Are patterns of vacant lots random? Evidence from empirical spatiotemporal analysis in Chiba prefecture,east of Tokyo

Abstract

According to the Japanese government, vacant lots are randomly generated and accumulated (without being rebuilt after demolition) in the process of increasing vacant lots, a phenomenon called urban perforation. Urban perforation in urban areas may trigger a high degree of inefficiency in public infrastructure management. However, this observation lacks theoretical and empirical foundations, a lacuna to which this paper will focus on. Consequently, our research objectives are to confirm: (1) whether or not vacant lots are randomly generated and (2) whether or not vacant lots are randomly accumulated as a result of random generation. The methodology includes a consistent and bottom-up approach to delineate urban areas (alongside statistical spatial analysis). Through theoretical and empirical analyses in Chiba Prefecture (situated in the eastern part of the Tokyo metropolitan region), we find that the random generation of vacant lots does not tend to continue in the same urban areas. Rather, in most urban areas, this process is a temporary phenomenon. Subsequently, phase transition generally shifts from random to clustered generation or vice versa. Nevertheless, once vacant lots are randomly accumulated in an urban area, this process tends to continue. The contributions of this article are not only to provide important spatiotemporal findings regarding the generation and accumulation patterns of vacant lots, but also to discuss how to apply policy for urban perforation where phenomena are significantly pronounced.

Keywords

Shrinkage perforation vacant demolition lot

Introduction

The process of population loss and its related problems, collectively termed urban shrinkage, constitutes one of the most pertinent issues in developed countries (Haase et al., 2014, 2017; Hollander, 2011; Hollander et al., 2017; Martinez-Fernandez et al., 2012). The places of urban shrinkage are named shrinking cities and face challenges of ever-increasing demolitions of buildings and vacancies of building lots (Frazier et al., 2013; Haase et al., 2017; Hackworth, 2016; Hollander, 2010; Schetke and Haase, 2008). Since demolition and vacancy have various structural impacts such as a decrease of the quality of the built environment (Frazier et al., 2013; Hollander et al., 2017), policies related to the management of areas subject to urban shrinkage have been discussed and proposed (e.g. Asami, 2014, 2017; Haase et al., 2014, 2017; Hollander, 2010, 2011; Hospers, 2014; Kubo and Yui, 2020; Radzimski, 2016). Although shrinking cities and an increase in vacant lots are not always related, the non-replacement of demolished buildings often remains a reliable marker for detecting areas undergoing urban shrinkage (Lee and Newman, 2017; Sakamoto et al., 2017; Usui, 2019). An increase in vacant lots in urban areas creates perforations of urban areas, which may trigger various problems, such as a high degree of inefficiency in terms of public infrastructure and services (maintenance of road networks, sewerage system, refuse collection, public transportation, etc.) (Haase et al., 2012; Reis et al., 2016; Schwarz et al., 2010).

Japan, which is at the forefront of an increase in vacant lots in urban areas, is also characterized by a proportion of urban population that is still increasing (expressed as a percentage of the total population; 91.8% in 2020, projected to 92.7% in 2030¹). Anticipating and planning a sustainable shrinkage in some spaces and a successful regeneration in others has become a priority for the local and national authorities. In this respect, the Japanese government is trying to promote a compact city policy to mitigate the problems and inefficiencies related to shrinkage. In this respect, the Act on Special Measures Concerning Urban Reconstruction was enforced in June 2002 to guide urban development. It was then amended in May 2014 to constitute proper location plans. Following the enforcement of the amendment in August of the same year, many municipalities started to institute their proper location plans to redevelop and manage their existing urban areas. As discussed in Asami (2017), in such plans, habitation promotion areas, also known as built clusters, are delineated within the present urban areas as future urban areas by considering existing building locations.

However, an increase in vacant lots does not necessarily proceed from the periphery of existing urban areas. According to a report issued in August 2017 by the Division of City Planning of the Panel on Infrastructure Development of the Ministry of Land, Infrastructure, Transport and Tourism² (MLIT), vacant lots tend to be randomly generated and accumulated in the process of increasing vacant lots without being reconstruction after demolition, a phenomenon called urban perforation. The closest description in the scientific literature is the one provided by Schwarz et al. (2010) that, without considering the notion of randomness, highlights the perforation of low-density counter settlements following decay and destruction of buildings. Hereafter, urban perforation is considered as the random generation and accumulation of vacant lots only. In this regard, urban perforation is not considered in the delineation process of present and future urban areas.

While dealing with urban perforation issues, at least three topics should be formerly addressed: (1) definition of randomness; (2) observation window’s size and shape; and (3) spatial data of vacant lots. First, the concept of ‘randomness’ in the MLIT report is vague and not quantitatively defined, rendering it difficult to understand the linkages with random generation and accumulation of vacant lots. However, from the point of view of statistical spatial analysis, randomness is in fact clearly defined. Assuming that the locations of vacant lots are modelled as a set of points, if these points are uniformly and independently distributed on a two-dimensional plane, then they are randomly distributed. This is called completely spatial randomness (CSR). CSR represents a baseline hypothesis against which to assess whether observed patterns are regular, clustered or random (Bailey and Gatrell, 1995; Cressie, 1993; Illian et al., 2008; Okabe and Sugihara, 2012). Several methods for testing CSR have been proposed, including the quadrat test, the nearest neighbour distance (NND) method and the K-function method. Therefore, the concept of randomness needs to be clearly defined based on CSR. Second, in testing for CSR, we encounter boundary effects. In the literature, predetermined census units and uniform grid cells have been adopted as standard observation windows for testing CSR. However, test results for CSR are contingent on the selected observation windows (Bailey and Gatrell, 1995; Frazier et al., 2013; Getis and Boots, 1978, Illian et al., 2008). Moreover, the pattern of vacant lots can be regarded as the conditional pattern, given that it is generated from the lot on which a building is located (Illian et al., 2008). In this case, it is wise to adapt the observation window’s size and shape to the pattern of building locations in an appropriate manner in order to avoid a potentially large, empty boundary zone around the pattern (Illian et al., 2008). This adaptation enables us to compare with more consistency present and future urban areas considering building demolitions. Third, spatial data regarding vacant lots are unavailable, as few government agencies specifically collect such information (Hollander, 2018). Furthermore, spatial data on the shape of building lots per se are difficult to obtain (Bradley and Behnisch, 2018; Gao and Asami, 2005; Usui, 2018). Thus, in the literature CSR regarding the generation and accumulation patterns of vacant lots has yet to be tested. As a result, an alternative method for specifying the generation and accumulation of vacant lots when their spatial data are unavailable is required.

