For one, the water table discharges best through porous materials so while soil 
and sand allow for a lot of permeability, bedrock and clay are more impermeable and 
will generally restrict the water table. Thus, not all areas of a coastline will have the 
same fresh water discharge. In addition, the cliffy coastline makes it certain areas less 
accessible than others.   
 Ocean access would have made available valuable resources for the ancient 
Rapanui, not only water but also whatever food resources they could glean from it 
through fishing and harvesting.30 It thus makes sense to mark and “protect” such 
locations, to identify and keep them for yourself and your immediate clan. Based off of 
diatoms in skeletons, there is evidence of regional geographic variability in the water 
sources of the ancient Rapanui that could be explained by a differing reliance on water 
sources, with individuals from the north and west coasts of the island relying on more 
                                                        29 Cook and Furneaux. 1777 30 Arana, "Ancient Fishing Activities Developed in Easter Island." 2014 
 
 
22  
temporary rainwater drinking sources while south coast individuals had more 
diversified sources.31 My paper examines the relationship between ahu and coastal 
access in an attempt to shed further light on this relationship and support the idea of 
coastal water table discharge as an important resource for those Rapanui living on the 
south coast of the island.  
 
Topography 
If ahu are to be considered costly signals related to fresh water access within the 
ancient Rapanui culture, topography could play an important factor as to why these 
structures were placed in certain locations. There are a couple reasons for this. One, as 
partially explained in the prior section on costly signaling and water, has to do with 
access to the water table. If a structure is placed at a lower elevation it is placed close to 
the water table. Additionally, a steep decline would not only bring people closer to the 
water-table, but these downhill areas, specifically if there is an abrupt drop in elevation, 
would remain slightly damper. This is because as water percolates through the ground, 
gravity still acts upon it, pulling it in a downhill direction in addition to simply down 
into the ground. If the land falls sharply, it is possible for this percolating water to 
emerge in these locations, perhaps not in large quantities, but more so than in areas 
where the land is flat or uphill.  
Ahu location and distance to resources could also result in a spatial relationship. 
It is generally expected that features related to one-another would be located closer                                                         31 Dudgeon and Tromp. “Diet, Geography and Drinking Water in Polynesia: Microfossil Research from Archaeological Human Dental Calculus, Rapa Nui (Easter Island).” 2012 
 
 
23  
together rather than being placed far apart. This is especially true when the resource 
being marked is an environmental one, such as land, food, or water. Thus, if ahu were 
markers of some locational based resource they should be placed near that resource. 
Topography can relate to a number of other factors relevant to ahu placement as 
well, particularly slope and visibility. 
Effect of Slope on Walking 
Creatures use energy every day, to breath, to grow, to move; life requires 
energy. The more energy a creature uses, the more energy it needs to put in. Thus, to be 
efficient, many creatures will avoid doing unnecessary work. As animals move through 
a landscape they lose energy but how much they lose depends upon the terrain; the 
more incline, the more energy spent in trying to ascend.32 This is illustrated by 
Naismith’s Rule, a rule of thumb proposed in the 1800s as a way to calculate 
walking/hiking times. In its initial form it suggested to allow 20 minutes for every mile 
and 30 minutes for every 1000 feet of rise. Overtime, it has been modified and 
expanded to take into account various terrain and slope gradation since steeper, rougher 
trails are harder, requiring more time and energy to ascend than the work needed for 
walking along flat ground.33  
Observed animal behavior in the wild supports the idea that slope dictates 
movement, especially for larger animals. This makes sense since, while it may take the 
same amount of energy to lift 1kg of body weight for all animals, a larger body means 
more energy expenditure. Thus, the fight against gravity is much more taxing on                                                         32 Hausken. 2014 33 Norman. 2004 
 
 
24  
animals of larger size. This is why the smaller mouse can more easily run uphill and at 
greater inclines than the larger chimpanzee.34 This behavior is also seen in elephants 
who avoid steep slopes since their large body sizes make going up even minor hills a 
large energy cost.35 Cattle too seem to generally prefer grazing on land with less slope, 
and only venture to hiller locals when food is scarce.36 With humans, in order to 
compensate for steepness, we ascend in a zigzagging manner as a way to reduce 
steepness (even though it increases distance). This phenomena is made permanent in the 
switchbacks of roads and trails.37 We, like many animals of decent size, rarely scale 
straight up a slope; rather we traverse it.  
Given this understanding, if people were regularly accessing the coastline, it 
would be advantageous to position one’s self so as to avoid steep coastal terrain that 
would require excessive work. When access is desired, a lesser slope is preferred 
because it would make the work less strenuous and more efficient. Thus, it would make 
sense for ahu structures related to coastal/resource access to be situated near areas 
where there is a gentler coastal slope. Even without taking into account possible fresh 
water resources at the tideline, the ocean also had a number of other resources that the 
ancient Rapanui utilized like fish and other forms of sustenance. Since ahu, particularly 
image ahu, seem to be tied to community gathering areas it seems reasonable that they 
would be placed in areas where that limited the amount of work required; rather than 
                                                        34 Taylor, et. al. "Running Up and Down Hills: Some Consequences of Size.". 1972 35 Wall, et. al. "Elephants Avoid Costly Mountaineering." 2006 36 Harris, et. Al. “The Effect of Topography, Vegetation, and Weather on Cattle Distribution at the San Joaquin Experimental Range, California.” 2002 37 Llobera and Sluckin. "Zigzagging: Theoretical Insights on Climbing Strategies." 2007; Gilks and Hague. "Mountain Trail Formation and The Active Walker Model." 2009 
 
 
25  
locate their communities near areas where steep cliffs would impede access to the 
ocean’s resources and require more energy expenditure, areas with less steep coastal 
slopes seem more ideal. There were no beasts of burden on the island prehistorically, so 
the Rapanui walked everywhere; it makes sense that they would be efficient about it. 
Visibility in Archeology 
Visibility and elevation go hand in hand. Everyone knows that generally, the 
higher you are, the further you can see. Usually, for an observer standing on the ground, 
the true horizon (the horizon resulting due to the curvature of the earth) is about three 
miles in the distance, but as they ascend in elevation, that distance increases. In reality, 
though, an observer cannot always see to the horizon, if features are in the way, be it a 
building, a mountain, or something else, an observer’s visibility is blocked as sight 
beyond the object is restricted. Conversely, if there are features of greater elevation far 
away their tops can be seen by an observer if they rise above the horizon line even if 
their bases below the horizon are not visible. This is why visibility is closely tied to 
topography, not only does the elevation of the observer matter, but the elevation of the 
landscape around them as well. Standing on a lone hill surrounded by plane has a 
different visibility result than standing on a hill surrounded by mountains. 
Viewsheds are a way of analyzing how much area is visible from a single 
location. This term to describe calculating areas of visibility was first introduced by 
Tandy in 1967 who compared the idea to that of a watershed.38 The concept grew in 
popularity when a computer program was developed that could quantifying visible areas 
                                                        38 Tandy. "The Isovist Method of Landscape Survey." 1967 
 
 
26  
across a plane in 1968.39 The modern versions of this program are now used in a wide 
variety of fields from architecture to the military. 
In archeology, visibility and intervisibility (seeing between features) have long 
been considered important factors in the initial construction of archeological 
landscapes, especially in regards to settlements. While there are many elements that 
factor into the location of settlements, visibility is often analyzed as a means of 
assessing defensiveness. Before the advent of computers, this was hard to quantify and 
generally noted qualitatively though words such as ‘hidden’ and ‘prominent’ but with 
GIS, quantitative analysis allows for a more thorough study of this phenomena.40 In 
their 1996 paper, Lock and Harris examined settlements in prehistoric southern England 
and used the amount of visible land from each location as a measure of defensibility, 
with the assumption that the greater the visible area from a settlement location, the 
easier it would be to defend.41 Jones also used visibility as a way to examine the 
placement of Onondaga Iroquois settlements and found that while settlements were not 
placed for a maximum field-of-view, they were generally within line-of-sight from one 
another making communication and mutual defense easier.42 
Though a large focus has been on settlements, there have also been a number of 
viewshed analyses of other archeological features. One very extensive study was 
conducted by Fraser on the Neolithic stone cairns of Orkney where visibility was 
                                                        39 Amidon et. Al. “Delineating landscape view areas...a computer approach.” 1969 40 Wheatly and Gillings. “Spatial Technology and Archeology.” 2002 41 Lock, Gary R., and Trevor M. Harris “Danebury Revisited: An English Iron Age Hillfort in a Digital Landscape.” 1996 42 Jones. “Using Viewshed Analysis to Explore Settlement Choice: A Case Study of the Onondaga Iroquois.” 2006 
 
 
27  
studied as a significant factor for the position of the territory delineating cairns.43 
Another study by Paliou used a visibility analysis to show how Bronze Age Aegean 
wall paintings would appear to the public outside looking in through windows as a way 
to show status.44 All these studies show how the informative light visibility sheds on the 
relationship between features and their surrounding topography allows for a better 
archeological understanding of civilizations. 
This idea is especially important if these ancient structures are meant to be a 
form of costly signaling. Costly signals are a way of communicating, they are meant to 
be seen, which means that visibility should have been an important factor considered by 
those constructing the ahu, particularly the image ahu.  
                                                        43 Fraser. Land and Society in Neolithic Orkney. 1983 44 Paliou. “The Communnicative Potential of Theran Murals in Late Bronze Age Akrotiri.” 2011 
 
 
28  
Part 3: Field Work and Data Analysis Methods 
Data Collection 
For the first two and a half weeks in January of 2015, I traveled to Rapa Nui 
with Drs. Lipo and Lee from California State University Long Beach (CSULB) in order 
to help them collect data on the island. Their main goal for the trip was to collect visual 
imagery for the south coast of the island using drones and I assisted them in this 
endeavor. We flew a Trimble UX5 Aerial Imaging Rover to collect our data (fig. 8) that 
was generally able to capture areas of about 1km2 per flight. Over the course of a week 
and a half we flew about 26 flights, capturing the whole of the southern coast as well as 
some more inland areas (fig. 9).  
Inside the body of the drone was a camera that took pictures at regular intervals 
during the plane’s flight path; Dr. Lee would adjust the settings to account for the sun 
brightness of each individual flight. The unmanned aircraft was controlled via a touch-
screen Yuma tablet computer onto which a map of the current location was loaded 
beforehand. With this map and our location locked on thanks to satellites, I would then 
indicate the wind direction (important for take-off, flight planning, and landing), decide 
the coverage location and area for the flight, and confirm and adjust the take-off and 
landing approaches for the drone by establishing boxes, points, and vectors in relation 
to map seen on the handheld computer screen. Before launch, we would go through a 
pre-flight checklist to ensure readiness and that everything was set (fig. 10). To launch, 
the UAV was placed upon a slingshot-like launch rail and at takeoff, a loaded bungee 
would fling the plane into the air. This speed and increase in altitude would start the 
 
 
29  
motor and the plane would begin to fly along its directed path. If the flight area was far 
away, Dr. Lipo and I would follow the drone in a car with the Yuma computer to ensure 
a signal between the devices was maintained. When the drone reached the area 
designated for imaging, it would begin to travel along designated flight paths that 
transected the square area. In general we would try to have the flight lines perpendicular 
to the wind so as to reduce velocity change as the plane traversed across the set 
coverage area taking pictures. When the plane reached the end of each flight line, it 
would continue for a little bit not taking photos before banking into a turn and coming 
back to align itself for its next flight path. When it was done flying all the flight paths 
for the assigned area, the drone would return to a set rally point near its landing 
location, circle to lose altitude and await confirmation to land. We would position 
ourselves to observe the area and descent approach before I gave the OK to land and 
watch then watch for any danger that would have me abort the landing (horses, cars, a 
bad angle, etc.).  
There was a variable range as to the exact landing spot due to satellite 
inaccuracy and the wind could also seriously affect the landing location. Since the UX5 
has no wheels and instead lands on its belly, its last few moment of flight are glide with 
one sharp reverse from the propeller. This means that a large, clear, rock-free area is 
needed for landing and given the rocky nature of Rapa Nui that is an uncommon thing, 
but we were able to find three decent locations for landings from which we launched 
most of our flights to cover the south coast. We had brought along extra bodies for the 
drone in case of accidents and had several very close calls with rocks but only ended up 
crashing once on a landing bad enough to warrant a body change. 
 
 
30  
 Each flight produced several hundred overlapping photos along with data from 
the flight path, which, among other things, included GPS coordinates for each picture. 
After each flight we would download the pictures from the camera’s memory card onto 
a laptop as well as extract the flight log from the drone’s body onto the Yuma computer. 
This data we would then bring back to the house we were staying at for further saving 
and processing.  
In general, we flew about two to three flights per day and towards the end when 
we were most efficient, we had a day where we flew four flights, and a day we flew 
five. Ideally we could have gotten more flights off but we were limited by only having 4 
lithium batteries (1 needed per flight) and the fact that we were not allowed to fly from 
around 11am until 1pm because we weren’t allowed to fly when the airport was busy 
and that’s when the flight from Chile arrived and departed. It also took us a while to 
achieve the necessary local permission to start flying so despite arriving on January 2nd, 
we didn’t start drone flights until January 9th. Despite that, we were still able to get 
imagery for the whole south coast as well as some other flights people requested. 
The data from each flight was taken and processed using the laptops set up back 
at the house we were renting. First, the pictures from the camera had to be matched with 
their flightlog metadata that was recorded by the drone as the pictures were being taken. 
Then, we input the photos into a structure from motion program called Agisoft 
PhotoScan which would analyze the images and composites them together. The result 
of this process is an Orthophoto with approximately 3cm resolution, a Digital Elevation 
Model (DEM), and a three-dimensional point-cloud model of the area. 
 
 
31  
 
Figure 8. The UX5 Drone 
Photo taken by author 
 
Figure 9. UX5 Flight Areas 
Each green square represents the area flown for one flight. While the flight area was a 
rectangle, the resulting orthophotos have jagged edges resulting from the corners and 
edges of the actual camera pictures taken and that the drone often flew at an angle to 
account for the crosswind. Generally these flights covered about 1km but the two flown 
at a higher altitude (225m) covered more.  
 
 
32  
 
Figure 10. Final Checks of the Drone before a Launch 
Photo taken by Dr. Lipo 
 
Structure from Motion and Topography 
Structure from motion (SfM) is a form of photogrammetry, the science of 
getting ground measurements by measuring angles and lengths in photographs.45 While 
the concept of photogrammetry is about as old as the technology of photography itself, 
a more modern subset of the field is structure from motion. SfM, like traditional 
photogrammetry, uses location triangulation between matching features to create the 
geometry of the scene and estimate three-dimensional features from two-dimensional 
images. To do this, SfM needs three or more overlapping images of an area or object 
taken from slightly different angles, unlike traditional photogrammetry’s two. Those                                                         45 Slama. Manual of Photogrammetry. 1980 
 
 
33  
points that are similar align and match up, their slightly different angles and the images 
are then composited together. This allows for the creation of perceived depth and three-
dimensional effects just like how human, and many animal, eyes work (Fig. 11). With 
SfM, however, feature recognition algorithms incorporated into the programs allow for 
larger camera angle and photograph scale changes than the traditional methods and the 
processing power of the computer means than many images can be used to achieve 
more accurate results.46 The end result of this is a composite three-dimensional model. 
The shape arise from the angle and points that allow a structure to be made and the 
texture and color detail come from the combined photo imagery which is “draped” over 
the shape. 
 
Figure 11: Structure from Motion Diagram 
How structure from motion uses various different camera angles to recreate 3D 
features. Source: https://www.jvrb.org/past-issues/8.2011/2822 
                                                         46 Johnson et. al. "Rapid Mapping of Ultrafine Fault Zone Topography with Structure from Motion." 2014 
 
 
34  
Topography 
While SfM can be used to create 3D images of objects or buildings, if a large 
number of images are taken from above looking straight down at the ground, it can also 
be used to make 3D models of the landscape topography. For creating a topographic 
model, the more points of data there are, the more accurate the resulting image. If done 
by hand, acquiring topographic data requires a large number of man-hours to conduct 
in-the-field surveys and does not result in a very dense data set of points. With SfM 
software, however, the matching and creation of points between photos can be done 
more quickly and with a much greater density than before.    
This is not to say that collecting data by hand is the only way to amass 
topographic data. One other way is the use of Light Detection and Ranging (LiDAR). 
LiDAR works via active remote sensing. With LiDAR, a laser beam is sent down to the 
earth’s surface from a platform (usually a plane for high resolution data) and the 
reflected light is analyzed upon its return; the varying elevations create different return 
times which can be used to determine elevation given the known location GPS location 
of the platform. A DEM can then be derived from the data via triangulation.47  
LiDAR and SfM both have their pros and cons in when it comes to mapping 
topographic data. One advantage of LiDAR is that it can collect topographic data of the 
ground even if there is tree or other vegetation cover in an area. This is because the laser 
beam can have different returns (ie. One for the tree canopy, one for a smaller plant in 
the understory, and one for the ground) each of which is analyzed. Unlike LiDAR, SfM 
                                                        47 Johnson et. al. 2014 
 
 
35  
uses just a simple camera and passive remote sensing (ie. the sunlight that is bouncing 
off the earth’s surface) to collect data. Thus, it is only useful in collecting topographic 
data if there is minimal to no vegetation. Luckily this is the case on Rapa Nui.  
Given that a plane is required to fly the LiDAR equipment, data can be collected 
for large areas, this can also be a good thing, but in some instances it can be a 
disadvantage. For one, a pilot and plane are needed which are expensive, especially 
when only a small area is needed to be surveyed. While we used a fixed wing drone to 
collect our SfM imagery data, it required little skill to use and if we did not have that 
option, we could have used a balloon to elevate the camera.48 Additionally, because 
SfM data collection generally takes place closer to the ground, it generally has a better 
resolution than LiDAR data. Studies have shown that SfM data results in a higher cloud 
density of data points and is generally quite accurate (when there is little to no 
vegetation) and that it results in a higher cloud density of data points than conventional 
airborne LiDAR.49 
Another advantage to SfM is that of the orthophoto. Since SfM uses regular 
photographs to get its elevation data, it has visible-light imagery for the area as well. 
During processing, the images are matched and meshed together so that all the images 
are incorporated into one, large area image that has been geometrically rectified, an 
orthophoto. The result if that features can be visually looked at and analyzed whereas 
with LiDAR datasets, there is only the density data with which to base ones 
                                                        48  Johnson et. al. 2014 49 Harwin and Lucieer. 2012; Fonstad et. al. 2013 
 
 
36  
understanding of the area. This easy visual visual accompaniment to the topographic 
data is a big advantage when trying to determine features on a landscape.    
 When collecting areal imagery for topographic analysis using SfM, there 
are a few key concepts that factor into the collection method and process. The major 
one is that it all depends on the quality of photos being used. While the computer does 
most of the processing work and can create 3D models rich in information and detail as 
outputs, it is still the job of the individual to acquire the photos needed for the 
reconstruction. Given the essential nature of having overlap between the images it is 
important to be methodical when taking pictures of the subject. Just as important, 
however, is to ensure that the subject and its varying aspects are captured from different 
angles. The easiest way to ensure this is to move while taking photos of the feature, 
hence the name structure from motion.   
With the drone, we were able to get regular pictures of the ground with 
systematic overlap and height (fig. 12) and as it moved across the landscape, it captured 
the ground features at different angles. This made it easier to get quality reconstructions 
of the landscapes.  
 
 
37  
 
Figure 12: Camera Angles and Overlap in Photoscan 
An image showing the camera location (black line and blue rectangle) for each photo 
taken and used in the reconstruction with the created ground topography beneath 
 
Ground Calibration 
An important aspect of SfM is the ground calibration. This is what gives the 
orthophotos, DEMs, and other outputs location in the real world. Just as the object of 
study was located in the world, so must the SfM outputs be able to align with that 
location. The outputs must be able to exist in a correct relationship to those areas around 
it as well as to the other points within the model. One way to ensure this is to collect 
GPS data at certain identifiable site. These locations or Ground Control Points (GCPs) 
can then be marked in the model and, by having certain points with known locations, 
the model becomes, in a sense, anchored. Thus, the more known points, the more 
precise and accurate the model. In more technical terms, GCPs register a model to a 
coordinate system.  
 
