1 Landscapes of the ‘Coast of Death’: topographies of NW Iberia 2 3 Gail Higginbottoma,d, A. César González-Garcíaa, Miguel Carrero-Pazosb, Benito Vilas-Estévezc, and Víctor 4 López-Lópeza. 5 6 a. Instituto de Ciencias del Patrimonio, (Incipit), Consejo Superior de Investigaciones Cientificas (CSIC), Avda. de s/n, 15705 de Com- 7 postela, A Coruña, España, b. Institute of Archaeology, University College London, 31-34 Gordon Square, London WC1H 0PY, United Kingdom. c. 8 University of Vigo - Campus, Circunvalación ao Campus Universitario, 36310 Vigo, Pontevedra, A Coruña, España. d. Corresponding 9 author [email protected]. 10 11 12 We ought (to) approach the project of building an Archaeology of Perception.

13 Criado-Boado, F; and Villoch-Vázquez, V. (2000: 188‐216). 14

15 Abstract

16 This paper investigates the landscapes of communities found within Costa da Morte (Coast 17 of Death), . Its goal is to uncover whether or not the megalithic monuments of a particular and 18 coherent area of the south-eastern side of the Atlantic Façade are situated in relation to complex loca- 19 tional variables. In particular, in this paper we explore the entirety of their surrounding topography. 20 For the very first time, we were able to demonstrate that very specific natural landscapes surrounding 21 the of this region in Iberia were likely selectively drawn upon, expanding our understanding 22 of the Neolithic of this area and the peoples’ relationship with their natural world. 23 Keywords: , Iberia, GIS, Landscapes, Community Practice, Cultural Astronomy

24 Introduction

25 Studies that systematically analyse the entire topographic landscape pattern surrounding all monu- 26 ments of a certain class within or across regions to determine the architectural and social systems they 27 might share remain uncommon (e.g. Cummings and Whittle’s 2004 work on Wales, Fraser’s 1988 28 work on the chambered cairns on and ), and it is very rare for such projects to include an 29 astronomical perspective (Higginbottom et al 2001, 2015; Higginbottom 2020a,b works in ). 30 For the very first time, however, this paper will show convincingly that very specific natural landscapes 31 surrounding the Neolithic dolmens of Costa da Morte in Galicia (Figs. 1 & 2), were selectively chosen.

32 We will demonstrate that individual Neolithic dolmens of this part of Galicia appear to be associated 33 with particular topographical shapes of the landscape. To do so, our main research objective is to show 34 that the shape of the horizon visible from each individual site appeared to be selected according to 35 shared criteria, and thus dolmen orientation by itself is not the only significant connection of these 36 monuments to the natural world. To carry out our investigation, we use 2D/3D GIS and immersion 37 technologies to uncover the considerations of those people who created the megalithic monuments in 38 the Neolithic in Costa da Morte, Galicia, in doing so we reveal that people who built the monuments 39 shared some kind of cultural ideology related to the relevance of place, along the Atlantic Façade.

1

40 Figure 1. Costa da Morte is outlined in the NW region of this map, beginning along the coast. The Study area is defined by 41 the entire watershed that incorporates Costa da Morte and defines a clear geographical region within Galicia. This study 42 area can be seen as lighter shades (yellows and blues for coloured publications). The dots designate the locations of the 43 exposed dolmens. Galicia is indicated by the small black square on the insert.

44 Background 45 Brief chronology 46 Costa da Morte is a county within the municipality of A Coruña. Within Costa da Morte, we can see 47 some of the most important Galician dolmens, such as Casa dos Mouros, Arca da Piosa, Pedra Cub- 48 erta, Parxubeira or Dombate (Fig. 3). The dates we have for the dolmens within our study area are 49 few and even fewer are high-precision determinations from recent excavations. So, for some consid- 50 eration of the dates for the construction and use of dolmens within Costa da Morte, we have con- 51 structed a table that includes some sites within the north of A Coruña and one from central A 52 Coruña, the latter is Chousa Nova (Supplementary Tables 1 & 2). Where there are several dates for 53 one site available, only those from the original database that were equal-to and over 95% Cal (2σ)

2

54 BC probability are included. These were then recalibrated with the most recent calibration curves 55 (IntCal 20). Currently, there are two sites with very early dates situating their possible construction 56 in the middle of the 5th Millennium BC (Chousa Nova, Silleda, and Forno dos Mouros no 5a, Or- 57 tigueira). The general consensus is that dolmens were constructed within the 5th millennium to early 58 late 4th Millennium, and that the various necropoleis were likely used from ca 4300 BC to around 59 2000 BC over millennia, with possible new tomb building occurring from second half of 5th millen- 60 nium up to early beginning of 3rd millennium (see Supplementary Table 1). Whilst not millennia 61 apart in construction, Dombate is a good working example of the incorporation of a previous monu- 62 ment plus the reuse of dolmens through millennia. (Cebrián del Moral et al 2011: 167). The more re- 63 cent and superlative dolmen of Dombate (the mound of which incorporates the first), was built very 64 soon after the first, possibly causing the likely single use of the first dolmen (Cebrián del Moral et al 65 2011: 168).

66 Landscape Approaches 67 The consideration of landscapes and monuments has been widely undertaken along the Atlantic Fa- 68 çade, especially by scholars in Ireland and Britain (Ruggles and Martlew 1992; Tilley 1994; Richards 69 1996a, 2013a, b, c; Bradley 1998; Cooney 2000; Fraser 1983, 1988; Fraser, S 1996, 2004; Cummings 70 2002; Cummings and Whittle 2003, 2004; Higginbottom 2003; 2020a,b). We know that the people of 71 Late Neolithic and Bronze Ages of Scotland, on the north-western façade, erected standing stones and 72 other monuments in very considered landscapes (Burl 1993, 2000; Richards 1996a,b, 2013c; Ruggles 73 1984, Higginbottom et al 2015). In the case of standing stones in western Scotland, there is a clear and 74 consistent choice of particular landscape types that surround each of these monuments. There are two 75 major horizon shapes or patterns, discovered through modelling and confirmed by statistics, surround- 76 ing them (Higginbottom 2020a, 2003; Higginbottom et al 2018; Higginbottom and Clay 2016), and 77 each of these patterns is made up of several variables (Higginbottom 2020a). Cummings and Whittle’s 78 work on tombs in Wales and SW Scotland also discovered that, similar to Higginbottom’s work on 79 standing stones: “a whole range of different landscape features were referenced from each monument. 80 90% of monuments have a restricted view in one direction, whilst 74% have a view of mountains and 81 59% of the sea. Thus sites were frequently positioned in order to have a … combination of features” 82 (Cummings and Whittle 2004: 88). It seems clear that these landscapes were already familiar to, and/or 83 inhabited by, the builders (Jones et al 2011; Ashmore et al 2016; Card et al 2018).

