Macroecological factors shape local-scale spatial patterns in

agriculturalist settlements

Tingting Taoa#, Sebastián Abadesb#, Shuqing Tengc, Zheng Y.X. Huangd, Luís Reinoe,f,g, Bin J.W. Chenh, Yong Zhangh, Chi Xua,1, Jens-Christian Svenningc

a School of Life Sciences, Nanjing University, Nanjing 210023, China,b Center for Genomics and Bioinformatics, Faculty of Science, Universidad Mayor, Camino La Pirámide 5750, Huechuraba, Chile,c Section for Ecoinformatics & Biodiversity, Department of Bioscience, Aarhus University, Ny Munkegade 114, DK-8000 Aarhus C, Denmark,d College of Life Science, Nanjing Normal University, Nanjing 210046, China,e CIBIO/InBIO, Research Centre in Biodiversity and Genetic Resources, University of Porto, 4485-661 Vairão, Portugal,f CIBIO/InBIO, Research Centre in Biodiversity and Genetic Resources, Universidade de Évora, 7004-516 Évora, Portugal,g CEABN/InBIO-Centro de Estudos Ambientais ‘Prof. Baeta Neves’, Instituto Superior de Agronomia, Universidade de Lisboa, Tapada da Ajuda, 1349-017 Lisboa, Portugal,h College of Biology and the Environment, Nanjing Forestry University, Nanjing 210037, China

1 To whom correspondence should be addressed: Chi Xu, E-mail: xuchi@nju.edu.cnSchool of Life Sciences, Nanjing University, Nanjing 210023, ChinaTel & Fax: +86-(0)25-89684396

#Authors contributed equally to this work

1

1

2

3456

789

101112131415161718192021

22232425

2627

Abstract

Macro-scale patterns of human systems ranging from population distribution to linguistic

diversity have attracted recent attention, giving rise to the suggestion that macroecological

rules shape the assembly of human societies. However, in which aspects the geography of

our own species is shaped by macroecological factors remains poorly understood. Here, we

provide a first demonstration that macroecological factors shape strong local-scale spatial

patterns in human settlement systems, through an analysis of spatial patterns in

agriculturalist settlements in eastern mainland China based on high-resolution Google Earth

images. We used spatial point pattern analysis to show that settlement spatial patterns are

characterised by over-dispersion at fine spatial scales (0.05-1.4 km), consistent with territory

segregation, and clumping at coarser spatial scales beyond the over-dispersion signals,

indicating territorial clustering. Statistical modelling shows that, at macroscales, potential

evapotranspiration and topographic heterogeneity have negative effects on territory size, but

positive effects on territorial clustering. These relationships are in line with predictions from

territory theory for hunter-gatherers as well as for many animal species. Our results help to

disentangle the complex interactions between intrinsic spatial processes in agriculturalist

societies and external forcing by macroecological factors. While one may speculate that

humans can escape ecological constraints because of unique abilities for environmental

modification and globalised resource transportation, our work highlights that universal

macroecological principles still shape the geography of current human agricultural societies.

Keywords: human macroecology, territory, scale, energy, topography, point pattern

2

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

1. Introduction

Among all organisms on Earth, our own species (Homo sapiens) is unique in many senses.

One remarkable characteristic of humans is the extraordinarily high sociocultural diversity,

despite comparatively little genetic variation [1]. Understanding sociocultural traits in

particular human communities has been a traditional focus in social sciences. However, recent

studies now start to put together the pieces for the ‘big picture’ of human societies, having led

to the discovery of a number of conspicuous large-scale patterns. Interestingly, a portion of

these anthropogenic patterns parallel classical patterns documented in the field of

macroecology [2], as seen e.g., in the congruence of latitudinal gradients in cultural and

species diversity [3-6]. It is increasingly suggested that human life histories are constrained

by ecological factors, and large-scale patterns of sociocultural traits may be explained by

macroecological processes that are analogous to those underlying biological patterns, giving

rise to the emerging field ‘human macroecology’, which aims to understand human-

environment relationships through quantifying statistical patterns of anthropogenic systems at

large spatiotemporal scales [7]. This emerging field has provided exciting findings showing

that macroecological rules underpin a wide range of anthropogenic patterns, e.g., population

distributions [8, 9], cultural distributions and dynamics [10-13], disease and public health [14,

15], as well as economy and sustainability [16, 17].

Macroecological approaches have provided valuable insights on many aspects of

ecological systems [18]. Spatial pattern is a fundamental characteristic of ecological systems;

importantly, it provides useful signatures for elucidating ecological interactions and processes

that structure the ecosystems [19-21]. However, the link between local spatial

patterns/processes and macroecology is not well quantified or understood. An important

example that illustrates such link are the so-called ‘Turing patterns’, a special class of patterns

periodically extended in space, resembling regular spots, stripes or labyrinths, that have been

found in a wide range of ecosystems such as dryland vegetation and coastal mussel beds [22,

23]. It has been suggested that Turing vegetation patterns are mainly governed by micro-scale

interactions, represented as coupled short-range positive feedbacks and long-range negative

3

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

feedbacks in general [24]. However, a recent worldwide analysis has shown that the

occurrence of these patterns are also well correlated with macroecological factors including

climate, topography and soil [25]. Such cross-scale correlations imply that at broad spatial

scales, relevant macroecological factors can substantially determine the local biotic and

abiotic factors underpinning spatial patterning. In this sense, linking local-scale spatial

patterns to large-scale determinants not only contributes to identifying the mechanisms

underlying pattern formation, but also help to scale up to better understand system

functioning and dynamics at macroscopic scales. So far, it remains unclear if and how

macroecological factors shape local spatial patterns in most systems.