 
38  
While having specific GPS points is one way to register a model, there are also 
others. One, is that if you have the GPS locations for where the individual pictures were 
taken, they can be used to help provide a location for the whole emerged image. 
Another way to ensure a correct data output is, if a DEM of the location at a different 
resolution exists, the same area of the two can be compared and used to correct each 
other. These various methods are not mutually exclusive and can be used in conjunction 
with one another in order to ensure the greatest accuracy for the data. 
For this project, we did not collect GCP data, rather, we used the GPS data 
associated with the various camera positions to provide locations for the images. As the 
drone was flying, it oriented itself based on its triangulation from several satellites. 
Since the drone knew its position, every time it took a picture, the drone’s position at 
that time was recorded. This data could then be used to help Photoscan align the photos 
as well as provide locational grounding for the output model, Orthophoto, and DEM.  
Unfortunately, this consumer-grade GPS on the UAV has error associated with it that 
can translate into registration error for the model. This is discussed in further detail in 
the sub-section “Errors” (page 45). 
Processing 
While there are a few free, open source SfM software options available, we used 
program Agisoft Photoscan to process the data and goes through the several steps to 
create the final products. There are several advantages to using Photoscan, the first 
being that the software can incorporate GPS camera positions into its calculations. This 
helps to reduce processing time as well as georeference the final files without relying on 
ground control points. Another perk to the Photoscan software is that it allows for batch 
 
 
39  
processing. This means that all of the processing steps can be set up before-hand and 
left to run rather than having to manually initiate each step. With an average of 853 
pictures taken per flight area (table 1), it could easily take 12 hours or more to process 
one flight, so the batch processing was key. 
To process a new flight’s worth of photos, first, the photos from the camera 
were matched up with their associated metadata (ie. GPS data) that was collected by the 
drone as it flew and extracted onto the handheld Yuma computer at the end of each 
flight. Once back at a computer, Agisoft Photoscan is opened and the photos are 
imported into a new project. A batch process is then set up to process the photos in the 
following order. First, the photos are then aligned with each other and since they have 
associated GIS data, this can be used to help speed up the alignment process. Next the 
geometry is constructed; first a sparse point cloud (fig. 13), and then a dense point cloud 
(fig. 14) that illustrates the points of overlap. It is during the construction of the point 
cloud that the various qualities (from low to very high) can be selected for the final 
product. This affects both processing time and the output because while a lower quality 
takes less time to construct, the end result is not as accurate as one constructed at a 
higher quality. While we were in the field we processed the data at a medium setting 
and the results were fairly good. Once we got back in the US however, Drs. Lipo and 
Lee used their resources at CSULB to reprocess the images at a high resolution. To do 
this processing myself would have been lengthy and redundant so I used the outputs of 
these re-runs as my data. 
After creating the dense point cloud, the next step is for Photoscan to create a 
wire frame (fig. 15) and then a solid surface (fig. 16) using these points. This helps to 
 
 
40  
create a dynamic, continuous, 3D surface and fill in any gaps that may occur in the data. 
Shading and texture (more imagery coloration) are then overlaid on top of the solid 
surface (fig. 17 and 18). The result of these steps is the final 3D model. During the 
model construction process, the images are processed in a way that creates a composite 
image of the photos as well as a topographic landscape this means that an orthophoto 
(fig. 19) and DEM (fig. 20) for the area can easily be exported and saved as their own 
files. This process was repeated for the data from all 26 flights. 
One of the trade-offs that must be considered when preforming SfM is 
resolution and time. The closer to the ground an image is taken, the higher the 
resolution will be but conversely, the picture will be smaller and much more time must 
be taken to get imagery of the whole desired area. The opposite is also true, pictures that 
are taken farther away from the earth’s surface cover more area and less of them are 
needed but the resolution is not as good.50 For our flights we flew at a 100 meters off 
the ground and the resulting resolution was about 3cm2.  
 
                                                        50 Johnson et. al. 2014 
 
 
41  
 
Figure 13. Photoscan Processing: Sparse Point Cloud 
 
Figure 14. Photoscan Processing: Dense Point Cloud 
 
 
 
42  
 
Figure 15. Photoscan Processing: Wireframe 
 
 
Figure 16. Photoscan Processing: Solid 
 
 
 
43  
 
Figure 17. Photoscan Processing: Shaded 
 
Figure 18. Photoscan Processing: Texture 
 
 
44  
 
Figure 19. Photoscan Processing: Orthophoto Output 
 
Figure 20. Photoscan Processing: DEM Output 
In this DEM the lighter color corresponds to higher elevation. If desired, this coloring 
can be changed. 
 
 
45  
Errors 
While SfM is a highly useful tool to process data, the process is not without its 
share of room for possible errors. The major possibility for error in the output features is 
tied to the errors in the drone’s GPS location. This arises because GPS, especially 
consumer-grade GPS, has a margin of error; it’s only accurate up to a certain point. This 
margin of error is due to the way the system works. GPS devices receive signals from 
different satellites and use the transmitted information to carry out trilateration 
calculations in order to pinpoint their location on earth. The error results from a the fact 
that most GPS devices are only accurate to ten nanoseconds which is important since 
knowing when and where a satellite transmitted its data from is important critical to the 
calculation. Additionally, the transmitting microwave signals travel at the speed of light 
and this in conjunction with the limited timing accuracy equates to about a three meter 
error (in x, y, and z directions) for all GPS locations.51  
As we collected our data over a number of different days, different satellites in 
different positions were used for trilateration and locating purposes. While the resulting 
position trilateration is still accurate for GPS, it may be slightly different than 
information provided by the satellites being used a day prior. Another possibility for 
error is due to the atmosphere. As the microwave signals are transmitted between the 
satellites and earth, molecules and particulates in the atmosphere distort the signals 
being sent. Though mathematical equations can correct most of this error, certain 
varying locational atmospheric disturbances cannot be accounted for.    
                                                        51 "GPS Accuracy and Error Sources." Mio Technology Corporation 
 
 
46  
 As errors in GPS location can affect the data, the result can come in the 
form of the output topographic data being shifted or tilted, it can also include bending, 
shrinking, stretching.52 This data warping can also result from the picture itself. When a 
2D image is taken of a 3D, curving landscape, distortion occurs. If an aerial picture is 
taken looking vertically straight down at the ground, some distortion occurs, especially 
around the edges whose light has the farthest distance to travel to be captured by the 
camera. Most of this error can be corrected for by algorithms and is in fact the 
definition of an orthophoto, a photograph that has been geometrically corrected so that 
it can be used like a map. Since the images used for this project were taken relatively 
close to the ground compared to using satellite imagery or photos taken from a maned 
plane, this should help to limit the distortion. Additionally, combining all the photos 
together with Photoscan should help to decrease error as there are a number of visual 
references for every location; the randomness of the error associated with each 
individual photo should help to counteract the random error in other photos when 
meshed together so the final composite datasets have less than 3m of error in all 
directions.  
 Recently, studies have found that a doming effect occurs in the final 
resulting composite images of topographic SfM DEMs. In essence, this means that the 
middle of the DEM bulges up slightly. This is thought to be a result of combining near-
parallel imaging directions in tandem with radial lens distortion. Thjs error can be 
                                                        52 Johnson et. al.  2014 
 
 
47  
reduced by the inclusion of oblique images into the mix of analyzed photos.53 I did not 
test for, nor correct for this doming affect however, so it could be influencing the data. 
 
ArcGIS Processing 
ArcGIS is a Geographic Information System (GIS) developed by ESRI 
(Environmental Systems Research Institute). ArcMap is the main program of the 
ArcGIS software suite and allows the user to view and edit geospatial information in 
order to create maps from datasets. While there are open-source GIS options available, I 
chose to use ArcGIS because the University of Oregon has the software licensed and 
available to use on most of its PCs and it also has the extensions I needed to process the 
data. I have also taken some classes that have used it and so am familiar with the 
program.    
There are two main types of data that can be used in/processed by ArcGIS. One 
is vector data, vectored polygons and points that generally come in the form of 
shapefiles. These various shapes can be visualized and modified and also have 
associated data in attribute tables. The other form of data is raster data. Rasters are a 
grid of cells/pixels where each cell has value that represents some form of information. 
Digital imagery falls under this category and thus so do the orthophotos and DEMs 
created as outputs from Photoscan. 
While we flew a total of 26 flights while on Rapa Nui, for my analysis I only 
used the 23 that took place along the south cost (the images from two flights had been 
                                                        53 James and Robinson, 2014. 
 
 
48  
combined on island to provide a composite output for a local so I actually used only 22 
files). We flew our flights on the island in a rather patchwork manner so there was no 
correlation between the flight numbers and the images’ position along the coast. Thus, 
the first thing I did was order and rename the files providing each with a number (N) so 
that there was a general sense of moving from west to east along the south coast as N 
increased (table 1). 
With some organization established, I then imported the orthophotos and DEMs 
into ArcMap. While these files had geographic extents (longitude and latitude points) 
associated with them due to the camera GPS data for each image, the coordinate system 
they were associated with was a Geographic Coordinate System (GCS). This places 
points based on a global/spherical surface which can be a problem when viewing data 
on a flat surface. When there is no consistency amongst the coordinate systems used 
problems arise when comparing and joining data features. Thus, I re-projected these 
files into a Projected Coordinate System (PCS). A PCS is created through the 
mathematical conversion, or projection, of our three-dimensional world onto a flat two-
dimensional plane. As a result there are constant areas, lengths and angles across the 
surface. One side effect from the 3D to 2D transformation is that there is some 
distortion, especially in areas that are closer to the edges. In order to minimize this, a 
projection is generally chosen that places the desired area at or near the center. There 
are many different projections, but I chose to transform my files so they were in the 
ESRI designated PCS, “WGS 1984 UTM Zone 12S.” UTM zones are based on the 
Universal Transverse Mercator projection and encircle the globe with each zone being 
about six degrees of longitude wide. UTM zone 12S is the one associated with the area 
 
 
49  
where Rapa Nui is located. Since the earth’s curved surface is transformed into a flat 
plane, a coordinate system is superimposed onto the surface. Except for the polar 
regions, the UTM coordinate system is used to “flatten” all the zones using the 
Transverse Mercator projection. The “WGS 1984” refers the World Geodetic System of 
1984 which is the datum the projection is based on.54 Datums reference specifications 
of a measurement system and are important factors in a projection because the projected 
coordinates are based on geographic coordinates which in-turn reference a datum.55      
Ahu and Coastline Identification 
Once the files were projected, I opened up all the orthophotos and used them to 
identify and mark all the ahu along the south coast. While the major ahu along the coast 
have been noted by many, I felt that it was necessary to go through myself because the 
aerial view and resolution of our orthophotos gave me a unique perspective to identify 
the long low shapes of the ahu. Additionally, this would ensure consistency throughout 
the data. To mark the ahu, I created a new shapefile in ArcCatalog and then used the 
Editor tool in ArcMap to added features to it. Wherever I saw an ahu, I placed a point 
on it and added information about in in the attribute table (fig. 21) (table 2). I 
recognized ahu as low, long stone structures, generally with a fatter, denser, or taller 
center and thinner, less dense, or lower “wings.” This recognition resulted from a 
number of factors. One was that I used an already established dataset of moai and ahu 
locations (though the ahu in this dataset were not categorized). I checked out these 
markers and ensured they corresponded with my orthophoto locations in my newly                                                         54 "UTM - Universal Transverse Mercator." UTM: Projection 55 "Support." Datum. ESRI 
 
 
50  
created vector file. From these already established ahu locations, I extrapolated out to 
include further similar features that seemed to follow these established shapes and 
guidelines. I also used the knowledge I had gained on Rapa Nui about how to recognize 
ahu (the strong retaining wall in the back, etc.) to help identify these features for the 
data set. 
As I was going through, I classified each ahu as one of four categories. These 
categories were “image ahu,” “ahu,” “ahu?,” and “image ahu?” Those classified as an 
“image ahu” had moai clearly associated with them, this generally meant there were 
fallen statues (or parts) near the platform (fig. 22). To be classified as “ahu” there had to 
be a clear ahu shape and no nearby moai (fig. 23). There were a number of features, 
however, that seemed close to an ahu shape but I had some doubt about whether it truly 
was an ahu or was actually just a bit of wall or extended lithic-mulching area so I 
classified these as “ahu?” (fig. 24). This is the only way I categorized the ahu and I did 
not attempt to include notation or distinction to ahu in regards to construction 
chronology or other features that may differentiate them. As a result, my spatial 
analyses are pure ones, as they assume time is not a factor. 
Another initial step I took was to modify the shapefile of the Rapa Nui coastline. 
The initial file I downloaded was a series of vector lines indicating the shoreline for all 
of Rapa Nui. As I only needed to look at the south coast, I edited out the northerly 
sections as well as the small islands and bits of rock off the coast. In addition, I went 
along and, using the orthophotos as a guide, edited the coastline to match what the 
images were showing me. The initial outline was a bit courser than what I could see 
with my imagery and sometimes the features indicated didn’t quite match. Thus, going 
 
 
51  
through and manually adjusting helped to improve the resolution and accuracy of the 
coastline. This was important as I would use the vector file later for processing data. 
It should be noted, however, that in reality a shoreline, is an area, not a line; as 
tides move in and out the shore changes. Thus, while better, my modified coastline still 
has some error in it. I tried to get the line to match up with what the tideline according 
to the orthophotos but in many cases it is still a general match. If there was water right 
up against a cliff, I would leave room to ensure I wasn’t cutting off any data associated 
with the cliff and if there were low rock formations jutting out into the ocean I generally 
included them.  
 
Figure 21. South Coast Orthophotos and Ahu Points 
All 22 orthophotos placed along the south coast of Rapa Nui. The red dots indicate ahu. 
 
 
 
52  
 
Figure 22. Identifying “Image Ahu”  
The large grey stone in the middle of the platform are identifiable as the body of fallen 
moai 
 
Figure 23. Identifying “Ahu” 
This ahu has the very clear oblong shape and no nearby moai 
 
 
53  
 
Figure 24. Identifying “Ahu?” 
A series of images of ahu that I considered to be questionable. I was unsure if these 
were actual ahu or historic walls with rocks piled around. The enclosed areas in the 
bottom right picture are clearly manavai. 
 
Batch and Model Data Processing 
Processing this data was time-consuming. This was mainly due to the fact that I 
was dealing with large raster files that require a long time to analyze due to their size 
and the amount of information they hold. In addition, though I was examining features 
along the continuous coast, the data I had for the coast was broken up into twenty-two 
pieces. Though I tried to join the raster features together to create one raster, the files 
were too big to make this reasonable so each step needed to be repeated around twenty-
two times in order to get data for the whole coastline (if it involved the coastline, this 
 
 
54  
number was nineteen because three flights were more inland and contained no actual 
coastal data). 
To make this process easier I used the ModelBuilder feature in ArcMap to help 
automate the process. ModelBuilder allows you to build a visual flow-chart of your 
process rather than running each tool by hand. The major advantage to this is that it 
allows the user to iterate through datasets, applying the same processing to all the files 
indicated (fig. 25). This made it much easier for me because certain processes I could 
set to run and then just wait until the computer finished processing all the files. 
Unfortunately, I was not always able to use models in helping me process data 
due to the limits on iterators. In ModelBuilder, iterators allow the same tool/function to 
be run on multiple files within a database but the program only allows for one iterator at 
a time. This was fine when I was using a shapefile that contained data for the entire 
coastline (ie. vector data associated with ahu or the coastline) because I would simply 
extract the data within the set parameters associated with that area from each raster. 
Some situations, however, called for two iterators. This was when data was processed 
using information previously extracted from the individual rasters and thus existed as its 
own file rather than as one continuous one. In these instances it was possible to use 
batch processing. This was not quite as convenient because each file must have all its 
information entered manually into a table (input file, output file, extents, etc.) rather 
than have a computer automate the process. Once it is done though, all the files can be 
completed in one run rather than having to repeat the process after each one finishes 
running. 
 
 
55  
Additionally, as each of the flight files were processed separately, they also had 
separate outputs. This is not as important when the outputs are shapefiles or rasters 
because they can be viewed in ArcMap at one time and any future processing can occur 
in the same manner as before. When this becomes a problem is when the outputs are 
data tables. With each flight area being processed separately, the data associated with 
one tool can end up spread amongst 19-22 different data tables. In order to fix this, I 
would merge all of the data table files together in order to create one big table. The 
problem is that while the resulting table would have all of the combined output data, it 
wouldn’t have the ahu category information I had assigned to each point. It was 
important that the tables have this material so that I could use it to separate out the data 
associated with each type of ahu (ahu, image ahu, etc.). In order to attach a category 
column to each data table, I would perform a join. In ArcMap, a join allows data to be 
added to an attribute table based off of a common field. This means that additional 
information is not just tacked onto a table, but systematically matched so that the 
correct associations remain.56 For these tables, as the general identification information 
remained with the output datasets, I was able to use that as a shared column to bring all 
of the information into one table. The table would then be paired down as some of the 
information was redundant and unnecessary. The final tables are the ones that can be 
seen in Appendix B. This process was followed for most of the output data tables. 
Another problem that arose from processing what should be a contiguous area 
artificially broken up into flight areas is repetition. The flight areas overlapped each 
                                                        56 "ArcGIS Help 10.1." ArcGIS Resources. ESRI 
 
 
56  
other in order to ensure data was collected for the whole coast but the side effect was 
that some ahu were captured multiple times. When it came time to process the data, 
these features would then receive associated information from each flight area they fell 
within. Thus some features had more than one set of associated data. This caused 
problems because some datasets were cutoff if they were close to the edge of the flight 
area making them incomplete, while when overlap occurred data was repeated within 
the two outputs.  
A singular feature should only have one set of associated data so I processed the 
data to achieve this. How I went about doing this depended on the data in question as 
well as how the overlap affected each data set. For some, it was easy because a feature 
that lay on the edge of one flight might not be on the edge of the other, this meant that 
the data associated with the area not cut off by an edge was used while the incomplete 
data was rejected. Other times, the complete data set was not contained in any one flight 
area. To make sure all the appropriate data was associated with each feature, I would 
clip down the resulting data so that it didn’t include the overlapped area (this data 
remained in the unclipped section of data) and then would add this data together to get a 
single complete data value associated with each figure.  
 
 
57  
 
Figure 25. ModelBuilder Example 
This model is set up to run the tool “Zonal Statistics as Table.” The data being 
analyzing comes from the rasters being iterated through within the “SCoast_Slope” 
geodatabase. The data provided by these rasters is examined within the area parameters 
dictated by the file “SCoat_whole_50mBuff” and this information is processed and 
results in a saved output file “%Name%_coast_50m.” the %Name% means that the first 
part of the saved file name is the individual file name from the extracted raster. I 
created other models like this one to run other tools and iterate through other datasets. 
 
Removing Extraneous Data 
In order to get more accurate data regarding the topography, it was necessary to 
remove the data associated with the ocean. In order to capture the coastline, and given 
its rectangular flight paths, the drone flew over a good deal of water. This data was 
encompassed in the original orthophoto and DEM outputs but if it was left included in 
the data, the elevations associated with the varying wave height would add inaccuracies 
to the terrain (fig. 26). I went through several steps in order to remove this part of the 
data. 
 
 
58  
First I used the Feature to Polygon tool on all of the DEMs. This created a 
polygon of the area encompassed by each flight area raster (I used the DEM data to 
ensure that the outline was that of the images and did not include the black areas around 
the orthophotos as a result of trying to make them a rectangle). Then I used the merge 
tool to combine each new polygon with the coastline vector. When this was done, the 
newly created output files contained a polygon outline of the flight area that was split 
into two pieces along the path of the shoreline. This allowed me to go in and edit out the 
area associated with the ocean. The result was a set of polygons that matched the flight 
areas on their land side, but their southern side conformed to the coastline (fig. 27). 
These polygons were then used in conjunction with the Clip Raster tool to extract a 
reduced raster area so that each raster only contained terrestrial data. With ocean 
associated data gone, when analysis tools were run, no irrelevant ocean data would be 
included in the results. 
 
 
59  
 
Figure 26. Removing Extraneous Data Before 
This image of a flight’s DEM shows how much of the data is water. The red line is the 
coastline and everything below it is data resulting from the ocean and thus not useful in 
assessing topography. 
  