84 Much work has also been done in the landscape studies of Galician megalithic monuments, notably 85 spatial networks & visibility (Criado-Boado & Villoch Vázquez 2000, Llobera 2015), visibility, intra- 86 site visibility, astronomy & GIS (González-García et al 2017), and GIS & spatial statistics to investi- 87 gate further ideas about locational qualities of megaliths in Galicia, with large raw datasets (Carrero- 88 Pazos 2018; Carrero-Pazos & Rodríguez Casal 2019). All papers from this millennium included 3

89 Figure 2 a-c. Examples of dolmens found in Costa da Morte. (a) Pedra da Arca (Regoelle), also known as Casa dos Mouros, 90 at top; (b-c) Pedra da Arca (Malpica de Bergantiños) bottom two images. Images by Gail Higginbottom.

4

91

92 Figure 3 a-c. (a-b) Further examples of corridor-chamber dolmens located in Costa da Morte (A: Dombate; B: 93 Parxubeira), and (c) C: field plans of several of them. Plan 3 is the tomb seen in Fig 2 (a), and Plan 5 is the tomb dis- 94 cussed in Fig. 14 (b-c) (C: Modified after Rodríguez Casal, 1990).

95 LiDAR data (post-2000). There are also papers that combine cultural and social elements of the meg- 96 alithic builders in Galicia for interpretive analyses (Criado-Boado & Villoch-Vázquez 1998; Criado 97 Boado et al. 2006; Gianotti et al. 2011, González-García 2018; González-García et al 2019). Criado- 98 Boado & Villoch-Vázquez´s 1998 work, based on the Barbanza peninsula, specifically engages with 99 the landscape in which the dolmens are located. Barbanza, like Costa da Morte, is also located along 100 the Atlantic coast. It contains a sharp inclining landscape upon which sits a sierra, with a highest alti- 101 tude of 680 metres above sea level, has a flat plateau near the top at around 550 metres above sea level. 102 Criado-Boado & Villoch-Vázquez concluded that the dolmens of Barbanza were deliberately posi- 103 tioned in order to indicate the best route to cross the sierra and that particular dolmens were intervisible 104 as you made your way across the sierra. These results were later confirmed using GIS analyses by 105 Llobera (2015). Finally, it should be noted that in their visits they noticed a location trend of the 106 mounds, namely that in general ‘the horizon is closer and higher towards West and North while it is

5

107 lower and far towards East and South’ (Criado-Boado and Villoch-Vázquez 1998 in González-García 108 et al 2017: 96-97, González-García 2018).

109 Interestingly, the orientation of dolmens in Galicia, considered as the inside-out direction towards the 110 symmetry axis of the corridor, appeared to systematically face a quite restricted area of the horizon, 111 from the east towards slightly south of the winter solstice sunrise (Hoskin 2001; González-García & 112 Belmonte 2010; Vilas-Estevez 2016; González-García et al 2017), a moment that might have been 113 rather interesting at particular monuments, such as Dombate (González-García et al. 2019). Within 114 this horizon range, the sunrise after the autumn equinox and before spring equinox can also be viewed, 115 but conversely it could be consistent with the area for moonrise (Silva and Pimenta 2012; Vilas-Esté- 116 vez 2016).

117 Method

118 As stated above, our research question is to investigate if the shape of the horizon visible from each 119 individual site appeared to be selected according to shared criteria. To do so we use a number of digital 120 terrain models with various resolutions (5m, 25m and 90m), and we employ the Horizon software (see 121 below for a detailed description of this software) to build 2D horizon profiles and 3D landscape pano- 122 ramas. We also use statistical tools (described below) to discern if the horizons observed from the sites 123 with megalithic monuments in our study area are systematically different from random samples spread 124 across the same area. In summary, we would like to verify if the sites with megalithic monuments were 125 selected to have specific horizon profiles.

126 To do so, and similar to Higginbottom’s previous works (Higginbottom et al 2001, 2002, 2015; Hig- 127 ginbottom 2020a,b, Higginbottom and Mom in press) and Criado-Boado and Villoch-Vázquez’s re- 128 search, through our work, we ‘approach the project of building an Archaeology of Perception’ (Criado- 129 Boado and Villoch-Vázquez 1998: 63), which uses a clear systematic strategy for examining the visual 130 features of past landscapes such that we can ‘approach a phenomenology of prehistoric perception 131 without falling in mere subjective solutions’ (Criado-Boado and Villoch-Vázquez 1998: 63).

132 Choosing the study area: Costa da Morte 133 Costa da Morte is characterised by several valleys and depressions intersected by rivers which drain 134 to the coast, through rugged terrains. The gentle elevations (e.g. O Facho, 312 m.a.s.l.) of this area 135 are surrounded by valleys such as the Vimianzo or Dumbría (Lema Suárez 2010). For our studies of 136 Galician dolmens, we chose to focus on one core dolmen district and its immediate periphery at a 137 time. Here we have chosen the area of Costa da Morte, a historically well-known core area of Gali- 138 cian megaliths in North-western Spain (Rodríguez Casal 1990, 2000; Lema Suárez 1999). For the 139 clear determination of the geographical research area, we adopted the mound clustering models cre- 140 ated by Carrero-Pazos & Rodríguez Casal for Galicia, using Kernel density analyses, (2018: figs. 6

141 6A&B). Figure 4 shows the different density clumps, where we have chosen the ‘Costa da Morte’ 142 one. There, we have analysed the 32 sites where there are exposed dolmens.

143 144 Figure 4. Kernel Density Analysis for the megalithic monuments of Galicia (After Carrero-Pazos & Rodríguez Casal 145 2018: figures 6B).