Here, we connect the two emerging themes, human macroecology and cross-scale spatial

pattern formation, by assessing the role of macroecological factors in shaping local-scale

spatial patterns of anthropogenic systems. We are particularly interested in rural human

settlements, because they are among the most conspicuous imprints that humans created on

the Earth’s land surfaces, and are home to about half world’s populations. A glimpse at rural

settlements from airplanes or remotely sensed images may provide the impression that their

spatial patterns mimic those of other species’ colonies (termites for instance) at proportionate

scales (see Fig. 1 for some snapshots). This said, there is no doubt that spatial patterns of

humans are likely to differ from those of termites, because humans are arguably not a eusocial

species [26], and human settlements are generally shaped by more diverse and complex

displays of social behaviours. In social sciences, there have been extensive studies suggesting

the important role of diverse sociocultural factors in the formation and spatial patterning of

rural settlements, ranging from general theories of spatial location [27, 28] to specific

sociocultral processes [29-31]. However, irrespective of sociocultural context, settlement

formation and associated land use intensification involve emergent processes resulting from

interactions between human individuals, groups, and the environments [32]. If these human

interactions are governed by coarse-grained ecological constraints as in other species, one

would expect that human settlement patterning follows general macroecological rules.

Indeed, recent work has revealed macroecological underpinnings (mostly related to

water-energy patterns) of human collective behaviour and social structures, providing

4

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

compelling clues for exploring settlements patterning in this framework. In hunter-gatherer

societies, territory size of foraging groups has been found to be a function of energy (food)

supply and population density [33, 34]. A corollary is that the dependence on these factors

should also hold for the spacing distance between camps as a surrogate measure of territory

size. When foraging groups strongly compete for food resources, it is reasonable to expect

regularly distributed (i.e., over-dispersed) camps as a result of intensive territory segregation.

In contrast, if positive relationships (e.g., cooperative defense and food sharing) are prevalent,

group territories would tend to overlap [35, 36], likely resulting in clustered patterns.

Little macroecological work has been done on agrarian societies, despite their

importance. In parallel with hunter-gatherers, an individual agriculturalist settlement can be

seen as a group that mostly utilise resources (agricultural production) from surrounding lands

as its territory. An important transition from hunter-gatherer to agrarian societies is land use

intensification (farming), which makes resources more abundant and predictable [37]. This

could have fundamental consequences on collective behaviour and the settlement system.

First, denser resources would allow for higher population densities and smaller territories

[35]. Second, high-quality resources and more sedentary groups would intensify territory

defense, as predicted by the theory of economic defendability [37, 38]. Third, rich resources,

dense populations, and sedentary groups could favour the development of complex, long-

range social networks [37] between settlements. These characteristics have major implications

for spatial patterning of agriculturalist settlements, leading us to propose the following

hypotheses:

(1) Significant territory segregation occurs between neighbouring agriculturalist settlements,

resulting in over-dispersed settlement patterns at local scales. At coarse spatial scales,

agriculturalist settlements could have clustered territories and hence under-dispersed

settlement patterns due to environmental constraints or positive social interactions (e.g.,

within ethnic groups).

(2) At macroscopic scales, high resource availability (energy measured as PET, soil quality,

hydrological density) has a negative influence on the territory size of agriculturalist

5

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

settlements, and also favours territory clustering.

(3) In mountainous terrains, steep topography result in smaller territory sizes, while also

forcing territorial clustering.

(4) Resource availability and topographic forcing can produce indirect influences on

settlement patterning via mediating settlement size and population density.

To test these hypotheses, we used high-resolution (0.6 m) remotely sensed images to

scrutinize rural settlements across mainland China, where traditional agriculture has a long

history and has produced strong landscape imprints. We conducted spatial point pattern

analysis to study spatial associations between the settlements and thereafter multiple

regression and structural equation modelling (SEM) to examine the effects of

macroecological factors on the detected patterns.

2. Methods

(a) Study area and settlement data

China has a history of thousands of years as an agrarian society. Despite rapid urbanisation in

recent decades, over 700,000 rural settlements are home to c. 60% of the Chinese population

[39]. A clear national-scale differentiation has been observed across mainland China, with

most of the population and settlements distributed in the south-east part in contrast with the

north-west part where is characterised by drylands and alpine plateaus. This divide roughly

follows a straight line stretching diagonally across China, from the Heihe City to the

Tengchong County (known as the ‘Heihe-Tengchong Line’ or ‘Hu Huanyong Line’ [40]). We

restricted our study area to area southeast of this line, as this is where relatively dense

settlements accommodating c. 96% of total Chinese population are found (Fig. 1a).

We divided the study area into 490 grid cells sized of 100 km×100 km, within each of

which we systematically sampled a 10 km×10 km plot using high-resolution remote sensing.

In this way, the sampling plots were relatively evenly distributed across the study area. The

plots were selected following the criteria: 1) Each plot was capable of representing a

relatively homogeneous agricultural landscape, i.e., no apparent non-agricultural land use

6

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

features were found within the plot. 2) Areas that are in proximity to (< 10 km) or contain

towns/cities were avoided. 3) Grid cells intersected with metropolitan regions or urban

agglomerations were excluded to avoid areas under substantial urban land use. When multiple

candidate remote sensing plots existed in a given 100 km×100 km cell, we selected the one

closest to the cell centroid. We identified 320 sampling plots for the 490 grid cells (Fig. 1).

