Preservation and utilisation of historic buildings in old district of Guangzhou from the perspective of space syntax

The old district of Guangzhou was chosen as the research scope. ‘Old district’ was given a detailed description in Technical Regulations on Urban and Rural Planning of Guangzhou Municipality (2012): the urban scope built over 30 years, including the regions of ‘Huanshi Road; Hengfu Road; South from Yongfu Road, West from Guangzhou Avenue, Changgang Road; North from Xingang Road, Zhujiang watercourse of Bai’e’tan; Hedong Bridge; East from Tong Deyong, with an area of about 54 sq.km.’ (Seen in Figure 1). The historic regions delimited by the city wall in the Ming-Qing Dynasty play an important role in the old district and affect the layout of the old district deeply, and therefore is specially marked.

Fig. 1

Urban forms and historic buildings of the old district of Guangzhou were studied based on the space syntax analysis of the road network structure. The road data of the old district were from Amap. Since the road network obtained in the modelling is prone to failure of sDNA calculation due to the inevitable fracture, the paper drew the road network of the old district manually according to the high-definition street map, and there were 16545 roads in the region. Historical buildings refer to buildings and structures that have certain protection values, can reflect historical features and local characteristics, and are not announced as cultural relics protection units or registered as immovable cultural relics after being determined and announced by the people's governments of cities and counties. Under normal circumstances, historical buildings are not evenly distributed in the city, but are affected by the economic and social development of the city, and are closely related to the history of the city. The historic buildings were obtained by the spider tool from the open API of Amap; a list of the historic buildings published by Guangzhou Municipal Bureau of Planning and Natural Resources six times were referred. Data were summarized and integrated using comparison between the ancient and modern map, analysis of historic effects and literature and history, and information retrieval. The included old district has certain cultural deposits and representativeness, and there were 150 regions of the old district in total, with the distribution shown in Figure 2.

Fig. 2

The research includes two aspects: one is the analysis of space syntax of the old district with space syntax as the theory and tool. The old district was analysed from the perspective of urban form based on space statistics, closeness, and betweenness so as to explore the spatial layout. The other is thinking about the preservation and utilisation of the historic buildings. According to the distribution features of the historic buildings in the spatial layout, suggestions and strategies were put forward for the preservation and utilisation. The framework and idea of the research are seen in Figure 3.

Fig. 3

Integration, also called closeness, is the core parameter of the space syntax theory. It represents how closely a small-scale space in a certain radius is connected to other space [4]. High integration means better centrality, accessibility, and integration, and often represents an area that is more attractive to the gathering of people and traffic in the region. In a city, areas with high closeness correspond to urban cores at a certain scale, and such areas are the first consideration in urban studies and perception of the urban structure [6]. In sDNA, closeness is denoted by NQPDA(x) (Network Quantity Penalised for Distance) with the following operational formula: NQPDA(x)=∑y∈Rxp(y)d(x,y) NQPDA\left( x \right) = \sum\limits_{y \in Rx} {{p(y)} \over {d(x,y)}} of which, P(y) is the weight of node y within the search radius R. The weight is up to 1, P(y) ∈ [0,1], and d(x, y) refers to the shortest topological distance from node x to node y. The different radii of 800 m, 1500 m, 3000 m, and 5000 m are chosen to analyse the closeness of the old district, and the urban layout is studied as per the different radii.

Fig. 4

Fig. 5

A distance of 800 metres is usually considered the radius of a walking trip in the city. Compared with other scales, the urban structure is more dispersed in the closeness analysis of an 800 m radius. The fishing net graph (aggregating the values onto a raster) can clarify the visualisation of the arithmetic results. In the east of the historic area of the city, there emerges a significant core of closeness, and the crossroad consisting of Dongfeng Road, Xiaobei Road, and Cangbian Road expands around; the second core consisting of Dezheng Road and Wenming Road expand horizontally in east and west. Both cores are connected through Cangbian Road, Dezhengbei Road, and Yuexiubei Road. Another significant core is nearby Sun Yat-sen University, forming a well-shaped structure with high closeness through Jiefangzhong Road, Huifuxi Road, Haizhuzhong Road, and Zhongshanliu Road, but no compact network structure is formed inside, nor a core of closeness for the active scope. The above characteristics of the spatial layout can explain the integrated urban core of the old district within the walking distance. In terms of spatial function, the above area is the gathering area of the community, and the spatial pattern is suitable for activities within the community [8].

