GIS-Based Visibility Network and Defensibility Model to Reconstruct Defensive System of the Han Dynasty in Central Xinjiang, China

International Journal of Geo-Information

Article GIS-Based Visibility Network and Defensibility Model to Reconstruct Defensive System of the Han Dynasty in Central Xinjiang, China

Jianfeng Zhu 1,2 ID , Yueping Nie 1,2, Huaguang Gao 2,3, Fang Liu 1,2 and Lijun Yu 1,2,*

1 Institute of Remote Sensing and Digital Earth (RADI), Chinese Academy of Science (CAS), Beijing 100101, China; [email protected] (J.Z.); [email protected] (Y.N.); [email protected] (F.L.) 2 Joint Laboratory of Remote Sensing Archaeology, Beijing 100101, China; [email protected] 3 National Museum of China, Beijing 100006, China * Correspondence: [email protected]; Tel.: +86-10-6485-8122

Received: 25 June 2017; Accepted: 10 August 2017; Published: 13 August 2017

Abstract: The Silk Road opened during the Han Dynasty, and is significant in global cultural communication. Along this route in the central part of Xinjiang, the archaeological sites with defensive characteristics once provided a safeguard for this area. Reconstructing the defensive system is an important way to explore the ancient culture’s propagation and the organizational structure of these sites. In this research, the compound visibility network with complex network analysis (CNA) and the least-cost paths based on the defensibility models from linear and logistic regression methods constitute the principle defensive structure. As possible transportation corridors, these paths are considered to be mostly fitted to each other in general, and are different from normal slope-based paths. The sites Kuhne Shahr and Agra play important roles for information control according to the CNA measures, while the sites Kuhne Shahr and Kuyux Shahr are considered to be crucial cities due to their positions and structural shapes. Some other sites, including Uzgen Bulak, Shah Kalandar, Chuck Castle, Caladar, and Qiuci, as well as some beacons, have important effects on defending the transportation corridors. This method is proven efficient for the study of the historical role of archaeological sites with defensive characteristics.

Keywords: defense system; visibility network; defensibility model; regression analysis; least-cost path; transportation corridor

1. Introduction The Silk Road has been an influential cultural path for communication in economy, politics, civilization, and many other aspects since it was opened during the Han Dynasty, after Zhang Qian was sent on a diplomatic mission to the Western Regions (west of Yumenguan, including today’s Xinjiang and parts of Central Asia) in 138 B.C. To defend against invasion and maintain political governance, a battery of military fortifications, including the Western Region’s Frontier Command, were established at that time. They are considered to be parts of the corporate defensive system to protect the Silk Road [1,2]. Among the discovered relics of ancient fortifications, the city sites were mostly used for expending cultivated land and the population, while the beacons were for safety surveillance and signal transmission [3–5]. Their location preference is related to the intrinsic characteristics for military defense. Therefore, the reconstruction of the defensive system in the Han Dynasty is a critical means to explaining the ancient organizational structure of the remaining archaeological sites and their cultural connotation along the Silk Road. The distribution of ancient sites is the result of human preference as well as complex decision-making processes. From a military perspective, the ancient sites observed some spatial

ISPRS Int. J. Geo-Inf. 2017, 6, 247; doi:10.3390/ijgi6080247 www.mdpi.com/journal/ijgi ISPRS Int. J. Geo-Inf. 2017, 6, 247 2 of 19 ISPRS Int. J. Geo-Inf. 2017, 6, 247 2 of 19 constitute a well-organized system. As an important social organization form, the visibility network canregularities indicate defensive to constitute structure a well-organized among sites system. [6,7]. Through As an important visibility socialanalysis, organization some archaeological form, the characteristicsvisibility network and can defensive indicate roles defensive can be structure explored among [8–10]. sites However, [6,7]. Through it cannot visibility explain analysis, the somespatial defensibilityarchaeological in the characteristics landscape with and defensivespecial preference roles can beto build explored defensive [8–10]. sites. However, While it the cannot defensibility explain modelthe spatial itself defensibilityis an effective in way the landscapeto solve this with problem special (e.g., preference some defensive to build defensive indexes are sites. used While to build the simpledefensibility models model with itselfsome is possible an effective influencing way to solve factors this problem[11,12]), (e.g.,however, some with defensive the drawback indexes are that used the methodsto build simpleof choosing models variables with some and possible defining influencing their coef factorsficients [11 in,12 the]), however,equation withare strongly the drawback subjective. that Athe more methods precise of model choosing of variablesspatial defensibility and defining is their necessary. coefficients The inspatial the equation distribution are strongly characteristics subjective. for buildingA more precisean archaeological model of spatial prediction defensibility model iscould necessary. provide The references, spatial distribution but most characteristicsof them put more for emphasisbuilding on an archaeologicalthe geographical prediction or environmental model could aspe providects (slope, references, aspect, butdistance most to of water, them put elevation, more etc.)emphasis in the on location the geographical analysis or under environmental environmen aspectstal (slope,determinism, aspect, distanceand some to water, spatial elevation, relations etc.) of archaeologicalin the location analysissites are under successfully environmental discovered determinism, [13–16]. and Other some intrinsic spatial relationsfactors, ofsuch archaeological as political, economic,sites are successfully or military discoveredconsiderations, [13–16 also]. Other contribute intrinsic to factors, the location such asdecision, political, but economic, they are or difficult military to quantifyconsiderations, [17–21]. also The contribute studies are to therare location on buildi decision,ng a defensibility but they are difficultmodel towith quantify quantified [17–21 military]. The variables.studies are As rare a matter on building of fact, a defensibilitythe comprehensive model with analysis quantified of a visibility military variables.network and As aa matterprecisely- of qualifiedfact, the comprehensivedefensibility model analysis could of ahelp visibility to explai networkn the ancient and a precisely-qualified defensive considerations defensibility [22–24]. model couldThis help study to explain reconstructed the ancient a defensivedefensive considerations system with [ 22a –24compounded]. visibility network and defensibilityThis study models reconstructed to speculate a the defensive cultural system linkage with behind a compounded military purpose. visibility The networktransportation and corridorsdefensibility based models on least-cost to speculate analysis the of cultural defensibility linkage models behind were military calculated purpose. to simulate The transportation military and culturalcorridors routes based during on least-cost the Han analysis Dynasty of defensibility along the Silk models Road. were The calculated organizational to simulate structure military and historicaland cultural roles routes of the during archaeological the Han Dynastysites were along explored, the Silk focusing Road. The on organizationalthe observation structure network and and historical roles of the archaeological sites were explored, focusing on the observation network and along the transportation routes. along the transportation routes.

2.2. Study Study Area Area and and Historical Historical Background Background TheThe study study area area is is located located in in the southern foothillsfoothills ofof thethe TianshanTianshan Mountains Mountains and and the the northern northern 2 fringefringe of of the the Tarim Tarim basin basin with with 135,744 kmkm2 of totaltotal area,area, combiningcombiningwith with ten ten counties counties from from Eastern Eastern YuliYuli to to Western Western Bytown Bytown and and Xinhe Xinhe (see (see Figure Figure 1),1), of of which which the the latitude latitude is is from from 39.7° 39.7 N◦ Nto to43.1° 43.1 N◦ andN longitudeand longitude from 84.4° from E 84.4 to ◦90.4°E to E. 90.4 This◦ E. area This has area an hasarid an climate arid climate but rich but water rich resources. water resources. It contains It threecontains main three types main of landforms: types of landforms: the mountainous the mountainous region, the region, desert the area, desert and area, the and piedmont the piedmont alluvial plain.alluvial Luntai plain. County, Luntai County, situated situated in the in center, the center, is documented is documented as asthe the location location of of WesternWestern Region’s Region’s FrontierFrontier Command. Command. Ninety-four Ninety-four archaeological archaeological sites inin thethe HanHan DynastyDynasty are are located located there, there, including including twenty-fourtwenty-four city city sites, sites, twenty twenty beacon towers,towers, thirty-ﬁve thirty-five tombs, tombs, and and ﬁfteen fifteen general general relics relics that that contain contain a a stonestone inscription,inscription, some some Buddhist Buddhist temples, temples, irrigation irrigation ditches, ditches, smelting smelting relics, relics, and and so on.so on.

FigureFigure 1. 1. SiteSite distribution distribution of Han DynastyDynasty inin thethe study study area. area.

ISPRS Int. J. Geo-Inf. 2017, 6, 247 3 of 19

This area has a famous history and played a crucial role in cultural communication between China and Central Asia, as well as the European continent, after the Silk Road opened. Since 133 B.C., the Han Empire had taken several large-scale military actions to ﬁght against the North Huns. In 101 B.C. (the fourth year of Taichu reign by Emperor Wudi in Han Dynasty), Quli, in today’s Yuli County [25,26], and Luntai were chosen to set up the ﬁrst military fortress to begin populating wasteland and garrisoning the border region [1]. In 60 B.C. (the second year of the Shenjue reign period of Emperor Xuandi of the Han Dynasty), the Western Region’s Frontier Command was established to guard the north route (after 1 D.C. of the Emperor Pingdi period, it was called the central path due to a new passageway constructed further north [1]) of the Silk Road [27,28]. Until the Later Han Dynasty, the governed territory changed repeatedly and the Silk Road was closed three different times. The garrison command once moved to Qiuci City, in the current Kuqa County [29]. The geospatial distribution of existing relics can reveal the history during this period.

3. Methodology and Results

3.1. Data Collection DEM (Digital Elevation Model) imagery is necessary for a spatial analysis of archaeological sites. Using a lower spatial resolution of 90 m, SRTM (Shuttle Radar Topography Mission) DEM can provide acceptable average accuracy for this study, in contrast to ASTER GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model) which has a higher resolution of 30 m but whose elevation values contain severe systematic errors, producing faulty results for visibility analysis in some areas [30]. The initial data of the archaeological sites are from the third cultural relic census, which are recorded with coordinates and attributive descriptions by field survey. For the purposes of this study, the sequence numbers of some important archaeological sites are defined as in Table A1.