For each sampling plot we collected and geo-referenced (using the UTM projections)

high-resolution (0.6 m) QuickBird images covering 2005-2010, freely available from Google

Earth. The distribution of rural settlements was derived using a supervised classification (with

the maximum likelihood method) in combination with visual inspections (manual tracing was

used afterwards to correct classification errors). Each individual settlement was represented as

a vector polygon with its centroid representing the settlement’s spatial position. The centroids

with geographic coordinates were taken as input data for the subsequent spatial point pattern

analysis (Fig. S1). The image processing was done using ENVI 4.8 and ArcGIS 10.2.

(b) Spatial point pattern analysis

We applied the univariate point pattern analysis (PPA) based on the second-order pair-

correlation function g(r) to test if the settlements points present significant signals of under-

dispersion or over-dispersion in contrast with complete spatial randomness (CSR, as the null

model) over a range of observational spatial scales. A clear merit of the PPA approach for

testing our hypotheses is that it is capable of detecting nested hierarchical spatial structures at

multiple spatial scales with a single summary statistic (i.e., the g(r) statistic), facilitating

subsequent statistical modelling. To characterise any spatial association signal (over- or

under-dispersion) detected beyond the 99% CIs of CSR, we quantified 1) presence/absence of

the signal (as a binary variable), and 2) spatial scale range across which the signal arises (as a

continuous variable), calculated as the width of the scale domain presenting the non-CSR

signal (Fig. 2). For the detected signals we used the nomenclature indicating both association

properties and their relative scales, e.g., small-scale over-dispersion or large-scale under-

dispersion (see Results). The point pattern analysis was completed with the Programita

software [41]. See S1 for more detailed descriptions.

7

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

(c) Macroecological variables

We selected four environmental variables for testing our hypotheses. Mean annual potential

evapotranspiration (PET) was used to characterise energy availability. Elevation range

derived from Shuttle Radar Topography Mission (SRTM) data was used to characterise

topographic relief. Soil quality was characterised by total exchangeable bases (the sum of

HCl-soluble bases including Ca2+, Mg2+, K+ and Na+) as the major components of soil fertility

[42]. The length of rivers and major channels per unit area (river density for abbreviation)

derived from the China hydrological map was used to characterise the hydrological system. In

addition, our fourth hypothesis involves settlement size and regional population density. Mean

settlement size was calculated as the mean area of the settlement polygons derived from the

remotely sensed images (Fig. S1). Regional population density was calculated from a Chinese

gridded database on population density. The mean values of the abovementioned variables

were calculated for each sampling plot, except that the topographic variable was represented

by elevation range within the plot. In addition, as the broad geographic extent of China covers

a diversity of social/ethnic backgrounds, to corrected for this potential influence, we used a

categorical variable representing the major historical/cultural regions, i.e., Northern, Southern

and Western China. This region variable was derived from a national-scale regionalisation

map developed by cultural geographers to classify characteristics of traditional Chinese

settlements [43]. Data sources for the environmental variables are listed in Table S1.

(d) Statistical analyses

We used multiple regression to examine the relations between the macroecological variables

(as predictors) and the spatial association variables previously quantified from the point

pattern analysis (as response). For presence/absence of spatial association we fitted

generalised linear models (GLMs) with the binomial distribution and logit link (logistic

regression), while for the scale range of spatial association we fitted ordinary least square

(OLS) models. We checked for multicollinearity of the predictors based on the variation

inflation factor (VIF). The VIF value was less than 3 for all predictors, indicating no serious

multicollinearity. We started from a full model, then performed a stepwise backward model

8

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

selection to obtain a parsimonious model based on AIC. The resulting parsimonious models

are robust when all possible models are considered. We also checked the Moran’s I coefficient

(based on the inverse distance weight method) for the model residuals, and detected rather

low absolute values of Moran’s I (Table S2-S4), suggesting little potential for spatial

autocorrelation problems [44]. Before model fitting, elevation range and settlement size were

log-transformed and scale ranges of the pattern features were log(x+1)-transformed to avoid

highly skewed distributions. We used piecewise structural equation models (piecewise SEMs)

to test our fourth hypothesis, where population density and settlement size were included as

modulator variables. For the logistic regression, we used a random subset with equal presence

and absence data to avoid the data imbalance problem. The statistical analyses were

implemented in R 3.3.1 [45] with the ape [46], and piecewiseSEM [47] packages.

3. Results

Our results showed that local spatial associations between agriculturalist settlements in

eastern China are characterised by two classes of non-random signals, as indicated by the

summary statistic g(r) of the point pattern analysis. These two classes of signals are structured

within a spatial scale hierarchy (Fig. 2). One class of signals represents over-dispersion (as

indicated by g(r) values falling below the lower 99% confidence intervals of CSR) at the

lower end of the scale gradient (small-scale over-dispersion or SO hereafter). Within the scale

range of SO (SRSO for abbreviation; 0.05-1.4 km), there was a significantly lower probability

of finding a neighbouring pair of settlement points than expected by chance. Interpreted

intuitively, this scale range reflects the average distance at which the settlements are

‘repelling’ each other. SRSO can thus serve as a surrogate measure of territory size. The signal

of SO was found in all 320 sampling plots, suggesting that strong territory segregation among

neighbouring settlements is ubiquitous.

At larger spatial scales beyond SO (usually >0.2 km), we detected a second class of

signals representing under-dispersion (as indicated by g(r) values rising above the upper 99%

confidence intervals of CSR, large-scale under-dispersion or LU hereafter) in ~30% of the

plots, with the rest plots characterised by CSR (Fig. 1a). Note that for a given study plot,

9

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

over-dispersion was found to be always present at smaller spatial scales than under-

dispersion, but this is not necessarily the case when comparing between different plots. An

interpretation is that at any given distance within the scale range of LU (SRLU for

abbreviation, 0.1-2.6 km), there was a significantly higher probability of finding a pair of

settlement points than expected by chance, suggesting that settlements are ‘attracting’ each

other. SRLU can thus serve as a surrogate measure of cluster size.