146 To determine the boundaries of the periphery, we modelled the river catchment zones in this area 147 (Fig. 2). The underlying assumptions are that areas where water is captured and maintained within 148 and by the natural environment could potentially be considered as viable areas for human congrega- 149 tion and habitation, and megaliths would be a kind of territorial marker as they tend to locate close to 150 the watershed edges, areas which, naturally, have high visibility (Martinón-Torres 2001; Carrero- 151 Pazos et al. 2019). Comparing Figures 1 and 4, we can see that the river catchment seen in Figure 1 152 contains sites that fall into the high, medium and low clustering designations of Costa da Morte seen 153 in Figure 4.

154 Horizon profiles

155 The primary aim of the software Horizon is to combine topographical data with atmospheric and as- 156 tronomical calculations to produce accurate two and three dimensional landscape information to aid 157 archaeoastronomical surveys and data analysis (Smith 2013). This includes the ability to create 2D 158 horizon profiles and rendered 3D landscape panoramas for each site, along with an ascii data file con- 159 taining calculations of the distance and direction between the point of origin or monument to all points 160 on the horizon and the altitude of each of these horizon points. Apart from the location, Horizon allows 161 you to choose epochs, astronomical phenomena and dates far back in time. The paths of astronomical

7

162 phenomena like the sun on the longest day and the shortest day and those of the Moon at the times of 163 its most extreme rising and setting points in its cycle, are then mapped onto the landscapes.

164 Data for 2D and 3D Horizon creations 165 For this project, we use a be-spoke version of Horizon. It was created to use 25m and 5m LiDAR 166 data made freely available by the Galician Xunta. These data had to be converted and tested for use 167 for our be-spoke version of Horizon, by the software engineer and designer of the program, Andrew 168 Smith (University of Adelaide). We used the converted 25m LiDAR data for the 2D landscape pano- 169 ramas and the concomitant ascii data. The creation of the 3D landscape panoramas was done by 170 merging different elevation data sets: 5m data extending out from the viewing point to 2.5km, after 171 which it is replaced with 25m data to speed up the processing time. Then 90m STRM data are used 172 to fill all the areas not covered by the LiDAR data.

173 Analysis of profiles

174 We created 32 2D Horizon profiles and their concomitant ascii files, along with 32 3D Landscape 175 panoramas. These outputs were used as described below to do our analyses.

176 Analysis 1 – the mean group profile and 1-sigma dispersion 177 Using the extracted ascii data of the 2D horizon profiles we obtained display each horizon profile in 178 horizon of each monument. For the ease of viewing, we have calculated a ‘mean’ horizon profile for

179 180 Figure 5: Horizon profile for the 32 sites presented in this paper (thin solid lines). Measurements run from 0º (due North) 181 to 360º, due south would be 180º. The dashed line indicates the mean horizon profile and the dotted line the mean plus 1- 182 sigma dispersion of the 35 profiles. See method for description of the 1-sigma dispersion. From left to right the yellow 183 lines are: Sun rise at the summer solstice, equinox and winter solstice: then the Sun set at the winter solstice, equinox and 184 summer solstice. 185 the 32 sites. We calculate the mean altitude of every one of the azimuths around the horizon for these 186 32 sites (36,000 azimuth points taken for the creation of each horizon profile; Fig. 5, dashed-line). We

8

187 then calculate the standard deviation of these altitudes and add this to the mean azimuth of the profiles 188 to give us the 1-sigma dispersion of the 32 profiles.

189 Analyses 2 – clustering of individual profiles

190 Using the ascii data, we then performed a K-means analysis based both on the horizon profiles for each 191 of the sites and another one on the distance to that horizon from each of those sites, for the entire 360- 192 degrees around each site. In general, the K-means approach (Everitt 1995) groups the data into clusters 193 by comparing, in this case, the shape of the profile of each data group with a given seed. The seed is 194 normally chosen randomly once we set the number of clusters we would like to explore. However, in 195 this case we have used the approach by Arthur and Vassilvitskii (2006) who proposed an alternative 196 method for the initial randomization. The first cluster centre, c0, is purely random. Then the distances 197 of all other measurements from c0 are computed following the Euclidean distance. Now these distances 198 are used to define a probability distribution and the next cluster is chosen again randomly, but now we 199 used the probability distribution above (instead of a uniform distribution). The idea was to draw new 200 seeds so that they are far from the first seed. With this idea, the new clusters will not necessarily be at 201 the very edges of the measurement space and it is less likely that isolated outliers would be picked. In 202 each step, the method computes the distance of each distribution to the seeds and then performs the 203 grouping. The method tries to find by an iterative process, the optimal clustering which minimizes the 204 distances within each cluster, defined as the sum-of-squares. To do so, at each step of the iteration, the 205 groups define a new seed by calculating the one in that cluster that is closer to the centroid of the 206 cluster. The process is iterated until it reaches convergence, i.e. until further iterations result in finally 207 obtaining the same groups.

208 Analyses 3 – clustering of individual profiles

209 In this analysis we tested for a systematic difference between the observed horizon profiles for all of 210 our sites and the horizon profiles of 320 random locations within the area of Costa da Morte (this is 211 one order of magnitude larger than our sample). To calculate the p-values, we used the Student-t test 212 in the following way. For each azimuth, we compute the observed sample altitudes, and compare to 213 the sample of random sites altitudes of horizon. So considered, the two samples (for each azimuth), 214 could be modelled by two Gaussian distributions. The two-sided Student-t test allows verifying if the 215 two samples are drawn from a common parent distribution. The p-value presented in the following 216 figures is that calculated using this method. We employ a rather conservative approach and set a 217 limit of p<0.05 to estimate that the null hypothesis, both samples are drawn from the same parent 218 distribution, is not supported (this is indicated in the following figures with a horizontal dotted line). 219 This means that if p>0.05 we cannot discard the null hypothesis. In the following, we will display in 220 the different figures the mean horizon (both for the observed and the random samples) but it must be 221 clear that we consider all profiles when comparing the different samples described, not the means. 9

222 Results and Discussion 223 Analysis 1 – the mean group profile 224 In Figure 5, we can see illustrated the altitude of all of the 32 horizons (thin solid lines). We also 225 include the mean altitude (thick dashed line) considered together along with its standard deviation. The 226 dashed yellow lines are the paths of the sun at particular moments.

227 It is interesting to notice that there is a general tendency to show slightly higher altitudes towards the 228 western part of the horizon. Also, the 1-sigma dispersion of the 32 profiles indicates that there is a 229 larger variation of horizon shape for the western horizons found at our sites than found for the eastern 230 horizons, which are more similar to one another (Fig. 5, thick dotted line). This means that there are 231 likely very high horizons as well as very low horizons in the west, while such variations seem to be 232 smaller in the east.