Our substantive motivation is to provide theoretical and empirical foundations related to urban perforation based not on predetermined census units (a top-down approach), but on fine-scale fixed spatial elements such as buildings (a bottom-up approach) (Caruso et al., 2017; Masucci et al., 2015; Usui, 2019). Hence, our research objectives are to confirm: (1) whether or not vacant lots are randomly generated and (2) whether or not vacant lots are randomly accumulated as a result of random generation based on a consistent and bottom-up approach delineating urban areas alongside statistical spatial analysis. While the generation of vacant lots can be regarded as reflecting urban perforation between two time points from a dynamic point of view, the accumulation of vacant lots may be deemed the result of random generations of vacant lots from a static perspective of urban perforation. Answering these questions with a spatiotemporal statistical perspective shall improve the global understanding of the urban perforation phenomenon. Furthermore, by adopting random generation and accumulation of vacant lots as the null hypothesis, we can specify the urban areas where urban perforation phenomena are significantly pronounced. Hereafter, clustered generation and accumulation of vacant lots are not considered as urban perforation. The policy for urban perforation can be applied to places where urban perforation phenomena, a random process as suggested by the MLIT report, are particularly pronounced.

In the literature, factors related to the generation of vacant lots and housing have been investigated in terms of the concentration of elderly population in old suburban housing estates (e.g. Nordvik and Gulbrandsen, 2009; Yui et al., 2017), massive out-migration from post-socialist and old industrial cities (e.g. Haase et al., 2017; Radzimski, 2016) and the total reality of specific and generic factors (e.g. Kubo and Yui, 2020). In Japanese cities, given that increases in housing vacancies have been regarded as coinciding with urban shrinkage, such processes are related to several factors, such as a declining economic base, population loss or ageing, institutional reasons such as taxation systems on real property or a combination of these factors (Kubo and Yui, 2020). As for the spatial patterns of vacant housing, Akiyama et al. (2018) have developed a method for estimating the spatial distribution of vacant housing through rasterisation and usage of spatial data and sample field surveys. Furthermore, Morckel (2014) has investigated whether the probability of vacant housing is influenced by spatial factors. Using indicators of spatial autocorrelation (local Moran’s I statistic), this author found that neighbourhoods with high housing vacancy rates are located near to other neighbourhoods with the same issue, whereas neighbourhoods with low housing vacancy rates are located near to other neighbourhoods with low housing vacancy rates, a phenomenon called neighbourhood effects. Molloy (2016) has examined the distribution of long-term vacant housing across neighbourhoods in the United States by using data on the duration of vacancy, finding that most metropolitan areas have at least a few neighbourhoods with a high long-term vacancy rate. In addition, Immergluck (2016) has demonstrated that neighbourhoods with persistently high levels of long-term housing vacancy tend to be high-poverty, minority-dominated neighbourhoods.

As regards to vacant lots, Sakamoto et al. (2017) have investigated the spatial patterns of vacant lots in residential areas of Tottori city in Japan, identifying a narrow front road width as the key characteristic of neighbourhoods where high rates of vacant lots tend to exist. Sakamoto and Yokohari (2016) have also identified the dynamic characteristics of vacant housing and lots in residential areas of Utsunomiya city, noting that (1) narrow front road width is the key factor behind the emergence of both vacant housing and lots and (2) distance from the core railway station is influential regarding the former but not the latter. Yamada et al. (2016) have also examined the issue of vacant lots in outer suburban cities in the Tokyo metropolitan region, finding that areas near to stations present greater vacant lot rates. Sakai (2014) has analysed the causes of the increase or decrease in vacant lots in suburban areas of Tokyo and Osaka and suggests that good accessibility to urban centres by rail tends to help reduce the phenomenon. Ohsawa et al. (2009) have examined patterns of the emergence and persistence of vacant lots in residential areas of Kashiwa city, a suburban city of Tokyo, discovering that both are related to the age of a residential area.

In the literature, spatiotemporal analysis of the generation and accumulation of vacant lots is widely perceived as an essential area of future research (e.g. Frazier et al., 2013; Morckel, 2014; Reis et al., 2016; Sakamoto and Yokohari, 2016). According to Hollander et al. (2017), who tested whether building demolitions are clustered in row house neighbourhoods in Baltimore, demolition is treated as a proxy for vacant lots since demolished buildings are also popular measures of shrinkage and easily geocoded. By comparing building footprints (polygons that represent building shapes on a two-dimensional plane) in 1972 and 2010, Hollander et al. (2017) found that at the parcel level, demolitions were no more clustered than expected, while the opposite was true at the block group level. Moreover, Lee and Newman (2017) have developed a land transformation model to forecast vacant lots through rasterisation at a city scale. However, both articles adopted predetermined basic spatial units as the observation windows and their analytical scale ranges from a neighbourhood to a city scale. Thus, in terms of both spatiotemporal analysis and non-predetermined basic spatial units, whether vacant lots are randomly generated and accumulated has yet to be tested. If the analytical scale is predetermined, we encounter boundary effects because the observation window (the pattern of building locations) does not always correspond to a neighbourhood and a city. Rather, in order to avoid boundary effects, large analytical scale including a city (e.g. a prefecture) should be adopted. To this end, in this article, we propose a new method to test: (1) whether or not vacant lots are randomly generated and (2) whether or not vacant lots are randomly accumulated as a result of the random generation of vacant lots, without predetermining basic spatial units such as census units or grid cells. By comparing building footprints in 2003, 2008, 2013 and 2016, we attempt to confirm whether building demolitions are randomly distributed in Chiba Prefecture in the eastern part of the Tokyo metropolitan region.