A travel distance of 1500m generally indicates that people's travel mode would change and they prefer to bicycle or public transportation such as subway and bus. So, the core of closeness in the old district begins to gather towards the historical district, and the two cores on the 800m radius gradually merge to expand around the core of Beijing Road and Zhongshan Road, while the advantage of the high closeness in the south bank decreases. In terms of spatial functions, business offices have emerged along these axes, such as Pedestrian Street along Beijing Road, Commercial Centre at Dama Station, and Yinhu Building, etc., which coincides with the results of space syntax closeness on a larger radius.

When for a radius of 3000m and 5000m, the core of closeness centres on the historic district to further assemble and expand with a significant increase of the values, and the peak emerges at the banks of the Pearl River. This indicates that areas along the river, at a large scale, are the ‘golden area’ of the old district. The fishing net graph shows that the core of closeness at a radius of 5000m assembles more than that at a radius of 3000m. Both have a similar position with the historic district as the core of closeness. Detailed results show that the horizontal axis with the high closeness of the historic district includes Dongfeng Road, Zhongshan Road, Huifuxi Road, and Wenming Road; the vertical axis with the high closeness of the historic district includes Haizhuzhong Road, Liurong Road - Mishi Road, Jiefangzhong Road, Jixiang Road, Beijing Road, Wende Road, and Dezheng Road. The criss-crossed roads constitute the framework of the whole historic district. This indicates the dominant position of the historic district in the old district. It is worth mentioning that among the above high-value axes are there many arterial streets since the Ming-Qing dynasty, such as Huifu Road, Wenming Road, and Dezheng Road. These roads were all recorded in the local chronicles and other historical materials, manifesting their deep effects in that time. Besides, short and thin road networks with low values are distributed in the grid of axes, indicating that the area is accessible from the outside but ‘closed’ from the inside for ‘strangers’. In this way, it attracts people and protects internal communication from being interrupted. The area has its own social logic: easy to reach from the outside, but with its own internal network, which results from a long historical evolution.

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Fig. 10

Fig. 11

Fig. 12

The radius of n is the analysis of the over closeness. Here, no obvious core of closeness within the old district emerges, and the area along the river is more clustered than other areas, but no core is formed either. Globally, the linear pattern is significant, and the clustered core centred on the historic district becomes a long well-shaped axis extending to the whole area. In the front ground network of overall closeness (road network whose values are top 10%), this pattern is more significant. Unlike the regional closeness, the overall closeness has a larger scope of axis extending to east and west while the extension in south and north is a little fragmented. Yuexiu Garden – People Park – Haizhu Square – Jiangnanxi is the vertical long axis, and Dongfeng Road, Zhongshan Road, Huifu Road – Donghua Road is the horizontal long axis. As per the values, the average of regional closeness and overall closeness at a radius of 5000m is 14.011 and 21.873, respectively, with a great improvement. This indicates that the grid structure, at the global scale, replaces the urban structure with the historic district as the core, and the structure of the urban road network is better developed. However, the foreground network reveals that the grid structure in the eastern region and the southern bank of the Pearl River is less developed and no high-value closeness axes have emerged.

Fig. 13

Fig. 14

Intelligibility (comprehensibility) is another important index of space syntax theory for the interpretation of large-scale space configurations, and users of characterisation build up the whole space cognition by observing the local area. Intelligibility is defined as the degree of fitting of the local and the global analyses results, i.e., the degree of linear correlation between the two [3]. Results show that the value of R2 increases from 800m to 5000m as the radius increases, and is 0.20, 0.36, 0.47, and 0.53, respectively. If the value of R2 on the 5000m radius is greater than 0.5, i.e., significantly correlated, then the city has a higher level of comprehensibility on the public transportation to vehicular scale. Therefore, it can be inferred that the urban operation speed of the old district matches the level of urban development on the 5000m and larger radius scales.