233 This diagram, however, does not allow us to extract much further information. We could discuss the 234 mean horizon, but it would probably turn into meaningless conclusions. We therefore, preferred to 235 verify if there are certain traits within these profiles, and later compare them to those generated by a 236 random sample in the study area.

237 Analyses 2 - K-means analysis 238 K-means analysis on the horizon profiles – do sites share the same horizon profile shapes? 239 We tried three to seven initial seeds and four appeared to be optimal choice, for it appeared to capture 240 the variability of the groups most clearly. The largest group comprises 14 of the 32 sites (this is the 241 green group on Fig. 6, Cluster 1), while the second greatest in number has eight (red group: Cluster 4, 242 n=8). The others have 6 and 4 members. For the ease of understanding, we can see their average hori- 243 zon profile or shape for each group in Figure 7 and the regional distribution of these sites is the one 244 seen in Figure 8.

245 K-means analysis on the distance to the horizon – do sites share the same horizon distance 246 patterns? 247 A similar analysis was done on the horizon distance instead of the horizon profile or shape. The re- 248 sults are summarized in the Figures 9 and 10. There are two main groups, one with 12 members (in 249 green in Fig. 9) and one with 10 members (gold). Interestingly they seem to behave as complemen- 250 tary profiles. While the gold line seems to indicate a close profile to the observer towards the eastern 251 half of the horizon, it gets farther away along the western horizon. The gold group behaves exactly 252 the opposite, with the profile a further away in the east and closer in the west. These two groups rep- 253 resent the more typical profiles in Costa da Morte.

10

254 255 Figure 6: K-means clustering for the horizon profiles. Each group after the algorithm is shown in different colours. From 256 bottom to top are Clusters 1, 2, 3, and 4.

257 258 Figure 7: The ‘K-means’ mean profile for each group and number of sites in each group. From bottom to top, the data 259 comes from clusters 1, 2, 3, and 4 in Figure 6.

11

260 Figure 8: Distribution of sites studied for the K-mean analysis performed on the horizon profile of the site (Figure 7). 261 Group 1, n=14, Group 2=6, Group 3= 4, Group 4=8.

262

263 Figure 9: K-means over the distance for the 32 sites. The plot shows the mean distance profile for each group and the 264 number of elements in each group. Note that the distance scale is logarithmic. There are two main groups, the green one 265 with 12 members and the gold with 10. Starting at zero azimuth we can see that the green group sits below the gold and 266 that close to south (180o) they change positions, with the green group sitting above the gold group. 12

267 Figure 10: K-means map for the distance analysis for the 32 sites. Group 1, n=11, Group 2=10, Group 3= 3, Group 4=7.

268 Analyses 3 - Random samples - pilot study

269 Random sample comparison with all horizon profiles

270 We produced the mean horizon profile for the 320 randomly selected locations within Costa de Morte. 271 This is shown in Figure 11 top panel as a thick solid line. The mean observed horizon profile (for all 272 of the exposed dolmens) is shown as a dotted line. The observed profiles are systematically lower than 273 the (mean) random sites profiles. Remember that these are the mean profiles, but they tend to indicate 274 the behaviour of the two samples. In the lower frame of Figure 11 is an estimate of the statistical 275 significance of the difference between the observed and random samples for each point observed 276 (every 0.01o). Thus, we are seeing the p-value for each azimuth value is plotted in the lower frame. 277 We find that the values are lower than p=0.05 for a broad range of directions eastwardly nearly from 278 north to south, while for the western part of the horizon the profile is not statistically different from 279 the random sample.

13

280 Figure 11: Top panel, random (solid) and observed (dotted) mean horizon profiles of the exposed dolmens in Costa da 281 Morte. The lower frame shows the statistical significance. In this same frame the dashed horizontal line shows the value 282 of p= 0.05.

283 Random sample comparison with the profiles in the dominant horizon clusters

284 We applied the same form of analysis for the 22 sites included in clusters 1 and 4 (red and green) of 285 Figure 6. We combined them because of their similarity in form, thereby also creating a larger ‘signal’ 286 of 22 sites. This similarity or equivalence was first evidenced by carrying out the K-means analyses. 287 When 3 seeds are used in the K-means analyses to define the data (where each seed ultimately equates 288 to the number of clusters used to define the data), all of the horizon profiles seen in clusters 1 and 4 289 become contained within a single cluster, signifying their similarity, whilst clusters 2 and 3 continue 290 to remain separate and contain the same profiles. We then performed a comparison of these dominant 291 horizon profiles with the 320 randomly selected locations within Costa da Morte (Fig. 12). We can see 292 that the mean observed profile here is even lower than the mean observed profile of all site horizons 293 seen in Figure 11. The lower panel indicates that the statistical significance in this case is also below 294 the limits for the eastern horizon found in Figure 11.

295 This is even clearer when we compare these results with those of the rest of our sample, where the 10 296 remaining sites shown in Figure 6 (Clusters 2 & 3) are used for the observed profile data (Fig. 13). 297 Here it is evident that, unlike the combination of clusters 1 and 4, the remaining sites do not differ 298 significantly from the random horizon shapes.

14

299 Figure 12: Random (solid) and observed (dotted) mean horizon profiles, where the observed profiles are the combined 300 profiles of Clusters 1 and 4 in Figure 6 (n=22 sites). From left to right the yellow lines are: Sun rise at the summer solstice, 301 equinox and winter solstice: then the Sun set at the winter solstice, equinox and summer solstice

302

303 Figure 13: Random (solid) and observed (dotted) mean horizon profiles for the sites NOT included in Clusters 1 and 4. 304 according to the horizon profile patterns, n=10.

305

15

306 Figures 14: Random (solid) and observed (dotted) mean horizon profiles for the sites with a further distance to the east- 307 ern side of the horizon.

308 Here it is evident that the remaining sites do not differ significantly from the random horizon shapes.

309 To investigate the possible nature of the horizons surrounding each monument further, we turned to 310 those profiles that were the result of the horizon distance assessment, namely the two largest clusters 311 in Figure 9, and compared their profiles with the profiles of the random locations. The results can be 312 found in Figures 14 and 15. In Figure 14 we can see that the profile is clearly lower than the average 313 random profile towards the eastern side, as might be expected due to the further distance on this side, 314 and that this difference is clearly significant. Whilst the western side on average seems closer to the 315 random profiles. Figure 15 shows how for those dolmens with a horizon further away in the west, we 316 do not find a significant difference with the random sample.