3.2. Visibility Analysis Different types of sites that command the ﬁeld of view in special directions or which are visible from much of the surrounding territory have been an important part of an ancient defensive system and are of particular note suggested in the context of military or cultural monuments [31]. In order to survey the inter-relation of the defensive system in this study area of the Han Dynasty, point-to-point visibility analysis was employed in ArcGIS 10.2. The recognition of the defensive system was ﬁrstly constructed through a compound visibility network combined with visibility lines across the DEM grid surface. It focused especially on the respective monitoring patterns of city sites and beacon towers, which constitute the organizational structure and control over the surrounding territory in different ways. The preconditions to constructing visibility lines observe the following rules:

• The city sites and beacon towers had obvious military function when being built. The tombs with almost no defensive role were ignored during visibility analysis in this research. The general relics—as the related part of city sites—were viewed of no observing function, but observed status in the defensive system. • According to the documentation and description of the remaining height, the average offset value of the city sites and general relics was assumed as 8 m, and the beacon tower as 12 m. • The beacon fire—as the signal for military watching and information transmitting—was defined as another 8 m in height. Therefore, the target offset value of an observed beacon is the sum of the tower and its fire height when calculating the visibility line from any site (see Table1). ISPRS Int. J. Geo-Inf. 2017, 6, 247 4 of 19

Table 1. Observing offset in the visibility analysis. B: Beacon tower; C: City site; R: General relic.

Observation Direction B to B B to C B to R C to C C to B Offset A 12 12 12 8 8 Offset B 20 8 8 8 20 ISPRS Int. J. Geo-Inf. 2017, 6, 247 4 of 19

The arid environment in this study area providesprovides good observation conditions between sites, excluding meteorological limitations. Nevertheless, Nevertheless, the the influe inﬂuencence of of the the Earth’s curvature and the border effecteffect shouldshould be be considered considered in in the the visibility visibilit analysis.y analysis. The The global global visibility visibility network network (Figure (Figure2) was 2) combinedwas combined with with an individual an individual visibility visibility line line connecting connecting C23 C23 to to C24 C24 and and three three divided divided local local networks.networks.

Figure 2. Visibility network for military defense in the Han Dynasty. It is combined with lines of sight and defense related sites, including city sites, beacon towerstowers andand generalgeneral sites.sites.

Complex network analysis (CNA) is a mathematical form of graph theory used to characterize Complex network analysis (CNA) is a mathematical form of graph theory used to characterize network structure and find the different importances of nodes. As a formalized methodology, it network structure and ﬁnd the different importances of nodes. As a formalized methodology, it delineates the social character of the visibility network among the archaeological sites [32–34], delineates the social character of the visibility network among the archaeological sites [32–34], including including how defensibility is organized through the nodes (archaeological sites) and links (visibility how defensibility is organized through the nodes (archaeological sites) and links (visibility lines) in lines) in the local region, and how important the sites are by the different measures in Table 2. the local region, and how important the sites are by the different measures in Table2. Table 2. The complex network analysis (CNA) measures in the visibility network [35–37]. Table 2. The complex network analysis (CNA) measures in the visibility network [35–37]. Density a Inclusivity a Degree b Betweenness b Closeness b a a b b b Density2⋅m Inclusivityv Degree Betweennessσ (i) Closeness1 j st 2·m  σ σst(i) 1d j (i) n⋅(n−1) v s≠i≠t∈∑N  ( ) n·(n−1) nn j stσst j≠∑i∈N dj i s6=i6=t∈N j6=i∈N mm:: number number of of vv:: number number of of j:j :number number of of σst(i):σ stnumber(i): number of shortest of shortest paths dj(dij):(i ):hops hops of ofthe visibilityvisibility lines; lines; visiblevisible nodes; nodes; connectingconnecting frompaths node from s to nodet via nodes to t ivia; σst: shortestthe shortest paths nn:: number number of of nn:: number number of of lineslines at at the the numbernode iof; σ thest: numbershortest of paths the frompaths the from node the i allall nodes. nodes. allall nodes. nodes. nodenode i.i . shortestfrom paths node from s to node t. s to t. nodeto inode to node j. j. a b a GlobalGlobal measures; measures; b Node-scaleNode-scale measures. measures. • • The density is deﬁneddefined as the ratio of actual connect connectionsions of visibility lines to potential connections, implicating integral redundancy oror multi-path transmissiontransmission of the information. The value closer to one represents moremore redundancyredundancy thanthan thatthat approachingapproaching zero.zero. • Inclusivity is another measurement of network density. It is defined as the proportion of visible nodes in the whole network. The value approximating one means a less isolated and more organized network. • The degree is introduced to quantify the number of nodes connected by visibility lines within one hop at the given node, referring to the local connectivity and the centrality of a node in the network. • Betweenness that scales with the number of pairs of nodes as implied by the summation indices

ISPRS Int. J. Geo-Inf. 2017, 6, 247 5 of 19

• Inclusivity is another measurement of network density. It is deﬁned as the proportion of visible nodes in the whole network. The value approximating one means a less isolated and more organized network. • The degree is introduced to quantify the number of nodes connected by visibility lines within one hop at the given node, referring to the local connectivity and the centrality of a node in the network. • Betweenness that scales with the number of pairs of nodes as implied by the summation indices is used to describe the degree of governmental control over the information exchange at the given node. It is suitable for detecting critical nodes and crossroads in the network. • Closeness is a length measure which is calculated as the sum of the shortest paths from a given node to all the others in the network. Thus, a small value of a given node means it has a central position to all. ISPRS Int. J. Geo-Inf. 2017, 6, 247 5 of 19 The global CNA measures in Table3 show a higher density and inclusivity of Region 2 and Regionis used 3. In addition,to describe each the site degree is described of governmental with three control indicators over in the both information scatter and exchange bubble forms at the in Figuregiven3. node. It is suitable for detecting critical nodes and crossroads in the network. • Closeness is a length measure which is calculated as the sum of the shortest paths from a given Table 3. CNA measures of global visibility network. node to all the others in the network. Thus, a small value of a given node means it has a central

position to all. Region 1 Region 2 Region 3 The global CNA measuresDensity in Table 3 show 0.003 a higher density 0.014 and inclusivity 0.013 of Region 2 and Region 3. In addition, each site is describedInclusivity with three 0.070 indicators in 0.228both scatter and 0.281 bubble forms in Figure 3.

(a) (b)

Figure 3. (a) Scatter plot and (b) bubble plot of node-scale CNA measures (degree, betweenness, and Figure 3. (a) Scatter plot and (b) bubble plot of node-scale CNA measures (degree, betweenness, closeness). and closeness).

Table 3. CNA measures of global visibility network. 3.3. Defensibility Models Region 1 Region 2 Region 3 3.3.1. Interpretation of PossibleDensity Variables 0.003 0.014 0.013 Inclusivity 0.070 0.228 0.281 Quantiﬁed defensibility could not only explain the relative importance of every site and the landscape, but also reveal the cultural communication pattern between different regions in sociology 3.3. Defensibility Models and anthropology and its inﬂuencing factors [21,34]. The independent variables selection is crucial when building a defensibility model. The geographical factors are mostly utilized for spatial analysis 3.3.1. Interpretation of Possible Variables and location prediction in archaeological models [38], and they are also considered as valid variables in thisQuantified defensibility defensibility model, in could association not only with explain other defense-related the relative importance factors [39 ,of40 ].every The defensibilitysite and the landscape,was delineated but also with reveal ten independent the cultural communicati variables byon GIS pattern tools: proximitybetween different to road regions network, in proximitysociology andto water, anthropology cumulative and viewshed its influencing from citiesfactors and [21,34 from]. The beacons, independent density ofvariables visibility selection lines, density is crucial of when building a defensibility model. The geographical factors are mostly utilized for spatial analysis and location prediction in archaeological models [38], and they are also considered as valid variables in this defensibility model, in association with other defense-related factors [39,40]. The defensibility was delineated with ten independent variables by GIS tools: proximity to road network, proximity to water, cumulative viewshed from cities and from beacons, density of visibility lines, density of sites, distance to city, distance to beacon, as well as the geographical factors of slope and elevation. In particular, the first two variables mentioned are inversely proportional to defensive requirements. All variables were normalized before entering a model to remove the impact of different units when judging the importance by the coefficient value of a variable. The defensibility model was built with linear regression and logistic regression analysis, respectively.

ISPRS Int. J. Geo-Inf. 2017, 6, 247 6 of 19 sites, distance to city, distance to beacon, as well as the geographical factors of slope and elevation. In particular, the ﬁrst two variables mentioned are inversely proportional to defensive requirements. All variables were normalized before entering a model to remove the impact of different units when judging the importance by the coefﬁcient value of a variable. The defensibility model was built with linear regression and logistic regression analysis, respectively.