Figure 1 shows the densely inhabited districts (DID) and railway networks in Chiba prefecture as of 2010. DID meet the following three conditions: (1) population density in each census unit is equal to or greater than 40 persons per hectare; (2) census units within administrative boundaries are adjacent to one another; and (3) population in DID is equal to or greater than 5000 persons. Since the late 20th century, Chiba Prefecture has experienced rapid population growth as a result of the rising commuting and schooling population with the central business districts of Tokyo. In particular, in the north-west region of Chiba Prefecture, land uses at the waterfront of Tokyo Bay (specifically agricultural and forest areas) were altered for commercial and residential purposes, with a large amount of flats and detached houses developed by both private companies and the Japan Housing Corporation (a government organization) (Sorensen, 2001). However, since 2011, Chiba Prefecture has entered an era of depopulation. In the near future, unprecedented rapid depopulation will accompany urban perforation phenomena. Hence, an empirical study in Chiba is important for providing both developed and developing countries with information on how vacant lots are generated and accumulated. The answers to our research questions will contribute to a rationale for policy makers for the random generation and accumulation of vacant lots.

Figure 1.

DID in Chiba prefecture.

This article is organized as follows. In the subsequent section, the methodology for specifying the locations of vacant lots and the notations and spatial data used in this research is explained. In the third section, the question of whether vacant lots are randomly generated and accumulated or not is answered based on a statistical approach. In the fourth section, the results and the factors mentioned in the third section are investigated in greater details and discussed with references to the wider literature. In the final section, concluding remarks and suggestions for future research are provided.

Methodology for specifying vacant lots

In this section, we explain the methodology for specifying building demolitions and the spatial data used to do so. For the spatial data of buildings, we acquired Residential Maps by Zenrin, Co., Ltd. The Residential Maps used in our analysis were released in 2003, 2008, 2013 and 2016 (approximately every five years). Residential Maps consist of building shape data with and without walls (e.g. car parks, material storage sites or greenhouses). Given that building shapes without walls can be regarded as either non-residential, non-commercial or non-industrial use, they are omitted hereafter.

As mentioned in the previous section, spatial data regarding vacant lots are difficult to obtain in Japan. Therefore, we propose a method for specifying demolished buildings and deem their lots vacant by comparing the spatial data of buildings in any two dates. According to the Building Standard Law in Japan (Japanese building codes), a building must be basically constructed in a building lot in order for building sizes and shapes to be appropriately regulated, termed the one lot for one building rule (as discussed in Usui, 2018). This rule is assumed to have both advantages and disadvantages. Indeed, the rule does not always hold, as it is not applied when two buildings’ use on a lot are inseparably related to one another. Nevertheless, assuming that this rule operates enables us not only to overcome data limitations regarding building lots per se, but also to specify vacant lots through the following procedure.

First, for a concise and consistent explanation, the following terms and notations will be introduced. As illustrated in Table 1, a building and its footprint (a polygon that represents a building shape on a two-dimensional plane) in year k are denoted by B _i (k) and A(B _i (k)), respectively. The lot of B _i (k) is denoted by L(B _i (k)). Second, for any k and k + 5, if A(B _i (k + 5)) is empty, then B _i (k) is demolished (called a demolished building, NB _i (k + 5)) and L(B _i (k + 5)) is called a vacant lot, L(NB _i (k + 5)). Thus, the set of buildings in year k, {B _i (k)}, is decomposed into {B _i (k + 5)} and {NB _i (k + 5)} in year k + 5. Moreover, for any k + 5 and k + 10, if A(NB _i (k + 10)) is not empty, then L(NB _i (k + 5)) is not vacant; otherwise, L(NB _i (k + 5)) remains vacant. Third, by regarding the gravity location of NB _i (k + 5) as the representative location of L(NB _i (k + 5)), denoted by G(NB _i (k + 5)), we can statistically test whether or not vacant lots are randomly generated during any two consecutive dates using conventional methods for testing CSR.

Table 1.

Spatial data processing for specifying vacant lots.

Notes: Black polygons and polygons drawn by dashed lines represent A(B _i (k)) and A(NB _i (k)), respectively. Black points indicate G(NB _i (k)).

Table 2 presents the transition patterns of the state of buildings and those that have been demolished between two time points. The transition patterns can be categorised into four categories. Among the four categories, the following two transitions promote urban shrinkage phenomena: demolished and vacant. Rebuilt transition is expected to prevent a region from urban perforation phenomena. Therefore, whether an urban perforation phenomenon proceeds or not is largely contingent on whether demolished or vacant transition is more prominent than rebuilt transition.

Table 2.

Transition patterns of the state of buildings and demolished counterparts.