317

16

318 319 Figure 15: Random (solid) and observed (dotted) mean horizon profiles for the sites with a further distance to the western 320 side of the horizon. 321

322 Landscape choices in the ‘Coast of the Dead’

323 The profiles illustrate that the monuments are located at higher local elevations, for the horizon profiles 324 appear generally lower for the monuments than those of the expected or random profiles. These results 325 reinforce the findings of previous studies, which showed a general preference for dolmens to be located 326 at higher than average elevations in their local areas, topographic prominence areas (Criado-Boado & 327 Villoch-Vázquez 2000; Carrero-Pazos & Rodríguez Casal 2019).

328 It should be noted that if the general landforms in the region would favour the eastern or western 329 horizons or have a higher or lower profile on one side of the horizon, such tendencies should appear 330 in the random sample of 320 sites investigated above. However, the differences between the observed 331 and the random samples are clear and statistically significant, particularly for certain sections of the 332 horizon. Looking back at Figure 12, at the observed horizon profiles clustering together in relation to 333 shape (Clusters 1 and 4, n=22/32 sites), it seems that the most likely range that is not randomly selected 334 runs from about 15 degrees north, through until south and might have some marginal significance until 335 the west point. In other words, the characteristics associated with this part of the horizon are those 336 most sort after for their difference to the rest of the naturally occurring landscapes. Added to this, when 17

337 we observe just those sites with a further distance to the eastern side of the horizon, this pattern more 338 or less remains firm, especially in the eastern half (Figure 14). Thus there is a statistical consistency 339 for this distant eastern horizon. Also, whilst the significance of the shape of horizons seems to be 340 higher for the eastern horizon systematically, there is an exception for those cases when the west hori- 341 zon is farther away (Figure 15).

342 What is of added interest, here, is that the statistically significant difference between random and ob- 343 served horizons appears to contain the area for the solar sunrise range (Figs. 11, 12 & 14, indicated by 344 the dotted yellow lines). This could perhaps indicate a search for a preferred horizon shape that also 345 includes a connection to their orientation preferences, together accomplishing a raft of complex land- 346 scape characteristics for the ‘ideal’ position upon which to position a dolmen. Further research is re- 347 quired to discover what, apart from orientation preferences, might be significant about the remaining 348 areas along these significant horizon ranges.

349 Despite our findings about the eastern horizons, along with the fact that every single dolmen faces east 350 in Costa da Morte (Vilas-Estévez 2016), unlike Criado-Boado and Villoch-Vázquez’s 1998 study of 351 Barbanza, there is no unique choice regarding which type of horizon should be closer - the eastern or 352 western. The lack of this unique quality is clearly illustrated by 29 out of 32 horizons, where 12 had a 353 distinctly closer eastern horizon and the remaining 17 had a distinctly closer western horizon. More 354 specifically, Costa da Morte horizon choices do not appear to be predominantly ‘closer and higher 355 towards West and North (and) lower and farther towards East and South’ as they are in Barbanza 356 (Criado-Boado and Villoch-Vázquez 1998 in González-García et al 2017: 96-97). However, with the 357 revelation of Barbanza’s mean horizon profile of all mounds being systematically lower towards the 358 eastern part of the horizon than west (González-García 2016), we can at least see a similarity with one 359 of the two families of horizon profiles found for the dolmens in Costa da Morte.

360 Remaining questions to be considered for Costa da Morte include, ‘what of the lie of the land between 361 the horizons and the dolmens? Does the immediate topography tend to slope downwards in the east 362 perhaps and upwards towards the west, regardless of eventual horizon heights?’ More research into 363 the topography between the monuments and the horizons is clearly required in the future. Similarly, 364 until a much more thorough investigation is completed regarding the complex astronomical factors 365 that might be connected to these two types of horizon profile choices (as undertaken in Scotland, Hig- 366 ginbottom et al 2015, Higginbottom 2020), it is not possible to say what might underlie these distant 367 differences in the east or west in relation to astronomy.

368 We can hypothesize, though, about the influence of the possible surrounding vegetation as a visual 369 element surrounding the dolmens. Due to the lack of research done in Costa da Morte itself, the clear 370 determination of local land-use and vegetation cover is at this stage unclear. Nevertheless, Kaal et al 371 demonstrate that during the 5th to 4th millennium BC, mountainous or hilly non-sierra areas, like Costa 18

372 da Morte in Galicia, were primarily covered with mixed deciduous forest, dominated by oak (Quercus) 373 with a light under-story of shrubs and herbs (Kaal et al 2013: 1522, fig 4). Such Oak forests tend to be 374 open and lose their leaves in early winter, creating a very open aspect until late Spring-early Summer. 375 Added to this, the analysis of the fen of Chan da Cruz, Gañidoira, Xistral and Chao de Lagozas (Montes 376 de Buio, Serra do Xistral and Toxiza) show the first clear evidence of forest reduction, the preponder- 377 ance of the ruderance taxa and any degradative shrub formations at the beginning of the first half of 378 the 5th millennia B.C. In this we have evidence of the reduction of forest and the creation of pockets 379 of grasslands (López Saez et al 2010). The beginning of these activity coincides with the earliest dates 380 we have for dolmens in A Coruna, like Forno dos Mouros 5a (Ua 20009 5635±50 BP, 4590 - 4350 cal 381 BC, 95.4% - see Supplementary Table 1).

382 Their natural world – a symbol of cultural continuity

383 This paper has shown that very specific natural landscapes surround the Neolithic dolmens of Costa 384 da Morte in Galicia, which were selectively chosen. Thus, we have compelling evidence that the as- 385 tronomical orientations of dolmens in Galicia were not their only significant connection to the Natural 386 World. The majority of these monuments are united as symbols of cultural continuity, reflected by the 387 similarities of community practices. Something about which we now know more than before. We sug- 388 gest that the observable locational and visibility patterns for these dolmens are connected to more 389 intangible patterns, where entire landscapes could possibly contain cosmological belief systems and 390 we look forward to discovering these potentialities, expanding and deepening the various studies car- 391 ried out in the past, as well as our own.