Figure 27. Removing Extraneous Data After 
This is the same DEM as the one above in fig. 26 but the undesired ocean data has been 
removed at the coastline. 
 
 
60  
Elevation and Viewshed Analysis 
I started to analyze the ahu-topography relationship by looking at the elevation 
of each ahu. While I had points marked where each ahu was located, I decided not to 
use a single point to collect elevation data as it would be giving a lot of weight to one 
single point in the raster. In reality the ahu are large and the resolution of my data is 
very high so I used a uniform area to get a more representative measure of the local 
elevation rather than a rock I may or may not have placed the marker on. I created a 
five-meter buffer around each ahu point which I used as the polygon input when I ran 
the Zonal Statistics as Table tool (fig. 28). Having the buffer polygon was essential in 
order to extract the raster data for the area because extracting data from a raster is messy 
due to the large amount of it. In an attempt to minimize the processing work and time, 
the Zonal Statistics as Table tool requires a pre-defined area as determined by a 
shapefile and the tool then summarizes the raster data found within that space. This data 
is then outputted as a tables which I merged together.  
In addition to looking at general elevation relationships, I also wanted to analyze 
how each ahu was located in relationship to its surrounding topography in general. The 
tool I used to help me look at this was the Visibility spatial analysis tool. This tool is 
essentially a combination of two other tools, Viewshed and Observer Points, that 
analyze topography in relation to points in order to determine lines of sight/areas of 
visibility. When it runs, if there is a large structure or hill that can block visibility the 
tool will indicate in the output raster that the areas behind it are unable to see the 
indicated point. In order to run, all of these tools require a DEM (or elevation 
 
 
61  
equivalent) as well as a set of points to analyze the visibility of, in this case these points 
refer to the ahu locations.  
The Visibility tool allows you to collect data in one of two ways, one is by 
frequency (essentially the Viewshed tool) and the other is by observers (the Observer 
Points tool). The input data is the same for both and visually the outputs look the same 
(fig. 29) but the main difference lies in how the visibility data is recorded. With 
frequency, the number of observers who can see a specific location is calculated but no 
indication is given as to who those observers are. As a result, the attribute table for the 
created visibility raster is fairly small and only has a couple columns. With observer 
points, however, the output attribute table indicates exactly which observers can see 
which areas. For this sort of analysis, the table looks like a matrix. Each observer point 
is a column and the rows represent various groups of locations within the DEM. If an 
observation point can see a certain group of locations, a “1” is placed in that row for the 
observer. Thus, the areas that can only be seen by Observer 1 only have a “1” in the 
Observer 1 column. If there is a “1” placed in the columns of both Observers 1 and 2 
though, this means that both those points can see those associated areas.  
For my purposes, I preformed my visibility analysis using the observer points 
because I wanted to get a precise area of observation for each ahu point and the matrix 
data table was much more conducive to that. First, though, I had to set up my 
parameters. Visibility is affected by a number of factors and the ArcGIS tool allows for 
these to be factored in. Parameters can constrict the vertical angle and direction of sight 
as well as how far the visibility extends (earth curvature can be automatically factored 
in) but I left these at their default settings; the two that I adjusted were the offset A and 
 
 
62  
offset B. The amount of these offsets, indicate a vertical addition to the surface 
elevations (z value) with offset A referring to the observation point and offset B 
referring to the observed areas (the rest of the topography). These are important 
parameters to include in order to take into account the height associated with the people 
doing the observing. In order to incorporate it, two new columns must be created in the 
attribute table of the observation points. For me this was my shapefile of ahu points and 
for both offset columns I instructed an offset of 1.75, about the height of a person (the 
DEM is in meters). 
While the Observer Points tool preforms the same sort of analysis as the 
Visibility tool, I chose to use the visibility tool because it allows for the non-visible 
areas to be ignored which made the resulting matrixes simpler. One aspect of the tool, 
however, is that it can only analyze up to 16 observer points at a time so I split up the 
points into 6 smaller groups that conformed to the edges of the flight areas and made 
them their own shapefiles which I used to run the visibility tests. I then used these 
attribute tables to ensure that each observation point was correctly matched up with its 
associated ahu ID (the output tables had columns numbered OBS1-16). In order to get 
the total observable area for each of the ahu, I selected out all the areas associated with 
each ahu point (rows with “1”) and then summed this area together. This ensured that 
the area associated with an ahu was not just the area it alone could see but also the area 
that could be seen by multiple ahu. 
 
 
63  
 
   
Figure 28. Elevation within 5m of Ahu  
The relative elevation around each point is indicated by the color of the 5m circle with 
green indicting lower elevation and orange/red higher. The red dots are ahu, the yellow 
image ahu, and the grey unsure (ahu?). 
 
Figure 29. Visual Raster Output of Observation Points 
The colored areas represent the areas visible to the five ahu points located within the 
DEM. The dark green represents areas that are visible by many locations while the light 
green areas are areas of low visibility. The area that are not green cannot be seen by, or 
see any of the ahu. The gray-back background is the flight area DEM underneath the 
Observation Point raster layer.  
 
 
64  
 
Extracting Slope Data 
The first step I took towards collecting data about the slope along the South 
Coast was create slope rasters for all the flight areas. This was done using the DEM and 
the slope spatial analysis tool which looks at the elevation data associated with a single 
raster cell and then compares that to the data from the cells surrounding it in order to 
determine a slope gradient. The resulting files showed the slope for their respective 
areas in decimal degrees. 
I established a set area of 50 meters within which the slope data would be 
analyzed. I did this because I was interested in getting a general feel for slope in an area 
and at my high data resolution the slope data for a single point generally had little 
meaning. Fifty meters was a decision I made as it seemed to allow for the cliff area to 
always be included in the coastline buffer and gave some room on either side of the ahu 
for slope variation. Establishing a set area also helped to keep size consistency 
throughout the data which was important for later statistical analysis. I realized that 
comparing slope data from around ahu and the coast would not necessarily be that 
useful since a number of ahu are not right along the coast, so for each ahu I calculated 
its closest point along the coastline vector (table 3). The resulting table had X,Y 
coordinates for each of the resulting locations and I used those to create a new shapefile 
of vector points along the coast. I then created a 50 meter buffer around these points as 
well as a 50m buffer along my coastline. I also made 50 meter buffers around all of the 
ahu points as well.  
 
 
65  
Once I had these polygon areas, I used them to cut and extract slope data from 
the larger slope raster files. To visualize the data I used the Raster Clip tool to clip the 
slope rasters using the 50 meter buffer created around the closest coast points and I also 
clipped them with the 50 meter coast buffer. The fact that I could now look at the slope 
in these areas in isolation was more for my own personal viewing to see the area and 
data collected rather than for data collection but I found it helpful in understanding 
exactly what data was being analyzed (fig. 30, 31, 32). To actually collect the slope data 
for these modified area, I used the Zonal Statistics as Table tool in combination with the 
already created buffers to act as the constraining boundary for the statistical summary. I 
collected this data for the 50 meter buffered stretch of coastline (table 7), the 50 meters 
around each ahu (table 8) and its associated closest coast point (table 6) and processed 
the resulting sets of data tables.     
 
 
Figure 30. 50m Coast Buffer Isolating Slope Data 
This image shows one section of coast slope data. The darker brown areas indicate a 
steeper slope while the lighter areas show gentle to no incline. The red line is the 
coastline and the ocean data below that line has been removed so only the land data is 
analyzed. The green dots represent the closest coast points, the red dots ahu, the yellow 
dots image ahu, and the grey dots unsure ahu (ahu?). 
 
 
 
66  
  
Figure 31. 50m Closest Coast Points Buffer Isolating Slope Data 
This image shows the slope data surrounding 50m of the closest coast points for ahu. 
The darker brown areas indicate a steeper slope while the lighter areas show gentle to 
no incline. The red line is the coastline and the ocean data below that line has been 
removed so only the land data is analyzed despite that the buffers used to extract this 
data were circular. The green dots represent the closest coast points, the red dots ahu, 
the yellow dots image ahu, and the grey dots unsure ahu (ahu?). 
 
Figure 32. 50m Ahu Buffer Isolating Slope Data 
This image shows the slope data surrounding ahu for 50 meters. The darker brown 
areas indicate a steeper slope while the lighter areas show gentle to no incline. The red 
line is the coastline and the ocean data below that line has been removed so only the 
land data is analyzed meaning that some buffers are not analyzing a complete circle of 
data. This is also the result of the edge of the flight area. The green dots represent the 
closest coast points, the red dots ahu, the yellow dots image ahu, and the grey dots 
unsure ahu (ahu?) 
 
 
67  
Part 4: Data Analysis and Discussion 
I organized my data into several questions around which I ran a series of 
statistical tests. For the most part, these questions revolve around seeing if there is a 
topographic difference between regular ahu and image ahu through the various ways I 
used to measure the landscape (ie. elevation, slope, visibility) but there are some 
questions that are about comparing ahu associated features with the overall topography 
of the South Coast. 
1. Are image ahu at higher elevations than ahu? (Table 3, Appendix B) 
2. Do image ahu have a larger viewshed than ahu? (Table 4, Appendix B) 
3. Are image ahu closer to the coast than ahu? (Table 5, Appendix B) 
4. Is the coastal slope near image ahu less steep than the coastal slope near ahu? 
(Table 6, Appendix B) 
5. Is the coastal slope near image ahu less steep than the general slope of the 
coast? (Tables 6 & 7, Appendix B) 
6. Is the coastal slope near ahu less steep than the general slope of the coast? 
(Tables 6 & 7, Appendix B) 
7. Is the ground slope around image ahu steeper than the ground slope around 
ahu? (Table 8, Appendix B) 
Is the ground slope around image ahu steeper than the ground slope around ahu? 
In order to help answer these questions, I made use of the open source statistics 
computer software R. R is a programming language and software environment 
specifically tailored for dealing with statistical computing and graphics. The software 
provides a platform that makes analyzing data and computing statistics much easier. My 
use of the program mainly focused around running one type of test because, though the 
 
 
68  
questions and data varied, a t-test could be used to answer all of the topographic 
relationship questions I’d asked.  
A t-test is a statistical hypothesis test which uses the average of a group or 
population in relation to the variation in the data to determine significance. There are a 
couple different types of t-tests that can be used in different situations but I used a two-
sample t-test that compares the mean of two independent samples to each other. Like 
most statistical tests, a t-test operates on the assumption of a null hypothesis (H0). Often 
this is a general statement that assumes no effect or difference between groups. 
Opposite the null hypothesis is the alternative hypothesis (HA) which is the statement 
hoped to be proved true. It is important to have null and alternate hypotheses because it 
clarifies the goals of the statistics and helps to ensure that the statistics have purpose. 
The null and alternative hypotheses for my seven questions are: 
1. “Image ahu are not at significantly higher elevations than ahu” (null) and 
“image ahu are at higher elevations than ahu” (alternative). This alternative 
hypothesis is founded on the concept that image ahu are costly signals so higher 
elevation is preferred as discussed in the sections “Costly Signaling and Water” 
(p. 18), “Topography” (p. 22), and “Visibility in Archeology” (p. 25). 
2. “Image ahu do not have significantly larger viesheds than ahu” (null) and 
“image ahu have larger viewsheds than ahu” (alternative). This alternative 
hypothesis centers on the concept that image ahu are costly signals so locations 
with greater visibility are preferred as discussed in the sections “Costly 
Signaling and Water” (p. 18) and “Visibility in Archeology” (p. 25).  
3. “Image ahu are not significantly closer to the coast than ahu” (null) and “image 
ahu are closer to the coast than ahu” (alternative). This alternative hypothesis is 
based on the concept that image ahu are markers related to coastal water 
resources so distance to these resources would be a concern as discussed in the 
sections “Costly Signaling and Water” (p. 18) and “Topography” (p. 22). 
4. “The coastal slope near image ahu is not significantly less steep than the 
coastal slope near ahu” (null) and “the coastal slope near image ahu is less 
steep than the coastal slope near ahu” (alternative). This alternative hypothesis 
 
 
69  
is founded on the concept that image ahu are placed in relation to coastal access 
to water and that less steep slopes provide easier coastal access, as discussed in 
the sections “Costly Signaling and Water” (p. 18) and “Effect of Slope on 
Walking” (p. 23). 
5. “The coastal slope near image ahu is not significantly less steep than the 
general slope of the coast” (null) and “the coastal slope near image ahu is less 
steep than the general slope of the coast” (alternative). This alternative 
hypothesis is also based on the concept that image ahu are placed in relation to 
coastal access to water and that less steep slopes provide easier coastal access 
as discussed in the sections “Costly Signaling and Water” (p. 18) and “Effect of 
Slope on Walking” (p. 23). 
6. “The coastal slope near ahu is not significantly less steep than the general slope 
of the coast” (null) and “the coastal slope near ahu is less steep than the general 
slope of the coast” (alternative). This alternative hypothesis centers on the idea 
that ahu are placed in relation to coastal access to water and that less steep 
slopes provide easier coastal access as discussed in the sections “Costly 
Signaling and Water” (p. 18) and “Effect of Slope on Walking” (p. 23). 
7. “The groud slope around image ahu is not significantly steeper than the ground 
slope around ahu” (null) and “the ground slope around image ahu is steeper 
than the ground slope around ahu” (alternative). This alternative hypothesis is 
grounded in the ideas that ahu are markers of fresh water and that steeper 
slopes on land indicate better access to this water as discussed in the sections 
“Costly Signaling and Water” (p. 18) and “Topography” (p. 22). 
In statistics, in order for the null hypothesis to be rejected and the alternative accepted, 
statistical significance needs to be proven. This means showing that the difference in the 
results did not occur by random chance, which comes in the outputs of a t-test 
Normalizing Data 
Before running a t-test, one concern is whether the data being used in the test is 
normally distributed. Normally distributed data clusters around the mean of the data set 
and then symmetrically moves out away from the center in decreasing concentrations. 
The resulting graphic has a peaked middle and flaring sides which is where it gets the 
colloquial name “Bell Curve.” Having data that conforms to these general parameters of 
 
 
70  
normality is generally required for most statistical assessments because it is an 
underlying assumption for most parametric tests.57 Unfortunately, raw data often does 
not conform to a normal distribution. Rather, data can often be concentrated on one side 
of the graph rather than being evenly distributed in the middle. This is skewed data. If 
the concentration is more the left on the graph the data is positively skewed and if it’s 
more clustered on the right the data is negatively skewed.  
Skewness is important to consider when doing t-tests because t-tests are 
parametric tests. This means that one of the underlying assumptions of the test is that 
the data being compared is normally distributed. Thus, if the data is actually skewed, 
the results can be false and misleading. There are, however, a number of ways to assess 
the skewness of a dataset.  
One way is that there is an actual equation that can calculate skewness. There 
are also histograms which are a type of bar chart used to show statistical information. 
For these graphs, the dependent variable is graphed along the x-axis in numerical 
intervals of equal size or “bins” and the bars represent the frequency of data associated 
with each group. The resulting graph can provide a visual for the data’s distribution so 
that a person can assess for themselves. There are a number of other ways to help gauge 
skewness and luckily in the statistical computer software IBM SPSS (Statistical 
Package for the Social Sciences), they provide a number of them. Their descriptive 
statistics option made it easy to get such information for various different data sets. 
When assessing skewness, there are a couple things to go by. One is simply visually, by 
                                                        57 Laerd Statistics. “Testing for normality.” 
 
 
71  
looking at histograms and other graphic representations as seeing if the data looks off. 
There are also rules of thumb like normalized data should have a skew value between -1 
and 1. 
If data are very skewed, this does not necessarily preclude it from being a part of 
t-test of other parametric test because there are ways of transforming data so that it 
takes on a more normalized shape. There are two main methods for a normalization 
transformation, one is through the use of log10 and the other via square-root. By 
applying these functions to a data set, they help to change its distribution while still 
maintaining the essence and relationships of its original state. If it is desired to revert, 
these transformation functions can be inverted. Transformation is also different if the 
skew is positive or negative. With a positive skew the function can just be applied to the 
data set but if the skew is negative a reflected transformation needs to occur.  
This does not mean, though, that data must be perfectly normalized in order for 
a t-test to work. T-tests are fairly robust and there are a number of other factors 
involved like sample size that can affect the shape of the data. Additionally, assessing 
skewness in smaller sample sizes can be difficult depending on the data. This is what I 
found with my data. I looked at histograms and q-q plots as visuals but I sometimes 
found the patterns in the histograms hard to see so I generally went for a more 
quantitative rather than qualitative approach, using the rule of thumb that skew should 
be between -1 and 1 and the Shapiro-Wilk significance test should be greater than .05 to 
indicate normality.58  
                                                        58 Shapiro and Wilk. "An analysis of variance test for normality (complete samples)" 1965 
 
 
72  
As you can see in the table on skew indicators, it can be hard to get a 
transformation that works for both of the data sets involved. Even if one data set is 
normally distributed, if the other data set isn’t the same transformation needs to be 
applied to all the variables. This is because though the relative differences stay the 
same, the unit of measurement changes which affects the differences between the 
variables.59 From the data, only two sets seemed to already be normally distributed 
(both the ahu groups associated with Observable Data and Slope Around Ahu). No 
transformation adequately worked to help normalize the ahu data for Slope Around 
Closest Coast Points and there were also problems with transformations in conjunction 
with the Whole Coast Slope data. After having gone through the data and also looked at 
the accompanying graphics I decided to just leave the data as it was and not transform 
any of it.    
                                                        59 Francis and Field. Discovering Statistics Using SPSS. 2011. 
 
 
73  
Table: Skew Indicators Type of data and type of 
ahu Ahu Type Normal Log10 SquareRoot Skew Shapiro-Wilk skew Shapiro-Wilk skew Shapiro-Wilk Elevation  Ahu 0.043 0.217 -1.072 0.01 -0.478 0.13 Image Ahu 1.846 0.004 -0.787 0.25 0.615 0.411 Observable Area  Ahu 0.059 0.462  Image Ahu 0.373 0.278 Distance to Coast  Ahu 0.894 0.075 -0.174 0.45 0.439 0.326 Image Ahu 1.096 0.003 0.624 0.051 0.879 0.012 Slope Around Closest Coast Point  Ahu -1.666 0.011 -1.185 0.001 -2.042 0.002 Image Ahu -0.884 0.218 -1.476 0.012 -1.185 0.058 Whole Coast Slope  -1.016 0.09 -1.666 0.009 -0.201 0.178 Slope Around 
ahu Ahu 0.701 0.148  Image Ahu -0.443 0.179 
This table shows the different values for two tests of normalcy, skew and Shapiro-Wilk 
as they applied to different data sets under different conditions. Each row “block” is its 
own set of data and the subset rows indicate the subset divisions of data by ahu type. 
“Normal,” “Log10,” and “SquareRoot” refer to when that transformation was applied to 
the data. Normalcy for Skew is a value between -1 – 1 and for Shapiro-Wilk it’s any 
number above 0.05. The two areas are blanked out because their initial data values meet 
the normalcy requirement.  
 
The T-Test 
A t-test is a hypothesis test using the average of one or two data sets. There are a 
couple different types but I used a two-sample t-test because it tests whether or not there 
is a significant difference between the averages of two independent groups, which 
applies exactly to all of my research questions. Additionally, the t-test is fairly good 
with smaller sample sizes and even if a dataset is not normally distributed, a t-test is 
generally still valid unless the data is extremely skewed. 
 