	B _i (k + 5)	NB _i (k + 5)
B _i (k)	Remained	Demolished
NB _i (k)	Rebuilt	Vacant

In the previous studies, although the definition of long-term vacant lots was not discussed, how long-term housing vacancy should be defined has been discussed. Immergluck (2016) has characterized long-term vacancy as vacancy for a period of six months or longer. Molloy (2016) has defined long-term vacancy as housing units that are in roughly the top quartile of the distribution of length of vacancy: 6 months or longer for units for rent; 12 months or longer for units for sale; and 5 years or longer for other vacant units. By adopting these findings as the substitution of the long-term vacant lots, consequently, it is not possible to distinguish long-term vacancy from short-term vacancy among the generation of vacant lot layers between k – 5 and k. Although the generation of vacant lot layers between k – 5 and k includes short-term vacancies of under 12 months, they will be regarded as long-term vacancies instead.

Figure S1 shows the generation patterns of vacant lots in each two-year interval and their accumulation patterns in each year within Chiba Prefecture (see online supplementary material). The accumulation pattern of vacant lots in 2013 (shown as (c) in Figure S1) is the overlay of the generation of vacant lots from 2003 to 2008 (shown as (a) in Figure S1) and from 2008 to 2013 (shown as (b) in Figure S1). Rebuilt transition from 2008 to 2013 is taken into consideration. In the same way, the accumulation pattern of vacant lots in 2016 (shown as (e) in Figure S1) is the overlay of the accumulation pattern of vacant lots in 2013 (shown as (c) in Figure S1) and the generation pattern of vacant lots from 2013 to 2016 (shown as (d) in Figure S1). Rebuilt transition from 2008 to 2016 is also taken into consideration.

We can see that: (1) at the large spatial scale of Chiba Prefecture, vacant lots are more likely to generate and accumulate in the north-west region of Chiba Prefecture than in other areas. In the north-west, land use in what were previously agricultural and forest areas has been modified to commercial and residential use, with a large amount of detached and flat housing developed; and (2) at the local scale, vacant lots extensively generate near to railway stations and networks.

By focusing on the spatial pattern of G(NB _i (k)), we can statistically test whether or not vacant lots are randomly generated during any two time points using the conventional methods mentioned above for CSR testing.

Generation and accumulation patterns of vacant lots

Method of testing for CSR

As noted in the Introduction, several methods for CSR testing have been proposed: the quadrat test, the NND method and the K-function method. The quadrat test represents a simple means of testing for CSR using quadrat counts. However, one limitation of this method is that it takes no account of the relative position of quadrats or the relative position of events within a quadrat. The K-function method provides a more effective summary of spatial dependence over a wide range of scales than measures merely based on NND. Nevertheless, a general problem of this method is that it takes a considerable amount of time to construct the upper and lower envelopes for the significance level on the basis of simulation per study area under the hypothesis of CSR (Bailey and Gatrell, 1995). Given that the number of study areas is approximately 3000, time inevitably constitutes an issue. Among these methods, although the NND test for CSR uses distances only to the closest locations and ignores information on larger scales of pattern, this method is easy to implement and makes use of more precise locational information than the quadrat test (Bailey and Gatrell, 1995; Getis and Boots, 1978). Moreover, as our main objective is to test whether vacant lots are randomly distributed rather than clustered, the NND test can be deemed as the appropriate method to start testing for CSR.

Observation windows of testing for CSR regarding the generation and accumulation patterns of vacant lots

In applying the NND method to test (1) whether vacant lots are randomly generated or not and (2) whether or not vacant lots are randomly accumulated as the result of random generation of vacant lots, we face the boundary effect. As mentioned in the Introduction, the method’s results critically depend on how the observation window is selected (Bailey and Gatrell, 1995; Frazier et al., 2013; Getis and Boots, 1978, Illian et al., 2008). Moreover, the pattern of vacant lots can be regarded as the conditional pattern, being generated from the lot on which the building in question is located (Illian et al., 2008). As shown in Figure S2 (see online supplementary material), the pattern of buildings in Chiba Prefecture is clearly non-uniform and exhibits numerous urban built clusters (the way in which urban built clusters are to be delineated will be explained in the following section). In this case, it is appropriate to adapt the window’s size and shape to the pattern of buildings in some appropriate manner in order to avoid a potentially large empty boundary zone around it (Illian et al., 2008). In order to delineate groups of buildings as urban built clusters, we will adopt the method proposed by Usui (2019), which focuses on the NND from a building centroid (see online supplementary material). Hereafter, we test the above two hypotheses in urban areas as of 2003 delineated by the method.

CSR tests for generation and accumulation patterns of vacant lots

In order to test for CSR, regarding the generation and accumulation patterns of vacant lots, we adopt the NND method proposed by Clark and Evans (1954). The NND from a representative location of a vacant lot, G(NB _i (k)), is denoted by x_j or simply x. Then, the test statistic is defined as the following

\bar{x} \equiv \frac{\sum_{j = 1}^{n} x_{j}}{n}

(1)

where n denotes the number of vacant lots in an urban area. Theoretically, if a set of representative locations of vacant lots, {G(NB _i (k))}, is independently and uniformly distributed in an urban area of an infinite two-dimensional plane, then the probability density function of x_j can be rigorously derived as follows

f (x) = 2 π ρ x e^{- 2 π ρ x^{2}}

(2)

where ρ = n/S denotes the location density of vacant lots in an urban area S. The average and variance of x are

\frac{1}{2 \sqrt{ρ}}

and

\frac{4 - π}{4 ρ π n}

, respectively. By normalizing

\bar{x}

to z-score: z

\equiv (\bar{x} - \frac{1}{2 \sqrt{ρ}}) / \sqrt{\frac{4 - π}{4 ρ π n}}

, z follows the normal distribution:

z \sim N (0, 1)

. At a 5% significance level, if |z| ranges from 0 to 1.96, then {G(NB _i (k))} is uniformly and independently distributed in an urban area; otherwise, {G(NB _i (k))} is either clustered or regularly distributed. Equation (2) can be derived by a Poisson distribution, which can be derived as a limiting form of a binomial distribution. In the literature, it is known that if n is more than 10, equation (2) can be adopted as the approximation of a binomial distribution (Getis and Boots, 1978; Koshizuka, 1985). Therefore, the Clark and Evans test statistic, z, of urban areas, whose size is equal to or more than 10, is computed.