Our results from the regression models showed that mean territory size (as measured by

SRSO) had significantly negative correlations with PET and elevation range (Fig. 3, S2, Table

S2), suggesting that higher energy availability and steeper topography favour smaller

territories, or put in another way, finer-grained territory segregation. We also found a

significant effect of the region variable, suggesting that territory size is influenced by large-

scale historical/cultural factors. Soil fertility and river density did not have significant effects.

Our SEM analysis further suggested several plausible pathways through which population

density and settlement size transmit the effects of the macroecological factors on territory size

(Fig. S4). The two major pathways are: 1) higher PET and lower elevation ranges can favour

higher population densities, in turn leading to smaller territories; 2) both PET and elevation

range have negative effects on settlement size, while settlement size has a positive effect.

Our statistical models also showed that occurrence of clustered patterns had significantly

positive correlations with PET and elevation range, suggesting that higher energy availability

and steeper topographic relief favored clustered territories at larger scales (Fig. 3, Table S3).

Again, our SEM analysis suggested population density and settlement size could play as

mediators of the influences from PET and elevation range (Fig. S4). Both population density

and settlement size had negative effects on the occurrence of clustered territories. The

abovementioned direct and indirect effects were similar, but weaker for cluster size (as

measured by SRLU, where present), except for a more pronounced effect of river density on

cluster size (Fig. 4, S3, S4, Table S4).

10

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

4. Discussion

The point pattern analyses revealed marked regularities in spatial pattern features in

agriculturalist settlements across eastern mainland China. These features could be

decomposed in terms of hierarchical patterns at two spatial scales: ubiquitous over-dispersed

patterns at small spatial scales, supporting the hypothesis of significant territory segregation;

and a moderate level of under-dispersed patterns at larger spatial scales, in line with the

hypothesis of territorial clustering.

Are these local-scale spatial pattern features governed by macroecological factors at

broad scales? Our statistical models indeed suggest so. One of our major findings points to a

significantly negative effect of PET on territory size of agriculturalist settlements, supporting

our second hypothesis. This energy effect coincides well with its counterparts for hunter-

gatherers [34], as well as group-living animals in general [48]. It has been commonly

observed that animal territory size expands when resource abundance decreases, with

individuals maintaining per-capita intake rates unchanged [48, 49]. When it comes to the

patterning of agriculturalist settlements, a similar mechanism may operate because individual

settlements would need more arable lands when per unit agricultural production is limited by

relatively low water-energy availability.

Further comparisons strengthen the suggestion that ‘universal’ ecological principles

shape territoriality of humans and animal species (here the distinction between humans and

animals is arbitrary from a biological perspective). The territory theory of animals predicts

that a relaxed degree of population crowding or a larger group (colony) size can result in

enlarged territory sizes [48]. In parallel, an analogous prediction for agriculturalist settlements

is that territory size should be positively correlated with settlement (group) size, because a

larger settlement with more residents would need a larger territory; while regional population

density should have a negative correlation with territory size, because increasingly crowded

populations at a regional level would compress individual territories if territory overlapping is

absent [36]. Our results from SEM confirmed that it is indeed possible that population density

and settlement size can have such mediation effects, as driven by energy availability (Fig S4).

11

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

The transmission pathway mediated via population density can be interpreted as high energy

availability providing support for higher population densities, which in turn compress

territories. Indeed, this mechanism has also been widely observed in group-living animals,

such as birds and mammals [48, 50, 51]. The other pathway represents the scenario that low

energy availability favours larger settlements (groups), which need larger territories. The

negative effect of PET on settlement size is plausibly explained by the sociality of humans: in

resource-poor areas, aggregating into (appropriately) larger settlements can have obvious

benefits for the residents as a collective, in the sense that they can ameliorate the harshness

via cooperation. Such enhanced positive interaction is a common strategy for plants in

stressful environments (the so-called stress gradient hypothesis) [52, 53], and has been

observed in some group-living animals, e.g., Octodon degu [54]. However, generally strong

intra-specific competition for food resources may be an important reason why this trend is not

prevalent in animal communities [55, 56].

Territorial clustering is not uncommon in nature [57-60]. For animals the prevailing

explanations for this pattern, regardless of specific mechanisms, address the benefits in

various forms gained through conspecific attraction [57]. A particular case is that in resource

abundant areas, territorial animals may settle closely next to each other, because in this way

they can identify and thereby access the high-quality resources more efficiently [58-60]. This

perhaps also partly explains our observed positive correlation between settlement clustering

and PET. Another benefit can be gained through intraspecific cooperation in terms of e.g.

intruder defense and predator protection [57, 61]. Likewise, rural settlement clustering could

be a sign of strong cooperation between the settlements. Indeed, it has been suggested that in

higher-resource settings, human societies tend to develop higher levels of cooperation and

more complex social networks [37]. In addition, settlement size and population density can

again play as significant mediators as suggested by SEM. A possible explanation is that

between-settlement connections would be weakened when the settlements are large and

relatively far away from their neighbours (because of their large territories), while dense

populations may enhance social connections that can interlink most of the settlements across

the whole region and therefore destruct formation of local clusters.