 
74  
T-tests compare the mean of the two data sets in an attempt to show if there is a 
significant difference between the two, the likelihood of any difference occurring by 
chance. In equations, statistical significance is generally represented by “p” which 
stands for probability. Another common symbol when determining significance is alpha 
(α) which refers to random chance. The α value commonly referenced in significance 
statistics is 0.05. Often this number is used as a cut-off for significance if a result is less 
than this, it means that the chance of such a result occurring randomly is less than 1 in 
in 20 and this is considered to be unlikely enough to conclude that the result is not 
random. 
When preforming a t-test, or any test of significance, there are two ways to look 
at the results. One way is called a one-tailed test and the other is a two-tailed test. With 
a two-tailed test, significance and difference are being looked for at both extreme ends 
of the data. And the alpha is split between the two extremes. With a one tailed test, 
though, significance is only being looked for one side of the data. This means that the 
whole alpha does not have to be split up. The main deciding factor in whether a one or 
two-tailed test is run lies in the wording of the research and study question. If the 
question mentions “greater,” “less than,” or other ranking indicator, a one-tailed test is 
used but if the question simply refers to there being a difference, then the test is two-
tailed. All of my questions have words such as “higher,” “larger,” “closer,” and 
“steeper” so I ran one-tailed tests. 
Preforming a t-test within R is fairly simple. First the data needs to be organized 
into separate columns in a CSV (Comma-Separated Values) or a Tab-delineated file. 
This information is then brought in, or “read in,” to R via certain commands and given a 
 
 
75  
variable name. The columns of data within the table can then be accessed through the 
use of the files variable name, followed by a “$” and then the header of the column. To 
run a t-test, the command “t.test” is used in conjunction with the variables being 
compared, but this automatically runs a two-tailed t-test. In order to run a one-tailed 
test, alt=“greater” or alt=“less” is also typed as part of the t.test command. Whether 
“greater” or “less” is used depends on the alternative hypothesis being studied and is in 
relation from the first variable to the second variable. For example, (assuming a fake 
dataset named “Trial”): 
t.test(Trial$X, Trial$Y, alt=”greater”) 
 In this instance, the one-tailed test being run is comparing the mean variable “X” of the 
dataset “Trial” to variable “Y” and with the alternative hypothesis guessing that the 
mean of variable “X” is greater than that of “Y.”    
 
 
76  
In order to answer my seven questions I ran one-tailed t-tests on my data and these were 
my results: 
1. Are image ahu at higher elevations than ahu? 
 
Figure 33. Elevation t-Test Results 
The t-test p-value is 0.6905 which means that I must accept my null hypothesis that 
image ahu are not higher than ahu.  
 
2. Do image ahu have a larger viewshed than ahu? 
 
Figure 34. Area Visible by Ahu t-Test Results 
A p-value of 0.8678 means that I cannot reject my null hypothesis and thus image ahu 
do not have a significantly greater viewshed then ahu. The image ahu mean 
(13,676,827) is actually smaller than that of ahu (67,185,809) rather than greater so 
insignificance is expected.   
 
 
77  
3. Are image ahu closer to the coast than ahu? 
 
Figure 35. Distance to Closest Coast Point t-Test Results 
A p-value of 0.4104 indicates that image ahu are not significantly closer to the coast 
then ahu. 
 
4. Is the coastal slope near image ahu less steep than the coastal slope near ahu? 
 
Figure 36. Slope Surrounding Closest Coast Points t-Test Results 
A p-value of 0.5349 shows that the coastal slope close to image ahu is not less steep 
than that associated with ahu. Though the difference is not large, the image ahu mean 
(17.57) is actually greater than that of ahu (17.46) rather than less so insignificance is 
expected. 
   
 
 
78  
5. Is the coastal slope near image ahu less steep than the general slope of the 
coast? 
 
Figure 37. Slope Comparison between Image Ahu and Coastline t-Test Results 
The p-value is 0.4653 which means that the null hypothesis is accepted. This means that 
the coastal slope close to image ahu is not less steep than that associated with the 
general coastline. 
 
6. Is the coastal slope near ahu less steep than the general slope of the coast? 
 
Figure 38. Slope Comparison between Ahu and Coastline t-Test Results 
The p-value of 0.4027 is well above the 0.05 alpha value which means that the null 
hypothesis must be accepted which says that the coastal slope close to ahu is not less 
steep than that associated with the general coastline.   
 
 
79  
7. Is the ground slope around image ahu steeper than the ground slope around 
ahu? 
 
Figure 39. Slope around Ahu t-Test Results 
The p-value of 0.5408 signifies that the slope surrounding image ahu is not steeper than 
the coast surrounding ahu.  
 
Overall, my p-values were: 0.6905, 0.8678, 0.4104, 0.5349, 0.4653, 0.4027, and 
0.5408. All of these are well above the 0.05 alpha value so I accepted all of my null 
hypotheses. This means that none of my predictions about the relationship between ahu, 
image ahu, and topography are true.  
Discussion   
While the questions I was asking over the course of this research had to do with 
topography in regards to analysis, beneath it all were underlying questions about costly 
signaling and water resources; could topographic analysis reveal or prove something 
about ahu in relation to these concepts, concepts that could provide great insight into 
the ancient Rapanui culture. Thus, the questions about visibility and elevation were 
actually about costly signaling and the other questions about slopes and distance to 
coast were actually questions about proximity to water resources, whether image ahu 
 
 
80  
indicated locations with access to water. I had originally thought that image ahu, given 
their associated maoi and likely existence as costly signals would be placed in areas of 
higher elevation where they would be the most visible, where they would be able to 
signal the most. I had also believed that image ahu would be in areas closer to coastal 
resources as well as near areas where these resources would be the most accessible. 
Though my statistical results do not indicate that these topographic differences 
exist between the ahu features, this does not mean that the underlying theories about 
ahu signaling and demarcation of water resources is false. My results simply show that 
other factors could be more important in regards to why ahu were placed where they 
were on the landscape. One of these features could actually be the water data itself. My 
analysis of slopes and coastal distance was merely an attempt to see if any resource 
relationship manifested itself in a spatial on as well. In reality the hydrology and 
geography of Rapa Nui might result in coastal water outflows in locations that do not 
necessarily correspond with areas of easy coastal access. This question is currently the 
thesis subject of another student (Sadie Trush) in the Clark Honors College (currently 
set to defend in 2016) and I am interested to hear about the results of her analysis. 
Another aspect that could be a factor as to why these architectural structures are 
located in certain places is time. My analysis was a pure spatial analysis because I 
assumed that time played no factor. In reality, however, chronology could be important, 
with ahu built at different times having different purposes that resulted in different 
spatial relationships. It is known that some ahu were rebuilt during the time of the 
ancient Rapanui and there are also some structures that could have been built 
 
 
81  
historically. This data thus shows that perhaps factors other than topography are more 
important in terms of understanding ahu placement.   
Errors 
Unfortunately, there was some error in my analysis and data. Some of this error 
lies in the SfM data processing. As stated in the “Error” section in Part 3, no ground 
calibration points were used to help correct for error in the consumer-grade GPS and no 
testing or correction was done for any doming affect. All of this could have an impact 
on the data. There were other error sources, however, that were not just SfM. 
One large source of analysis error lies in how the viewshed data was processed. 
A continuous elevation surface is key to this analysis but my analysis was hampered by 
the restricting flight areas. Though the detail was fantastic, my rasters were too big to 
merge together which meant that the viewshed for each point was only calculated for 
within that limited area. This is problematic as key amounts of area may be excluded 
simply because they are located outside of the arbitrarily located flight zone. Those ahu 
that were located in the overlap between two flight zones had two different viewsheds 
that were combined together (with the redundant data due to the overlap removed) but 
not all ahu got this. A complete viewshed was not collected for every ahu point, even 
for those with a combined viewshed area due to overlap, so it is hard to accurately 
understand assess the results of this analysis (question #2).   
Another cause for analysis concern in in the size of the slope buffer area. For all 
of my slope analyses I used a buffer area with a radius of 50m from the feature in 
question. The reason for this was because in some places, the coastline was further 
away from the cliffs. Thus, in order to ensure that the cliff areas (ie. Where the livable 
 
 
82  
land began) were always included, I chose a 50m buffer area. Unfortunately, slope is a 
concept that is very sensitive to scale and the 50 meter buffer I used in my data 
collection could easily be too big to accurately show the topographic nuances of Rapa 
Nui’s south coast. There might be topographic differences there but my analysis simply 
aggregated too much data together to reflect it. 
Future Research 
As a result of these areas of error, there are a couple things I would do if I was to 
further this research. One of the first and easiest things would be to would be to try 
analyzing the slope with different buffer areas to see if there are any different results. If 
ground calibration points could be collected during another trip to Rapa Nui, they could 
be applied to the SfM results in order to lessen any error in the data. In order to correct 
for the visibility analysis error, preforming a viewshed analysis with a more complete 
DEM could produce more reliable results. Unfortunately, such a publically accessible 
dataset does not yet exist. From my understanding there is a 30m DEM from satellites 
but just as my 3cm SfM resolution was too small, this is too big. It won’t be able to get 
the nuances of the topographic data that distinguishes the ahu. Ideally this analysis 
would use a 1m LiDAR DEM but to my knowledge, such a dataset does not yet exist.  
Another analysis test that could be run is a least-coast path. This tool can take 
into account elevation and the work associated with it in order to calculate the path 
between points that will require the least amount of work and energy expenditure. This 
would be a particularly strong analysis tool to use because the 3cm resolution of the 
dataset is at a human, walking-scale level. If this analysis is performed with a DEM of a 
 
 
83  
few meters, the results are still useful, but for short distances and more realistic results, 
having the finer detail is a great advantage and could produce some interesting results.          
In addition to these topics, one major area for further study is the testing of 
random points in comparison to the ahu locations. This would be a true null hypothesis 
as it would see if there is any topographic relationship as to how the ahu are placed or if 
they are equivalent to points picked at random along the coast. Though no statistical 
difference was shown between the two types of ahu, I believe that preforming this true 
null hypothesis test would show that ahu locations are not just random points in regards 
to topography. I think this would be particularly obvious when comparing the distance 
to the coast. While there are a few ahu further inland than others, the majority are very 
near the coastline and this is true throughout the island, not just on the south coast (fig. 
1). This is one obvious spatial relationship and I think it would be reflected through the 
running of the true null hypothesis.    
 
 
84  
Part 5: Conclusion 
The main reason for conducting this study and topographic analysis was to 
further the research and understanding into the lives and society of the ancient Rapanui. 
It is obvious given the amount of effort put into the construction of the moai and ahu 
that they played an important part in Rapanui culture and by better understanding the 
purpose of these features, we better understand the lives of those who built them. My 
examination into the relationship between topography and ahu is just one attempt to 
understand this purpose. 
Given the possible errors and incomplete topographic data analysis as talked 
about in the previous Discussion section, I am hesitant to definitively say that there is 
absolutely no relationship between ahu and topography. That being said, however, none 
of the results of my tests were all that close to being significant. To me this means that 
if any topographic relationship exists for ahu, it exists on a smaller scale than I was able 
to capture with my slope areas, in the incomplete visibility areas, or in some manner 
that I did not think of. Rather, I think that this is simply an indication that other factors 
play a larger role in the purpose of these structures, factors that do not have a 
topographic manifestation.  
This study was just one in a number of studies, tests, and projects meant to 
contribute towards the research being conducted by Drs. Hunt and Lipo on Rapa Nui on 
relation to the prehistoric and current water resources there. Since this study was solely 
focused on the southern coast of the island, I hope that my work and methods can lay a 
groundwork for others and encourage the continued study of topographic relationships 
both on Rapa Nui and off it. There is undoubtedly much research still to be done in this 
 
 
85  
regard and much still to be discovered and learned from Rapa Nui. My study on 
topography has added to this growing knowledge and understanding about ahu, their 
purpose, and the people who built them.  
 
 
86  
Appendix A: Terms 
 
Ahu- long, low platforms built by the ancient Rapanui that are generally found near the coast. Generally “ahu” (in italics) refers to all such long low platforms. When I conducted my analysis, I classified those ahu without moai associated with them as “ahu” while those with moai became “image ahu”  
Alpha (α)- in statistics, it refers to the probability level of rejecting the null hypothesis when it is in fact true. To keep this from happening, a probability level of 95% or 0.05 is generally used to determine significance. If a p-value is less than .05 than the null hypothesis is rejected, but if the p-value is greater, the null hypothesis is accepted. 
Attribute Table- the information table that accompanies each GIS shapefile. Each row is associated with a vector feature within the shapefile. Columns contain information about various fields for each feature.     
DEM (Digital Elevation Model)- a graphical output/file that depicts the topography of the region in question  
Geographic Coordinate System (GCS)- a geographic locating system associated with a 3D spherical surface. This is how places are located on a globe. A grid is created over the surface with lines of latitude running parallel around the sphere and longitude lines beings meridians. Longitude and latitude refer to angles measured from the center of the earth to the surface and are generally described in degrees.  With this grid, every point on the surface can be referenced by a longitude and latitude value.  
Ground Control Points (GCPs)- Points on the ground with a known GPS location that are used to calibrate the output structure from motion features. They help ensure data accuracy as they ground the constructed models.  
Image ahu- long low platforms built by the ancient Rapanui that have moai associated with them (historically the stood on top of the platform) 
Kurtosis- A measure in statistics used to indicated the peakedness or flatness of a data set.  
 
 
87  
LiDAR- Light Detection and Ranging. A form of active remote sensing where a laser beam is emitted from an aerial platform and the differing return times are analyzed to determine the surface. The use of a laser allows for multiple returns which makes it possible for ground topography to be recorded even in the presence of vegetation. 
Lithic Mulching- a form of agriculture that places small to medium sized rocks in the desirered growing area in order to put nutrients back into the ground and help retain water moisture in the soil.  
Manavai- a small enclosure created by a low stone wall. These features were constructed by the ancient Rapanui as a strategy for growing crops. The enclosed area is more protected from the wind and the soil has more nutrients.  
Moai- the large stone statues built on ancient Rapa Nui 
Orthophoto- an aerial photograph that is geometrically corrected to have a uniform scale 
P-value- in statistics, the p-value is a function of the observed sample results  used for testing a statistical hypothesis. It represents the probability of the occurrence of a single event.  
Projected Coordinate System (PCS)- a geographic coordinate system that has been projected onto a flat surface. The process of changing 3D features to 2D creates distortions within the projection. There are a wide variety of projections and different projections have different advantages, fidelity of shape, area, and direction cannot exist all at once so projections are chosen based off of the desired purpose of the map. There is more accuracy towards the center of a projection so a singly type of projection may have multiple versions where the center is placed in different locations.  
Puna- A structure created by the ancient Rapa Nui considered to be a well. Though there is also some confusion about whether some were boat ramps. 
Ranu Raraku- the quarry where the moai statues were carved out of the hillside by the pre-historic Rapanui 
Rapa Nui- the local name for Easter Island 
 
 
88  
Skew- In statistics, if one side of a dataset is more heavily favored. It is a measure of symmetry and often used in determining whether a data set fits a normal distribution.  
Structure from Motion (SfM)- A way to process images in order to create a 3D model of the observed area. The process uses a computer program and many overlapping images taken from different angles to composite the final model together 
T-test- a statistical examination of the of the two different population means. 
Taheta- rocks carved into basins. Thought to have been for holding water.  
Trilateration- The calculation of a point in 3Demensional space. This is similar to triangulation but triangulation refers to working with angles whereas trilateration refers to working with working with distances. This is how GPS calculates position because it uses speed and time to determine distance.  
Trimble- name of the company that made the drone we used 
UAV-Unmanned Aerial Vehicle 
Yuma- name of the hand-held tablet computer used to program and communicate with the drone 
 
 
 
 
 
89  
Appendix B: Tables 
    
Nu
m
be
r (
N)
Fl
ig
ht
 D
at
e
Da
te
 F
lig
ht
 #
O
ve
ra
ll 
Fl
ig
ht
 #
# 
of
 P
ic
tu
re
s
De
sc
rip
tio
n
1
13
-Ja
n
2
10
36
7
Fu
rth
es
t w
es
t (
le
ft)
 o
f S
 c
oa
st
2
9-
Ja
n
1
1
86
1
Q
ua
rry
3
9-
Ja
n
2
2
86
9
rig
ht
 o
f q
ua
rry
4
13
-Ja
n
1
9
93
8
5
13
-Ja
n
3
11
39
1
Ha
ng
a 
Te
'e
6
15
-Ja
n
1
16
62
1
7
17
-Ja
n
3
25
82
1
lo
ng
, m
ai
nl
y 
no
n-
co
as
ta
l o
ve
r t
he
 to
p 
of
 a
nd
 
ov
er
la
pp
in
g 
N4
-6
8
15
-Ja
n
2
17
93
1
9
16
-Ja
n
2
19
93
5
10
16
-Ja
n
3
20
98
1
W
es
t s
id
e 
of
 A
ka
ha
ng
a
11
10
-Ja
n
1
2!
87
4
Ea
st
 s
id
e 
of
 A
ka
ha
ng
a.
#2
! b
/c
 a
cc
id
en
tly
 
pr
es
se
d 
re
fly
 fl
ig
ht
 fo
r C
IR
 o
f N
3 
an
d 
ca
n'
t j
um
p 
st
ra
ig
ht
 to
 fl
ig
ht
 4
12
16
-Ja
n
4
21
10
28
To
p.
 In
la
nd
 fl
ig
ht
 o
ve
rla
pp
in
g 
N1
0-
11
13
10
-Ja
n
2
4
89
4
La
un
ch
ed
 a
nd
 fl
ew
 n
ea
r A
ka
ha
ng
a.
 C
ra
sh
ed
 
du
rin
g 
la
nd
in
g 
b/
c 
sm
al
l l
an
di
ng
 zo
ne
 a
nd
 to
o 
w
in
dy
. C
am
e 
up
 ~
30
m
 s
ho
rt 
of
 la
nd
in
g 
zo
ne
14
11
-Ja
n
1
5
85
3
Fl
ew
 in
 th
e 
m
or
ni
ng
. L
ig
ht
 w
in
d 
&
 s
un
ny
15
11
-Ja
n
2
6
62
7
Ha
ng
a 
Te
te
ng
a.
 L
an
de
d 
sh
or
te
r t
ha
n 
ex
pe
ct
ed
 
b/
c 
m
or
e 
he
ad
w
in
d 
th
an
 1
st
 fl
ig
ht
16
12
-Ja
n
1
7
92
5
Li
ttl
e 
w
in
d.
 L
au
nc
h 
~9
am
 fr
om
 H
an
ga
 T
et
en
ga
. 
17
17
-Ja
n
2
24
93
4
To
p.
 In
la
nd
 fl
ig
ht
 o
n 
to
p 
of
 N
15
18
17
-Ja
n
1
23
87
8
To
p.
 In
la
nd
 fl
ig
ht
 o
n 
to
p 
of
 N
16
19
12
-Ja
n
2
8
83
1
W
in
d 
sl
ig
ht
ly
 s
tro
ng
er
 th
an
 fl
ig
ht
7.
 C
ra
ck
ed
 th
e 
le
ns
e 
fli
gh
te
r c
ov
er
 fo
r c
am
er
a 
on
 la
nd
in
g
20
14
-Ja
n
1
12
10
53
Ta
bl
e 
1:
 B
as
ic
 F
lig
ht
 D
at
a 
 
 
90  
   N
um
be
r (
N)
Fl
ig
ht
 D
at
e
Da
te
 F
lig
ht
 #
O
ve
ra
ll 
Fl
ig
ht
 #
# 
of
 P
ic
tu
re
s
De
sc
rip
tio
n
21
14
-Ja
n
2
13
11
24
22
3
14
-Ja
n
3&
4
14
&
15
14
=6
65
, 1
5=
37
4
To
ng
ar
ik
i  
co
m
bi
ne
d 
fli
gh
ts
. F
ar
 e
as
t (
rig
ht
) o
f 
So
ut
h 
co
as
t
10
1
9-
Ja
n
3
3/
2b
87
3
Sa
m
e 
ar
ea
 a
s 
fli
gh
t 2
 (N
3)
 b
ut
 w
ith
 th
er
m
al
 
(C
IR
) 
10
2
16
-Ja
n
1
18
57
6
La
rg
e 
ov
er
la
pp
in
g 
fli
gh
t. 
Ac
ci
de
nt
al
ly
 fl
ow
n 
at
 
22
5m
 n
ot
 1
00
m
10
3
16
-Ja
n
5
22
43
6
Ra
nu
 R
ar
ak
u.
 F
lig
ht
 H
ei
gh
t @
22
5m
10
4
17
-Ja
n
4
26
26
6
Te
pe
u 
ou
t o
n 
NW
 c
oa
st
 (l
as
t f
lig
ht
 o
f s
ta
y)
Th
is
 ta
bl
e 
sh
ow
s s
om
e 
of
 th
e 
ba
si
c 
da
ta
 a
bo
ut
 th
e 
va
rio
us
 d
ro
ne
 fl
ig
ht
s f
lo
w
n 
on
 R
ap
a 
N
ui
. T
he
 o
rd
er
 in
 w
hi
ch
 th
e 
fli
gh
ts
 
w
er
e 
flo
w
n 
w
er
e 
no
t l
in
ea
r i
n 
re
la
tio
n 
to
 th
e 
co
as
t, 
th
is
 is
 th
e 
O
ve
ra
ll 
Fl
ig
ht
 #
. F
or
 o
rg
an
iz
at
io
na
l p
ur
po
se
s, 
I r
e-
nu
m
be
re
d 
th
e 
fli
gh
t i
m
ag
es
 in
to
 a
 g
en
er
al
 o
rd
er
 m
ov
in
g 
fr
om
 w
es
t t
o 
ea
st
 a
lo
ng
 th
e 
so
ut
h 
co
as
t a
nd
 th
is
 is
 th
e 
N
um
be
r (
N
). 
N
um
be
rs
 1
-2
23
 
(a
 c
om
bo
 o
f 2
2 
an
d 
23
) a
re
 im
ag
es
 o
f t
he
 so
ut
h 
co
as
t a
nd
 th
e 
on
es
 th
at
 I 
us
ed
 fo
r m
y 
da
ta
. N
um
be
rs
 1
01
-1
04
 w
er
e 
ot
he
r 
fli
gh
ts
 w
e 
fle
w
 a
t d
iff
er
en
t h
ei
gh
ts
 a
nd
 o
f d
iff
er
en
t p
la
ce
s f
or
 re
se
ar
ch
 p
ur
po
se
s b
ut
 a
re
 n
ot
 re
la
te
d 
to
 th
e 
re
se
ar
ch
 I 
am
 
co
nd
uc
tin
g.
 T
he
 D
at
e 
Fl
ig
ht
 #
 re
fe
rs
 to
 th
e 
fli
gh
t’s
 o
rd
er
 in
 th
e 
da
y’
s l
au
nc
he
s s
in
ce
 m
or
e 
th
an
 o
ne
 fl
ig
ht
 w
as
 fl
ow
n 
pe
r d
ay
. 
Th
e 
# 
of
 P
ic
tu
re
s r
ef
er
s t
o 
ho
w
 m
an
y 
im
ag
es
 w
er
e 
ta
ke
n 
by
 th
e 
dr
on
e 
w
hi
le
 fl
yi
ng
 o
ve
r t
he
 a
re
a.
 F
or
 N
1-
22
3,
 th
e 
av
er
ag
e 
nu
m
be
r o
f p
ic
tu
re
s t
ak
en
 p
er
 fl
ig
ht
 w
as
 a
pp
ro
xi
m
at
el
y 
85
3.
 