Fig. 15

Betweenness(TPBt) is another important parameter provided by sDNA. It is highly correlated with the total depth of space and indicates the probability that a spatial node is passed. Higher TPBt represents the higher probability that the road network will be passed through and accordingly carries a higher calming capacity of the traffic [6]. Unlike closeness which emphasises reachability, TPBt represents passability and is calculated as: OD(y,z,x)={1, lies in the shortest path from y to z12,x≡y≢z12,x≢y≡z13,x≡y≡z0,others OD\left( {y,z,x} \right) = \left\{ {\matrix{ {1,} \hfill & {\;lies\;in\;the\;shortest\;path\;from\;y\;to\;z} \hfill \cr {{1 \over 2},} \hfill & {x \equiv y \ne z} \hfill \cr {{1 \over 2},} \hfill & {x \ne y \equiv z} \hfill \cr {{1 \over 3},} \hfill & {x \equiv y \equiv z} \hfill \cr {0,} \hfill & {others} \hfill \cr } } \right. TPBt(x)=∑y∈N∑z∈RyOD(y,z,x)P(z)Links(y) TPBt\left( x \right) = \sum\limits_{y \in N} \sum\limits_{z \in Ry} OD(y,z,x){{P(z)} \over {Links(y)}} of which, od(y, z, x) is the total number of nodes passing through node x within the search radius R, Links(y) is the total number of nodes within the search radius R for each node y, and P(z) is the weight of node z. The urban pattern is further interpreted by means of the above-mentioned radius.

Fig. 16

The figure shows that TPBt at a radius of 800m and 1500m does not show a coherent long axis, and it is basically irregular, mostly fine line segments. Therefore, it can be assumed that the old district has poor traffic circulation ability at the radius scale of 800m to 1500m. The shaped axis only emerges at a large radius scale of 3000m to 5000m. In terms of the foreground network, the network structure of TPBt is most obvious at the global scale, supporting almost the whole area of the old district. From the 3000m radius to the global scale, the results of TPBt analysis become more and more clustered, with the highest values located along the river areas. The horizontal axis includes Yanjiang Road, Zhongshan Road, and Dongfeng Road, and the vertical axis includes Jiefangzhong Road, Jiangnan Avenue, Donghu Road, Haiyin Bridge, and Dongxiao Road. But this structure is not balanced and perfect, and fractures emerge along the axis from the fringe areas to the central area; also, the axis along the connection of the old district and the new district also has fractures, indicating that the road access from outside to outside and from inside to inside are both relatively good, but the access from outside to inside is relatively poor. In addition, there may be some problems in fusion between the old and the new district.

Fig. 17

Fig. 18

Fig. 19

Fig. 20

Fig. 21

Fig. 22

Fig. 23

Fig. 24

In line with the above illustration, the fitting results also show that the radius below 3000m is almost uncorrelated with the global, while the 3000m radius and 5000m radius are highly correlated locally and globally, and TPBt results at the 5000m radius are even almost identical to the global results (R2 from 800m to 5000m is 0.11, 0.35, 0.74, and 0.91, respectively). As seen from the whole old district, the radius that best fits the urban TPBt is from 3000m to 5000m. Among the four radii, 3000m is a node for whether the passing structure appears, indicating that the scale suitable for the operation of the existing urban structure in the old district is no less than 3000m, while the fitting degree of 0.91 indicates that the limit of the operation scale is roughly around 5000m. Based on the basic perception of urban efficiency, this scale represents that the urban flow is also at a high level, and the urban structure has been elevated to the benchmark of a fast road network, while the high-value axis of TPBt has many overlaps with the above TPBt axis. Singly seen from the urban operation speed, we should consider how to match the high-speed urban movement with the low-speed activities of people, and how the urban flow of people and vehicles is compatible. Seen from the practical situations, the congestion problems that occur from time to time within the old district are consistent with the arithmetic results [7].

Fig. 25

Geostatistical analysis has many advantages over ArcGIS spatial analysis though they are often intertwined. Geostatistical analysis, often based on a variety of statistical inference methods, analyses the spatial pattern of ground objects, and the good statistical characteristics of the space syntax allow researchers to carry out various geostatistical analyses on the results of their calculations [10]. The results of space syntax of the old district – closeness and TPBt – have been described above, and preliminary analysis has been carried out based on the actual situations of the old district. The following will use the statistical tools of ArcGIS to further analyse the results and discuss the deeper urban pattern.