392 Acknowledgements

393 We are particularly indebted to Andrew Smith of the University of Adelaide. Andrew Smith con- 394 structed the 2D and 3D software, Horizon, as well as acting as a consultant in relation to the updating 395 of Horizon and conversion of LiDAR data for the use in this bespoke software. We are also indebted 396 to Felipe Criado-Boado for his support in carrying out this work. We would very much like to thank 397 Andy Michael Jones for his very helpful comments on the manuscript, acknowledging that all re- 398 maining errors are our own.

399 This work has received funding from the European Union’s Horizon 2020 research and innovation 400 programme under the Marie Skłodowska-Curie grant agreement SHoW No 800236.

19

401 References

402 Ashmore, P. J., 2016. Calanais survey and excavation 1979–1988. With contributions by T. Ballin, S. 403 Bohncke, A. Fairweather, A. Henshall, M. Johnson, I. Maté, A. Sheridan, R. Tipping & M. Wade 404 Evans. Edinburgh: Historic Environment Scotland.

405 Arthur, D. and Vassilvitskii, S. 2006. K-means ++: The Advantages of Careful Seeding. Technical 406 Report. Stanford. http://ilpubs.stanford.edu:8090/778/ last access: 02/10/2020

407 Belmonte, J.A. & Hoskin, M., 2002. Reflejo del Cosmos: Atlas de Arqueoastronomía del Mediterráneo 408 Antiguo. Madrid: Equipo Sirius.

409 Bradley, R., 1998. The Significance of Monuments. London: Routledge.

410 Burl, H. A. W. 1993. From Carnac to Callanish: The prehistoric stone rows and avenues of Britain,

411 Ireland and Brittany. New Haven and London: Yale University Press.

412 Card, N., Mainland, I., Timpany, S., Towers, R., Batt, C., Ramsey, C., Dunbar, E., Reimer, P., Bayliss, 413 A., Marshall, P., & Whittle, A., 2017. To Cut a Long Story Short: Formal Chronological Modelling 414 for the Late Neolithic Site of , Orkney. European Journal of Archaeology 21(2), 217- 415 263..DOI: 10.1017/eaa.2016.29

416 Carrero-Pazos, M. & Rodríguez Casal, A., 2019. General and local spatial trends in Galician mega- 417 lithic landscapes (North-Western Iberian Peninsula) in Megaliths, Societies and Landscapes: Early 418 Monumentality and Social Differentiation in Neolithic Europe, eds M. Hinz, J. Müller, M. Wunderlich 419 (Universität zu Kiel, Kiel). Bonn: Habelt, 641-645.

420 Carrero-Pazos, M., Bevan, A., & Lake, M. (2019). The spatial structure of Galician megalithic land- 421 scapes (NW Iberia): A case study from the Monte Penide region. Journal of Archaeological Science 422 108. DOI: https://doi.org/10.1016/j.jas.2019.05.004

423 Carrero-Pazos, M., 2018. Beyond the scale. Building formal approaches for the study of spatial pat- 424 terns in Galician moundscapes (NW Iberian Peninsula). Journal of Archaeological Science, Reports 425 19, 538–551.

426 Cebrián del Moral, F. et al., 2011. El dolmen de dombate: arqueología, arquitectura y conservación. A 427 Coruña: Diputación de A Coruña.

428 Cooney, G. 2000. Landscapes of Neolithic Ireland. London: Routledge.

429 Criado Boado, F., Mañana Borrazás, P., & Gianotti García, C., 2006. Before the Barrows: Forms of 430 Monumentality and Forms of Complexity in Iberia and Uruguay, in Archaeology of Burial Mounds, 431 ed. L. Šmejda. Plzeň: University of West Bohemia, 38-52. 20

432 Criado-Boado, F. & Villoch Vázquez, V. 1998. La Monumentalización del Paisaje: Percepción y Sen- 433 tido Original en el megalitismo de la Sierra de Barbanza, Trabajos de Prehistoria 55(1), 63–80.

434 Criado-Boado, F; & Villoch-Vázquez, V., 2000. Monumentalizing landscape: a formal study of Gali- 435 cian Megalithism (NW Iberia). European Journal of Archaeology 3(2), 188‐ 216.

436 Cummings, V. & Whittle, A., 2004. Places of special virtue: Megaliths in the Neolithic landscapes of 437 Wales. Oxford: Oxbow.

438 Cummings, V. & Whittle, A., 2003. Tombs with a view: landscape, monuments and trees. Antiquity, 439 77, 255-266. doi:10.1017/S0003598X00092255

440 Cummings, V., 2002. Between mountains and sea: A reconsideration of the Neolithic monuments of 441 Southwest Scotland. Proceedings of the Prehistoric Society, 68, 125–146.

442 Everitt, B.S., 1995. Cluster Analysis. London: Arnold.

443 Fraser, D.,1983. Land and society in Neolithic Orkney. BAR British Series 117. Oxford: British Ar- 444 chaeological Reports.

445 Fraser, D., 1988. The orientation of visibility from the chambered cairns of Eday, Orkney, in Records 446 in stone: Papers in memory of Alexander Thom, ed. C. L. N. Ruggles. Cambridge: Cambridge Univer- 447 sity Press, 325–336.

448 Fraser, S. M., 1996. Physical, social and intellectual landscapes in the Neolithic: Contextualizing 449 Scottish and Irish Megalithic architecture. Unpublished PhD thesis, University of Glasgow.

450 Fraser, S. M., 2004. Metaphorical journeys: Landscape, monuments and the body in a Scottish Neo- 451 lithic. The Proceedings of the Prehistoric Society 70, 129–151.

452 Gianotti, C., Mañana-Borrazás, P., Criado-Boado, F., & López-Romero, E., 2011. Deconstructing Ne- 453 olithic Monumental Space: The Montenegro Enclosure in Galicia (Northwest Iberia). Cambridge Ar- 454 chaeological Journal 21(3), 391-406. doi:10.1017/S0959774311000436

455 González-García, A.C., 2018. Light and Shadow Effects in Megalithic Monuments in the Iberian Pen- 456 insula, in The Oxford Handbook of Light in Archaeology. Currently online only. DOI: 10.1093/ox- 457 fordhb/9780198788218.013.6

458 González-García, A. C., Criado-Boado, F. & Vilas, B., 2017. ‘Megalithic Skyscapes in Galicia’, The 459 Marriage of Astronomy and Culture, a special issue of Culture and Cosmos 21(1) and 2, pp. 87–103.