12

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

We did not observe the hypothesised effects of soil fertility and river density (Fig. S2,

S3). One possibility is that, humans can increasingly temper the limitation of soil quality and

surface water resource with the aid of technologies that have been developed through the

long-term agricultural practices during the past millennia, while rivers can act as another

environmental forcing that tends to obstruct the development of clustered patterning through

segregation. It is also possible that sometimes settlements developing along rivers form linear

patterns, but such river-settlement association cannot be well detected using the PPA. The

effects of strong macroecological forcing on settlement patterning might be dependent on the

environmental modification ability of our own species. A possible scenario in the future is that

advancement of technology would make humans escape macroecological constraints,

eventually reshaping settlements patterning and persistence. Still, it remains to be studied if,

when and how human settlements start to become sytematically unconstrained by

macroecological rules, pointing to an important area for future research on human

macroecology, i.e., focusing on regions most strongly characterised by the expected

Anthropocene societal development trends.

Topography as a strong environmental forcing usually poses marked limitations on

human behaviours [62, 63], increasing the difficulty for agricultural land use, settlement

establishment and human mobility. Our analysis suggested that steep topography restricts

territory size directly as well as indirectly via restricting settlement size. While steep

topography could also have an indirect positive effect via limiting population density, though

any such effect was outweighed by the overall negative effects. Another important

consequence of topographic relief is territorial clustering, in the sense that only some

(relatively flat) areas are suitable or attractive for settlements in complex terrains. This direct

forcing coincides with the indirect effects that small settlements and low population densities

in complex terrains also favour clustered patterns.

Quantifying sociocultural and historical effects is an important future challenge for

improving our understanding of settlement patterning [64]. The coarse-scale regionalisation

variable used in this study was not able to well reflect the LU signals (Table S3, S4),

suggesting that more complex processes are likely playing a role. This is especially the case

13

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

for understanding the reason why territorial clustering is mostly present in the southernmost

part of mainland China (Fig. 1a). A promising approach for future studies would be using

phylogenetic information (based on cultural lineages, rather than genetic lineages of humans)

to account for cultural/historic similarities, given that reliable cultural/historical phylogenetic

trees can be constructed.

Taken together, our results help to disentangle the complex interactions between intrinsic

spatial processes in agriculturalist societies and forcing by macroecological factors (see Fig. 4

for a synthesis of the direct and indirect effects). Settlement spatial patterns provide us an

opportunity to unravel the interesting similarities in the territoriality of agriculturists, hunter-

gatherers and animals, and to understand the ecological mechanisms that are plausibly

‘universal’ to humans and other species. Although territory theory has long been established

in ecology, successfully explaining territory patterning of many animal species, our work is,

to our knowledge, the first extension of territory theory to the geography of human settlement

patterns. Macroecological factors have played a strong role in shaping current local-scale

spatial patterns in agriculturalist settlement systems despite the fact that humans have the

ability to temper ecological constraints through environmental modification and globalised

resource transfer. Our work demonstrates the value of a macroecological perspective for

understanding human collective behaviour and settlement systems, highlighting the ecological

nature of human societies. While our focus is on ecological interactions, further consideration

of historical-geographic context would facilitate comprehensive understanding of variability

of settlement patterning. Importantly, macro-scale sociopolitical factors (e.g., governmental

planning) can play a substantial role in settlement patterning [65], leaving it an open question

when and how sociopolitical factors lead to settlement systems that deviate from those

generated by spontaneous intrinsic socio-ecological dynamics and extrinsic macroecological

forcing.

14

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

Funding

This work is supported by The National Key R&D Program of China (2017YFC0506001),

the National Natural Science Foundation of China (NSFC 31770512) and the Fundamental

Research Funds for the Central Universities (0208-14380068) awarded to C.X. J.-C.S.

considers this work a contribution to his VILLUM Investigator project “Biodiversity

Dynamics in a Changing World” funded by VILLUM FONDEN and the Danish National

Research Foundation Niels Bohr professorship project Aarhus University Research on the

Anthropocene (AURA). S.A. acknowledges support from FONDECYT 1170995. S.T.

acknowledges support from China Scholarship Council (201406190179) and the European

Research Council (ERC-2012-StG-310886-HISTFUNC). L.R. received support from

the Portuguese Ministry of Science and Higher Education and the

European Social Fund, through FCT, under POPH – QREN – Typology 4.1

(post-doc grant SFRH/BPD/93079/2013). Z.Y.X.H. acknowledges support from

NSFC (31500383). B.J.W.C. acknowledges support from NSFC (31600328) and Natural

Science Foundation of Jiangsu Province, China (BK20160924).

Acknowledgements

We thank Marten Scheffer and Guochun Shen for valuable discussion. We also thank

Melodie McGeoch and two anonymous reviewers for constructive comments on an earlier

version of this manuscript.

Author contributions

C.X. designed the study. T.T., S.A. and C.X. collected and carried out spatial analyses, S.A.

and C.X. carried out statistical analyses, C.X. wrote the first draft, all authors interpreted the

results and revised the manuscript. L.R., C.X. and J.-C.S. approved the final manuscript. T.T.

and S.A. contributed equally to this work.

Competing interests

The authors declare no competing interests.