3.3.2. Linear Regression Analysis Linear regression analysis can describe most of the issues with causalities, in practice [41]. As for the dependent variable, the relative importance (RI) of a landscape is not a directly observable value, but can be described with weighted and classified indexes. The definition is expressed in the following formula: RI = Pi · Ai (1) where Ai is the area of an archaeological site, and Pi is the importance of an ancient site according to ISPRSthe protection Int. J. Geo-Inf. level, 2017, as 6, 247 defined by the national cultural relics department (national level: 4; provincial6 of 19 level: 3; municipal or prefectural level: 2; unidentified level: 1). The RI grade of these ninety-four value, but can be described with weighted and classified indexes. The definition is expressed in the archaeological sites was classified into four ranges according to the defensive contribution of each following formula: archaeological type as in Table4, and defined as four corresponding grade values as in Figure4. The = ⋅ relative importance of a site was used to describeRI itsPi surroundingAi landscape in this linear model.(1) The higher the importance grade of a site, the more defensibility is required in the landscape. where Ai is the area of an archaeological site, and Pi is the importance of an ancient site according to the protectionTable 4. Classification level, as defined of relative by the importance national in cultural the defensive relics system. department The importance (national is level: ranked 4; inprovincial four level:ranges 3; municipal by the sequence or prefectural of city, beacon, level: general 2; unidentified relics, and tomblevel: according 1). The toRI the grade defensive of these contribution, ninety-four archaeologicaland then each sites archaeological was classified type into is classified four ranges into four according grades. to the defensive contribution of each archaeological type as in Table 4, and defined as four corresponding grade values as in Figure 4. The relative importance of a site was used toCity describe Beacon its surrounding General landscape Relic Tombin this linear model. The higher the importanceClassified grade gradeof a site, the 7–10 more defensibility 5–8 is required 3–6 in the 1–4 landscape.

Figure 4. DefinitionDeﬁnition of relative importance grade ba basedsed on defensive function and site area.

TableA backward 4. Classification regression of methodrelative importance was used inin linearthe defensive model building.system. The All importance the independent is ranked variables in werefour entered ranges into by the theinitial sequence model, of city, and beacon, then the general one withrelics, a minimumand tomb ac correlationcording to tothe the defensive dependent variablescontribution, was deleted and then first. each The archaeological remaining variablestype is classified repeated into thisfour stepgrades. in the updating model until no independent variables could be eliminated.City Beacon Four variables General were RelicTomb chosen in the final model, where three had a positive effectClassified on thegrade defensibility 7–10 and5–8 only the slope3–6 variable1–4 had negative impact (in Table5). The statistical results are shown in Table6. An adjusted R2 value of 0.427 represents about 42.7%A ofbackward defensive regression importance method explainable was inused this in model. linear The model Koenker building. (BP) andAll robustthe independent probability (Robust_Pr)variables were were entered statistically into the significant, initial model, meaning and that then all the the one coefficients with a minimum were significant correlation to represent to the dependentan actual relationship variables was between deleted each first. explanatory The remaining variable variables and therepeated dependent this step variable. in the Theupdating Wald model until no independent variables could be eliminated. Four variables were chosen in the final model, where three had a positive effect on the defensibility and only the slope variable had negative impact (in Table 5). The statistical results are shown in Table 6. An adjusted R2 value of 0.427 represents about 42.7% of defensive importance explainable in this model. The Koenker (BP) and robust probability (Robust_Pr) were statistically significant, meaning that all the coefficients were significant to represent an actual relationship between each explanatory variable and the dependent variable. The Wald statistic significance means that this overall model is statistically acceptable. If the variance inflation factors (VIF) are small in the result, it indicates less redundancy among the explanatory variables (when VIF < 7.5). With the Jarque-Bera statistic rejecting the significance test, predictions were unbiased and the residuals were normally distributed. All the statistical results proved that the performance in this model was robust and significant. The final defensive importance in the landscape is normalized in Figure 5.

ISPRS Int. J. Geo-Inf. 2017, 6, 247 7 of 19 statistic significance means that this overall model is statistically acceptable. If the variance inflation factors (VIF) are small in the result, it indicates less redundancy among the explanatory variables (when VIF < 7.5). With the Jarque-Bera statistic rejecting the significance test, predictions were unbiased and the residuals were normally distributed. All the statistical results proved that the performance in this model was robust and significant. The final defensive importance in the landscape is normalized in FigureISPRS Int.5. J. Geo-Inf. 2017, 6, 247 7 of 19

TableTable 5.5. IndependentIndependent variablesvariables statisticalstatistical results.results. Robust_Pr:Robust_Pr: robust probability;probability; VIF:VIF: variancevariance inﬂationinflation factor.factor. Variables Intercept ProxtoRoad DistoCity Slope Line_Den Variables Intercept ProxtoRoad DistoCity Slope Line_Den Coefficient −16.928 18.183 5.022 −7.745 3.221 CoefﬁcientProbability − 0.00216.928 * 18.183 0.001 * 5.0220.000 * 0.046−7.745 * 0.020 3.221 * ProbabilityRobust_Pr 0.0020.001 * 0.001 0.001 ** 0.000 0.000 * * 0.000 0.046 * * 0.010 0.020 * * Robust_PrVIF 0.001 -- * 0.001 1.119 * 0.000 1.175 * 0.0001.059 *1.332 0.010 * VIF – 1.119 1.175 1.059 1.332

* An* An asterisk asterisk indicates indicates a a statistically signiﬁcant significantp-value p-value (p < ( 0.05).p < 0.05). Table 6. Overall model statistic results. Table 6. Overall model statistic results.

NumberNumbe ofr of Observations Observations 94 (AICc) (AICc) 416.186 416.186 MultipleMultiple R-Squared R-Squared 0.452 Adjusted Adjusted R-Squared R-Squared 0.427 0.427 JointJoint F-Statistic F-Statistic 0.000 0.000 * KoenKoenkerker (BP) (BP) Statistic Statistic 0.0110.011 * * JointJoint Wald Wald Statistic Statistic 0.000 0.000 * Jarque-Bera Jarque-Bera Statistic Statistic 0.755 0.755 * An asterisk* An asterisk indicates indicates a statistically a statistically significant signiﬁcant p-value ( p(p< < 0.05). 0.05). AICc: AICc: corrected corrected Akaike Akai informationke information criterion. criterion.

Figure 5. Linear regression of defensive importanceimportance inin thethe HanHan Dynasty.Dynasty.

3.3.3.3.3.3. Logistic RegressionRegression AnalysisAnalysis TheThe logisticlogistic regressionregression analysisanalysis waswas mostlymostly usedused toto delineatedelineate thethe probabilityprobability ofof predictingpredicting thethe locationlocation ofof an an archaeological archaeological site site by by a binarya binary dependent dependent variable variable (0: nonoccurrence;(0: nonoccurrence; 1: occurrence) 1: occurrence) and independentand independent variables variables that affect that event affect occurrence event occu [42rrence–44]. In [42–44]. this research, In this the research, incident occurrencethe incident of occurrence of an archaeological site being located in a landscape (marked as {y = 1}) and independent an archaeological site being located in a landscape (marked as {y = 1}) and independent variables Xj relatedvariables to X defensivej related to factors defensive are expressed factors are as expressed below: as below: 1 P (y = 1) = , (2) 1-zk P(y = 1) =1+ e , (2) −zk = β + β 1⋅+ e + ε zk 0  j x jk k , (3) j zk = β0 + ∑ βj·xjk + εk, (3) where variable β0 and regression coefficients βj reflectj the impact of independent variables Xj on an incident of P(y = 1). Furthermore, the presupposition of building a logistic model is that the where variable β and regression coefﬁcients β reﬂect the impact of independent variables X on relationship between0 independent and dependenj t variables is non-linear, whereas the unionj an incident of P(y = 1). Furthermore, the presupposition of building a logistic model is that the exponent Zk is a linear combination in the hypothetical function. The observation groups are set as that Group 1 with ninety-four archaeological sites and Group 0 with ninety-four non-site random points. Forward stepwise regression (Likelihood Ratio) is a classic variable selection method that stipulates the criteria of a variable entering into the model that its statistical probability value of likelihood ratio should be less than the given significant level in partial quasi-maximum likelihood estimation. For the overall model, the Cox & Snell R2 test value was 0.410 (>0.4 best), and Nagelkerke R2 test value was 0.547 (>0.5 best). Another important indicator—the Hosmer–Lemeshow estimation (Chi-square = 17.073, sig. = 0.000)—showed less bias between

ISPRS Int. J. Geo-Inf. 2017, 6, 247 8 of 19 relationship between independent and dependent variables is non-linear, whereas the union exponent Zk is a linear combination in the hypothetical function. The observation groups are set as that Group 1 with ninety-four archaeological sites and Group 0 with ninety-four non-site random points. Forward stepwise regression (Likelihood Ratio) is a classic variable selection method that stipulates the criteria of a variable entering into the model that its statistical probability value of likelihood ratio should be less than the given significant level in partial quasi-maximum likelihood estimation. For the overall model, the Cox & Snell R2 test value was 0.410 (>0.4 best), and Nagelkerke R2 test value was 0.547 (>0.5 best). Another important indicator—theISPRS Int. J. Geo-Inf. Hosmer–Lemeshow 2017, 6, 247 estimation (Chi-square = 17.073, sig. = 0.000)—showed less8 biasof 19 between estimation of incident {y = 1} and practical observations. The event occurrence is viewed withestimation more possibilityof incident when{y = 1} P(y) and is practical greater thanobserv 0.5.ations. As a result,The event the classificationoccurrence is accuracyviewed with in Group more 0possibility was 84%, when and in P(y) Group is greater 1 it was than 87.2%. 0.5. As The a result, general the accuracy classification of the accuracy prediction in Group was 85.6%. 0 was Three 84%, variablesand in Group contributed 1 it was to the87.2%. final The logistic general model accuracy and were of significant the prediction in Wald was statistics 85.6%. as Three in Table variables7. The probabilitycontributed distribution to the final of logistic archaeological model and sites were which significant are related in to Wald defensive statistics factors as inin theTable landscape 7. The isprobability shown in distribution Figure6. of archaeological sites which are related to defensive factors in the landscape is shown in Figure 6. Table 7. Variables in the equation. Table 7. Variables in the equation. 95% C.I. for EXP(B) B S.E Wals df Sig. Exp (B) 95% C.I. for EXP(B) B S.E Wals df Sig. Exp (B) LowerLower Uppe Upperr Site_DenSite_Den 11.054 11.054 2.1452.145 26.54626.546 11 0. 0.000000 63,205.608 63,205.608 943.035 943.035 4,236,267.459 4,236,267.459 ProxtoRoadProxtoRoad 12.90112.901 3.8083.808 11.47811.478 1 0.001 0.001 400,730.185400,730.185 229.897229.897 698,507,800.021698,507,800.021 DistoCityDistoCity −2.552−2.552 0.8570.857 8.873 8.873 1 1 0.003 0.003 0.078 0.078 0.015 0.015 0.418 0.418 ConstantConstant −11.464−11.464 3.5653.565 10.343 10.343 1 1 0.001 0.001 0.000 0.000 -- – -- –

Figure 6. Logistic regression of defensive related sites in the Han Dynasty.