Standard Deviational Ellipse is often used to identify the central and directional trends of spatial elements and to characterise the core and direction of distribution of spatial elements. If the Basic spatial pattern of the elements is a spatial normal distribution (scattered from the centre to the periphery), a standard deviational ellipse surface will contain about 68% of the elements in the cluster. The formula is as follows: σ1,2=((∑i=1nx˜i2+∑i=1ny˜i2)±(∑i=1nx˜i2−∑i=1ny˜i2)2+4(∑i=1nx˜iy˜i)22n)12 {\sigma _{1,2}} = {\left( {{{\left( {\sum\nolimits_{i = 1}^n {\tilde x_i^2} + \sum\nolimits_{i = 1}^n {\tilde y_i^2} } \right) \pm \sqrt {{{\left( {\sum\nolimits_{i = 1}^n {\tilde x_i^2} - \sum\nolimits_{i = 1}^n {\tilde y_i^2} } \right)}^2} + 4{{\left( {\sum\nolimits_{i = 1}^n {{{\tilde x}_i}{{\tilde y}_i}} } \right)}^2}} } \over {2n}}} \right)^{{1 \over 2}}} of which, n is the total number of elements, x and y are the coordinates of element i, and {x¯,y¯} \{ \bar x,\bar y\} denotes the mean centre of the elements. The results of the operations of overall closeness and overall TPBt are imported into the calculation, and the results are as follows:

Fig. 26

Results show that the standard deviational ellipses of closeness and TPBt have a high contact ratio, which is consistent with the above analysis, and characteristics of the directional distribution of both are more obvious, roughly in east-west, slightly inclined along the northwest-southeast direction. Both ellipses include the entire range of the historic urban area, in which characteristics of the pattern of the old urban road network in TPBt are more clustered. This indicates that the historic district and its east areas are the reachability centre and high-value core of passability within the whole area of the old district. This once again verifies the profound effect of the historic district on the urban pattern, and also shows that the construction of eastern areas from the historic district has good integration with the historic district after the founding of New China. But the standard deviational ellipse with a high contact ratio still suggests that solving the contradiction between passability and reachability in terms of the crow and vehicle flow is still a consideration.

The above description summarises the spatial pattern of the old district. Further analysis of the internal spatial pattern requires more accurate identification of spatial heterogeneity, which requires verification of spatial autocorrelation, i.e., spatial dependence, which refers to the potential interdependence between observations of spatial elements in the same distribution area. Moran's I coefficient is an important tool to measure spatial autocorrelation, and its formula is as follows: I=nS0∑i=1n∑j=1nwi,jzizj∑i=1nzi2 I = {n \over {{S_0}}}{{\sum\nolimits_{i = 1}^n {\sum\nolimits_{j = 1}^n {{w_{i,j}}{z_i}{z_j}} } } \over {\sum\nolimits_{i = 1}^n {z_i^2} }} of which, z_i is the deviation of the property and the average (x_i − barX) of element i, w_i,j is the spatial weight of elements i and j, n is the total number of elements, and S₀ is the cluster of all the spatial weights. With overall closeness and overall TPBt imported, the inverse-distance matrix is used as the spatial weight. p-value of both is 0, indicating that the spatial data is statistically significant, and the spatial autocorrelation of overall closeness, from the perspective of the z value, is higher. This indicates that both have significant high/low clustering in space.

Fig. 27

Fig. 28

With the existence of spatial autocorrelation clarified, we hope to further analyse the clustering pattern. Local Moran's I is an important measurement for spatial clustering and spatial outliers. Local Moran's I values, z scores, pseudo p values, and codes indicate the type of clustering for each statistically significant element. Z scores and pseudo p values indicate the statistical significance of the calculated index. The formula is as follows: Ii=xi−X¯Si2∑j=1,j≠inwi,j(xj−X¯) {I_i} = {{{x_i} - \bar X} \over {S_i^2}}\sum\limits_{j = 1,j \ne i}^n {w_{i,j}}\left( {{x_j} - \bar X} \right) of which, x_i is property of element i, X¯ \bar X is the average of the corresponding property, w_i,j is the spatial weight of elements i and j, and n is the total number of elements. Similarly, reverse distance matrix and standardisation are set to determine the overall closeness and the overall TPBt, with the obtained results as follows.