Figure 2 shows the results of the NND test for CSR regarding the generation and accumulation pattern of vacant lots within urban areas of Chiba Prefecture, delineated by the aforementioned method proposed by Usui (2019). If z is less than –1.96 and more than 1.96, then urban areas are drawn in red and in blue, respectively; otherwise, urban areas are drawn in yellow. It can be stated that in urban areas drawn in yellow, vacant lots tend to be randomly generated and accumulated. It is found that: (1) the generation and accumulation of vacant lots tend to be clustered along railway networks in the north-west region of Chiba Prefecture and (2) in some urban areas, vacant lots are randomly generated and accumulated.

Table 3.

Comparison of the number of urban areas where vacant lots were randomly generated and accumulated continuously.

	Random generation	Random accumulation
From 2003 to 2013/from 2003 to 2008	212/923 = 0.23	737/923 = 0.80
From 2003 to 2016/from 2003 to 2013	32/212 = 0.15	677/737 = 0.92
From 2008 to 2016/from 2008 to 2013	54/575 = 0.09	1449/1632 = 0.89

Figure 2.

Value of z-score for CSR regarding the generation and accumulation patterns of vacant lots within urban areas of Chiba Prefecture (n ≥10). The number of these areas that have a clustered, random and regular distribution is provided. (a) Generation pattern from 2003 to 2008; (b) Generation pattern from 2008 to 2013; (c) Accumulation patterns in 2013; (d) Generation pattern from 2013 to 2016; (e) Accumulation patterns in 2016.

Table 3 presents the number of urban areas where vacant lots were randomly generated from 2003 to 2008 (923), from 2003 to 2013 (212), from 2003 to 2016 (32), from 2008 to 2013 (575) and from 2008 to 2016 (54). A comparison of the number of urban areas where vacant lots were randomly generated is shown as the proportion (e.g. the number of urban areas where vacant lots were randomly generated from 2003 to 2016 (32) and from 2003 to 2013 (212) is shown as 32/212). We can see that the proportion of continuous random generation of vacant lots ranges from 0.09 to 0.23. The left side of Figure S5 shows the urban areas where vacant lots were randomly generated continuously (see online supplementary material). Table 3 additionally presents the number of urban areas where vacant lots were randomly accumulated from 2003 to 2008 (923), from 2003 to 2013 (737), from 2003 to 2016 (677), from 2008 to 2013 (1632) and from 2008 to 2016 (1449). A comparison of the number of urban areas where vacant lots were randomly accumulated is shown as the proportion (e.g. the number of urban areas where vacant lots were randomly accumulated from 2003 to 2016 (677) and from 2003 to 2013 (737) is shown as 677/737). It can be seen that the proportion of continuous random accumulation of vacant lots ranges from 0.80 to 0.92. The right side of Figure S5 shows the urban areas where vacant lots were randomly accumulated continuously (see online supplementary material).

Therefore, it can be concluded that: (1) the random generation of vacant lots does not tend to continue in the same urban areas and (2) once vacant lots are randomly accumulated in an urban area, such random accumulation tends to continue. These are the answers to our research questions. These factors will be discussed in the following section.

Discussion

In this section, we first discuss the problem regarding the adoption of predetermined basic spatial units and the advantage of our methodology regarding this issue. Second, the results mentioned in the previous section are investigated in greater details and these factors are discussed based on findings in the literature. Third, the limitations regarding the methodology proposed in the previous section are debated. Finally, future policy for urban perforation is discussed.

In Japan, a small area typically of a few hundred people and tens to hundreds of hectares in size has been adopted as the basic spatial unit for applying urban planning policies (Asami, 2017). Figure 3 shows the patterns of buildings and locations of vacant lots around the capital of Chiba Prefecture in 2016. The left- and right-hand maps in Figure 3 are based on small areas and urban areas, respectively. It was found that urban areas do not always correspond to small areas. Moreover, within any small area, patterns of buildings are not uniform. Therefore, as long as small areas are adopted as the observation windows for testing CSR, we face a potentially large empty boundary zone around the patterns. This triggers the underestimation of ρ, which results in viewing patterns of vacant lots in small areas as clustered, otherwise those in urban areas (defined in the previous section) are random.

Figure 3.

Patterns of buildings and locations of vacant lots around the capital of Chiba Prefecture in 2016. Solid lines in the left-hand figure indicate the boundary of ‘small areas’. In the right-hand figure, values of the z-score for CSR regarding the accumulation patterns of vacant lots within urban areas (n ≥10) are shown in red (cluster), yellow (random) and blue (regular), respectively.

Given these results, urban perforation phenomena in urban areas and in non-urban areas should not be treated in the same way, because the former occur within the boundary of an urban area where the intervals of building centroids tend to be shorter than in the latter. In contrast, the latter occur in non-urban areas where the intervals of building centroids are generally longer and buildings are sparsely distributed. In terms of urban planning policy, the former is widely deemed the most appropriate urban perforation phenomenon in current urban areas with sufficient and efficient urban infrastructure such as road networks and sewage systems, while the latter is the result of urban sprawl, where built clusters of small size are randomly and sparsely distributed. Using our methodology, we can specify urban areas where urban perforation is significantly pronounced. In Figure 3 (right), the urban areas where vacant lots are randomly accumulated (yellow colour) are the places where urban perforation policies need to be applied.