Above all, the effective defensive factors that contribute to the defensibility models are Above all, the effective defensive factors that contribute to the defensibility models are reasonable reasonable interpretations of a defensive structure in a landscape. The factor of proximity to road interpretations of a defensive structure in a landscape. The factor of proximity to road networks had a networks had a vital influence on defensibility in both models. Given the z-score of 4.69 in Moran’s I vital inﬂuence on defensibility in both models. Given the z-score of 4.69 in Moran’s I auto-correlation auto-correlation analysis of the distance to road network from a site, there is less than 1% likelihood analysis of the distance to road network from a site, there is less than 1% likelihood that this clustered that this clustered pattern could be the result of random chance. This means that the distribution of pattern could be the result of random chance. This means that the distribution of sites is related to the sites is related to the road system, which coincides with the ancient roads to a certain degree, road system, which coincides with the ancient roads to a certain degree, representing the reciprocity of representing the reciprocity of culture and economy in different areas. Another factor that is culture and economy in different areas. Another factor that is important in each model is the distance important in each model is the distance to the city sites, due to their function of information to the city sites, due to their function of information transmission and territory defense. Moreover, transmission and territory defense. Moreover, slope is another factor with inverse relation to slope is another factor with inverse relation to defensibility, because steep terrain or a mountainous defensibility, because steep terrain or a mountainous area is difficult for normal marching, thus area is difﬁcult for normal marching, thus requiring less defending. However, it is not always like requiring less defending. However, it is not always like that. Some sites like C2, C21 and B18 are still that. Some sites like C2, C21 and B18 are still located in that region in case of emergency and provide located in that region in case of emergency and provide deterrent power in the surrounding deterrent power in the surrounding landscape. landscape.

3.4. Least-Cost Paths Terrain surface has an influence on anthropogenic preference and security strategy for defense, although sometimes the landscape has changed over such a long period [45]. The least-cost path analysis, based on weighted terrain surface, is considered to be a valid method to explore the ancient transportation paths and find an optimal route with minimum cumulative cost from a certain site to another, in practice [39,46]. Particularly, the weighted topography in least-cost analysis has a great impact on the result [47], such that the optimal path may not always follow the Euclidean direction (straight line is the shortest way). The traveling costs are mostly influenced by slope or river networks in terrain surface [48]. Nevertheless, some different cost simulations were verified effectively for rebuilding ancient routes, such as flow direction models and cumulative focal mobility networks [49,50].

ISPRS Int. J. Geo-Inf. 2017, 6, 247 9 of 19

3.4. Least-Cost Paths Terrain surface has an influence on anthropogenic preference and security strategy for defense, although sometimes the landscape has changed over such a long period [45]. The least-cost path analysis, based on weighted terrain surface, is considered to be a valid method to explore the ancient transportation paths and find an optimal route with minimum cumulative cost from a certain site to another, in practice [39,46]. Particularly, the weighted topography in least-cost analysis has a great impact on the result [47], such that the optimal path may not always follow the Euclidean direction (straight line is the shortest way). The traveling costs are mostly influenced by slope or riverISPRS Int. networks J. Geo-Inf. in 2017 terrain, 6, 247 surface [48]. Nevertheless, some different cost simulations were verified9 of 19 effectively for rebuilding ancient routes, such as flow direction models and cumulative focal mobility The least-cost path analysis is useful to explore cultural communication [51,52]. In this study, the networks [49,50]. The least-cost path analysis is useful to explore cultural communication [51,52]. In least-cost paths—based on an unusual cost surface weighted with defensive factors in the linear and this study, the least-cost paths—based on an unusual cost surface weighted with defensive factors in logistic defensibility models—were used to reconstruct the transportation corridor and are compared the linear and logistic defensibility models—were used to reconstruct the transportation corridor and with the common least slope cost paths in Figure 7. are compared with the common least slope cost paths in Figure7.

(a) from C1 to C22

(b) from C4 to C22

Figure 7. Least cost paths based on: linear model, logistic model, and slope model in the direction of (a) from C1 to C22 and the direction ofof ((b)) fromfrom C4C4 toto C22C22 (P1,(P1, P5: the locations of largest cumulative deviation between the paths of logistic modelmodel andand slopeslope model;model; P2,P2, P4:P4: the locations of largest cumulative deviation between the paths of linear model and slope model;model; P3, P6: the locations of largest cumulative deviation betweenbetween thethe pathspaths ofof linearlinear modelmodel andand logisticlogistic model).model).

The least-cost path from linear regression analysis is based on the definition of relative importance in defensibility and from the logistic regression analysis is based on probability distribution with optimum defensive conditions. They are different concepts. Yet, the defensibility in the landscape is in direct proportion to the possibility of ancient sites to pass through due to their defensive character. As a result, the least-cost paths were calculated from the linear model that focuses on the differing importance among all archaeological sites, and the logistic model that views all sites as the same occurrence probability. The starting points were chosen as C1 and C4 in different path directions, while C22 is viewed as the destination due to its highest relative importance grade in the west (see Figure 4). This moving from east to west represents the Han Culture propagating to the Western Region. Of course, the inverse traveling direction will not affect the final results.

ISPRS Int. J. Geo-Inf. 2017, 6, 247 10 of 19

The least-cost path from linear regression analysis is based on the deﬁnition of relative importance in defensibility and from the logistic regression analysis is based on probability distribution with optimum defensive conditions. They are different concepts. Yet, the defensibility in the landscape is in direct proportion to the possibility of ancient sites to pass through due to their defensive character. As a result, the least-cost paths were calculated from the linear model that focuses on the differing importance among all archaeological sites, and the logistic model that views all sites as the same occurrence probability. The starting points were chosen as C1 and C4 in different path directions, while C22 is viewed as the destination due to its highest relative importance grade in the west (see Figure4).

ThisISPRS moving Int. J. Geo-Inf. from 2017 east, 6, to 247 west represents the Han Culture propagating to the Western Region.10 of 19 Of course, the inverse traveling direction will not affect the ﬁnal results. 4. Discussions 4. Discussions 4.1. Defensive Structure 4.1. Defensive Structure The defensive system of the Han Dynasty is considered as combining the visibility network, transportationThe defensive paths, system and special of the Hansites with Dynasty defensiv is considerede functions. as The combining statistics in the Figure visibility 8 show network, that transportationthe average length paths, of and sight special lines sitesbetween with beacons defensive in the functions. visibilityThe network statistics is 15.2 in km, Figure similar8 show to the that theaverage average length length of of 14.4 sight km lines between between city beaconssites. However, in the visibility the maximum network visi isbility 15.2 km,distance similar varies. to the averageBeacon length towers of 14.4have km stronger between forewarning city sites. However, and wa thetching maximum ability.visibility Besides, distancethey still varies. have Beacon the towersadvantages have stronger of low-cost forewarning building and and high-efficiency watching ability. signal Besides, transmission. they still As in have Figure the 9, advantages the strong of low-costmonitoring building ability and of high-efﬁciency each beacon tower signal is also transmission. shown through As in high Figure cumulative9, the strong frequency, monitoring while ability the of eachcity site beacon focuses tower on integral is also shownassociated through defensibilit high cumulativey, contributing frequency, to the surv whileeillance the city due site to the focuses lower on integralcumulative associated frequency defensibility, of each site contributing and larger to cumulative the surveillance area observed due to the from lower the whole cumulative city group. frequency In of eachaddition, site andthe signal larger transmission cumulative between area observed beacon fromand city the is whole also an city important group. defensive In addition, precaution. the signal transmissionIt connects betweenthe sites beaconwith longer and citylinkages is also for an fast important signal delivery defensive and precaution. constitutes Ita connectsmultifunctional the sites withmonitoring longer linkages system. for fast signal delivery and constitutes a multifunctional monitoring system.

(No.) B-B B-B 35 B-B B-B B-B B-B Length B-B 30 B-B B: Beacon tower B-B B-B C: City site B-B B-B 25 B-C B-C B-C B-C B-C 20 B-C B-C B-C B-C B-C 15 B-C C-C C-C C-C C-C 10 C-C C-C C-C C-C C-C 5 C-C C-C C-C C-C C-C (km) 0 0 1020304050

FigureFigure 8. 8.Length Length of of visibility visibility lines. lines. TheThe observingobserving site sitess are classified classiﬁed into into three three observing observing types. types.

The CNA analysis method describes the speciﬁc relationship of every single site in the local 1.0 visibility network. In Region 1 of Figure3, sites B16, B17, and C20 all have large closeness of CNA measure in the small local visibility0.9 network. They also share almost the same high central importance. In other words, the equal importance of the sites1.0 means that they are all less important in this simple network. By the analysis of0.8 density and inclusivity in Table3, Region 2 and Region 3 both have a larger redundancy than Region 1, meaning a better social0.8 organization network and important defensive regions. This structure has0.7 the advantage of stable information exchange to maintain the governance Beacon and integral defense. The special site C16 in the local visibility networkCity of Region 2, with the three

Cumulative Area 0.6 0.6

highest indexes, is explicitly Frequency Cumulative the central site and in the dominant position. It has a strong capacity

0 5 10 15 20 0.5 Observing Points

0 5 10 15 20 Observing Points

Figure 9. Cumulative viewshed of city sites and beacon tower.