Local Moran's I can reveal four patterns of HH, HL, LH, and LL, representing the local high values around the high values, the low values around the high values...and so on. It can be seen that closeness and TPBt are two different patterns. Analysis of local Moran's I of closeness has significant structure extending from the centre to outside; HH pattern mainly locates within the historic district, and has a high contact ratio; LH pattern, although has spatial outliers, is surrounded by high values, thus providing a guarantee to the internal social communication of areas of high reachability, that is, the “social logic” of the historic district mentioned above. At the same time, the spatial difference within the historic district is small, and HH and HL are complementary. With a long time of natural evolution and modern construction, the historic district has formed a good layout. HL pattern mainly emerges around the marginal area, indicating that the road network in the non-centre area of the old district has a certain centre, but is poorly integrated with the surrounding areas and fails to improve its spatial pattern. Analysis of local Moran's I of overall TPBt is more fragmented with inconspicuous clustering and rules. The original high-value axis of TPBt fails to maintain the continuous HH pattern and fractures in many places emerge. In addition, the LL pattern and the road network structure without significant clustering are scattered in the whole area, and the spatial distribution is more random with many spatial outliers. The originally obvious network of high-value TPBt may be interrupted by too many nodes around, thus losing characteristics of the spatial clustering. Optimisation of the TPBt pattern of the old district needs more consideration, and this paper will not discuss it for the time being.

Fig. 29

Fig. 30

Fig. 31

The historic buildings around Yuexiu Mountain, People Park and Haizhu Bridge have a high level of clustering, which can improve the accessibility of the whole axis by adding urban greenbelt and reducing nodes that has weak functions. Based on the features of the historic buildings, the natural scene in the north – Yuexiu Mountain, and the revolutionary sites – Ye Jianying's uprising site, Guangzhou Uprising Memorial Hall, and the historic buildings in the south – Haizhu Bridge, and Zhaoqing Hall can be integrated. With the historic buildings along the river as supplementary, the cultural axis of the old district can be created, echoing with the middle axis of the new district, and enriching the cultural deposits of the modern city.

According to Bill Hiller's research, traditional communities have a different ‘spatial DNA’ from modern communities, and the spatial grouping is a reflection of the social connection. For historic buildings, fenced-in preservation can lead to a gradual loss of vitality, and integration into the surrounding social ecology is necessary to maintain such vitality [1]. These communities are based on neighbourhoods, and their form of social association is different from that of modern cities, and it is their unique social logic that highlights the ‘urban fabric’. As mentioned above, the spatial pattern of the historic district has its own unique ‘social logic’: the exterior is easily accessible while the interior activities are not easily disturbed. Based on this feature, the urban fabric can be protected through the optimisation of the pattern of the public service, so that visitors and residents of the community do not interfere with each other. For example, community activity centrer, service centrer, small commercial sites, community health services and education services are provided within the walking distance of the community residents to fulfil their living needs, and meanwhile, space is provided to support their social behaviour. Large-volume urban complexes such as commercial centres should be avoided. This kind of urban function tends to attract pedestrians and vehicles, exacerbating the contradiction between aggregation and reachability. This strategy can be adopted around Guangxiao Temple, Beijing Road, Zhongshan Road, and Wende Road.

This strategy is applicable to historic buildings located outside the historic district or the high-closeness core, such as Xiguandawu and Yongqingfang in the western part of the old district [14]. According to the analysis of spatial structure, if it is not in the core of the city, the road adjustment plan can be prepared according to the field research to improve the accessibility of the area. These areas are located in regions with low-value closeness and TPBt in the old district and are in a weak position in the spatial network of the whole area, but local Moran's I reveal that spatial heterogeneity exists around, i.e. HL pattern. This indicates that it is not that difficult to improve the regional spatial pattern, and reasonable planning, as well as adjustment of the road network structure, can topologically reduce the degree of isolation and improve the spatial pattern; it is also integrated into the spatial network in a manner of ‘darning’.