15

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

References

[1] Laland, K.N., Odling-Smee, J. & Myles, S. 2010 How culture shaped the human genome: bringing genetics and the human sciences together. Nat Rev Genet 11, 137-148. (doi:10.1038/nrg2734).[2] Brown, J.H. & Maurer, B.A. 1989 Macroecology: the division of food and space among species on continents. Science 243, 1145. (doi:10.1126/science.243.4895.1145).[3] Collard, I.F. & Foley, R.A. 2002 Latitudinal patterns and environmental determinants of recent human cultural diversity: do humans follow biogeographical rules? Evol Ecol Res 4, 371-383.[4] Laitin, D.D., Moortgat, J. & Robinson, A.L. 2012 Geographic axes and the persistence of cultural diversity. Proc Natl Acad Sci USA 109, 10263-10268. (doi:10.1073/pnas.1205338109).[5] Pagel, M. & Mace, R. 2004 The cultural wealth of nations. Nature 428, 275-278. (doi:10.1038/428275a).[6] Sutherland, W.J. 2003 Parallel extinction risk and global distribution of languages and species. Nature 423, 276-279. (doi:10.1038/nature01607).[7] Burnside, W.R., Brown, J.H., Burger, O., Hamilton, M.J., Moses, M. & Bettencourt, L.M.A. 2012 Human macroecology: linking pattern and process in big-picture human ecology. Biol Rev 87, 194-208. (doi:10.1111/j.1469-185X.2011.00192.x).[8] Xu, C., Chen, B.J.W., Abades, S., Reino, L., Teng, S.Q., Ljungqvist, F.C., Huang, Z.Y.X. & Liu, M.S. 2015 Macroecological factors explain large-scale spatial population patterns of ancient agriculturalists. Global Ecol Biogeogr 24, 1030-1039. (doi:10.1111/geb.12343).[9] Benito, B.M., Svenning, J.-C., Kellberg-Nielsen, T., Riede, F., Gil-Romera, G., Mailund, T., Kjaergaard, P.C. & Sandel, B.S. 2016 The ecological niche and distribution of Neanderthals during the Last Interglacial. J Biogeogr. (doi:10.1111/jbi.12845).[10] Amano, T., Sandel, B., Eager, H., Bulteau, E., Svenning, J.-C., Dalsgaard, B., Rahbek, C., Davies, R.G. & Sutherland, W.J. 2014 Global distribution and drivers of language extinction risk. Proceedings of the Royal Society B: Biological Sciences 281. (doi:10.1098/rspb.2014.1574).[11] Botero, C.A., Gardner, B., Kirby, K.R., Bulbulia, J., Gavin, M.C. & Gray, R.D. 2014 The ecology of religious beliefs. Proc Natl Acad Sci USA 111, 16784-16789. (doi:10.1073/pnas.1408701111).[12] Capitán, J.A., Bock Axelsen, J. & Manrubia, S. 2015 New patterns in human biogeography revealed by networks of contacts between linguistic groups. Proceedings of the Royal Society B: Biological Sciences 282. (doi:10.1098/rspb.2014.2947).[13] Foley, R.A. & Mirazón Lahr, M. 2011 The evolution of the diversity of cultures. Philosophical Transactions of the Royal Society B: Biological Sciences 366, 1080. (doi:10.1098/rstb.2010.0370).[14] Guernier, V., Hochberg, M.E. & Guégan, J.-F. 2004 Ecology drives the worldwide distribution of human diseases. PLoS Biol 2, e141. (doi:10.1371/journal.pbio.0020141).[15] Stephens, P.R., Altizer, S., Smith, K.F., Alonso Aguirre, A., Brown, J.H., Budischak, S.A., Byers, J.E., Dallas, T.A., Jonathan Davies, T., Drake, J.M., et al. 2016 The macroecology of infectious diseases: a new perspective on global-scale drivers of pathogen distributions and impacts. Ecol Lett 19, 1159-1171. (doi:10.1111/ele.12644).[16] Burger, J.R., Allen, C.D., Brown, J.H., Burnside, W.R., Davidson, A.D., Fristoe, T.S., Hamilton, M.J., Mercado-Silva, N., Nekola, J.C., Okie, J.G., et al. 2012 The macroecology of sustainability. PLoS Biol 10, e1001345. (doi:10.1371/journal.pbio.1001345).[17] Brown, J.H., Burger, J.R., Burnside, W.R., Chang, M., Davidson, A.D., Fristoe, T.S., Hamilton, M.J., Hammond, S.T., Kodric-Brown, A., Mercado-Silva, N., et al. 2014 Macroecology meets