The CNA analysis method describes the specific relationship of every single site in the local visibility network. In Region 1 of Figure 3, sites B16, B17, and C20 all have large closeness of CNA

ISPRS Int. J. Geo-Inf. 2017, 6, 247 10 of 19

4. Discussions

4.1. Defensive Structure The defensive system of the Han Dynasty is considered as combining the visibility network, transportation paths, and special sites with defensive functions. The statistics in Figure 8 show that the average length of sight lines between beacons in the visibility network is 15.2 km, similar to the average length of 14.4 km between city sites. However, the maximum visibility distance varies. Beacon towers have stronger forewarning and watching ability. Besides, they still have the advantages of low-cost building and high-efficiency signal transmission. As in Figure 9, the strong monitoring ability of each beacon tower is also shown through high cumulative frequency, while the city site focuses on integral associated defensibility, contributing to the surveillance due to the lower cumulative frequency of each site and larger cumulative area observed from the whole city group. In addition, the signal transmission between beacon and city is also an important defensive precaution. It connects the sites with longer linkages for fast signal delivery and constitutes a multifunctional monitoring system.

(No.) B-B B-B 35 B-B B-B B-B B-B Length B-B 30 B-B B: Beacon tower ISPRS Int. J. Geo-Inf. 2017, 6, 247 B-B 11 of 19 B-B C: City site B-B B-B 25 B-C B-C B-C B-C to control the information transmission to other sites.B-C The site C12 plays a major role in information 20 B-C B-C B-C control and has a higher centrality levelB-C due to high betweenness and degree. Sites B12 and B13 are B-C 15 B-C located in the high value zone in a scatter-plotC-C map in Figure3, representing strong dominant power C-C C-C for communication to adjacent cities. InC-C this local defensive system, these two beacon towers have the 10 C-C C-C relatively important function of surveillance andC-C control. In Region 3, the beacons align. Particularly C-C C-C from B2 to B7, the visibility5 lines are combined inC-C a multiple cross pattern that is the result of good C-C C-C C-C visibility and closeness in geographicalC-C space, and more importantly, the(km) requirement of a compact observation structure to guard0 the transportation path. Sites B1, B4, and B7 have an important control 0 1020304050 function in this local network, while site B11 is beyond visible range from other beacons and acts as an independentFigure tower 8. Length in the of defensivevisibility lines. system. The observing sites are classified into three observing types.

1.0

0.9 1.0

0.8

0.8 0.7 Beacon City

Cumulative Area 0.6 0.6 Cumulative Frequency Cumulative

0 5 10 15 20 0.5 Observing Points

0 5 10 15 20 Observing Points

FigureFigure 9. 9.Cumulative Cumulative viewshedviewshed of city sites sites and and beacon beacon tower. tower.

4.2. TransportationThe CNA analysis Corridors method describes the specific relationship of every single site in the local visibility network. In Region 1 of Figure 3, sites B16, B17, and C20 all have large closeness of CNA The transportation corridors have an important influence on economic and cultural communication which is controlled by geographical factors and defensive considerations. With the least-cost path analysis that is based on defensibility models, some progress was made to interpret the internal relationship of archaeological sites and some cultural or political connections. The two transportation directions starting from C4 in the orientation of Turpan and from C1 in the orientation of Loulan represent different cultural communication pathways which coincide with the trend of the Silk Road in Han Dynasty [26]. Transportation corridor analysis is compared in the way of cumulative difference that represents the total deviation of two paths during the movement and cumulative difference ratio that delineates the real-time discrepancy proportion in the traveling process. For analysis and validation, 2281 and 1697 sampling points were separately extracted in 200 m intervals in an east-to-west direction along the paths from C1 to C22 and from C4 to C22. As is shown in Figures7 and 10, the cumulative difference derived from each defensibility model is quite larger than paths from a slope model, which takes the geographical reason as the only influence. The discrepancy varies with different starting points. The two paths from C1 to C22 in linear and logistic models are highly differentiated, with slope-based paths according to the large maximum disparity located at both P1 and P2, in contrast to locations P4 and P5 along the paths from C4 to C22 in linear and logistic models. This is caused by both geographical conditions and regional defensive requirements. With a less cumulative difference from a slope-based route, the paths connecting C4 and C22 do not fit well as those from C1 to C22 in the two defensibility models, because the maximum difference ratio which reflects the variation in the intermediate processing increases to 0.0823 at P6 from 0.0487 at P3. The most different tendency of the paths happens in the section of C6–C11, which is affected by their high probability density based on ISPRS Int. J. Geo-Inf. 2017, 6, 247 12 of 19 sites, majorly according to the geographical law. The sites C7, C8, C18, C19, C23 that align in the surroundingISPRS Int. J. Geo-Inf. of the2017 paths, 6, 247 are scientifically designed to save physical cost when traveling 12between of 19 them in ancient times. In contrast, the paths from the defensibility model were reconstructed in differentthe logistic ways. model They in focus its entirety, more on differentiating defensive consid with theeration, region not of only C6–C12 on the under geographical single-emphasis conditions. for Indefensive the defensibility importance models, in the linearthe logistic model, regression and the turning method points takes in the all traveling sites in progressthe study correspond area as an integrationwith the position with equal of C11 importance and C12 in grade Figures in7 the and de 10fensive. The difference system, so ratio the is paths weakened would in pass terms through of a thelong sites traveling concentrated distance regions. from C1 While to C22. the However, linear regr for describingession method total variation,is applied the by average differentiating deviation the defensivederived fromimportance cumulative of each differences site, so thatthe arepaths assigned go through to every the sampling most important point along site. the Some transportation critical sites influencepaths was the 2.80 tendency km from C1of tothe C22, traveling similar toways, the average unlike deviation the paths of 2.89based km on from geographical C4 to C22. It canfactor influencingbe concluded the that location the paths of sites from in different the slope defensibility model. So, models the coincideanalysisin of general, the defensive but with role a gap of in the importantdetail along sites these is needed. corridors.

(a) (b)

FigureFigure 10. 10. CumulativeCumulative deviation deviation between defensibilitydefensibility models models and and slope slope model. model.

4.3. Shape Analysis and Defensive Role The transportation corridors play an important role in connecting the Han Empire and the Western Regions.Shape Theanalysis slope-based aims at paths discussing are obviously cultural different patterns from and thatdefensive based onroles the in defensibility the whole models,system by anespecially experiential in the judgment. direction ofBy C1 general to C22. consensus, They are useful the square to find city some site inter-relations in Han Dynasty of the is sites, from majorly Central Plainaccording Culture, to thewhile geographical the round law. shape The is sites a classic C7, C8, style C18, in C19, the C23local that Western align in Region the surrounding [25]. The ofcultural the fusionpaths and are scientificallymilitary conflict designed exist to as save a unit. physical On costthe whenone hand, traveling the betweenculture-related them in sites ancient signify times. the communicationIn contrast, the between paths from each the other. defensibility On the modelother werehand, reconstructed the troops were in different stationed ways. in a They fortified focus site tomore defend on against defensive invasion. consideration, Sites C7, not C8, only C18, on theC19, geographical and C23 are conditions. located along In the the defensibility southern path models, from C1the to logisticC22 as transfer regression stations method in takesFigure all 7a. sites Their in the inherent study area relations as an integrationare reflected with in shape equal importanceand structure withgrade the in classic the defensive character system, of Han so Culture, the paths such would as pass the throughsquare shape the sites of concentratedC7, C19 and regions.the complicated While compositionthe linear regression of C18 shown method in is Figure applied 11. by differentiatingSites C8 and C23 the defensiveare not included importance in ofthe each shape site, analysis, so the paths go through the most important site. Some critical sites influence the tendency of the traveling because the ruin of C8—of which the location is inferred from the record of the second heritage ways, unlike the paths based on geographical factor influencing the location of sites in the slope model. census— is not currently visible. Location C23 was investigated in 1954 and recorded as a square city So, the analysis of the defensive role of the important sites is needed. with a length of 69 m from south to north and 74 m from east to west, which is now disappearing due to4.3. occupation Shape Analysis by farmland. and Defensive Site RoleC18, with a large city and special pattern, is considered to be the captain city of cultivating wasteland from the archaeological discovery in early times. So, this Shape analysis aims at discussing cultural patterns and defensive roles in the whole system by an transportation path going through these sites is proven influential to cultural communication under experiential judgment. By general consensus, the square city site in Han Dynasty is from Central Plain the supervisory control of the Han Empire, but its strategic position is not important, as seen by the Culture, while the round shape is a classic style in the local Western Region [25]. The cultural fusion and lack of beacon towers needed in a complete defensive system. In addition, sites C3, C12, C16, C17, military conflict exist as a unit. On the one hand, the culture-related sites signify the communication andbetween C20 are each together other. On considered the other hand,to guard the troops the ke werey junction stationed connecting in a fortified east site and to defend west along against the northerninvasion. path, Sites based C7, C8, on C18, the C19,slope and and C23 defensibility are located alongmodels. the Through southern pathanalysis from of C1 shape to C22 and as transfer structure instations Figures in11 Figure and 12,7a. these Their sites inherent have relationsclassic styles are reflected of square in or shape irregular and structure polygon withshapes, the and classic some withcharacter a large of mound Han Culture, inside, such which as the manifests square shape the infl of C7,uence C19 of and the the Central complicated Plain compositionCulture to the of C18 West Regionshown (site in Figure C20 was 11. Sitesthe historically C8 and C23 famous are not includedQiuci City, in thethe shape cultural analysis, center because that was the necessary ruin of C8—of for Silk Roadwhich passage, the location and now is inferred is mostly from gone the except record for of thepartial second walls heritage in the census—isnorth ruin,not making currently it unavailable visible. Location C23 was investigated in 1954 and recorded as a square city with a length of 69 m from south