As shown in Figure S5, there are only 32 urban areas in Chiba Prefecture where vacant lots were randomly generated continuously from 2003 to 2016. Moreover, there were only 21 urban areas where vacant lots were not only randomly generated but also randomly accumulated continuously from 2003 to 2016. Figure S6 displays one of the urban areas where vacant lots were randomly generated and accumulated continuously from 2003 to 2016. The location of this area is more than 1 km away from the nearest railway station and was developed as a place of detached housing. The other 20 areas generally share the same characteristics as this area. As mentioned in the previous section, the random generation of vacant lots does not continue in the same urban areas (except for the small number of urban areas). Rather, in most urban areas, the random generation of vacant lots is a temporary phenomenon. Before and afterwards, phase transition tends to occur from random to clustered generation or vice versa. On the other hand, once vacant lots are randomly accumulated in an urban area, their random accumulation tends to continue. Thus, urban perforation should be redefined as phenomena where vacant lots are randomly accumulated.

These factors will be discussed based on findings in the literature. In urban areas drawn in red, generating a vacant lot tends to lead to the demolition of the adjacent building during any two consecutive time points, resulting in an accumulation or clustering of vacant lots. On the other hand, in urban areas drawn in yellow, generating a vacant lot does not stimulate the demolition of the adjacent building during any two consecutive time points. One factor is represented by neighbourhood effects. As noted in the Introduction, neighbourhoods with high housing vacancy rates are located near to other neighbourhoods with high housing vacancy rates, while neighbourhoods with low housing vacancy rates are located near to other neighbourhoods with low housing vacancy rates (Morckel, 2014). Furthermore, narrow road width has been identified as one of the key factors behind the generation of vacant lots, because building construction is not permitted if the front road is narrower than four metres (Sakamoto and Yokohari, 2016; Sakamoto et al., 2017). In urban areas drawn in red, irrespective of distance from railway stations, road width tends to be narrow. Hence the generation and accumulation of vacant lots tends to be clustered along narrow road networks. On the other hand, urban areas are not marked by any distinct factors that lead to the generation of vacant lots, such as narrow road width. Rather, whether a vacant lot is generated is contingent on the specific site’s locational and socioeconomic factors, leading to the random generation and accumulation of vacant lots.

However, one limitation should be considered. In order to specify the urban areas where the generation and accumulation of vacant lots are clustered, we must investigate the second-order effects that result from the spatial correlation structure or the spatial dependence. In particular, the NND method cannot distinguish between two types of clustered patterns: a small number of large clusters and a large number of small clusters. The latter type can be regarded as a kind of urban perforation if small clusters are randomly generated and accumulated. In comparison, the K-function test for CSR can both differentiate between them and provide a more effective summary of spatial dependence over a wide range of scales. Nevertheless, as mentioned above, a general problem of this method is that simulation under the hypothesis of CSR takes a considerable amount of time.

Finally, policy for urban perforation should be discussed. First, appropriate policies need to be applied by considering the present urban physical conditions, such as the built environment of each urban area. If the building density is low, continuous urban perforation may result in inefficient urban infrastructure usage. In fact, in urban areas with a random accumulation of vacant lots – as shown in Figure S5(c) and (e) and Figure S6 – building density (the number of buildings per unit area) is lower than the criterion for identifying densely built-up areas (60 buildings per hectare) and average building density (gross) is approximately 30 buildings per hectare. In the report cited in the Introduction, the random accumulation of vacant lots prevents us from managing them proactively. Hence, community-based activities and administrative interventions are important to ensure a liveable residential environment and a continuous supply of vacant lots suitable for new housing development (Kubo and Yui, 2020; Sakamoto and Yokohari, 2016). In particular, in urban areas located far from railway stations and central business districts, it is implausible to expect market mechanisms to reduce vacant lots without administrative interventions. On the other hand, if the building density is higher than 60 buildings per hectare, such as in residential hyper-compact neighbourhoods (Perez et al., 2019), continuous urban perforation is to be welcomed because vacant lots play an important role in reducing risks such as the spread of fire. Nevertheless, urban areas with a high building density tend to correspond to those where the accumulation of vacant lots is not random but clustered. This is not urban perforation. If vacant lots are ordinarily accumulated in a region, they can be proactively managed through policies focusing, for example, on urban redevelopment projects. As mentioned above, given that vacant lots tend to be accumulated along narrow roads near railway stations, other policies such as urban redevelopment projects could be adopted. Furthermore, in order to apply appropriate policies, continuous monitoring of the generation and accumulation of vacant lots is crucial. The methodologies proposed in this article may provide the relevant monitoring tools.

Concluding remarks and future works

Our research objectives were to confirm: (1) whether or not vacant lots are randomly generated and (2) whether or not vacant lots are randomly accumulated as a result of random generation based on a consistent and bottom-up approach to delineate urban areas alongside statistical spatial analysis.

From the results of our theoretical and empirical analyses in Chiba Prefecture, we have found that the random generation of vacant lots rarely continues in the same urban areas. Rather, in most urban areas, the random generation of vacant lots represents a temporary phenomenon. Before and afterwards, phase transition tends to occur from random to clustered generation or vice versa. On the other hand, once vacant lots are randomly accumulated in urban areas, their random accumulation tends to continue. Thus, following the MLIT’s suggestion, urban perforation could be redefined as phenomena where vacant lots are randomly accumulated. Perforation policies shall be applied in these places only since the accumulation of vacant lots can be proactively managed through different strategies where accumulation is not random. Prior to this study, generation and accumulation patterns of vacant lots were not investigated separately, while the relationship between them was not studied in any detail. Therefore, this article contributes to existing literature by offering important spatiotemporal findings regarding the generation and accumulation patterns of vacant lots based on a bottom-up rather than a top-down approach of predetermined basic spatial census units.