16

411

412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453

macroeconomics: Resource scarcity and global sustainability. Ecol Eng 65, 24-32. (doi:DOI 10.1016/j.ecoleng.2013.07.071).[18] Smith, F.A., Lyons, S.K., Ernest, S.M. & Brown, J.H. 2008 Macroecology: more than the division of food and space among species on continents. Prog Phys Geog 32, 115-138. (doi:10.1177/0309133308094425).[19] Levin, S.A. 1992 The problem of pattern and scale in ecology. Ecology 73, 1943-1967. (doi:10.2307/1941447).[20] McIntire, E.J. & Fajardo, A. 2014 Facilitation as a ubiquitous driver of biodiversity. New Phytol 201, 403-416. (doi:10.1111/nph.12478).[21] Xu, C., Holmgren, M., Van Nes, E.H., Maestre, F.T., Soliveres, S., Berdugo, M., Kefi, S., Marquet, P.A., Abades, S. & Scheffer, M. 2015 Can we infer plant facilitation from remote sensing? a test across global drylands. Ecol Appl 25, 1456-1462. (doi:10.1890/14-2358.1).[22] Liu, Q.X., Herman, P.M.J., Mooij, W.M., Huisman, J., Scheffer, M., Olff, H. & van de Koppel, J.V. 2014 Pattern formation at multiple spatial scales drives the resilience of mussel bed ecosystems. Nature Communications 5. (doi:10.1038/ncomms6234).[23] Rietkerk, M., Dekker, S.C., De Ruiter, P.C. & Van de Koppel, J. 2004 Self-organized patchiness and catastrophic shifts in ecosystems. Science 305, 1926-1929. (doi:10.1126/science.1101867).[24] Rietkerk, M. & Van de Koppel, J. 2008 Regular pattern formation in real ecosystems. Trends in Ecology and Evolution 23, 169-175. (doi:10.1016/j.tree.2007.10.013).[25] Deblauwe, V., Barbier, N., Couteron, P., Lejeune, O. & Bogaert, J. 2008 The global biogeography of semi-arid periodic vegetation patterns. Global Ecol Biogeogr 17, 715-723. (doi:10.1111/j.1466-8238.2008.00413.x).[26] Gintis, H. 2012 Clash of the Titans. Bioscience 62, 987-991. (doi:10.1525/bio.2012.62.11.8).[27] Christaller, W. 1966 Central places in southern Germany, Prentice-Hall.[28] Hudson, J.C. 1969 A location theory for rural settlement. Annals of the Association of American Geographers 59, 365-381.[29] Chisholm, M. 1973 Rural settlement and land use, Transaction Publishers.[30] Curtis, D.R. 2013 The emergence of concentrated settlements in medieval western Europe: explanatory frameworks in the historiography. Canadian Journal of History 48, 223-251. (doi:10.3138/cjh.48.2.223).[31] Lichter, D.T. & Johnson, K.M. 2006 Emerging rural settlement patterns and the geographic redistribution of America's new immigrants. Rural Sociology 71, 109-131. (doi:10.1526/003601106777789828).[32] Ellis, E.C. 2015 Ecology in an anthropogenic biosphere. Ecol Monogr 85, 287-331. (doi:10.1890/14-2274.1).[33] Freeman, J. & Anderies, J.M. 2015 The socioecology of hunter–gatherer territory size. Journal of Anthropological Archaeology 39, 110-123. (doi:10.1016/j.jaa.2015.03.002).[34] Hamilton, M.J., Milne, B.T., Walker, R.S. & Brown, J.H. 2007 Nonlinear scaling of space use in human hunter-gatherers. Proc Natl Acad Sci USA 104, 4765-4769. (doi:10.1073/pnas.0611197104).[35] Freeman, J. 2016 The socioecology of territory size and a "work-around" hypothesis for the adoption of farming. PLoS ONE 11, e0158743. (doi:10.1371/journal.pone.0158743).[36] Hamilton, M.J., Burger, O., DeLong, J.P., Walker, R.S., Moses, M.E. & Brown, J.H. 2009 Population stability, cooperation, and the invasibility of the human species. Proc Natl Acad Sci USA 106, 12255-12260. (doi:10.1073/pnas.0905708106).

17

454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497

[37] Marean, C.W. 2016 The transition to foraging for dense and predictable resources and its impact on the evolution of modern humans. Philosophical Transactions of the Royal Society B: Biological Sciences 371. (doi:10.1098/rstb.2015.0239).[38] Brown, J.L. 1964 The evolution of diversity in avian territorial systems. The Wilson Bulletin 76, 160-169.[39] Tian, G.J., Qiao, Z. & Zhang, Y.Q. 2012 The investigation of relationship between rural settlement density, size, spatial distribution and its geophysical parameters of China using Landsat TM images. Ecol Model 231, 25-36. (doi:10.1016/j.ecolmodel.2012.01.023).[40] Hu, Z., Wang, Y., Liu, Y., Long, H. & Peng, J. 2016 Spatio-Temporal Patterns of Urban-Rural Development and Transformation in East of the “Hu Huanyong Line”, China. ISPRS International Journal of Geo-Information 5, 24.[41] Wiegand, T. & Moloney, K.A. 2014 Handbook of spatial point-pattern analysis in ecology, Chapman and Hall.[42] Huston, M.A. 2012 Precipitation, soils, NPP, and biodiversity: resurrection of Albrecht's curve. Ecol Monogr 82, 277-296. (doi:10.1890/11-1927.1).[43] Liu, P., Liu, C., Deng, Y., Shen, X., Li, B. & Hu, Z. 2010 Landscape division of traditional settlement and effect elements of landscape gene in China. Acta Geographica Sinica 65, 1496-1506.[44] Dormann, C.F., McPherson, J.M., Araujo, M.B., Bivand, R., Bolliger, J., Carl, G., Davies, R.G., Hirzel, A., Jetz, W., Kissling, W.D., et al. 2007 Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography 30, 609-628. (doi:DOI 10.1111/j.2007.0906-7590.05171.x).[45] R Core Team. 2016 R: a language and environment for statistical computing. Vienna, Austria., R Foundation for Statistical Computing.[46] Paradis, E., Claude, J. & Strimmer, K. 2004 APE: analyses of phylogenetics and evolution in R language. Bioinformatics 20, 289-290. (doi:10.1093/bioinformatics/btg412).[47] Lefcheck, J.S. 2016 PIECEWISESEM: Piecewise structural equation modelling in R for ecology, evolution, and systematics. Methods in Ecology and Evolution 7, 573-579. (doi:10.1111/2041-210x.12512).[48] Adams, E.S. 2001 Approaches to the study of territory size and shape. Annu Rev Ecol Syst 32, 277-303. (doi:10.1146/annurev.ecolsys.32.081501.114034).[49] Johnson, D.D.P., Kays, R., Blackwell, P.G. & Macdonald, D.W. 2002 Does the resource dispersion hypothesis explain group living? Trends Ecol Evol 17, 563-570. (doi:10.1016/s0169-5347(02)02619-8).[50] Enoksson, B. & Nilsson, S.G. 1983 Territory size and population density in relation to food supply in the nuthatch Sitta europaea (Aves). J Anim Ecol 52, 927-935. (doi:10.2307/4464).[51] Jedrzejewski, W., Schmidt, K., Theuerkauf, J., Jedrzejewska, B. & Kowalczyk, R. 2007 Territory size of wolves Canis lupus: linking local (Bialowieza Primeval Forest, Poland) and Holarctic-scale patterns. Ecography 30, 66-76. (doi:10.1111/j.0906-7590.2007.04826.x).[52] Callaway, R.M. 2007 Positive interactions and interdependence in plant communities, Springer, Dordrecht.[53] Holmgren, M. & Scheffer, M. 2010 Strong facilitation in mild environments: the stress gradient hypothesis revisited. Journal of Ecology 98, 1269-1275. (doi:10.1111/j.1365-2745.2010.01709.x).[54] Ebensperger, L.A., Villegas, Á., Abades, S. & Hayes, L.D. 2014 Mean ecological conditions modulate the effects of group living and communal rearing on offspring production and survival. Behav Ecol 25, 862-870. (doi:10.1093/beheco/aru061).