ISPRS Int. J. Geo-Inf. 2017, 6, 247 13 of 19 to north and 74 m from east to west, which is now disappearing due to occupation by farmland. Site C18, with a large city and special pattern, is considered to be the captain city of cultivating wasteland from the archaeological discovery in early times. So, this transportation path going through these sites is proven inﬂuential to cultural communication under the supervisory control of the Han Empire, but its strategic position is not important, as seen by the lack of beacon towers needed in a complete defensive system. In addition, sites C3, C12, C16, C17, and C20 are together considered to guard the key junction connecting east and west along the northern path, based on the slope and defensibility models. Through analysis of shape and structure in Figures 11 and 12, these sites have classic styles of square or irregular polygon shapes, and some with a large mound inside, which manifests the influence of the Central Plain Culture to the West Region (site C20 was the historically famous Qiuci City, the culturalISPRS center Int. J. Geo-Inf. that was 2017, necessary 6, 247 for Silk Road passage, and now is mostly gone except for partial13 of 19 walls in the north ruin, making it unavailable for shape analysis). They are connected in a geographically for shape analysis). They are connected in a geographically optimal way for convenient optimaltransportation, way for convenient which greatly transportation, contributes to which the ancient greatly cultural contributes fusion. to the ancient cultural fusion.

(a) C3 (b) C4 (c) C7

(d) C10 (e) C11 (f) C12

(g) C13 (h) C14 (i) C17

(j) C18 (k) C19 (l) C22

Figure 11. Archaeological planar graphs of some city sites with remains on the surface. The materials Figure 11. Archaeological planar graphs of some city sites with remains on the surface. The materials are from the third cultural relic census data. are from the third cultural relic census data. While based on defensibility paths, the city sites C6, C13, as well as beacons B8 to B11 and B13 to B15, also have a great effect on the military activity. Particularly, C6 is destroyed and not able to be judged for the overall shape, but the inevitable traffic place indicates its important role. Site C13 was confirmed as the military castle to guard the transportation corridor, with similar shape and function to C17 (its incomplete square is caused by destruction). The two sites—C13 and C17—are in close relation to the neighboring beacon towers as outposts to defend against attack from the west. Another approximate site, C11, was concluded to be a castle constructed by the Han Empire. Site C10, 4 km away, with a rounder shape, may be the auxiliary city of C11, or the earlier city abandoned after occupation by Han Empire due to its local cultural style in configuration. Along the same path in the

ISPRS Int. J. Geo-Inf. 2017, 6, 247 14 of 19 east, the site C9, with more than twenty housing ruins, is considered to be for residential living. Its

ISPRSsize and Int. J.shape, Geo-Inf. however,2017, 6, 247 is difficult to confirm. The cities from C9 to C13 form a linear defensive14 ofand 19 residential group together with beacon B12.

FigureFigure 12.12. ((aa)) MultispectralMultispectral andand ((bb)) hillshadehillshade imageryimagery ofof KuyuxKuyux ShahrShahr (C15);(C15); ((cc)) MultispectralMultispectral andand ((dd)) hillshadehillshade imageryimagery ofof KuhneKuhne ShahrShahr (C16).(C16). TheyThey areare fromfrom aerialaerial photosphotos byby unmannedunmanned aerialaerial vehiclevehicle (UAV)(UAV) inin 2016.2016.

Another group, with four sites from C14 to C17 in the central position, constitutes a cooperative While based on defensibility paths, the city sites C6, C13, as well as beacons B8 to B11 and B13 defensive unit to guard transportation lines. As a dominant site to control information exchange in to B15, also have a great effect on the military activity. Particularly, C6 is destroyed and not able the visibility network, C16 also has a high structural hierarchy with anomalous shape and quite large to be judged for the overall shape, but the inevitable traffic place indicates its important role. Site scale. It has 22 horse-faces (the city defense facility in the walls in ancient China against side attack) C13 was confirmed as the military castle to guard the transportation corridor, with similar shape and along the city wall, shown in Figures 12 and 13b. The defensive role is obvious. Only 8 km away in function to C17 (its incomplete square is caused by destruction). The two sites—C13 and C17—are in the south, C15 is another high-grade city whose shape is between round and square and with a large close relation to the neighboring beacon towers as outposts to defend against attack from the west. mound inside (see Figure 13a). Sites C15 and C16 meet some criteria of the Western Region’s Frontier Another approximate site, C11, was concluded to be a castle constructed by the Han Empire. Site C10, Command in the following aspects: high defensive importance, vital location along the transportation 4 km away, with a rounder shape, may be the auxiliary city of C11, or the earlier city abandoned after lines, large scale, and classic architectural shape of the Han Culture. As well, site C17, as a military occupation by Han Empire due to its local cultural style in configuration. Along the same path in the castle, guarded the west of Luntai, and C14 may be the ancient satellite city or captain city for east, the site C9, with more than twenty housing ruins, is considered to be for residential living. Its cultivating wasteland and providing logistical supplies with lower defensive importance in spatial size and shape, however, is difficult to confirm. The cities from C9 to C13 form a linear defensive and position, in spite of its large dimension. What should be noted is that the mound in the east side of residential group together with beacon B12. C14 could provide far horizon to defend against invasion from the east. Site C14 is the first gate to Another group, with four sites from C14 to C17 in the central position, constitutes a cooperative pass through when traveling further west. It can be inferred that this region had cooperative defense defensive unit to guard transportation lines. As a dominant site to control information exchange and prospered during the Han Dynasty for these intensive sites. in the visibility network, C16 also has a high structural hierarchy with anomalous shape and quite In the field investigation, more archaeological details were acquired about C15 and C16. In the large scale. It has 22 horse-faces (the city defense facility in the walls in ancient China against side city site C15, the commanding height of the large mound in city and its adobe buildings prove the attack) along the city wall, shown in Figures 12 and 13b. The defensive role is obvious. Only 8 km local authority controlling the whole city. The burning indications of the ancient buildings inside the away in the south, C15 is another high-grade city whose shape is between round and square and city are proof of being destroyed from warfare. This site is considered an unidentified official with a large mound inside (see Figure 13a). Sites C15 and C16 meet some criteria of the Western institution or kingdom in the Western Region. Many remains inside C16 were found, including Region’s Frontier Command in the following aspects: high defensive importance, vital location along numerous pottery shards, iron slag, and coins, demonstrating frequent human activities and the transportation lines, large scale, and classic architectural shape of the Han Culture. As well, site

C17, as a military castle, guarded the west of Luntai, and C14 may be the ancient satellite city or captain city for cultivating wasteland and providing logistical supplies with lower defensive importance in ISPRS Int. J. Geo-Inf. 2017, 6, 247 15 of 19

ISPRS Int. J. Geo-Inf. 2017, 6, 247 15 of 19 spatial position, in spite of its large dimension. What should be noted is that the mound in the east sideimportant of C14 residential could provide and military far horizon roles. to It defend was governed against invasion by the Han from Empire. the east. By Sitethe comprehensive C14 is the ﬁrst gateanalysis to pass of defensive through when function traveling and field further investigation, west. It can the be important inferred that historical this region roles had of the cooperative two sites defensewere verified and prospered by these archaeological during the Han characteristics. Dynasty for these intensive sites.

(a) (b)

FigureFigure 13.13.( a(a)) The The large large mount mount inside inside the the site site of Kuyux of Kuyu Shahrx Shahr city withcity anwith 8-meter an 8-meter height; height; (b) the ( cornerb) the ofcorner the site of the Kuhne site ShahrKuhne in Shahr the north-west in the north- withwest a horse-facewith a horse-face for defending for defending attack. attack.