 
 
91  
    
 Ahu_ID
Pi
c_
O
rd
er
 (N
)
Po
in
tP
ic
ID
Ca
te
go
ry
N
ot
es
1
2
2.
1
ah
u
Al
so
 in
 P
ic
 #
1
2
2
2.
2
ah
u
Cl
os
e 
to
 co
as
t/
cl
iff
. R
am
p 
to
 ri
gh
t
3
2
2.
3
ah
u
Sm
al
l. 
De
pr
es
se
d 
ar
ea
 b
eh
in
d 
it/
on
 o
ce
an
 si
de
. C
IR
 sh
ow
ed
 w
at
er
?
4
2
2.
4
ah
u?
Pa
rt
 w
al
l?
 M
ay
be
 a
ll 
w
al
l?
 N
ea
r d
irt
 ro
ad
5
2
2.
5
im
ag
e 
ah
u
co
ve
re
d 
m
oa
i b
/c
 cl
os
e 
to
 q
ua
rr
y.
 C
lo
se
 to
 ro
ad
 w
he
re
 it
 b
en
ds
6
1
1.
1
ah
u?
Co
ul
d 
ea
si
ly
 ju
st
 b
e 
ro
ck
s/
ol
d 
w
al
l b
its
. C
lo
se
 to
 cl
iff
7
4
4.
1
ah
u?
Q
ue
st
io
na
bl
e 
ah
u.
 R
ig
ht
 o
n 
co
as
t/
cl
iff
8
4
4.
2
ah
u?
Q
ue
st
io
na
bl
e 
ah
u.
 sm
al
l
9
4
4.
3
ah
u?
Q
ue
st
io
na
bl
e 
ah
u.
 n
ea
r c
lif
f. 
ta
rp
 o
ve
r p
ar
t o
f i
t
10
4
4.
4
im
ag
e 
ah
u
La
rg
e.
 w
/ f
al
le
n 
m
oa
i. 
w
al
l i
nc
lu
de
 a
hu
 a
s o
ne
 p
ar
t o
f e
nc
lo
se
m
en
t. 
M
ay
be
 sm
al
l o
ne
 b
eh
in
d?
11
4
4.
5
ah
u
sm
al
le
r a
nd
 o
ff
 to
 th
e 
SE
 si
de
 o
f t
he
 b
ig
 im
ag
e 
ah
u 
(4
.3
)
12
5
5.
1
ah
u
sm
al
le
r. 
To
w
ar
d 
ba
ck
 o
f b
ay
. M
ai
n 
id
en
tif
ie
r p
la
ce
d 
st
on
es
/"
po
rc
h"
13
5
5.
2
im
ag
e 
ah
u
bi
g.
 F
al
le
n 
m
oa
i. 
Sc
at
te
rd
 h
at
s.
 N
ea
r b
ay
/i
nl
et
 a
nd
 o
ce
an
.
14
5
5.
3
ah
u?
M
ay
be
. g
ra
ss
 b
eh
in
d,
 cl
os
e 
to
 cl
iff
 a
nd
 d
irt
 ro
ad
 g
oi
ng
 d
ow
n
15
6
6.
1
ah
u
lo
ng
. M
id
w
ay
 b
tw
 ro
ad
 a
nd
 co
as
t
16
6
6.
2
ah
u
x2
? 
La
rg
e 
ba
ck
 w
al
l s
to
ne
s.
 A
t b
ac
k 
of
 b
ay
/i
nl
et
. 2
nd
 a
hu
 a
t d
ia
go
na
l, 
di
ff
er
en
t a
ge
s?
17
6
6.
3
ah
u?
sm
al
l. 
ba
ck
 a
nd
 w
es
t o
f 6
.4
. I
n 
gr
as
sy
, r
oc
ky
 a
re
a 
ba
ck
 fr
om
 w
at
er
. 
cu
rv
ed
? 
no
 w
in
gs
18
6
6.
4
ah
u
la
rg
e.
 m
ay
be
 im
ag
e?
 N
ea
r r
oc
ky
, f
la
t c
oa
st
19
6
6.
5
ah
u
lo
ng
 ri
gh
t a
lo
ng
 co
as
t. 
no
t c
lif
fy
. p
an
in
su
la
 w
/ g
ra
ss
y 
fli
gh
t f
ie
ld
20
6
6.
6
ah
u
lo
ng
. c
lo
se
 to
 sh
or
e.
 n
ea
r 6
.5
. p
an
in
su
la
 w
/ g
ra
ss
y 
fli
gh
t f
ie
ld
21
6
6.
7
ah
u?
M
ay
be
? 
or
 m
ay
be
 w
al
l. 
ve
ry
 fa
in
t o
ut
lin
e
22
3
3.
1
im
ag
e 
ah
u
Im
ag
e 
ah
u.
 O
n 
"p
oi
nt
" o
f l
an
d.
 N
ot
 cl
iff
y,
 e
sp
ec
ia
lly
 b
ac
k 
of
 n
ea
rb
y 
in
le
t. 
Fa
lle
n 
m
oa
i a
nd
 h
at
s
Ta
bl
e 
2:
 B
as
ic
 A
hu
 D
at
a 
 
 
92  
   A
hu
_I
D
Pi
c_
O
rd
er
 (N
)
Po
in
tP
ic
ID
Ca
te
go
ry
N
ot
es
23
8
8.
1
ah
u?
lo
ng
, w
al
l b
ui
lt 
ov
er
 p
ar
t, 
m
ak
es
 it
 sl
ig
ht
ly
 u
nc
er
ta
in
. N
ea
r b
ac
k 
of
 in
le
t. 
ge
nt
le
 co
as
ta
l s
lo
pe
24
8
8.
2
ah
u
sm
al
l. 
on
 o
th
er
 si
de
 o
f r
oa
d,
 b
ut
 n
ea
r b
ac
k 
of
 in
le
t. 
ba
ck
 w
al
l s
to
ne
s 
cl
ea
r/
in
lin
e
25
8
8.
3
ah
u
lo
ng
, m
ay
be
 2
in
1/
bu
ilt
 o
nt
op
 o
f e
ac
ho
th
er
. N
ea
r c
oa
st
 a
nd
 ro
ad
. g
en
tle
 
sl
op
e.
 n
o 
m
oa
i?
26
9
9.
1
ah
u?
sm
al
l. 
M
ay
be
 a
hu
. N
ea
r r
oa
d 
an
d 
co
as
t. 
no
t t
oo
 st
ee
p 
ac
es
s t
o 
w
at
er
? 
Ro
ck
y
27
9
9.
2
ah
u?
m
ay
be
. o
n 
ot
he
r s
id
e 
of
 ro
ad
 b
ut
 st
ill
 cl
os
e 
to
 w
at
er
. n
ea
r w
al
l. 
no
t t
oo
 
bi
g
28
10
10
.1
im
ag
e 
ah
u
2i
n1
 a
hu
. M
ay
be
 m
or
e.
 ca
n 
se
e 
di
ff
 a
ng
le
s.
 fa
lle
n 
m
oa
i a
nd
 h
at
s.
 n
ea
r 
sl
op
in
g 
w
at
er
 a
cc
es
s
29
10
10
.2
im
ag
e 
ah
u
1 
fa
lle
n 
m
oa
i. 
no
t m
uc
h 
le
ft
. D
ire
ct
ly
 b
eh
in
d 
= 
ea
sy
 sl
op
e 
to
 w
at
er
. 
Pu
na
? 
ba
ck
 o
f i
nl
et
30
10
10
.3
im
ag
e 
ah
u
Fa
lle
n 
m
oa
i a
nd
 h
at
s.
 la
rg
e.
 n
ea
r 1
0.
2 
bu
t m
or
e 
ou
t o
n 
po
in
t o
f l
an
d
31
10
10
.4
ah
u?
sm
al
le
r. 
M
ay
be
 a
hu
. t
o 
rig
ht
 o
f 1
0.
3.
 m
or
e 
cl
iff
y
32
10
10
.5
ah
u
la
rg
er
. a
w
ay
 fr
om
 co
as
t, 
ot
he
r s
id
e 
of
 ro
ad
. B
ig
 b
ac
k 
st
on
es
.
33
10
10
.6
ah
u?
M
ay
be
 a
hu
. h
as
 th
e 
lo
ng
 sh
ap
e,
 b
ut
 m
ay
be
 a
 ch
ic
ke
n 
ho
us
e 
re
co
ns
tu
ct
io
n 
th
in
g?
34
10
10
.7
ah
u?
m
ay
be
 a
hu
? 
M
ai
n 
as
pe
ct
 is
 th
e 
lo
ng
 fo
rm
 o
f r
oc
ks
 w
/ s
qu
ar
ed
 e
nd
s.
 n
o 
bi
g 
st
on
es
35
10
10
.8
ah
u?
m
ay
be
. 1
 si
de
 o
f b
ac
k 
co
nt
in
ue
s a
s w
al
l. 
on
e 
en
d 
sq
ua
re
d.
 S
ee
m
s w
id
e 
at
 p
la
ce
s
36
11
11
.1
ah
u
lo
ng
, i
nt
ac
t w
/ t
ap
er
ed
 e
nd
s a
nd
 b
ul
ge
 in
 m
id
dl
e.
 n
o 
bi
g 
w
al
l s
to
ne
s.
 
N
ea
r e
as
y 
sh
or
e 
ac
ce
ss
, l
ef
t
37
11
11
.2
im
ag
e 
ah
u
sq
ua
ris
h.
 si
de
 w
al
ls
/e
dg
es
 m
os
t d
is
tin
ct
. f
al
le
n 
m
oa
i. 
O
n 
ot
he
r s
id
e 
of
 
ro
ad
 b
ut
 n
ea
r b
ac
k 
of
 in
le
t
 
 
93  
   A
hu
_I
D
Pi
c_
O
rd
er
 (N
)
Po
in
tP
ic
ID
Ca
te
go
ry
N
ot
es
38
13
13
.1
im
ag
e 
ah
u
Cp
t. 
Co
ok
's 
ah
u.
 O
th
er
 si
de
 o
f r
oa
d 
bu
t n
ea
ris
h 
w
at
er
. F
al
le
n 
m
oa
i
39
13
13
.2
ah
u
pa
rt
 o
f l
ar
ge
 b
ac
k 
ro
ck
s c
le
ar
, f
ai
nt
 o
ut
lin
g 
of
 sq
ua
ris
h 
fo
rm
 st
re
tc
hi
ng
 
ou
t f
ro
m
 it
. C
lif
fy
40
13
13
.3
ah
u
2i
n1
? 
re
ct
an
gu
la
r o
ut
lin
e 
pr
et
ty
 cl
ea
r. 
bi
gg
er
 o
ne
 "o
nt
op
" o
f s
m
al
le
r?
41
13
13
.4
ah
u?
ah
u?
 F
ar
 fr
om
 co
as
t. 
1 
fa
lle
n 
m
oa
i n
ea
r. 
fa
in
t s
tr
ai
gh
t l
in
es
 o
f s
to
ne
s.
 
ro
un
de
d 
"p
or
ch
" s
to
ne
s?
42
14
14
.1
ah
u?
m
ay
be
? 
ju
st
 re
ct
an
gu
la
r o
ut
lin
e.
 n
ot
 ra
is
ed
 1
la
ye
r o
f s
to
ne
s?
 fa
r f
ro
m
 
se
a,
 cl
os
e 
to
 o
th
er
 fe
at
ur
es
43
14
14
.2
ah
u?
m
ay
be
? 
lo
ng
 ce
nt
er
 w
al
l?
 w
/ m
an
av
i s
ur
ou
nd
in
g.
 W
al
l w
/ 2
 si
de
s a
nd
 
ce
nt
er
. f
ar
 fr
om
 o
ce
an
44
14
14
.3
im
ag
e 
ah
u
fa
lle
n 
m
oa
i. 
2p
ar
ts
? 
ba
ck
 a
nd
 fr
on
t. 
1 
w
al
l r
ig
ht
 a
lo
ng
 cl
iff
/e
dg
e 
go
in
g 
do
w
n 
to
 th
e 
se
a.
 n
ea
r p
un
a?
46
14
14
.5
ah
u
st
ub
by
/s
qu
ar
is
h.
 1
 b
ig
ge
r, 
1 
sm
al
le
r. 
fa
r f
ro
m
 o
ce
an
. d
on
't 
re
al
ly
 se
e 
hi
nt
 o
f "
w
in
gs
"
47
15
15
.1
im
ag
e 
ah
u
lo
ng
. 2
 p
ar
ts
? 
bu
lg
e 
in
 b
ac
k.
 fa
lle
n 
m
oa
i. 
Ri
gh
t a
lo
ng
 co
as
t w
/ g
en
tle
 
sl
op
e
48
15
15
.2
ah
u?
M
ay
be
 if
 m
or
e 
ov
er
-g
ro
w
n?
 h
as
 h
as
 ri
gh
t s
ha
pe
 b
ut
 ro
ck
s n
ot
 v
er
y 
cl
ea
r. 
ne
ar
 e
as
y 
sl
op
e 
to
 se
a
49
16
16
.1
im
ag
e 
ah
u
sm
al
l t
o 
ha
ve
 fa
lle
n 
m
oa
i (
1?
). 
st
ru
ct
ur
e 
no
t w
el
l d
ef
in
ed
. b
it 
of
 b
ac
k 
w
al
l. 
al
on
g 
co
as
t b
ut
 cl
iff
y
50
16
16
.2
ah
u?
m
ay
be
? 
sm
al
l. 
ha
s w
in
gs
 b
ut
 in
 o
dd
 o
si
tio
n 
w
/ b
ed
 ro
ck
 ru
nn
in
g 
un
de
r. 
N
ot
 q
ui
te
 p
er
p 
to
 co
as
t
51
16
16
.3
ah
u
lo
ng
. t
ap
pe
re
d 
en
ds
. n
ea
r s
lo
pi
ng
 o
ce
an
 a
cc
es
s o
n 
le
ft
, r
ig
ht
, f
ro
nt
(?
)
52
16
16
.4
ah
u
lo
ng
, m
or
e 
sq
ua
re
d.
 ri
gh
t b
y/
be
hi
nd
/f
ro
nt
 o
f s
lo
pe
/r
am
p 
to
 th
e 
oc
ea
n
53
16
16
.5
ah
u?
M
ay
be
? 
de
fin
ite
ly
 m
an
av
ai
 b
ui
lt 
ar
ou
nd
 ce
nt
er
 w
al
l(?
) m
ay
be
 m
an
av
ai
 
on
 si
de
 o
f a
hu
?
 
 
94  
   
 Ahu_
ID
Pi
c_
O
rd
er
 (N
)
Po
in
tP
ic
ID
Ca
te
go
ry
N
ot
es
54
16
16
.6
ah
u?
m
ay
be
? 
ne
ar
 co
as
t (
cl
iff
is
h)
. n
ot
 v
er
y 
cl
ea
r s
ha
pe
, m
or
e 
bl
ob
bi
sh
 w
/ 
ta
il.
55
18
18
.1
ah
u?
ve
ry
 a
hu
 in
 sh
ap
e 
w
/ t
ap
er
ed
 e
nd
s b
ut
 fa
r f
ro
m
 se
a 
an
d 
w
/ m
an
av
ai
 o
n 
ei
th
er
 si
de
56
19
19
.1
im
ag
e 
ah
u
lo
ng
 p
ile
 o
f r
oc
ks
 w
/ t
ap
er
ed
 e
nd
s &
 su
rr
ou
nd
in
g 
sq
ua
re
 o
ut
lin
e.
 n
ea
r 
co
as
t
57
19
19
.2
ah
u?
m
ay
be
? 
Co
ul
d 
be
 w
al
l w
/ s
ur
ou
nd
in
g 
ro
ck
s?
 D
oe
s h
av
e 
bi
t o
f t
ap
er
. 
Cl
os
e 
to
 co
as
t
58
19
19
.3
im
ag
e 
ah
u?
m
ay
be
 a
hu
? 
1 
m
oa
i. 
rig
ht
 p
os
iti
on
 &
sh
ap
e 
bu
t o
dd
 w
al
l?
 in
te
rs
ec
ts
 
pe
rp
en
di
cu
la
r
59
19
19
.4
ah
u?
m
ay
be
 a
hu
 o
r j
us
t w
al
l w
/ s
to
ne
s p
ile
d 
on
 1
 si
de
. n
ea
r l
es
s c
lif
fy
 co
as
t 
ac
ce
ss
60
19
19
.5
ah
u?
Pr
et
ty
 su
re
 n
ot
 a
hu
 b
ut
 m
ay
be
? 
m
or
e 
in
la
nd
 w
/ m
an
av
ai
 &
 1
 fa
lle
n 
m
oa
i
61
20
20
.1
ah
u
ta
pe
re
d 
en
ds
. n
ea
r c
oa
st
 e
as
y 
ac
ce
ss
 to
 o
ce
an
?
62
20
20
.2
im
ag
e 
ah
u
2i
n1
. c
an
 se
e 
ov
er
la
pp
in
g 
ah
u 
st
uc
tu
re
s.
 se
ve
ra
l f
al
le
n 
m
oa
i. 
ea
sy
 o
ce
an
 