18

498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541

[55] Barrio, I.C., Hik, D.S., Bueno, C.G. & Cahill, J.F. 2013 Extending the stress-gradient hypothesis - is competition among animals less common in harsh environments? Oikos 122, 516-523. (doi:10.1111/j.1600-0706.2012.00355.x).[56] Kittle, A.M., Anderson, M., Avgar, T., Baker, J.A., Brown, G.S., Hagens, J., Iwachewski, E., Moffatt, S., Mosser, A., Patterson, B.R., et al. 2015 Wolves adapt territory size, not pack size to local habitat quality. J Anim Ecol 84, 1177-1186. (doi:10.1111/1365-2656.12366).[57] Stamps, J.A. 1988 Conspecific attraction and aggregation in territorial species. Am Nat 131, 329-347. (doi:10.1086/284793).[58] Dias, R.I., Kuhlmann, M., Lourenco, L.R. & Macedo, R.H. 2009 Territorial clustering in the blue-black grassquit: reproductive strategy in response to habitat and food requirements. Condor 111, 706-714. (doi:10.1525/cond.2009.090142).[59] Stamps, J.A. 1987 Conspecifics as cues to territory quality - a preference of juvenile lizards (Anolis aeneus) for previously used territories. Am Nat 129, 629-642. (doi:10.1086/284663).[60] Ward, M.P. & Schlossberg, S. 2004 Conspecific attraction and the conservation of territorial songbirds. Conserv Biol 18, 519-525. (doi:10.1111/j.1523-1739.2004.00494.x).[61] Port, M. & Johnstone, R.A. 2013 Facing the crowd: intruder pressure, within group competition, ‐and the resolution of conflicts over group membership. ‐ Ecology and evolution 3, 1209-1218. (doi:10.1002/ece3.533).[62] Odgaard, M.V., Bøcher, P.K., Dalgaard, T., Moeslund, J.E. & Svenning, J.-C. 2014 Human-driven topographic effects on the distribution of forest in a flat, lowland agricultural region. Journal of Geographical Sciences 24, 76-92. (doi:10.1007/s11442-014-1074-6).[63] Sandel, B. & Svenning, J.-C. 2013 Human impacts drive a global topographic signature in tree cover. Nature Communications 4. (doi:10.1038/ncomms3474).[64] Brewer, J., Gelfand, M., Jackson, J., MacDonald, I., Peregrine, P., Richerson, P., Turchin, P., Whitehouse, H. & Wilson, D. 2017 Grand challenges for the study of cultural evolution. Nature Ecology & Evolution 1, 0070.[65] Batistella, M., Robeson, S. & Moran, E.F. 2003 Settlement design, forest fragmentation, and landscape change in Rondônia, Amazônia. Photogrammetric Engineering & Remote Sensing 69, 805-812. (doi:10.14358/PERS.69.7.805).

19

542543544545546547548549550551552553554555556557558559560561562563564565566567568569570

Figure Legends

Figure 1. Distribution of the sampling plots and snapshots of typical rural settlement systems

from Google Earth images across mainland China. (a) The plots locations are represented by

squares (blue and red indicate presence and absence of large-scale under-dispersion of

settlement pattern, respectively); the dashed line is known as the ‘Hu Huanyong Line’ with

sparse human populations on the left side and dense populations on the right side. (b, c) The

red points represent the centroids of individual settlements in two typical rural landscapes as

examples. The high-resolution remotely sensed images were acquired from Google Earth (the

original source: DigitalGlobe).

Figure 2. Illustrations of settlement pattern features revealed by the spatial point pattern

analysis. The typical point maps in (a, c) present the two classes of pattern features detected

by the g(r) function (b, d). Small-scale over-dispersion is found ubiquitously, while large-

scale under-dispersion is found only in a portion of the settlement systems (e.g., present in the

map c whereas absent in the map a). The scale ranges of small-scale over-dispersion and

large-scale under-dispersion are calculated as the widths of the cyan and orange bars (scale

domains) in (b, d), as indicated by the g(r) values (red solid lines) falling below and above the

99% confidence intervals (dashed lines), respectively.

Figure 3. Effects of potential evapotranspiration (PET), elevation range (TOPO), soil fertility

(SOIL) and river density (RIVER) on the scale range of small-scale over-dispersion (a),

occurrence of large-scale under-dispersion (b), and scale range of large-scale under-dispersion

(c). Standardized coefficients with standard errors from the multiple regression models are

shown (***P<0.001, **P<0.01, *P<0.05, •P<0.1).

Figure 4. Synthesis of the direct and indirect effects of energy availability (measured by

potential evapotranspiration) and topographic relief as strong predictors on settlement spatial

pattern features characterised by (a) small-scale over-dispersion as a measure of territory size,

and (b) large-scale under-dispersion as a sign of territorial clustering. Blue: positive effects,

red: negative effects, solid: direct effects, dashed: indirect effects.

20

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597