Through this research, some defensive roles of sites along the transportation corridors have been In the field investigation, more archaeological details were acquired about C15 and C16. In the confirmed, though the paths based on defensibility analysis in the Han Dynasty still need more city site C15, the commanding height of the large mound in city and its adobe buildings prove the local verification. It affords researchers feasible methods to analyze the organizational pattern of the authority controlling the whole city. The burning indications of the ancient buildings inside the city archaeological sites and find ancient transportation corridors. These corridors are useful to confirm are proof of being destroyed from warfare. This site is considered an unidentified official institution or the cultural connections and provide reference to discover unknown ancient routes. The defensibility kingdom in the Western Region. Many remains inside C16 were found, including numerous pottery distributions of both the linear and logistic regression models can be recognized as efficient cost shards, iron slag, and coins, demonstrating frequent human activities and important residential and surfaces to explore optimal routes under cultural and political purposes. The military defensive military roles. It was governed by the Han Empire. By the comprehensive analysis of defensive system is also represented, with the visibility network constituted by special sites. More verification function and field investigation, the important historical roles of the two sites were verified by these from archaeological information will be helpful to explain specific internal cultural relationships. As archaeological characteristics. recorded in documentation, the Han Empire opened up wasteland and built the Western Region’s Through this research, some defensive roles of sites along the transportation corridors have Frontier Command to govern the early Xinjiang and protect the Silk Road from invasion by the Huns. been confirmed, though the paths based on defensibility analysis in the Han Dynasty still need These discussions can provide some clues to locate the Western Region’s Frontier Command of the moreHan Dynasty verification. among It affords important researchers sites and feasible find methodsthe missing to analyze beacons the along organizational the Silk Road pattern in ofmajor the archaeologicaldefence area. The sites ancient and find paths ancient were transportationalso simulated corridors.in anotherThese way, based corridors on the are slope useful model. to confirm They theare culturalconsidered connections to be the andearly provide routes referencebefore large to discovernumbers unknown of sites were ancient built routes. to guard The the defensibility Silk Road. distributionsWhen defensive of bothrequirements the linear became and logistic important, regression the paths models transferred can be to recognized other ways as with efficient maximum cost surfacesdefensive to importance explore optimal in the routeslandscape under or culturalhigh-probability and political density purposes. places with The site military clustering. defensive The systemleast physical is also represented,cost is a basic with consideration the visibility for networkmatching constituted or traffic at byany special time, sites.so the Moreslope-based verification path fromis an archaeologicalalternative option. information The transportation will be helpful line to in explain the south specific of the internal study culturalarea goes relationships. along the river As recordedbasin of Tarim, in documentation, and thus is easily the Han affected Empire by openedchanges up in wastelandthe watercourse. and built The the historical Western correlation Region’s Frontierbetween Command path transition to govern and environmental the early Xinjiang variation and protect is worthy the Silkof more Road discussion from invasion in the by future. the Huns. These discussions can provide some clues to locate the Western Region’s Frontier Command of the Han5. Conclusions Dynasty among important sites and find the missing beacons along the Silk Road in major defence area. The ancient paths were also simulated in another way, based on the slope model. They are consideredIn this toresearch, be the earlythe defensive routes before system large in numbersthe Han Dynasty, of sites were which built is directly to guard associated the Silk Road. with Whenmilitary defensive authority requirements in the governed became area important, of the sout thehern paths Tianshan transferred Mountains to other in waysCentral with Xinjiang, maximum was defensiverebuilt using importance SRTM DEM. in the landscapeIn history, or the high-probability military authority density of the places Han with Empire site clustering. contributed The to least the physicalprosperity cost and is astability basic consideration along the Silk for Road. matching By constructing or traffic at anythe time,compound so the visibility slope-based network path is and an alternativesimulating option.transportation The transportation corridors linebased in theon souththe quantified of the study defensibility area goes alongmodel, the the river defensive basin of Tarim,function and of thusancient is easily sites affectedwas discussed by changes together in the with watercourse. the pattern The of their historical information correlation transmission between pathand cultural transition communication. and environmental The anal variationysis results is worthy show of that more the discussion visibility innetwork the future. constituted three local parts, two of which in the east are the important defensive regions. The beacons with stronger observing ability play a more important role in this network. The conspicuous defensive functions are mainly reflected in the eight city sites (Uzgen Bulak, Shah Kalandar, Agra, Chuck Castle, Kuyux

ISPRS Int. J. Geo-Inf. 2017, 6, 247 16 of 19

5. Conclusions In this research, the defensive system in the Han Dynasty, which is directly associated with military authority in the governed area of the southern Tianshan Mountains in Central Xinjiang, was rebuilt using SRTM DEM. In history, the military authority of the Han Empire contributed to the prosperity and stability along the Silk Road. By constructing the compound visibility network and simulating transportation corridors based on the quantified defensibility model, the defensive function of ancient sites was discussed together with the pattern of their information transmission and cultural communication. The analysis results show that the visibility network constituted three local parts, two of which in the east are the important defensive regions. The beacons with stronger observing ability play a more important role in this network. The conspicuous defensive functions are mainly reflected in the eight city sites (Uzgen Bulak, Shah Kalandar, Agra, Chuck Castle, Kuyux Shahr, Kuhne Shahr, Caladar and Qiuci) and some Beacons (Turkish, Caleta, Akeerdick, Karaya, West Layisu, QiuFutur, and Dunmaili Tur). Respectively speaking, the sites of Agra and Kuhne Shahr were proven to have more controlling ability in their local defensive region in a visibility network. They have important effects on the surveillance and governance of the transportation corridors. The sites of Kuyux Shahr and Kuhne Shahr have been confirmed as significant in the defensive system as the alternative site of the Western Region’s Frontier Command through defensibility and shape analysis. Meanwhile, the southern sites (Chalcedo Ursa, Kajaskule, Big city, Aksi and Tashton) may serve as military stations and residential cities along the slope-based path and Tarim River valley. The route passing through these southern sites had less defensive importance during the Han Dynasty because no strategic beacons in that area can combine to form a comprehensive defensive system. In addition, the transportation corridors based on the linear defensibility model fit well with that based on the logistic regression method, in general. They were the main historical defensive transportation corridors. The southern least-cost path, based on a slope model, differs greatly with other paths from Yingpan to Muscat. It provided an optional traffic path with less defensive requirements and played a less important role in cultural communication after the Silk Road was opened. All the paths constitute the main arterial traffic in the Han Dynasty. The reconstructed defensive system in this research shows the different defensive roles of the ancient sites and possible transportation lines for military activity or cultural communication. It can be applied to provide some insights into social organization and human behavior preference under defensive considerations. Furthermore, using this methodological framework, the spatial connotation of ancient sites can be explored, and anthropological archaeology can be recognized more broadly. More meaningful work in the future is needed, such as seeking out the Western Region’s Frontier Command or the lost beacon towers along the Silk Road. The conclusions from the defensive system analysis are expected to be verified by the some more archaeological discoveries.

Supplementary Materials: The SRTM DEM data set was provided by the International Scientific & Technical Data Mirror Site, Computer Network Information Center, the Chinese Academy of Sciences, http://www.gscloud.cn. Acknowledgments: This research is funded by the national key technology R & D program of the project “Research on the Application of Remote Sensing Technology in Exploring the Origin of Chinese Civilization” (No. 2013BAK08B06) and “Remote sensing and geophysical comprehensive in archaeological research of China classical archaeological site” (No. 2015BAK01B01). We sincerely thanks Luntai administration of cultural heritage, Huarui Zhang and the staff of Luntai museum in Xinjiang. We also thank Xinjiang Cultural Relics and Archaeology Institute and the staff for providing the data of archaeological sites, especially Jianjun Yu, Xingjun Hu for giving advises of archaeological analysis and Zhihao Dang for providing help for field investigation. Author Contributions: J.Z. Contributed to the article writing and experiment. F.L. and H.G. analyzed the archaeological data and the historical document. Y.N. and L.Y. designed the study methods. Conflicts of Interest: The authors declare no conflicts of interest. ISPRS Int. J. Geo-Inf. 2017, 6, 247 17 of 19

Appendix A Table A1. Archaeological site comparison table.

Yingpan (C1) Shahr Tokai (C2) Uzgen Bulak (C3) Bogurda (C4) Tacroli (C5) Shah Kalandar (C6) Chalcedo Ursa (C7) Kajaskule (C8) Yeyungou Mazar (C9) Aziz (C10) Wuliponk (C11) Agra (C12) Chuck Castle (C13) Drow Kurt (C14) Kuyux Shahr (C15) Kuhne Shahr (C16) Caladar (C17) Big city (C18) Aksi (C19) Qiuci (C20) Kuonaxiehair (C21) Muscat (C22) Tashton (C23) Egmery Ruddock (C24) Turkish (B1) West Turkish (B2) Keyak Kuduk (B3) Caleta (B4) Gumush (B5) Saluwak (B6) Akeerdick (B7) Saqit (B8) Sunji (B9) Ark Shimron (B10) Sukati (B11) Karaya (B12) West Layisu (B13) QiuFutur (B14) Dunmaili Tur (B15) KuokongBazi (B16) Kezier Garha (B17) Yabuyi (B18)

References

1. Abdukirim, M. On the troops wasteland-opening up and its effect in the western regions during the Han-Tang Dynasty. J. Kashgar Teach. Coll. 2010, 31, 46–49. [CrossRef] 2. Jiang, N. Research on the System of Frontier in Han Dynasty. Ph.D. Thesis, Central China Normal University, Wuhan, China, 2013. 3. Song, C. Establishment of the Western Regions Frontier Command and Competition for Qiuci between Han Empire and Huns during the Western and Eastern Han, analyzing from Wulei to Tagan. Qiuci Stud. 2012, 5, 58–72. 4. Zhou, H. Jurisdiction and management of Han and Tang Dynasty in Xinjiang from the archaeological data. J. Xinjiang Norm. Univ. 2000, 21, 40–49. [CrossRef] 5. Zhang, A.F.; Tian, H.F. The remains of the site and the layout of the defense cities of the Han Dynasty in the Western Regions. J. Chin. Hist. Geogr. 2015, 30, 47–55. 6. Garcia-Moreno, A. To see or to be seen...is that the question? An evaluation of palaeolithic sites’ visual presence and their role in social organization. J. Anthropol. Archaeol. 2013, 32, 647–658. [CrossRef] 7. Rua, H.; Goncalves, A.B.; Figueiredo, R. Assessment of the Lines of Torres Vedras defensive system with visibility analysis. J. Archaeol. Sci. 2013, 40, 2113–2123. [CrossRef] 8. Déderix, S. A matter of scale. Assessing the visibility of circular tombs in the landscape of Bronze Age Crete. J. Archaeol. Sci. Rep. 2015, 4, 525–534. [CrossRef] 9. Ogburn, D.E. Assessing the level of visibility of cultural objects in past landscapes. J. Archaeol. Sci. 2006, 33, 405–413. [CrossRef] 10. Sevenant, M.; Antrop, M. Settlement models, land use and visibility in rural landscapes: Two case studies in Greece. Landscape Urban Plan. 2007, 80, 362–374. [CrossRef] 11. Kyle Bocinsky, R. Extrinsic site defensibility and landscape-based archaeological inference: An example from the Northwest Coast. J. Anthropol. Archaeol. 2014, 35, 164–176. [CrossRef] 12. Martindale, A.; Supernant, K. Quantifying the defensiveness of defended sites on the Northwest Coast of North America. J. Anthropol. Archaeol. 2009, 28, 191–204. [CrossRef] 13. Danese, M.; Masini, N.; Biscione, M.; Lasaponara, R. Predictive modeling for preventive archaeology: Overview and case study. Cent. Eur. J. Geosci. 2014, 6, 42–55. [CrossRef] 14. Carrer, F. An ethnoarchaeological inductive model for predicting archaeological site location: A case-study of pastoral settlement patterns in the Val di Fiemme and Val di Sole (Trentino, Italian Alps). J. Anthropol. Archaeol. 2013, 32, 54–62. [CrossRef] 15. Turrero, P.; Domínguez-Cuesta, M.J.; Montserrat Jiménez-Sánchez, M.; García-Vázquez, E. The spatial distribution of Palaeolithic human settlements and its inﬂuence on palaeoecological studies: A case from Northern Iberia. J. Archaeol. Sci. 2013, 40, 4127–4138. [CrossRef] 16. Jaroslaw, J.; Hildebrandt-Radke, I. Using multivariate statistics and fuzzy logic system to analyse settlement preferences in lowland areas of the temperate zone: An example from the Polish Lowlands. J. Archaeol. Sci. 2009, 36, 2096–2107. [CrossRef] 17. Garcia, A. GIS-based methodology for Palaeolithic site location preferences analysis. A case study from Late Palaeolithic Cantabria (Northern Iberian Peninsula). J. Archaeol. Sci. 2013, 40, 217–226. [CrossRef] 18. Ullah, I. A GIS method for assessing the zone of human-environmental impact around archaeological sites: A test case from the Late Neolithic of Wadi Ziqlâb, Jordan. J. Archaeol. Sci. 2011, 38, 623–632. [CrossRef] 19. Jones, E.E. An analysis of factors inﬂuencing sixteenth and seventeenth century Haudenosaunee (Iroquois) settlement locations. J. Anthropol. Archaeol. 2010, 29, 1–14. [CrossRef] ISPRS Int. J. Geo-Inf. 2017, 6, 247 18 of 19