ac
ce
ss
63
21
21
.1
ah
u?
m
ay
be
? 
sm
al
l, 
co
ul
d 
be
 w
al
l w
/ r
oc
ks
 o
n 
1 
si
de
. n
ea
r c
oa
st
64
21
21
.2
ah
u?
m
ay
be
? 
sm
al
l, 
so
rt
 o
f t
ap
er
ed
 e
nd
s?
 m
ay
be
 ju
st
 w
al
l (
sp
lit
s w
/ t
ap
er
). 
co
as
t a
cc
es
s
65
21
21
.3
im
ag
e 
ah
u?
bi
g 
ah
u.
 m
or
e 
su
re
 a
hu
 b
ut
 st
ill
. b
ac
k 
w
al
l s
om
e 
ne
ar
 fa
lle
n 
m
oa
i
66
21
21
.4
ah
u?
ah
u?
m
ay
be
 2
in
1.
en
ds
 se
em
ed
 sq
ua
re
d 
of
f, 
co
up
le
 m
an
av
ai
. n
ot
 q
ui
te
 
rig
ht
 a
ng
le
 to
 co
as
t
 
 
95  
   Ahu_
ID
Pi
c_
O
rd
er
 (N
)
Po
in
tP
ic
ID
Ca
te
go
ry
N
ot
es
67
21
21
.5
ah
u?
m
ay
be
? 
w
or
ki
ng
 o
ff
 n
at
ur
al
 ro
ck
 fe
at
ur
es
? 
ju
st
 o
ne
 e
nd
. n
ea
r c
oa
st
 
ac
ce
ss
 (b
eh
in
d&
to
 si
de
)
68
22
22
.1
im
ag
e 
ah
u
To
ng
ar
ik
i. 
N
um
er
ou
s r
e-
er
ec
te
d 
ah
u.
 e
as
y 
oc
ea
n 
ac
ce
ss
. a
t b
ac
k 
of
 in
le
t
69
23
23
.1
im
ag
e 
ah
u
sm
al
l a
hu
 w
/ f
al
le
n 
m
oa
i. 
ca
n 
se
e 
on
e 
sq
ua
re
d 
en
d.
 B
it 
fa
r f
ro
m
 w
at
er
 
bu
t e
as
y 
co
as
t a
cc
es
s
Th
is
 ta
bl
e 
sh
ow
s t
he
 d
at
a 
as
so
ci
at
ed
 w
ith
 th
e 
ah
u 
I i
de
nt
ifi
ed
 o
n 
th
e 
su
rf
ac
e.
 A
hu
_I
D
 re
fe
rs
 a
n 
in
di
vi
du
al
 ID
 n
um
be
r g
iv
en
 to
 e
ac
h 
ah
u 
an
d 
is
 u
se
d 
in
 o
th
er
 ta
bl
es
 to
 c
on
ne
ct
 d
at
a 
to
 th
ei
r a
ss
oc
ia
te
d 
ah
u.
 P
ic
_O
rd
er
 (N
) r
ef
er
s t
o 
th
e 
ge
ne
ra
l p
la
ce
m
en
t o
f t
he
 a
hu
’s
 
so
ur
ce
 im
ag
e’
s p
os
iti
on
 a
lo
ng
 th
e 
so
ut
h 
co
as
t i
n 
re
la
tio
n 
to
 th
e 
ot
he
r i
m
ag
es
 (s
ee
 ta
bl
e 
1)
 a
nd
 th
e 
Po
in
tP
ic
ID
 g
iv
es
 re
fe
re
nc
e 
to
 th
e 
so
ur
ce
 im
ag
e 
w
hi
le
 a
ls
o 
id
en
tif
yi
ng
 it
 in
 re
la
tio
n 
to
 th
e 
ot
he
r a
hu
 a
ss
oc
ia
te
d 
w
ith
 th
at
 im
ag
e.
 I 
pl
ac
ed
 e
ac
h 
po
in
t/l
oc
at
io
n 
in
 o
ne
 o
f 4
 
ca
te
go
rie
s:
 “
A
hu
” 
m
ea
ns
 th
at
 I 
am
 su
re
 th
at
 th
e 
fe
at
ur
e 
is
 a
n 
ah
u 
bu
t i
t h
as
 n
o 
m
oa
i, 
“I
m
ag
e 
ah
u”
 m
ea
ns
 th
at
 th
e 
fe
at
ur
e 
is
 a
n 
ah
u 
an
d 
th
at
 th
er
e 
ar
e 
m
oa
i o
n/
ne
ar
 it
, “
A
hu
?”
 m
ea
ns
 th
at
 I 
am
 n
ot
 su
re
 if
 th
e 
st
ru
ct
ur
e 
is
 a
n 
ah
u 
(ie
. i
t c
ou
ld
 b
e 
so
m
e 
ot
he
r b
e 
an
 o
ld
 
w
al
l o
r j
us
t r
ub
bl
e)
 a
nd
 “
Im
ag
e 
ah
u”
 re
fe
rs
 to
 st
ru
ct
ur
es
 w
ith
 n
ea
rb
y/
as
so
ci
at
ed
 m
oa
i b
ut
 I 
ca
n’
t b
e 
su
re
 if
 th
is
 is
 a
cc
id
en
ta
l o
f t
he
 
st
ru
ct
ur
e 
re
al
ly
 is
 a
n 
ah
u.
 T
hi
s i
s t
he
 sm
al
le
st
 c
at
eg
or
y.
 
 
 
96  
 
Ta
bl
e 
3:
 A
hu
 E
le
va
tio
n 
D
at
a 
Ahu_ID
Catego
ry
COUNT
 (pixels
) ARE
A
MIN
MAX
RANGE
MEAN
STD
1
ahu
5258
78.170
52985
35.770
54596
38.970
62302
3.2000
77057
37.442
82974
0.7508
44709
2
ahu
5257
78.155
66288
20.219
80667
21.805
11475
1.5853
08075
20.905
96939
0.3915
22837
3
ahu
5259
78.185
39682
15.374
93896
17.173
73276
1.7987
93793
16.467
17508
0.3672
14494
4
ahu?
5250
78.051
59408
29.432
26624
30.588
23967
1.1559
73434
30.163
13582
0.2395
5855
5im
age ah
u
5263
78.244
8647
40.988
1897
43.957
16095
2.9689
71252
42.793
00643
0.9140
0665
6
ahu?
10874
78.153
37235
22.323
30132
26.713
71651
4.3904
15192
25.003
21591
1.2577
70847
7
ahu?
14573
78.117
63711
14.068
81905
15.723
85597
1.6550
36926
14.921
61981
0.3872
86827
8
ahu?
14566
78.080
11405
9.1146
54541
10.461
66801
1.3470
13474
9.5476
80665
0.2552
13843
9
ahu?
14580
78.155
16016
3.1313
50517
3.8471
53187
0.7158
0267
3.4326
46997
0.1593
51965
10
image 
ahu
14589
78.203
40409
2.7678
88546
4.8349
0324
2.0670
14694
3.6311
77232
0.5145
06547
11
ahu
14579
78.149
79972
0.8013
6013
2.6770
35809
1.8756
75678
1.7990
51096
0.2517
7592
12
ahu
17659
78.158
27909
-1.5768
96667
1.1741
06598
2.7510
03265
0.5260
74197
0.5484
69785
13
image 
ahu
11685
78.129
52953
-4.2040
84396
-1.4114
96997
2.7925
87399
-2.8606
6489
0.7332
01911
14
ahu?
17649
78.114
01935
5.5243
64471
7.0937
70981
1.5694
06509
6.1846
84635
0.3624
83912
15
ahu
17659
78.158
27909
15.191
66946
17.515
83099
2.3241
6153
16.248
42639
0.5021
46447
16
ahu
16437
78.166
79441
3.7871
25349
6.2881
25992
2.5010
00643
5.2493
79122
0.4679
10793
17
ahu?
16427
78.119
23902
0.1125
29993
1.0767
05217
0.9641
75224
0.5075
91373
0.2114
64418
18
ahu
16443
78.195
32764
-0.2936
37991
2.0754
10366
2.3690
48357
0.7559
78393
0.6045
13494
19
ahu
16427
78.119
23902
0.3042
61923
2.0332
38411
1.7289
76488
1.2943
043
0.3209
7155
20
ahu
16425
78.109
72794
2.0256
30951
4.0058
32672
1.9802
01721
2.9217
42529
0.5410
35149
21
ahu?
18549
78.162
78035
5.7666
9693
7.7069
86427
1.9402
89497
6.4790
44049
0.4740
87426
22
image 
ahu
15168
78.111
95432
10.264
97841
12.332
26776
2.0672
89352
11.243
30811
0.4691
34663
23
ahu?
18547
78.154
35264
8.9498
39592
12.500
40531
3.5505
6572
10.270
94318
0.7882
74246
24
ahu
18543
78.137
49723
8.6014
45198
10.365
73696
1.7642
91763
9.3461
8184
0.3110
50543
25
ahu
19624
78.119
84435
10.022
13669
13.378
67928
3.3565
42587
11.780
43277
1.1508
16137
26
ahu?
19635
78.163
6335
4.7948
15063
6.5660
35271
1.7712
20207
5.7884
79132
0.3574
15533
27
ahu?
19627
78.131
78685
7.8927
07348
9.3132
72476
1.4205
65128
8.5300
46856
0.2822
89441
28
image 
ahu
21249
78.159
44336
2.3526
89743
4.6074
76711
2.2547
86968
3.6080
67121
0.4241
0477
Ahu an
d Buffe
r
Elevati
on (m)
 
 
97  
 Ahu_ID
Catego
ry
COUNT
 (pixels
)ARE
A
MIN
MAX
RANGE
MEAN
STD
29
image 
ahu
4684
78.237
80392
-1.9470
60943
0.0349
6033
1.9820
21272
-1.1036
4924
0.4625
57153
30
image 
ahu
4678
78.137
5847
3.8645
70379
7.2818
21251
3.4172
50872
5.3963
67391
0.8422
49916
31
ahu?
4679
78.154
2879
6.8528
33271
8.4231
28128
1.5702
94857
7.8561
78488
0.4424
94983
32
ahu
21243
78.137
37377
23.445
6749
27.999
99619
4.5543
21289
25.946
76597
1.4377
17227
33
ahu?
4679
78.154
2879
9.4600
72517
11.108
94775
1.6488
75237
10.216
09396
0.3742
69712
34
ahu?
4677
78.120
8815
5.2754
79317
5.8090
27672
0.5335
48355
5.5717
41895
0.1362
02442
35
ahu?
21242
78.133
69551
-0.8338
46092
1.2984
15661
2.1322
61753
0.3968
86331
0.6019
97672
36
ahu
4679
78.154
2879
8.5307
69348
9.8959
47456
1.3651
78108
9.1380
55875
0.3655
81586
37
image 
ahu
4677
78.120
8815
21.228
24669
23.527
91595
2.2996
69266
22.351
33288
0.5123
11614
38
image 
ahu
5182
78.163
45831
20.966
7511
23.182
81174
2.2160
60638
21.844
12381
0.4031
66924
39
ahu
5188
78.253
96019
16.009
92012
17.760
12421
1.7502
04086
17.082
40119
0.3842
59015
40
ahu
5178
78.103
12372
11.916
65649
13.772
26448
1.8556
07986
12.891
51498
0.4223
73723
41
ahu?
5173
78.027
70549
27.003
85857
27.856
16302
0.8523
04459
27.286
61498
0.1807
38245
42
ahu?
4815
78.208
72694
33.903
41187
34.099
13635
0.1957
24487
34.018
77025
0.0467
41405
43
ahu?
4805
78.046
29968
25.893
00919
28.543
95294
2.6509
43756
27.031
42863
0.5756
30944
44
image 
ahu
4810
78.127
51331
10.156
951
2.3053
7987
2.1484
29871
11.253
39158
0.5584
99378
45
ahu
4813
78.176
24149
11.786
97014
13.463
78613
1.6768
15987
12.596
34131
0.3343
69351
46
ahu
19513
78.142
17837
9.4368
17169
11.253
18718
1.8163
7001
10.371
24635
0.4769
47202
47
image 
ahu
19502
78.098
12753
1.3047
90497
3.7911
17668
2.4863
27171
2.5603
97601
0.6519
73325
48
ahu?
19508
78.122
15526
-1.3887
78687
4.1118
77441
5.5006
56128
2.0340
32132
1.4380
38752
49
image 
ahu
19509
78.126
15988
4.6252
86102
7.0747
36595
2.4494
50493
5.5840
55971
0.5860
35523
50
ahu?
19385
78.142
19406
7.5972
30911
9.0054
49295
1.4082
18384
8.2779
96634
0.2965
70446
51
ahu
19385
78.142
19406
7.1886
2772
9.3853
05405
2.1966
77685
8.1503
53672
0.4853
39449
52
ahu
18320
77.246
89766
8.2552
61421
9.8676
65291
1.6124
0387
9.0938
31172
0.2479
12105
53
ahu?
19391
78.166
38045
34.149
80698
36.457
48138
2.3076
74408
35.340
35635
0.5620
06252
54
ahu?
19385
78.142
19406
4.5315
7568
6.5415
06767
2.0099
31087
5.5505
9201
0.3929
26823
55
ahu?
32286
78.118
10503
23.997
13898
26.033
8974
2.0367
58423
25.044
06288
0.5694
39142
56
image 
ahu
18533
78.145
01934
9.5108
54721
11.773
0093
2.2621
54579
10.874
44894
0.5583
95379
57
ahu?
18534
78.149
23588
-0.2738
30414
1.7585
51598
2.0323
82011
0.7595
69471
0.4469
15147
Ahu an
d Buffe
r
Elevati
on (m)
 
 
98  
Ahu_ID
Catego
ry
COUNT
 (pixels
)ARE
A
MIN
MAX
RANGE
MEAN
STD
58i
mage a
hu?
18535
78.153
45241
0.4667
11044
2.8349
00856
2.3681
89812
1.5760
06969
0.7023
91363
59
ahu?
16121
78.091
30284
9.4167
2802
11.545
56274
2.1288
34724
10.190
61797
0.5258
57568
60
ahu?
16132
78.144
58765
18.854
96521
21.640
41901
2.7854
53796
19.661
82602
0.5837
37191
61
ahu
16143
78.197
87245
3.6537
02736
5.7631
71196
2.1094
6846
4.6559
40417
0.5503
95289
62
image 
ahu
16126
78.115
52321
9.3040
86685
13.039
99519
3.7359
08508
10.271
77079
0.8047
85464
63
ahu?
16784
78.132
11141
6.4425
01068
7.8335
58559
1.3910
57491
7.0286
41689
0.2958
50092
64
ahu?
16783
78.127
45626
7.3163
51414
9.4251
89972
2.1088
38558
8.2372
34116
0.5693
10422
65i
mage a
hu?
16780
78.113
4908
8.9232
36847
11.631
0091
2.7077
72255
10.475
0103
0.5147
47284
66
ahu?
16786
78.141
42172
11.063
591
12.957
88383
1.8942
92831
11.552
18533
0.3135
66774
67
ahu?
16794
78.178
66296
1.0148
77319
6.9120
35465
5.8971
58146
4.6481
00508
1.4705
21723
68
image 
ahu
4136
78.189
96521
-1.6781
09646
-0.9667
03296
0.7114
0635
-1.3786
5741
0.1916
36427
69
image 
ahu
4139
78.246
6794
3.8739
42137
5.4145
298
1.5405
87664
4.5467
18179
0.3177
81661
Ahu an
d Buffe
r
Elevati
on (m)
Th
is
 ta
bl
e 
sh
ow
s t
he
 d
at
a 
su
m
m
ar
y 
fo
r t
he
 e
le
va
tio
n 
su
rr
ou
nd
in
g 
ea
ch
 a
hu
. T
hi
s t
ab
le
 is
 a
 re
su
lt 
of
 ru
nn
in
g 
th
e 
Zo
na
l S
ta
tis
tic
s a
s 
Ta
bl
e 
to
ol
 w
ith
 a
 5
m
 b
uf
fe
r a
ro
un
d 
ea
ch
 o
f t
he
 a
hu
. T
he
 ra
st
er
 b
ei
ng
 u
se
d 
to
 e
xt
ra
ct
 th
e 
da
ta
 is
 th
e 
D
EM
. T
he
 e
le
va
tio
n 
da
ta
 is
 a
 
su
m
m
ar
y 
of
 th
e 
el
ev
at
io
n 
ra
st
er
 d
at
a 
fo
un
d 
w
ith
in
 th
e 
5m
 b
uf
fe
r. 
Fo
r t
ho
se
 a
hu
 th
at
 h
ad
 m
or
e 
th
an
 o
ne
 se
t o
f a
ss
oc
ia
te
d 
el
ev
at
io
n 
da
ta
 d
ue
 to
 o
ve
rla
pp
in
g 
fli
gh
t a
re
as
, t
ho
se
 re
pe
tit
io
ns
 w
er
e 
re
m
ov
ed
 fr
om
 th
e 
ta
bl
e 
w
ith
 n
o 
ad
di
tio
na
l w
or
k 
be
ca
us
e 
th
e 
5m
 b
uf
fe
r 
ar
ea
 w
as
 sm
al
l e
no
ug
h 
th
at
 e
ac
h 
co
ul
d 
be
 fu
lly
 c
ov
er
ed
 b
y 
at
 le
as
t o
ne
 o
f t
he
 o
ve
rla
pp
in
g 
fli
gh
ts
. 
  
99  
Table 4: Total Observation Area per Ahu 
Ahu_ID Category Total area 1 ahu 27270207 2 ahu 41267371 3 ahu 6416704 4 ahu? 32789215 5 image ahu 38590412 6 ahu? 28978536 7 ahu? 126231516 8 ahu? 78724656 9 ahu? 61513165 10 image ahu 87982982 11 ahu 96667308 12 ahu 55351865 13 image ahu 28039359 14 ahu? 86308918 15 ahu 152277219 16 ahu 62193056 17 ahu? 49362420 18 ahu 56283049 19 ahu 62274463 20 ahu 62997228 21 ahu? 160892525 22 image ahu 72199192 23 ahu? 91371303 24 ahu 68968654 25 ahu 164217474 26 ahu? 123914565 27 ahu? 104118609 28 image ahu 65125910 29 image ahu 15094927 30 image ahu 66526067 31 ahu? 105714152 32 ahu 105382698 33 ahu? 63155748 34 ahu? 57577396 35 ahu? 41362115 36 ahu 16059996 37 image ahu 46694941 38 image ahu 27849872 39 ahu 30543048 40 ahu 27190978 41 ahu? 24913031 42 ahu? 17032999 43 ahu? 23966681 44 image ahu 21335856 
 
 
100  
Ahu_ID Category Total area 46 ahu 66748032 47 image ahu 46700787 48 ahu? 25951425 49 image ahu 81685118 50 ahu? 80862453 51 ahu 57357014 52 ahu 123892964 53 ahu? 114549902 54 ahu? 70756356 55 ahu? 74438721 56 image ahu 104562093 57 ahu? 54411863 58 image ahu? 49872018 59 ahu? 138399675 60 ahu? 153886299 61 ahu 101919250 62 image ahu 98893844 63 ahu? 67521589 64 ahu? 82310538 65 image ahu? 101766741 66 ahu? 111177724 67 ahu? 45927297 68 image ahu 19201560 69 image ahu 38346306 
 
A summary table showing the total amount of area from which each ahu can be seen. 
These numbers were acquired from the point data observations. For each ahu point, the 
area from all the rows that had a 1 in it (indicating visibility) were added together and 
then associated with the basic ahu information from table 3. For those ahu located in 
more than one flight rea, they had two different viewsheds. One of these viewsheds was 
clipped by the other in order to eliminate overlap and redundant data. The two 
viewsheds were then added together to get a more complete assessment of land that was 
not a part of the previous flight area, and thus makes the overall area for these ahu more 
accurate. Unfortunately, not all the ahu have this position, and most ahu probably 
indicate having less observable area than they do in reality because they are only being 
analyzed wthin their specific flight area and not the whole coast.  
   