Several points are suggested here for future research. First, the one lot for one building rule does not hold where two buildings’ use on a lot are inseparably related to one another. Moreover, as mentioned above, spatial data on the shape of building lots per se are difficult to obtain. Therefore, in the future, where feasible, the effect of assuming this rule on the result of the NND test for CSR regarding the generation and accumulation patterns of vacant lots needs to be investigated. Second, a further detailed investigation of the characteristics of these areas (e.g. road width, population change and socioeconomic neighbourhood characteristics) will provide urban planners with important implications. Given that statistical data regarding population and socioeconomic change are aggregated based on predetermined basic spatial units, some adjustments are necessary in order to apply these data to the urban areas defined in this article. Third, in order to specify the urban areas where the generation and accumulation of vacant lots are clustered, we must investigate the second-order effects that result from the spatial correlation structure or the spatial dependence. In particular, the NND method cannot distinguish between the two types of clustered patterns: a small number of large clusters and a large number of small clusters. In contrast, the K-function test for CSR can both do this and provide a more effective summary of spatial dependence over the wide range of scales involved. Fourth, the method for delineating future urban areas by taking into consideration patterns of the generation and accumulation of vacant lots is crucial when discussing how to promote a compact city policy. The findings of this article are expected to contribute to the development of such methods.

Supplemental Material

sj-pdf-1-epb-10.1177_2399808320956656 - Supplemental material for Are patterns of vacant lots random? Evidence from empirical spatiotemporal analysis in Chiba prefecture, east of Tokyo

Supplemental material, sj-pdf-1-epb-10.1177_2399808320956656 for Are patterns of vacant lots random? Evidence from empirical spatiotemporal analysis in Chiba prefecture, east of Tokyo by Hiroyuki Usui and Joan Perez in EPB: Urban Analytics and City Science

Footnotes

Acknowledgements

The authors are grateful to Professor Yasushi Asami and two anonymous referees for their extremely valuable comments and suggestions. This research was the result of the joint research with CSIS, the University of Tokyo (No. 785) and used the following data: Residential Maps provided by Zenrin, CO., Ltd.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Japan Society for the Promotion of Science (16H01830 and 17K12978).

ORCID iDs

Hiroyuki Usui

Joan Perez

Supplemental material

Supplemental material for this article is available online.

Notes

Hiroyuki Usui received PhD from the University of Tokyo in 2014 and now is an assistant professor of Department of Urban Engineering, the University of Tokyo. His research interests are city planning, spatial analysis, urban mathematical models and urban form. He studies the relationship of urban density and urban form. He published several papers in spatial analysis related journals, such as Computers, Environment and Urban systems, International Journal of Geographical Information Science, International Regional Science Review and Journal of Geographical Systems. He teaches urban planning and geographical information systems (GIS) to undergraduate and graduate students.

Joan Perez is a French researcher that owns a PhD in geography from the University of Avignon. His research interests are urban morphologies, fabrics and urban form in general. He carried out several international academic projects in urban geography which led him to publish on urban fabrics, morphologies and urban systems of Japan, France, Congo, India and China. After holding positions at the University of Tokyo, the French institute of Pondicherry and São Paulo State University, he joined the laboratory ESPACE-CNRS at the end of 2018. He now lives at the very center of the French Riviera in Nice where Université Côte d’Azur is located.

References

Akiyama

Ueda

Ono

, et al. (2018) Monitoring of spatial distribution of vacant houses using municipal public data in Kagoshima city, Kagoshima prefecture. Journal of Architecture and Planning (Transactions of AIJ) 83(744): 275–283.

Asami

(2014) Thinking Vacant Land and Vacant Houses in Cities. Shinjuku, Japan: Progress.

Asami

(2017) Planning support models in an era of shrinking population: Recent planning trends and research developments in Japan. In: Yamu

Poplin

Devisch

, et al. (eds) The Virtual and the Real in Planning and Urban Design: Perspectives, Practices and Applications. Abingdon, UK: Routledge, pp.170–185.

Bailey

Gatrell

(1995) Interactive Spatial Data Analysis. Harlow, UK: Prentice Hall.

Bradley

Behnisch

(2018) Heavy-tailed distributions for building stock data. Environment and Planning B: Urban Analytics and City Science 46(7): 1281–1296.

Caruso

Hilal

Thomas

(2017) Measuring urban forms from inter-building distances: Combining MST graphs with a local index of spatial association. Landscape and Urban Planning 163: 80–89.

Clark

Evans

(1954) Distance to nearest neighbor as a measure of spatial relationships in populations. Ecology 35(4): 445–453.

Cressie

(1993) Statistics for Spatial Data Revised Edition. Hoboken, NJ: John Wiley & Sons.

Frazier

Bagchi-Sen

Knight

(2013) The spatio-temporal impacts of demolition land use policy and crime in a shrinking city. Applied Geography 41: 55–64.

10.

Gao

Asami

(2005) Estimating the boundary lines of land lots with a multiobjective optimization approach. Environment and Planning B: Planning and Design 32: 581–596.

11.

Getis

Boots

(1978) Models of Spatial Processes: An Approach to the Study of Points, Lines and Area Patterns. Cambridge, UK: Cambridge University Press.

12.

Haase

Nelle

Mallach

(2017) Representing urban shrinkage—The importance of discourse as a frame for understanding conditions and policy. Cities 69: 95–101.

13.