460 Gonzalez-Garcia, A. C., 2016. Presentation at The Marriage of Astronomy and Culture: Theory and 461 Method in the Study of Cultural Astronomy: The 24th Conference of the European Society for Astron- 462 omy in Culture, Bath, UK, 12th–16th September, 2016.

21

463 González-García, A. C, Vilas-Estévez, B., López-Romero, E. & Mañana Borrazás, P. (2019) Domes- 464 ticating Light and Shadows in the Neolithic: The Dombate Passage Grave (A Coruña, Spain). Cam- 465 bridge Archaeological Journal 29(2), 327–343.

466 Higginbottom, G., 2020a. The World ends here, the World begins here: megalithic mon- 467 uments in western Scotland. Journal of World Prehistory 33, 25-134.

468 Higginbottom, G. 2020b Perception creates worlds: meaning and experience in the erection of free- 469 standing in Yachay Wasi: a collection of papers in honour of Ian S. Farrington, eds 470 L. Solling, T. Knight, C. Gant-Thompson, D. Tybussek & R. Parkes. Oxford: Archaeopress/Hardrian.

471 Higginbottom, G. and Mom, V. In press Illuminating Time: the visibility of temporality in prehistory 472 in The Oxford Handbook of Light in Archaeology, eds. C. Papadopoulos & H. Moyes.

473 Higginbottom, G. and Clay, R. 2016 Origins of Standing Stone Astronomy in Britain: New quantitative 474 techniques for the study of archaeoastronomy. Journal of Archaeological Science: Reports, 9, 249- 475 258.

476 Higginbottom, G., Vilas-Estévez, B., González-Garcia, C., Carrero-Pazos, M. & Lopez-Lopez, V. in 477 preparation, Ideologies and lifestyles of NW megalithic Iberia.

478 Higginbottom, G., Clay, R., Voisin, F. & Nguyen, P., 2018.Testing landscape as cultural expression, 479 in Selected Papers of the INSAP X – Oxford XI – SEAC 25th Joint Conference ‘ROAD TO THE 480 STARS’, eds. Frank Prendergast, A. César González-García, Gary Wells and Juan A. Belmonte. Med- 481 iterranean Archaeology and Archaeometry, 18(4), 441-451.

482 Higginbottom, G., Smith, A. G. K., & Tonner, P., 2015. A re-creation of visual engagement and the 483 revelation of world views in Bronze Age Scotland. Journal of Archaeological Method and Theory 22, 484 584–645. https ://doi.org/10.1007/s1081 6-013-9182-7. Published online 2013.

485 Higginbottom, G., 2003. Interdisciplinary study of megalithic monuments in western Scotland. Un- 486 published PhD thesis, University of Adelaide.

487 Higginbottom, G., Simpson, K., & Clay, R. (2002). Using viewsheds wisely: Developing sound meth- 488 odologies from spatial analyses of megalithic monuments in western Scotland. In G. Burenhult & J. 489 Arvidsson (Eds.), Archaeological informatics: Pushing the envelope. BAR International Series 1016 490 (pp. 53–62). Oxford: Archaeopress

491 Higginbottom, G., Smith, A. G. K., Simpson, K., & Clay, R., 2001. Incorporating the natural environ- 492 ment: Investigating landscape and monument as sacred space, in One land, many landscapes, eds. M. 493 Gojda & T. Darvill. (BAR International Series 987). Oxford: Archaeopress, 97–104.

22

494 Hoskin, M., 2001. Tombs, Temples and their Orientations: A New Perspective on Mediterranean Pre- 495 history. Oxford: Ocarina Books.

496 Jones, A. M., Freedman, D., O’Connor, B., Lamdin-Whymark, H., Tipping, R., & Watson, A., 2011. 497 An animate landscape: Rock art and the prehistory of Kilmartin, Argyll, Scotland. Oxford: Windgather 498 Press.

499 Kaal, J., Criado-Boado, F., Costa-Casais, M., López-Sáez, J-A., López-Merino, L., Mighall, T., Car- 500 rión, Y., Silva Sánchez, N., Martínez Cortizas, A. 2013. Prehistoric land use at an archaeological hot- 501 spot (the rock art park of Campo Lameiro, NW Spain) inferred from charcoal, synanthropic pollen and 502 non-pollen palynomorph proxies, Journal of Archaeological Science, Volume 40: 3, 1518-1527.

503 Lema Suárez, X. M., 1999. Arquitectura megalítica na Costa da Morte (Antas e Mámoas). Asociación 504 Neria/Deputación de A Coruña: A Coruña.

505 Lema Suárez, X. M. (Coord)., 2010. A , no corazón da Costa da Morte. Vigo: Xerais.

506 Llobera, M., 2015. Working the Digital: Some Thoughts from Landscape Archaeology, in Material 507 Evidence: Learning from Archaeological Practice, eds. R. Chapman and A. Wylie. London: 508 Routledge, 173–88.

509 López Saez, J. A., López Merino,L. and Pérez Días, S. 2010. Neolitización, Megalitismo y Antropiza- 510 ción del Paisaje en Galicia entre el VII y el IV milenio cal. BC, Munibe. 32: 488-49.

511 Martinón-Torres, M., 2001. Os monumentos megalíticos despois do Megalitismo. Arqueoloxía e His- 512 toria dos megalitos galegos a través das fontes escritas (S. VI-S. XIX). Valga: Concello de Valga.

513 Richards, C., 1996a. Monuments as landscape: Creating the centre of the world in Late Neolithic Ork- 514 ney. World Archaeology 28, 190–208.

515 Richards, C., 1996b. and water: Towards an elemental understanding of monumentality and 516 landscape in Late Neolithic Britain. Journal of Material Culture 1(3), 313–336.

517 Richards, C. (ed.), 2013a. Building the great stone circles of the north. Oxford: Windgather Press.

518 Richards, C. 2013b. Interpreting stone circles, in Building the great stone circles of the north, ed. C. 519 Richards. Oxford: Windgather Press, 2–30.

520 Richards, C., 2013c. Wrapping the hearth: Constructing house societies and the tall Stones of , 521 Orkney, Building the great stone circles of the north, ed C. Richards. Oxford: Windgather Press, 64– 522 89.