20. Moskal-del Hoyo, M.; Litynska-Zajac, M.; Korczynska, K.; Cywa, K.; Kienlin, T.L.; Cappenberg, K. Plants and environment: Results of archaeobotanical research of the Bronze Age settlements in the Carpathian Foothills in Poland. J. Archaeol. Sci. 2015, 53, 426–444. [CrossRef] 21. Menze, B.H.; Ur, J.A. Mapping patterns of long-term settlement in Northern Mesopotamia at a large scale. Proc. Natl. Acad. Sci. USA 2012, 778–787. [CrossRef][PubMed] 22. Canosa-Betés, J. Border surveillance: Testing the territorial control of the Andalusian defense network in center-south Iberia through GIS. J. Archaeol. Sci. Rep. 2016, 9, 416–426. [CrossRef] 23. Lanen, R.J.; Kosian, M.C.; Groenewoudt, B.J.; Spek, T.; Jansma, E. Best travel options: Modeling Roman and early-medieval routes in the Netherlands using a multi-proxy approach. J. Archaeol. Sci. Rep. 2015, 3, 144–159. [CrossRef] 24. Fletcher, R. Some spatial analyses of Chalcolithic settlement in Southern Israel. J. Archaeol. Sci. 2008, 35, 2048–2058. [CrossRef] 25. Lin, M.C. Research on today’s location of Regions Frontier Command site from archaeological perspective. Hist. Res. 2013, 6, 43–58. 26. Xinjiang Bureau of Surveying Mapping and Geoinformation. The Map of China Xinjiang, 1st ed.; Sino Maps Press: Beijing, China, 2015; pp. 6–9. 27. Zhou, J. Study on the Xiyu Duhu of the Western Han Dynasty. Master’s Thesis, Northwest Normal University, Lanzhou, China, 2010. 28. Hong, T. A brief comment on the research on the safeguard city in the Western Region in Han Dynasty. J. Xinjiang Norm. Univ. 2007, 28, 5–10. [CrossRef] 29. Shang, Y.B. Several economical problems of Qiuci in Han and Tang Dynasty. J. Xinjiang Norm. Univ. 1989, 4, 31–37. 30. Hu, Z.H.; Peng, J.W.; Hou, Y.L.; Shan, J. Evaluation of Recently Released Open Global Digital Elevation Models of Hubei, China. Remote Sens. 2017, 9, 1–16. [CrossRef] 31. Fisher, P.; Farrely, C. Spatial Analysis of visible areas from the Bronze Age Cairns of Mull. J. Archaeol. Sci. 1997, 24, 581–592. [CrossRef] 32. Scott, J. Social Network Analysis: A Handbook, 3rd ed.; Sage Publications: London, UK, 2012; pp. 32–97. 33. Wasserman, S.; Faust, K. Social network analysis in the social and behavioral sciences. In Social Network Analysis Methods and Application, 1st ed.; Cambridge University Press: Cambridge, UK, 1994; pp. 3–27. 34. Earley-spadoni, T. Landscapes of warfare: Intervisibility analysis of Early Iron and Urartian ﬁre beacon stations (Armenia). J. Archaeol. Sci. Rep. 2015, 3, 22–30. [CrossRef] 35. De Montis, A.; Caschili, S. Nuraghes and landscape planning: Coupling viewshed with complex network analysis. Landscape Urban Plan. 2012, 105, 315–324. [CrossRef] 36. Mullins, P. Webs of defense: Structure and meaning of defensive visibility networks in Prehispanic Peru. J. Archaeol. Sci. Rep. 2016, 8, 346–355. [CrossRef] 37. Wikipedia. Available online: https://en.wikipedia.org/wiki/Centrality (accessed on 4 November 2016). 38. De Reu, J.; Bourgeois, J.; De Smedt, P.; Zwertvaegher, A.; Antrop, M.; Bats, M.; De Maeyer, P.; Finke, P.; Van Meirvenne, M.; Verniers, J.C.; et al. Measuring the relative topographic position of archaeological sites in the landscape, a case study on the Bronze Age barrows in northwest Belgium. J. Archaeol. Sci. 2011, 38, 3435–3446. [CrossRef] 39. Siart, C.; Eitel, B.; Panagiotopoulos, D. Investigation of past archaeological landscapes using remote sensing and GIS: A multi-method case study from Mount Ida, Crete. J. Archaeol. Sci. 2008, 35, 2918–2926. [CrossRef] 40. Marsh, E.J.; Schreiber, K. Eyes of the empire: A viewshed-based exploration of Wari site-placement decisions in the Sondondo Valley, Peru. J. Archaeol. Sci. Rep. 2015, 4, 54–64. [CrossRef] 41. Falconer, L.; Hunter, D.C.; Telfer, T.C.; Ross, L.G. Visual, seascape and landscape analysis to support coastal aquaculture site selction. Land Use Policy 2013, 34, 1–10. [CrossRef] 42. Fernandes, R.; Geeven, G.; Soetens, S.; Klontza-Jaklova, V. Deletion/Substitution/Addition (DSA) model selection algorithm applied to the study of archaeological settlement patterning. J. Archaeol. Sci. 2011, 38, 2293–2300. [CrossRef] 43. Perakis, K.; Moysiadis, A. Geospatial predictive modeling of the Neolithic archaeological sites of Magnesia in Greece. Int. J. Digit. Earth 2011, 4, 421–433. [CrossRef] 44. Espa, G.; Benedetti, R.; De Meo, A.; Ricci, U.; Espa, S. GIS based models and estimation methods for the probability of archaeological site location. J. Cult. Herit. 2006, 7, 147–155. [CrossRef] ISPRS Int. J. Geo-Inf. 2017, 6, 247 19 of 19

45. Carballo, D.M.; Pluckhahn, T. Transportation corridors and political evolution in highland Mesoamerica: Settlement analyses incorporating GIS for northern Tlaxcala. Mexico. J. Anthropol. Archaeol. 2007, 26, 607–629. [CrossRef] 46. Sakaguchi, T.; Morin, J.; Dickie, R. Defensibility of large prehistoric sites in the Mid-Fraser region on the Canadian Plateau. J. Archaeol. Sci. 2010, 37, 1171–1185. [CrossRef] 47. Lanen, R.J.; Kosian, M.C.; Groenewoudt, B.J.; Jansma, E.; Rowin, J.; Menne, C.; Bert, J.; Esther, J. Finding a way: Modeling landscape prerequisites for Roman and early-Medieval routes in the Netherlands. Geoarchaeology 2015, 30, 200–222. [CrossRef] 48. Güimil-Farina, A.; Parcero-Oubina, C. “Dotting the joins”: A non-reconstructive use of least cost paths to approach ancient roads. The case of the Roman roads in the NW Iberian Peninsula. J. Archaeol. Sci. 2015, 54, 31–44. [CrossRef] 49. Fábrega-Álvarez, P. Moving without destination. A theoretical, GIS-based determination of routes (optimal accumulation model of movement from a given origin). Archaeol. Comput. Newsl. 2006, 64, 7–11. 50. Llobera, M.; Fábrega-Álvarez, P.; Parcero-Oubiña, C. Order in movement: A GIS approach to accessibility. J. Archaeol. Sci. 2011, 38, 843–851. [CrossRef] 51. Field, J.S.; Petraglia, M.D.; Lahr, M.M. The southern dispersal hypothesis and the South Asian archaeological record: Examination of dispersal routes through GIS analysis. J. Anthropol. Archaeol. 2007, 26, 88–108. [CrossRef] 52. Yubero-Gómez, M.; Rubio-Campillo, X.; Javier López-Cachero, F.; Esteve-Gràcia, X. Mapping changes in late prehistoric landscapes: A case study in the northeastern Iberian Peninsula. J. Archaeol. Sci. 2015, 40, 123–134. [CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).