Haase

Rink

Grossmann

, et al. (2014) Conceptualizing urban shrinkage. Environment and Planning A: Economy and Space 46(7): 1519–1534.

14.

Haase

Kabisch

, et al. (2012) Actors and factors in land-use simulation: The challenge of urban shrinkage. Environmental Modelling & Software 35: 92–103.

15.

Hackworth

(2016) Demolition as urban policy in the american rust belt. Environment and Planning A: Economy and Space 48(11): 2201–2222.

16.

Hollander

(2010) Moving toward a shrinking cities metric: Analysing land use changes associated with depopulation in flint. Cityscape 12(1): 133–151.

17.

Hollander

Johnson

Drew

, et al. (2017) Changing urban form in a shrinking city. Environment and Planning B: Urban Analytic and City Science 46(5): 963–991.

18.

Hollander

(2011) Can a city successfully shrink? Evidence from survey data on neighborhood quality. Urban Affairs Review 47(1): 129–141.

19.

Hollander

(2018) A Research Agenda for Shrinking Cities. Cheltenham, UK: Edward Elgar Publishing.

20.

Hospers

(2014) Policy responses to urban shrinkage: From growth thinking to civic engagement. European Planning Studies 22(7): 1507–1523.

21.

Illian

Penttinen

Stoyan

, et al. (2008) Statistical Analysis and Modelling of Spatial Point Patterns. Hoboken, NJ: John Wiley & Sons.

22.

Immergluck

(2016) Examining changes in long‐term neighborhood housing vacancy during the 2011 to 2014 US national recovery. Journal of Urban Affairs 38(5): 607–622.

23.

Koshizuka

(1985) On the relation between the density of urban facilities and the distance to the nearest facility from a point in a given area. Journal of the City Planning Institute of Japan 20: 85–90.

24.

Kubo

Yui

(2020) The Rise in Vacant Housing in Post-Growth Japan: Housing Market, Urban Policy, and Revitalizing Aging Cities. Singapore: Springer.

25.

Lee

Newman

(2017) Forecasting urban vacancy dynamics in a shrinking city: A land transformation model. ISPRS International Journal of Geo-Information 6(4): 124.

26.

Martinez-Fernandez

Audirac

Fol

(2012) Shrinking cities: Urban challenges of globalization. International Journal of Urban and Regional Research 36(2): 213–225.

27.

Masucci

Arcaute

Hatna

, et al. (2015) On the problem of boundaries and scaling for urban street networks. Journal of the Royal Society, Interface 12(111): 20150763.

28.

Molloy

(2016) Long-term vacant housing in the United States. Regional Science and Urban Economics 59: 118–129.

29.

Morckel

(2014) Spatial characteristics of housing abandonment. Applied Geography 48: 8–16.

30.

Nordvik

Gulbrandsen

(2009) Regional patterns in vacancies, exits and rental housing. European Urban and Regional Studies 16(4): 397–408.

31.

Ohsawa

Yokohari

Amemiya

(2009) Relationships between topography and emergence and persistence patterns of vacant lots in suburban residential zones. Journal of the Japanese Institute of Landscape Architecture 72(5): 683–686.

32.

Okabe

Sugihara

(2012) Spatial Analysis along Networks: Statistical and Computational Methods. Hoboken, NJ: John Wiley & Sons.

33.

Perez

Araldi

Fusco

, et al. (2019) The character of urban Japan: Overview of Osaka-Kobe’s. Urban Science 3(4): 105.

34.

Radzimski

(2016) Changing policy responses to shrinkage: The case of dealing with housing vacancies in Eastern Germany. Cities 50: 197–205.

35.

Reis

Silva

Pinho

(2016) Spatial metrics to study urban patterns in growing and shrinking cities. Urban Geography 37(2): 246–271.

36.

Sakai

(2014) A research of the reality of generation and diminishment of vacant parcels and the mechanism in suburban housing area: Case studies of 2 suburban housing areas in Capital area and Kinki area. Journal of the City Planning Institute of Japan 49(3): 1035–1040.

37.

Sakamoto

Yokohari

(2016) Dynamic characteristics of emerging vacant houses and abandoned lands in residential neighborhoods: A case study in the residential neighborhoods of utsunomiya, Tochigi. Journal of the City Planning Institute of Japan 51(3): 854–859.

38.

Sakamoto

Iida

Yokohari

(2017) Spatial emerging patterns of vacant land in a Japanese city experiencing urban shrinkage. Urban and Regional Planning Review 4: 111–128.

39.

Schetke

Haase

(2008) Multi-criteria assessment of socio-environmental aspects in shrinking cities: Experiences from Eastern Germany. Environmental Impact Assessment Review 28(7): 483–503.

40.

Schwarz

Haase

Seppelt

(2010) Omnipresent sprawl? A review of urban simulation models with respect to urban shrinkage. Environment and Planning B: Planning and Design 37(2): 265–283.

41.

Sorensen

(2001) Building suburbs in Japan: Continuous unplanned change on the urban fringe. Town Planning Review 72(3): 247–274.

42.

Usui

(2018) Statistical distribution of building lot frontage: Application for Tokyo downtown districts. Journal of Geographical Systems 20(3): 295–316.

43.

Usui

(2019) A bottom-up approach for delineating urban areas minimizing the connection cost of built clusters: Comparison with top-down-based densely inhabited districts. Computers, Environment and Urban Systems 77: 101363.

44.

Yamada

Terada

Tanaka

, et al. (2016) Directions for vacant lot management in the outer suburbs of the Tokyo metropolitan region. Urban and Regional Planning Review 3: 66–84.

45.

Yui

Kubo

Miyazawa

(2017) Shrinking and super-aging suburbs in Japanese metropolis. Sociology Study 7(4): 195–204.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.23 MB