 
 
101  
Table 5: Closest Coast Points to Ahu Associated Ahu Data Closest Coast Point Ahu_ID Category NEAR_DIST (m) NEAR_X NEAR_Y 1 ahu 217.2275975600 659494.542301 6993805.84591 2 ahu 35.6039123486 659483.429778 6993803.30591 3 ahu 74.8716587650 659667.262646 6993759.01457 4 ahu? 145.7052238760 659792.552462 6993910.04158 5 image ahu 159.7561154090 659803.931983 6993922.54071 6 ahu? 49.5485065790 659009.802026 6993910.94393 7 ahu? 48.5971014531 661167.722098 6994020.17422 8 ahu? 60.0561808506 661445.124572 6993922.45018 9 ahu? 59.1630980914 661658.849100 6993842.50013 10 image ahu 83.7072109411 662030.149142 6994009.44967 11 ahu 36.3323681853 662034.774878 6994001.24360 12 ahu 16.1397631470 662239.121522 6994207.14176 13 image ahu 33.6507134615 662258.982838 6994101.75168 14 ahu? 33.2078723512 662725.679458 6994254.37529 15 ahu 115.4937263430 662855.652218 6994351.00132 16 ahu 31.4781502604 662958.887363 6994410.72593 17 ahu? 52.0368184976 663100.415723 6994337.87101 18 ahu 31.8053701825 663130.186245 6994327.18084 19 ahu 48.6100742748 663406.043135 6994257.20690 20 ahu 55.5172763245 663468.284945 6994313.45846 21 ahu? 53.7486735677 663524.780867 6994377.26015 22 image ahu 59.1449327589 660502.665871 6993957.78562 23 ahu? 51.0607194931 663591.841036 6994749.19759 24 ahu 90.1874834291 663517.086635 6994753.34493 25 ahu 93.7450148079 663944.970040 6995072.38428 26 ahu? 51.0566937832 664049.112876 6995220.36852 27 ahu? 123.4481060930 664182.020355 6995290.43646 28 image ahu 44.8810859066 664586.513995 6995810.90176 29 image ahu 35.5910981939 664763.789925 6995984.96916 30 image ahu 42.6636718265 664838.516820 6995932.66733 31 ahu? 41.6711975387 665052.444241 6996011.01616 32 ahu 281.1863909640 664760.922767 6995984.39619 33 ahu? 119.1496077010 664769.601619 6995983.33286 34 ahu? 85.4942758100 664847.513656 6995930.88607 35 ahu? 51.7313752519 664692.440731 6995891.48609 36 ahu 46.6483533891 665397.522160 6996201.04867 37 image ahu 137.4171584260 665501.834890 6996392.26597 38 image ahu 183.6820164540 665868.345652 6996620.68169 39 ahu 60.2796169566 665893.313946 6996653.53237 40 ahu 35.6362636393 666165.376364 6996788.07144 41 ahu? 303.2488911810 666235.395650 6996844.65230 42 ahu? 312.3470079380 666288.173304 6996884.61438 43 ahu? 183.7133399190 666329.133711 6996893.37530 
 
 
102  
Associated Ahu Data Closest Coast Point Ahu_ID Category NEAR_DIST (m) NEAR_X NEAR_Y 44 image ahu 31.8517616434 666631.909768 6996887.76435 45 ahu 36.3399356277 666927.298177 6996977.32036 46 ahu 256.1859019310 667172.850592 6997148.74032 47 image ahu 43.7039928550 667235.332898 6997154.00352 48 ahu? 27.7703706420 667326.431511 6997151.71344 49 image ahu 44.0301275507 667753.798914 6997153.42072 50 ahu? 72.5546343842 668209.128493 6996820.13463 51 ahu 30.8559034373 668431.259745 6996878.83157 52 ahu 51.5795743990 668470.532582 6996914.22597 53 ahu? 413.7235394280 668104.536855 6996915.17842 54 ahu? 37.1403624253 667909.466572 6996986.14110 55 ahu? 506.2277938750 668693.468820 6997028.07596 56 image ahu 50.4988756399 668646.266140 6996977.74866 57 ahu? 83.8409202663 668697.966829 6997029.53889 58 image ahu? 37.1446319874 668906.918763 6997016.32622 59 ahu? 132.0229110560 669225.040314 6996692.37845 60 ahu? 282.9764710180 669523.913768 6996920.41297 61 ahu 91.6472449230 669889.170376 6996834.21607 62 image ahu 60.6691312905 670185.848395 6997070.72083 63 ahu? 53.1627908877 670441.450314 6997439.84993 64 ahu? 62.5352616575 670613.256435 6997563.53451 65 image ahu? 107.6961487150 670729.660894 6997712.34357 66 ahu? 78.3144750139 670830.555134 6997808.20049 67 ahu? 44.4738161174 671098.706639 6997979.12795 68 image ahu 82.7930983721 670849.631377 6998425.27019 69 image ahu 153.3925149840 671316.365271 6998701.88861  
This table shows data about the closest point on the coast for each ahu. This data does 
not take into account elevation, merely distance. NEAR_DIST refers to the distance 
between each ahu and its associated point on the coast and the NEAR_X and Y 
columns relate to coordinate points. These coordinate points were used to map the 
points in ArcMap.   
 
 
103  
  
Ta
bl
e 
6:
 D
at
a 
fo
r S
lo
pe
 w
ith
in
 5
0m
 o
f C
lo
se
st
 C
oa
st
 P
oi
nt
s o
f A
hu
 
Ahu_ID
Catego
ryC
OUNT
AREA (
m)
MIN
MAX
RANGE
MEAN
STD
1ahu
70920
9
5097.2
10206
0.0247
9816
87.832
06177
87.807
26361
19.060
85039
14.883
78769
2ahu
91076
654.57
92801
0.0063
56403
86.613
62457
86.607
26817
19.774
60163
15.313
73583
3ahu
68086
4
4893.4
8969
0.0032
79554
84.839
07318
84.835
79363
15.027
54326
10.830
65903
4ahu?
32541
1
4837.8
75673
0.1999
58712
87.466
07971
87.266
121
17.494
59308
16.265
9421
5imag
e ahu
10210
1
1517.9
32535
0.5397
70186
87.407
15027
86.867
38008
15.006
21585
14.306
83171
6ahu?
57387
0
4124.5
04935
0.0332
86132
88.775
58899
88.742
30286
30.594
475
16.901
52916
7ahu?
90468
4
4849.5
00886
0.0137
63064
87.996
47522
87.982
71216
24.300
01271
17.560
36123
8ahu?
76465
2
4098.8
6828
0.0073
88444
88.992
55371
88.985
16527
18.683
79489
17.141
91324
9ahu?
99511
1
5334.2
29052
0.0132
87202
87.844
17725
87.830
89004
16.714
53643
12.470
47077
10ima
ge ahu
83835
0
4493.9
21709
0.0073
76655
84.388
86261
84.381
48595
14.597
69239
13.174
81908
11ahu
90049
602.09
55075
0.0233
84731
81.747
28394
81.723
8992
19.171
56568
14.766
95757
12ahu
77909
2
5209.2
50443
0.0049
7158
86.184
68475
86.179
71317
8.5967
62515
9.7448
51263
13ima
ge ahu
83082
2
5555.1
33246
0.0126
17367
84.751
00708
84.738
38971
16.438
28625
12.725
11878
14ahu
?
10245
84
4534.7
8239
0.0115
25417
88.443
8858
88.432
36039
17.697
27902
16.214
9032
15ahu
11819
72
5620.9
13933
0.0253
23501
87.944
54193
87.919
21843
22.931
03343
16.429
16911
16ahu
10902
71
5184.8
26252
0.0112
13547
88.511
12366
88.499
91011
18.686
72535
14.192
16883
17ahu
?
10008
89
4759.7
66666
0.0076
00136
86.383
88824
86.376
28811
17.098
55154
11.881
55967
18ahu
45126
9
2146.0
27325
0.0160
22649
86.685
30273
86.669
28009
16.900
16892
14.250
76499
19ahu
11526
64
5481.5
38596
0.0062
02784
88.480
69763
88.474
49485
17.166
94183
16.813
82844
20ahu
86127
6
4095.8
31601
0.0060
26874
88.329
78821
88.323
76133
15.440
60108
14.349
53907
21ahu
?
86294
7
4103.7
78106
0.0153
88933
88.634
26971
88.618
88078
19.291
27891
16.581
15338
22ima
ge ahu
89110
9
4589.0
20669
0.0039
31223
87.863
57117
87.859
63994
21.141
32069
16.663
43303
23ahu
?
10576
13
4456.6
26913
0.0028
62258
88.427
52838
88.424
66612
24.446
48609
16.342
25932
24ahu
11481
30
4838.0
52348
0.0190
38264
85.147
23206
85.128
19379
16.390
5182
10.992
7908
25ahu
10137
36
4271.7
35635
0.0099
10981
87.288
45978
87.278
5488
19.879
05224
15.436
93353
26ahu
?
12110
13
4820.8
39129
0.0193
01662
88.082
72552
88.063
42386
22.564
29648
15.396
00415
27ahu
?
10872
67
4328.2
2711
0.0131
52015
86.394
06586
86.380
91384
20.253
32501
13.671
71979
28ima
ge ahu
12935
59
4758.0
52209
0.0089
18962
88.557
21283
88.548
29387
21.020
02451
16.666
2712
Ahu an
d Buffe
r
Slope (
degree
s)
 
 
104  
 
 Ahu_I
DC
ategor
yCO
UNT
AREA (
m)
MIN
MAX
RANGE
MEAN
STD
29ima
ge ahu
15690
48
5771.3
73631
0.0133
93601
88.252
22778
88.238
83418
17.740
96645
13.665
24753
30ima
ge ahu
45478
07
596.28
2764
0.0108
95002
89.608
9325
89.598
03749
24.027
71891
18.645
55348
31ahu
?
29401
7 4
911.02
5704
0.0345
72221
86.892
38739
86.857
81517
18.794
10263
15.009
05608
32ahu
57529
211.60
68811
0.0239
71824
72.709
27429
72.685
30247
18.524
03304
12.638
60812
33ahu
?
19891
332.24
34154
0.7551
0931
74.886
94763
74.131
83832
10.082
30328
12.349
73079
34ahu
?
54788
915.13
50985
0.1135
08373
69.886
56616
69.773
05779
11.294
32074
10.909
40488
35ahu
?
11025
50
4055.4
70576
0.0147
56956
85.357
34558
85.342
58862
18.256
62083
13.002
88429
36ahu
26912
54
495.24
9569
0.0734
42087
88.384
09424
88.310
65215
20.471
64886
14.687
96891
37 ima
ge ahu
32620
55
448.66
8409
0.0417
8717
88.761
80267
88.720
0155
21.229
39754
15.503
08675
38ima
ge ahu
29985
44
522.18
0095
0.0675
86184
85.896
55304
85.828
96686
19.082
01314
13.100
27719
39ahu
17423
12
627.62
5312
0.0549
60754
85.734
16901
85.679
20825
15.412
14711
14.749
77858
40ahu
25207
93
801.67
2268
0.0098
04519
85.066
94794
85.057
14342
14.127
25228
13.605
21569
41ahu
?
26904
34
369.89
9792
0.1059
97883
87.506
65283
87.400
65495
18.021
69157
14.535
92449
42ahu
?
23918
23
607.16
9088
0.0196
40561
87.825
90485
87.806
26429
16.151
37629
16.948
32768
43ahu
?
17727
0
2879.2
8746
0.0425
44022
87.674
31641
87.631
77238
15.072
42018
14.657
48836
44ima
ge ahu
23974
53
894.03
0418
0.0245
56693
86.355
0415
86.330
48481
15.573
05378
16.378
39365
45ahu
25863
2
4200.8
0033
0.0140
46267
88.577
23999
88.563
19372
16.871
1572
15.245
5355
46ahu
29931
1
4861.5
2428
0.0189
94343
83.536
78131
83.517
78697
16.030
49615
11.597
55282
47ima
ge ahu
63347
1 2
536.70
2109
0.0087
77293
86.765
61737
86.756
84008
19.138
51451
12.984
9413
48ahu
?
12404
86
4967.4
62525
0.0185
44413
88.478
90472
88.460
36031
19.688
5975
13.459
81044
49 ima
ge ahu
10255
27
4133.7
8815
0.0057
59308
89.237
05292
89.231
29361
20.138
19784
14.998
53821
50ahu
?
14121
92
5692.3
92843
0.0082
98002
87.015
22827
87.006
93027
15.242
23171
12.048
65852
51ahu
16660
56
6715.6
91104
0.0121
89744
88.199
31793
88.187
12819
18.783
00839
13.635
28819
52ahu
97118
7 3
914.74
9502
0.0262
45711
88.651
6037
88.625
35799
22.472
10284
16.159
73082
53ahu
?
11939
71
4812.7
67652
0.0072
18701
86.655
22766
86.648
00896
15.657
83296
12.097
50008
54ahu
?
12124
73
4887.3
47208
0.0085
95416
86.992
63763
86.984
04222
15.928
33699
13.064
47584
55 ahu
?
11778
61
4966.4
7398
0.0049
09046
87.868
62183
87.863
71278
18.509
45772
13.440
16372
56 ima
ge ahu
10516
32
4434.2
26928
0.0064
78092
89.089
8056
89.083
32751
17.589
53436
14.552
93413
57 ahu
?
52424
221.04
68229
0.0427
34116
74.898
72742
74.855
9933
14.653
91266
10.650
26823
Ahu an
d Buffe
r
Slope (
degree
s)
 
 
105  
 Ahu_I
DC
ategor
yCO
UNT
AREA (
m)
MIN
MAX
RANGE
MEAN
STD
58ima
ge ahu
?116
0524
4893.3
7218
0.0107
10596
87.331
25305
87.320
54246
20.710
11742
15.073
50405
59ahu
?
13622
66
5744.0
21274
0.0097
02253
86.459
75494
86.450
05269
16.148
34867
11.983
77584
60ahu
?
10397
64
5036.3
29453
0.0339
99287
89.184
69238
89.150
6931
20.786
37942
15.537
08897
61ahu
10039
41
4862.8
12742
0.0037
10685
78.674
11041
78.670
39973
14.973
18804
10.950
62338
62ima
ge ahu
10426
61
5050.3
6172
0.0180
47499
88.693
71033
88.675
66283
20.363
9827
16.344
7975
63ahu
?
94099
64
380.37
7408
0.0108
12469
86.700
51575
86.689
70328
17.478
29287
13.437
38049
64ahu
?
12159
48
5660.2
90957
0.0049
41953
87.147
10999
87.142
16803
20.325
80168
15.251
03728
65ima
ge ahu
?83
9833
3909.4
59233
0.0049
17049
83.023
71216
83.018
79511
18.132
5362
12.337
8501
66ahu
?
12449
04
5795.0
82399
0.0191
8116
88.260
79559
88.241
61443
20.846
04179
16.214
1321
67ahu
?
10099
49
4701.3
5663
0.0058
5656
88.254
34113
88.248
48457
16.176
06232
13.520
83285
68ima
ge ahu
33743
36
378.39
9686
0.0039
92422
72.507
04956
72.503
05714
9.6642
8812
9.3694
42555
69ima
ge ahu
22587
54
269.65
0654
0.0257
96553
88.156
41785
88.130
62129
8.4122
11847
7.5685
28765
Ahu an
d Buffe
r
Slope (
degree
s)
Th
is
 ta
bl
e 
sh
ow
s s
lo
pe
 d
at
a 
(in
 d
eg
re
es
) f
ro
m
 w
ith
in
 5
0m
 o
f e
ac
h 
ca
te
go
riz
ed
 a
hu
’s
 c
lo
se
st
 c
oa
st
 p
oi
nt
. T
he
 sl
op
e 
fr
om
 th
e 
fli
gh
t 
im
ag
es
 (m
in
us
 th
e 
oc
ea
n 
ar
ea
s)
 w
er
e 
us
ed
 to
 c
al
cu
la
te
 th
is
 in
fo
rm
at
io
n.
 S
in
ce
 th
e 
fli
gh
ts
 o
ve
rla
p 
in
 p
la
ce
s, 
so
m
e 
po
in
ts
 h
ad
 
as
so
ci
at
ed
 d
at
a 
fr
om
 m
or
e 
th
an
 o
ne
 fl
ig
ht
 a
re
a.
 F
or
tu
na
te
ly
, e
ve
ry
 5
0m
 b
uf
fe
r z
on
e 
w
as
 a
lw
ay
s c
om
pl
et
el
y 
co
ve
re
d 
by
 a
t l
ea
st
 o
ne
 o
f 
th
e 
ov
er
la
pp
in
g 
fli
gh
ts
 so
 th
e 
as
so
ci
at
ed
 d
at
a 
fr
om
 th
e 
in
co
m
pl
et
e 
se
ct
io
ns
 w
er
e 
re
m
ov
ed
 fr
om
 th
e 
ta
bl
e.
   
 
 
 
106  
      
Ta
bl
e 
7:
 S
lo
pe
 d
at
a 
fo
r t
he
 W
ho
le
 C
oa
st
 W
ith
in
 5
0m
 
Picture
 (N)CO
UNT (p
ixels)A
REA (m
)
MIN
MAX
RANGE
MEAN
STD
1
10213
144
73403
.66779
0.0030
02785
89.595
12329
89.592
12051
22.196
75991
16.784
63958
2
40512
41
60229
.67963
0.0072
4162
89.633
9035
89.626
66188
17.059
09297
14.832
68349
3
13784
233
70985
.85038
0.0026
78186
88.926
12457
88.923
44639
18.688
70046
15.805
6167
4
17995
646
96464
.51271
0.0035
11513
89.043
61725
89.040
10574
18.032
11576
15.482
90414
5
10903
901
72906
.85967
0.0021
66135
88.012
11548
88.009
94934
14.604
43084
12.648
11992
6
17125
459
81440
.78803
0.0044
63901
88.634
26971
88.629
80581
17.566
39977
14.718
0999
7
80857
55
35787
.34333
0.0052
61089
88.660
3775
88.655
11641
17.403
10185
14.391
5717
8
16848
831
70998
.51618
0.0028
62258
88.670
95947
88.668
09721
19.684
55267
14.641
46545
9
13894
419
55311
.34578
0.0048
36777
88.467
27753
88.462
44075
18.364
75352
14.711
22008
10
14912
041
54850
.43174
0.0030
98811
89.245
59021
89.242
4914
18.930
47609
15.127
79689
11
51219
90
85553
.6399
0.0088
7517
89.650
00916
89.641
13399
17.666
93952
14.631
01314
13
44764
03
67509
.85661
0.0098
04519
89.524
43695
89.514
63243
16.224
53192
14.575
52222
14
46333
15
75256
.0827
0.0059
19636
88.577
23999
88.571
32035
15.204
11729
13.749
46146
15
11887
828
47604
.19714
0.0017
74575
89.248
38257
89.246
60799
19.537
18309
15.202
90634
16
20995
221
84629
.45957
0.0031
94968
89.237
05292
89.233
85795
17.247
20604
13.808
81859
19
19388
655
81752
.64361
0.0035
50106
89.089
8056
89.086
2555
18.950
18087
14.222
20431
20
23285
630
11278
9.1562
0.0008
12178
89.509
08661
89.508
27443
18.880
73006
14.401
90052
21
21970
143
10227
1.9736
0.0030
07423
88.477
64587
88.474
63845
18.326
36534
13.915
45019
223
45020
44
85100
.85272
0.0039
92422
88.499
5575
88.495
56507
11.280
82969
10.237
92344
Averag
es12
84609
9.95
74465
.62407
0.0042
13372
88.984
34609
88.980
13272
17.676
23514
14.415
22727
Coast A
rea
Slope (
degree
s)
Th
is
 is
 sl
op
e 
da
ta
 (i
n 
de
gr
ee
s)
 fo
r t
he
 a
re
as
 th
at
 a
re
 w
ith
in
 5
0m
 o
f t
he
 c
oa
st
 li
ne
. T
he
 5
0m
 sl
op
e 
da
ta
 d
oe
s n
ot
 in
cl
ud
e 
oc
ea
n 
da
ta
, 
on
ly
 te
rr
es
tri
al
. T
he
 d
at
a 
w
as
 c
ol
le
ct
ed
 b
y 
fli
gh
t s
o 
th
er
e 
is
 so
m
e 
re
pe
tit
io
n 
in
 th
e 
da
ta
 w
he
re
 th
e 
ed
ge
s o
f t
he
 fl
ig
ht
 im
ag
es
 o
ve
rla
p 
bu
t t
hi
s i
s m
in
im
al
.  
 