523 Rodríguez Casal, A. A., 1990. O Megalitismo. A primeira arquitectura monumental de Galicia. San- 524 tiago de Compostela: Universidade de Santiago de Compostela.

23

525 Rodríguez Casal, A. A., 2000. O mundo megalítico nas comarcas da Costa da Morte, in AA.VV. Nas 526 orixes da nosa identidade. Actas do II Simposio de Historia da Costa da Morte (Carnota, 16, 17 e 18 527 de xuño do 2000). Cee: Asociación Neria, 7-28.

528 Rodríguez Casal, A. A., 2003. Imagen apotropaica y espacio ritual funerario en el megalitismo gallego 529 (4000-2000 a.C), in Semata Nº 14. Profano y Pagano en el arte gallego, eds. M. A. Castiñeiras Gon- 530 zález & Díez Platas. Santiago de Compostela: Universidad de Santiago de Compostela, 26-28.

531 Ruggles, C. L. N. (1984). Megalithic astronomy: A new archaeological and statistical study of 300 532 western Scottish sites. BAR British Series 123. Oxford: British Archaeological Reports.

533 Ruggles, C., & Martlew, R., 1992. The north Mull project, 3: Prominent hill summits and their astro- 534 nomical potential. Archaeoastronomy 17, S1–S13.

535 Silva, F. & Pimenta, F., 2012. The Crossover of the Sun and the Moon, Journal for the History of 536 Astronomy. 43(2):191-208. doi:10.1177/002182861204300204 V

537 Smith, A. G. K. (2013). Horizon user guide and implementation notes. Documentation Version 0.12 538 728 December 3, 2013. http://www.agksmith.net/horizon

539 Tilley, C. 1994. A Phenomenology of Landscape: Places, Paths and Monuments. Oxford: Berg Pu- 540 blishers.

541 Vilas-Estévez, B., 2015. Estudio de las Orientaciones y Emplazamientos de los Túmulos de la Necro- 542 pólis de la Serra do Leboreiro en base a la Arqueología del Paisaje y la Arqueoastronomía. M.A. 543 Dissertation, University of Santiago.

544 Vilas-Estévez, B. 2016. Were the megaliths of Galicia (Spain) aligned with specific astronomical 545 events? M.A. Dissertation, University of Wales Trinity Saint David.

546

24

547

Laboratory Cal (2σ) BC Cal (2σ) BC BP Reference Nº OxCal 3.10 probability Chousa Nova, Beta-277240 5450 ± 4350-4240 Cal No information Domínguez-Bella, María José Bóveda 2011 Silleda 40 BP BC ** **calibración del laboratorio a partir de la curva de calibración INTCAL04, Reimer et al. 2004.

Dombate, second CSIC 939 4410±25 3100 - 2920 93.8% Bello 1995; Alonso y Bello 1995. 1997: 519. tomb Dombate second CSIC 962 4020±30 2620 - 2460 95.4% Bello 1995: Alonso y Bello 1995. 1997: 519; phase, second tomb Dombate, second CSIC 963 4380±35 3100 - 2900 95.4% Bello 1995: Alonso y Bello 1995. 1997: 519; tomb Dombate, second UtC 3201 3950±60 2620 - 2280 94.2% Bello 1995; Alonso y Bello 1997: 519; tomb Dombate, second UtC 3203 4950±70 3950 - 3630 95.4% Bello 1995; Alonso y Bello 1997: 519; tomb Dombate, first tomb CAMS 4900±40 3770 - 3630 95.4% Steelman. Carrera et al. 2005. 101903 Dombate, first tomb CAMS 4890±40 3770 - 3630 95.4% Steelman. Carrera et al. 2005. 101904

Forno dos Mouros Ua 20009 5635±50 4590 - 4350 95.4% Mañana 2005 5a Forno dos Mouros Ua 20010 5500±50 4460 - 4250 95.4% Mañana 2005 5a Forno dos Mouros Ua 21687 3565±40 2030 - 1860 76.4% Mañana 2005 5b Forno dos Mouros Ua 21688 4390±45 3120 - 2900 88.1% Mañana 2005 5b Pedra Cuberta CAMS 77923 5010±60 3960 - 3660 95.4% Carrera y Fábregas 2002; Steelman. Carrera et al. 2005.

548 Supplementary Data Table 1: The dating examples of dolmens in A Coruña. The information for the first three columns 549 and the last, except that for the site of Chousa Nova, is taken from Prieto Martínez et al 2012 and their contribution to the 550 volume’s major data table at the end of the book: for data on Galicia see pages 585-589. OxCal dates provided by Andy 551 M. Jones. 552

553

25

Dolmen Name UTMX UTMY Dombate 502487 4781930 Pedra Moura do Monte Carnio 501803 4769844 Dolmen Arca da Piosa 501573 4765872 Pedre da Lebre I 501300 4768580 Mina de Recesindes 494026 4777221 Mina Folias 494309 4774664 Mina d’Aquela banda 493685 4776030 Pedra da Arca (Malpica) 514780 4793766 Mina de Espiñaredo 515125 4749833 Dombate antiguo 502487 4781930 Seconde 503938 4774078 Reparada 503913 4773982 Mámoa do Rei Penide 526955 4689614 Mamoa do Rei 529981 4679203 Gándara da Barca 499533 4762080 Pedra Moura Aldemunde 532245 4777251 Forno dos Mouros Silvoso 539950 4783517 Pedra Cuberta 501256 4770760 Pedra da Arca de Regoelle 498823 4761922 Casota de Freans 492966 4766132 Arca de Rabós 498363 4764983 Arca de Ogas 495642 4769752 Necropole das Caxadas 496334 4769533 Pedra Vixía 504537 4776870 Dolmen de Fontemoureira 503700 4738409 Fornela dos Mouros 499409 4779427 Anta da Gandara de Baiñas 1 500077 4762161 Anta do Alterio 499541 4762067 Arquiña de Vilaseco 493916 4769065 Mina da Parxubeira 505178 4754164 Dolmen de Axeitos 498625 4716405 Casiña dos Mouros 502418 4744283 Dolmen de Cabaleiros 536590 4771671 Monte das Mamoas 514829 4747992 Parxubeira 2 505305 4754378

554 Supplementary Data Table 2: Site List of 35 monuments with UTM (N29) locations. 555

26