 
107  
    
Ta
bl
e 
8:
 D
at
a 
fo
r S
lo
pe
 w
ith
in
 5
0m
 o
f A
hu
 
A
hu
_I
D
C
at
eg
or
y
C
O
U
N
T
A
R
E
A
M
IN
M
A
X
R
A
N
G
E
M
E
A
N
S
TD
1ahu
10898
907
833.23
20
.00338
762
85.046
49353
85.043
10591
12.269
5277
9.6981
96248
2ahu
10096
49
7256.5
30.0
06356
403
87.832
06177
87.825
70536
14.223
72098
14.830
6929
3ahu
10899
64
7833.7
60.0
03279
554
69.312
67548
69.309
39592
8.3044
34674
6.1277
00979
4ahu?
10899
58
7833.7
20
.00455
752
85.260
71167
85.256
15415
11.319
24652
9.0258
79205
5imag
e ahu
52686
67
832.90
0.0112
03443
81.753
37219
81.742
16875
9.8847
23381
8.2091
44444
6ahu?
10895
09
7830.4
90.0
21009
127
88.775
58899
88.754
57986
21.501
98974
16.380
38355
7ahu?
14604
77
7828.7
90.0
07261
015
87.949
48578
87.942
22476
16.424
61398
15.368
50128
8ahu?
14613
03
7833.2
20.0
03079
966
77.560
52399
77.557
44402
8.7213
35047
7.7502
83146
9ahu?
14613
07
7833.2
40.0
03777
936
87.296
82922
87.293
05129
9.2214
15006
9.3490
7435
10ima
ge ahu
14613
13
7833.2
80.0
04100
627
87.053
48969
87.049
38906
9.0793
85294
11.057
34055
11ahu
67268
84
497.80
0.0063
8167
84.029
83856
84.023
45689
17.499
10292
13.873
34368
12ahu
94290
86
304.57
0.0049
7158
85.734
13849
85.729
16691
7.2229
07622
8.7455
49869
13ima
ge ahu
11239
28
7514.9
30.0
08088
565
84.751
00708
84.742
91852
10.908
0709
11.141
53921
14ahu
?
16090
37
7121.5
60.0
01057
015
88.443
8858
88.442
82879
13.343
98318
13.896
93408
15ahu
17698
37
7833.2
50.0
00821
331
81.344
58923
81.343
7679
10.480
98522
8.2054
11292
16ahu
14703
58
6992.3
40
.00555
006
88.501
34277
88.495
79271
13.285
90431
13.100
88179
17ahu
?
16471
78
7833.2
20.0
02445
308
83.456
3446
83.453
8993
10.562
02798
9.5190
90602
18ahu
70290
03
342.67
0.0173
84335
86.685
30273
86.667
9184
15.391
93819
12.306
70501
19ahu
16461
07
7828.1
30.0
05326
983
87.627
43378
87.622
10679
9.5324
2947
10.705
63033
20ahu
12364
50
5879.9
90
.00162
446
83.041
47339
83.039
84893
7.0135
31657
9.0699
1566
21ahu
?
11600
31
5516.5
70.0
01832
334
87.685
20355
87.683
37122
10.981
63468
12.820
63274
22ima
ge ahu
15211
22
7833.4
50.0
03931
223
87.721
68732
87.717
75609
13.050
11263
13.304
9359
23ahu
?
18590
04
7833.5
70.0
02088
532
88.427
52838
88.425
43985
15.470
79498
14.415
01112
24ahu
18589
02
7833.1
40.0
01840
576
83.617
11884
83.615
27826
7.4379
77819
7.9785
49755
25ahu
18590
03
7833.5
70.0
02948
002
83.587
90588
83.584
95788
7.6264
0414
8.2454
80119
26ahu
?
19677
82
7833.4
10.0
04253
726
86.988
92212
86.984
66839
15.196
45133
13.608
76208
27ahu
?
19677
76
7833.3
90.0
05158
016
82.229
37012
82.224
2121
9.3266
04866
8.8324
21229
Ahu an
d Buffe
r
Slope (
degree
s)
 
 
108  
  Ahu_ID
C
at
eg
or
y
C
O
U
N
T
A
R
E
A
M
IN
M
A
X
R
A
N
G
E
M
E
A
N
S
TD
28ima
ge ahu
20993
41
7721.9
30.0
05333
816
87.653
37372
87.648
0399
15.358
62984
15.018
00236
29ima
ge ahu
20545
26
7557.0
90.0
01815
925
88.252
22778
88.250
41186
14.541
46174
11.203
5512
30ima
ge ahu
44862
67
493.49
0.0331
94676
89.650
00916
89.616
81448
12.903
06988
14.241
89411
31ahu
?
45691
77
631.98
0.0203
60883
86.892
38739
86.872
02651
14.361
18971
13.688
12687
32ahu
21296
62
7833.4
60.0
00955
603
84.799
04175
84.798
08615
10.016
4364
10.762
09275
33ahu
?
20635
15
7590.1
50.0
03635
341
77.384
43756
77.380
80222
8.1218
86774
5.8797
65641
34ahu
?
11775
37
4331.2
90.0
05910
078
60.987
3085
60.981
39842
7.4095
98676
5.6171
62151
35ahu
?
20417
08
7509.9
40.0
02356
687
85.357
34558
85.354
98889
12.617
45557
11.220
76653
36ahu
46620
67
787.13
0.0794
52671
88.358
65784
88.279
20517
12.560
12468
12.625
9744
37ima
ge ahu
51942
37
833.56
0.0083
06252
68.661
30829
68.653
00204
5.5521
67217
6.1148
69421
38ima
ge ahu
51942
37
833.56
0.0099
61641
82.523
91052
82.513
94888
6.0262
20514
8.6437
24996
39ahu
51942
17
833.53
0.0289
84925
85.743
31665
85.714
33173
7.7581
50253
9.2618
05663
40ahu
46902
17
073.43
0.0098
04519
85.066
94794
85.057
14342
10.024
40936
11.841
57828
41ahu
?
51845
37
818.93
0.0063
70605
89.623
55042
89.617
17981
11.165
5658
18.806
26872
42ahu
?
48227
07
833.21
0.0025
72463
70.172
2641
70.169
69164
2.2891
31569
2.8845
26034
43ahu
?
48226
97
833.20
0.0935
8988
74.173
96545
74.080
37557
4.2355
687
4.3056
99911
44ima
ge ahu
42422
46
890.41
0.0086
98586
86.355
0415
86.346
34292
11.641
3578
14.510
53669
45ahu
44891
97
291.51
0.0140
46267
88.577
23999
88.563
19372
11.913
91483
13.386
92003
46ahu
19561
30
7833.2
20.0
00965
398
82.010
65826
82.009
69287
7.4077
01318
6.3144
58846
47ima
ge ahu
19041
30
7624.9
90.0
02135
231
87.295
4483
87.293
31307
13.247
39368
12.643
29663
48ahu
?
17761
46
7112.4
90.0
04056
538
88.055
39703
88.051
3405
15.299
60423
12.484
75057
49ima
ge ahu
19092
50
7695.9
80.0
05759
308
89.064
07166
89.058
31235
15.406
08921
13.076
3205
50ahu
?
19433
31
7833.3
60.0
04918
778
86.417
90009
86.412
98131
10.776
63538
8.3568
55164
51ahu
19070
44
7687.0
90.0
10328
256
88.199
31793
88.188
98968
15.962
9344
12.766
25468
52ahu
15405
52
6209.8
00.0
02996
836
88.651
6037
88.648
60686
12.877
41234
11.848
50101
53ahu
?
19433
39
7833.3
90.0
04303
206
84.684
68475
84.680
38155
10.208
84391
8.3838
57243
54ahu
?
18372
54
7405.7
70.0
07101
105
86.992
63763
86.985
53653
12.667
61115
10.464
54202
55ahu
?
32375
02
7833.3
50.0
00878
201
86.264
81628
86.263
93808
9.0367
65405
9.5995
55954
56ima
ge ahu
18577
75
7833.3
40.0
04462
641
89.089
8056
89.085
34296
11.126
83949
11.031
54974
Ahu an
d Buffe
r
Slope (
degree
s)
 
 
109  
 
Ahu_ID
Catego
ryC
OUNT
AREA
MIN
MAX
RANGE
MEAN
STD
57ahu
?
18577
77
7833.3
50.0
04293
764
74.940
41443
74.936
12066
9.6767
98802
7.1499
60846
58ima
ge ahu
?18
00035
7589.8
80.0
03719
733
87.331
25305
87.327
53332
14.027
61071
13.558
83707
59ahu
?
18577
77
7833.3
50.0
02611
693
74.473
31238
74.470
70068
8.5645
6625
6.7813
30456
60ahu
?
16172
19
7833.3
60.0
04003
315
83.038
36823
83.034
36491
10.358
21655
8.6111
99436
61ahu
16172
12
7833.3
30.0
00505
203
78.285
94971
78.285
4445
8.5742
63895
7.6940
23345
62ima
ge ahu
16172
22
7833.3
80
.00323
711
88.158
91266
88.155
67555
11.810
78933
13.893
90068
63ahu
?
16827
80
7833.4
10.0
01424
569
86.700
51575
86.699
09118
9.4224
9484
10.847
89773
64ahu
?
16827
87
7833.4
50.0
04941
953
86.668
12134
86.663
17939
9.0387
97706
10.217
59721
65ima
ge ahu
?16
82783
7833.4
30.0
11195
377
81.513
65662
81.502
46124
12.673
00479
10.510
65243
66ahu
?
16827
80
7833.4
10.0
03317
203
83.678
61176
83.675
29455
7.2873
86674
6.5994
44783
67ahu
?
16599
08
7726.9
40
.00585
656
88.254
34113
88.248
48457
10.558
4816
10.772
96052
68ima
ge ahu
41440
87
833.44
0.0177
17987
83.345
34454
83.327
62656
6.1242
88661
9.2719
56692
69ima
ge ahu
41441
77
833.61
0.0236
50032
73.323
28033
73.299
6303
4.7265
39565
4.1385
73959
Ahu an
d Buffe
r
Slope (
degree
s)
Th
is
 ta
bl
e 
sh
ow
s s
lo
pe
 d
at
a 
(in
 d
eg
re
es
) f
ro
m
 w
ith
in
 5
0m
 o
f e
ac
h 
ah
u.
 T
he
 sl
op
e 
fr
om
 th
e 
fli
gh
t i
m
ag
es
 (m
in
us
 th
e 
oc
ea
n 
ar
ea
s)
 w
er
e 
us
ed
 to
 c
al
cu
la
te
 th
is
 in
fo
rm
at
io
n.
 S
in
ce
 th
e 
fli
gh
ts
 o
ve
rla
p 
in
 p
la
ce
s, 
so
m
e 
po
in
ts
 h
ad
 a
ss
oc
ia
te
d 
da
ta
 fr
om
 m
or
e 
th
an
 o
ne
 fl
ig
ht
 a
re
a.
 
Fo
rtu
na
te
ly
, e
ve
ry
 5
0m
 b
uf
fe
r z
on
e 
w
as
 a
lw
ay
s c
om
pl
et
el
y 
co
ve
re
d 
by
 a
t l
ea
st
 o
ne
 o
f t
he
 o
ve
rla
pp
in
g 
fli
gh
ts
 so
 th
e 
as
so
ci
at
ed
 d
at
a 
fr
om
 th
e 
in
co
m
pl
et
e 
bu
ffe
r z
on
es
 w
as
 re
m
ov
ed
. 
  
110  
Bibliography 
Amidon, Elliot L.; Elsner, Gary H.  “Delineating landscape view areas...a computer 
approach.”   Res. Note PSW-RN-180. Berkeley, CA: U.S. Department of 
Agriculture, Forest Service, Pacific Southwest Forest and Range Experiment 
Station. (1968): 1-5 
"ArcGIS Help 10.1." ArcGIS Resources. ESRI, n.d. Web. 29 Sept. 2015. 
<http://resources.arcgis.com/en/help/> 
Arana, Patricio M. "Ancient Fishing Activities Developed in Easter Island." Latin 
American Journal of Aquatic Research Lajar 42.4 (2014): 673-89 
Beardsley, Felicia Rounds. Spatial Analysis of Platform Ahu on Easter Island. N.p.: 
n.p., 1990. Print. 
Cook, James, and Tobias Furneaux. A Voyage towards the South Pole and round the 
World. Performed in His Majesty's Ships the Resolution and Adventure, in the 
Years 1772, 1773, 1774 and 1775. Vol. 1. London: Printed for W. Strahan and 
T. Cadell ..., 1777. 276-97. Print. 
Corney, Bolton Glanvill, Felipe González De Haedo, and Jacob Roggeveen. The 
Voyage of Captain Don Felipe González: In the Ship of the Line San Lorenzo, 
with the Frigate Santa Rosalia in Company, to Easter Island in 1770-1: 
Preceded by an Extract from Mynheer Jacob Roggeveen's Official Log of His 
Discovery of and Visit to Easter Island in 1722. Cambridge: Printed for the 
Hakluyt Society, 1908. Print. 
Däniken, Erich Von. Chariots of the Gods?: Unsolved Mysteries of the past. New York: 
Putnam, 1970 
Diamond, Jared M. "Twilight at Easter." Collapse: How Societies Choose to Fail or 
Succeed. New York: Viking, 2005. 79-119. 
Dudgeon, John V., and Monica Tromp. "Diet, Geography and Drinking Water in 
Polynesia: Microfossil Research from Archaeological Human Dental Calculus, 
Rapa Nui (Easter Island)." International Journal of Osteoarchaeology Int. J. 
Osteoarchaeol. 24.5 (2012): 634-48. Web. 
"Easter Island Travel Guide." Travel Guide to Easter Island, South Pacific. N.p., n.d. 
Web. 15 May 2015. <http://easterisland.southpacific.org/index.html>.  
"About Rapa Nui." Easter Island Statue Project Official Website RSS. N.p., n.d. Web. 
15 May 2015. <http://www.eisp.org/category/outreach/aboutrapanui/>.    
 
 
111  
Fonstad, Mark A., James T. Dietrich, Brittany C. Courville, Jennifer L. Jensen, and 
Patrice E. Carbonneau. "Topographic Structure from Motion: A New 
Development in Photogrammetric Measurement." Earth Surf. Process. 
Landforms Earth Surface Processes and Landforms 38.4 (2013): 421-30. Web. 
Fraser, David. Land and Society in Neolithic Orkney. Oxford, England: B.A.R., 1983. 
Print. 
Gilks, S. J., and J. P. Hague. "Mountain Trail Formation and The Active Walker 
Model." Int. J. Mod. Phys. C International Journal of Modern Physics C 20.06 
(2009): 869-90. 
"GPS Accuracy and Error Sources." Mio: Explore More. Mio Technology Corporation, 
n.d. Web. < http://eu.mio.com/en_gb/global-positioning-system_gps-accuracy-
error-sources.htm> 
Harris, Norman R., Douglas E. Johnson, Melvin R. George, and Niel K. McDougald. 
“The Effect of Topography, Vegetation, and Weather on Cattle Distribution at 
the San Joaquin Experimental Range, California.” USDA Forest Service Gen. 
Tech. Rep. (2002): 53-63. 
Harwin, Steve, and Arko Lucieer. "Assessing the Accuracy of Georeferenced Point 
Clouds Produced via Multi-View Stereopsis from Unmanned Aerial Vehicle 
(UAV) Imagery." Remote Sensing 4.12 (2012): 1573-599. Web. 
Hausken, Kjell. "The Influence of Slope and Speed on Locomotive Power in Cross-
country Skiing." Human Movement Science 38 (2014): 281-92. 
Holton, Graham E. L. "Heyerdahl’s Kon Tiki Theory and the Denial of the Indigenous 
Past." Anthropological Forum 14.2 (2004): 163-81. 
Hunt, Terry L., and Carl P. Lipo. The Statues That Walked: Unraveling the Mystery of 
Easter Island. New York: Free, 2011. 
James, Mike R., and Stuart Robson. "Mitigating Systematic Error in Topographic 
Models Derived from UAV and Ground-based Image Networks." Earth Surf. 
Process. Landforms Earth Surface Processes and Landforms 39.10 (2014): 
1413-420. 
Johnson, K., E. Nissen, S. Saripalli, J. R. Arrowsmith, P. Mcgarey, K. Scharer, P. 
Williams, and K. Blisniuk. "Rapid Mapping of Ultrafine Fault Zone Topography 
with Structure from Motion." Geosphere 10.5 (2014): 969-86. 
Jones, Eric E. "Using Viewshed Analysis to Explore Settlement Choice: A Case Study 
of the Onondaga Iroquois." American Antiquity 71.3 (2006): 523-38. JSTOR. 
Web. 11 Nov. 2015. 
 
 
112  
Ladefoged, Thegn N., Christopher M. Stevenson, Sonia Haoa, Mara Mulrooney, Cedric 
Puleston, Peter M. Vitousek, and Oliver A. Chadwick. "Soil Nutrient Analysis 
of Rapa Nui Gardening." Archaeology in Oceania 45.2 (2010): 80-85. 
Llobera, M., and T.j. Sluckin. "Zigzagging: Theoretical Insights on Climbing 
Strategies." Journal of Theoretical Biology 249.2 (2007): 206-17. 
Lock, Gary R., and Trevor M. Harris. "Danebury Revisited: An English Iron Age 
Hillfort in a Digital Landscape." Anthropology, Space, and Geographic 
Information Systems. Ed. Mark S. Aldenderfer and Herbert D. G. Maschner. 
New York: Oxford UP, 1996. 214-40. 
Mulloy, William T., and Gonzalo Figueroa. The A Kivi-Vai Teka Complex and Its 
Relationship to Easter Island Architectural Prehistory. Honolulu: Social 
Science Research Institute, U of Hawaii at Manoa, 1978. 
Norman, J. M. "Running Uphill: Energy Needs and Naismith's Rule." Journal of the 
Operational Research Society J Oper Res Soc 55.3 (2004): 308-11. 
Paliou, Eleftheria. "The Communicative Potential Of Theran Murals In Late Bronze 
Age Akrotiri: Applying Viewshed Analysis In 3D Townscapes." Oxford Journal 
of Archaeology 30.3 (2011): 247-72. 
Shapiro, S. S. and Wilk, M. B. (1965). "An analysis of variance test for normality 
(complete samples)", Biometrika, 52, 3 and 4, pages 591-611. 
Slama, Chester C., Charles Theurer, and Soren W. Henriksen. Manual of 
Photogrammetry. Falls Church, VA: American Society of Photogrammetry, 
1980. 
Smith, Eric, and Rebecca Bliege Bird. "Costly Signaling and Cooperative Behavior." 
Moral Sentiments and Material Interests: The Foundations of Cooperation in 
Economic Life. Ed. Herbert Gintis. Cambridge, MA: MIT, 2005. 115-48. 
 "Support." Datum. ESRI, n.d. Web. 26 Nov. 2015. 
<http://support.esri.com/en/knowledgebase/GISDictionary/term/datum>. 
"Testing for Normality Using SPSS Statistics." When You Have Only One Independent 
Variable. Laerd Statistics, n.d. Web. 25 Oct. 2015. 
Tilburg, JoAnne Van. Among Stone Giants: The Life of Katherine Routledge and Her 
Remarkable Expedition to Easter Island. New York: Scribner, 2003. 
Taylor, C. R., S. L. Caldwell, and V. J. Rowntree. "Running Up and Down Hills: Some 
Consequences of Size." Science 178.4065 (1972): 1096-097. 
 
 
113  
Thompson, William J. Te Pito Te Henua or Easter Island. Washington: U.S. 
Government Printing Office, 1889. 
Trandy, C.R.V. "The Isovist Method of Landscape Survey." Methods of Landscape 
Analysis, London, 3 May 1967. Ed. A. C. Murray. S.l.: S.n., 1967. 9-10. 
Urish, Daniel W., and Thomas E. Mckenna. "Tidal Effects on Ground Water Discharge 
Through a Sandy Marine Beach." Ground Water 42.7 (2004): 971-82.  
"UTM - Universal Transverse Mercator." UTM: Projection. Geokov, n.d. Web. 26 Nov. 
2015. <http://geokov.com/education/utm.aspx> 
Wall, Jake, Iain Douglas-Hamilton, and Fritz Vollrath. "Elephants Avoid Costly 
Mountaineering." Current Biology 16.14 (2006): n. pag. 
Wheatley, David, and Mark Gillings. "Spatial Technology and Archaeology." (2002): 
201-14.