COMPARATIVE ANALYSES ON FACTORS OF THE PEDESTRIAN NUMBERS IN A DOWNTOWN AREA AT TWO POINTS OF TIME USING SPACE SYNTAX INDICATORS – IN THE CASE OF SAKAE-SOUTH, NAGOYA CBD IN JAPAN

Objective: The objective of this study is to investigate the factors of the estimated pedestrian numbers in Nagoya, Japan at two points of time, with the aim of clarify the impact from two Space Syntax indicators, intensity of space use, distance from the station and pavement width on the pedestrians at two points of time. Theoretical Framework: The factor ranking for pedestrians in some prior reports shows the Space Syntax indicators to be higher ranking. This suggests a possibility of pedestrian factor analysis in which the Space Syntax indicators are added to factors. However, no one has checked if the Space Syntax indicator has an influence as a factor of the pedestrian number even if the time is changed. Therefore, it is meaningful to conduct the factor analyses in multiple times and compare the models. Method: The methodology adopted for this research comprises the creation of the correlation matrix between pedestrian numbers and the candidate explanatory variables, conduction of the multiple linear regression analysis at two points of time and comparison of two models. Data collection was carried out through the survey of the pedestrian numbers, calculation of Space Syntax Indicators. Results and Discussion: The results obtained revealed that the multiple correlation coefficients of the models decreased from 0.701 at first point of time to 0.562 at second point of time, but the Space Syntax indicator satisfied a significant difference at a 5% level in a t-test for the two points of time. In the discussion section, as tasks to be tackled in the future, further examination of points of time along with understanding the likely influence of Space Syntax indicators if other factors significantly change will clarify the validity of the Space Syntax indicators. Originality/Value: This study contributes to the literature by showing that the Space Syntax indicator has a certain degree of influence as a factor of the pedestrian number even though a point of time and pedestrians are changed.


INTRODUCTION
Creation and simulation of the "hustle and bustle" found at the heart of the world's large cities is now a widely studied subject.For this study, factor analysis of hustle and bustle is important.As research that approaches hustle and bustle from the viewpoint of spatial configuration, an analysis incorporating Space Syntax (hereinafter, SS) Theory by the UCL Group is well-known (Hillier and Hanson, 1984).SS Theory is an approach that quantifies the urban form; however, the intensity of a factor expressed by each indicator and their validity are still a matter for debate.Particularly, for the validity of the SS indicators, no report understanding any changes to the indicators between two or more points of time has been made.
When dealing with a case where factors other than SS indicators are considered and social changes are made, understanding how the influence of the SS indicators may change could enable the further examination of the validity of the SS indicators.
In our research, based on a case study of the Sakae-South district, the busiest area in the heart of Nagoya City, in order to analyze hustle and bustle in each pavement and street, the factors of the estimated number of pedestrians (hereinafter, pedestrian number) in 2005 and 2011 were analyzed.Multiple linear regression analysis was conducted using the following candidate explanatory variables, intensity of space use (commercial or office use floor-space ratio), distance from a station, and pavement width, and two SS indicators, visible area and street network integration.In addition, the obtained 2005 and 2011 models were compared to explore the validity of the SS indicators.

THEORETICAL FRAMEWORK
Prior to our research, we examined earlier domestic reports published by the Architectural Institute of Japan, and the City Planning Institute of Japan, and an overseas report by the ISUF.These reports concerned factor analysis applying the SS indicators to pedestrians in the midtown areas.In the examination, the following framework was established: regression analysis of the pedestrian numbers was called "pedestrian factor analysis" and as candidate factor variables, SS indicators were added to the conventional variables, which are the distance from the nearest station (ticket gate), intensity of land use, width of pavement (street), and upper limit of a floor-area ratio.Focus was then given to the ranking of factor's intensity (Table 1).
With regard to the pedestrian numbers, in 2003 an overseas report by Desyllas et al. (2003) presented an analysis in which an SS indicator, visible areas, were included in addition to the distance from the nearest station, intensity of land use, and pavement width.Özer and Kubat (2007) analyzed the pedestrian movement level with street network integration as SS indicator, intensity of land use and safety in Istanbul.In Japan, in 2005 Araya et al. (2005) analyzed the pedestrian numbers using both an SS indicator, integration value, and the distance from the nearest station; the SS indicator was found to be the primary factor, and the distance from the nearest station the secondary factor.Subsequently, in a multilayered analysis conducted by Ueno et al. (2009), the SS indicator was also found to be the primary factor.In analyses by Ota et al. (2008) and Okamoto et al. (2013), the distance from the nearest station or ticket gate became the primary factor, and the SS indicators were found to be secondary or later factors; however, they were adopted in statistical tests.
The factor ranking in the prior reports shows a tendency for the distance from the nearest station (ticket gate) to be higher ranked; however, some reports do give the SS indicators a higher ranking.In addition, in all reports, the SS indicators were adopted in statistical tests; this suggests a possibility of pedestrian factor analysis in which the SS indicators are added to factors.

Table 1
Prior research on the factor analysis of pedestrians applying SS indicators

Location of Sakae-South District in Nagoya City
In the Sakae-South district, the opening of Nadya Park in 1996 was the start of its shopping and leisure facilities becoming a powerful customer magnet and an increasingly busy commercial area.Its predominate feature is a row of eight large-store buildings (Mitsukoshi Main and South Buildings, Matsuzakaya North, Main, and South Buildings, and Parco East, West, and South Buildings), all tightly wedged in the area between Otsu Avenue and Hisaya Main Street and from Hirokoji Street to Wakamiya Main Street (Fig. 2).In addition, with the increasing number of brand name shops facing onto the their street, the clustering of brand name and street-level shops such as PRADA and an Apple Store further confirms the growing success of this commercial district.

Figure 2
Summary of Sakae-South District

ESTIMATION OF PEDESTRIAN NUMBERS IN THE SAKAE-SOUTH DISTRICT
The pedestrian number, an indicator of the level of hustle and bustle, used in this research, is the all-day (10:00 a.m. to 9:00 p.m.) sectional pedestrian traffic volume surveyed or estimated at 61 points located in pavements and streets in 2005 and 2011(Fig.3).

All-day number of pedestrians
The data was created based on the surveys conducted on holidays by the Minami-Otsu Avenue Shopping District Promotional Association (1989Association ( -2011)), and adjustments for the weather and missing data were made to allow the data to be interpreted as a typical example of a holiday with fine weather.For the 2005 data, based on the pedestrian numbers at 39 points taken from the 2005 pedestrian traffic volume survey, Sugiura made weather adjustments because it was a rainy day.Moreover, using a regression equation where the obtained 39 points are an explained variable, and data from a shop-around behavior survey is an explanatory variable, the pedestrian numbers at another 22 points were estimated to obtain pedestrian number data at a total of 61 points.For the 2011 data, based on the pedestrian number data at 34 points taken from the 2011 pedestrian traffic volume survey, to complement this data, we conducted a gate count survey in November 2014, by measuring pedestrian traffic volume for 5 minutes at each of 61 points.The survey time slots were chosen between 13:00 and 16:00 on a holiday, because usually many strolling pedestrians would be observed.The correlation of data at 34 points, common to the previous survey, was 0.848.From these pieces of data, Sugiura's method was used to complement and estimate the data at 27 points as of 2011 so as to obtain the pedestrian numbers at a total of 61 points.
A comparison of the obtained pedestrian numbers in 2005 and 2011 showed the total at the 61 points decreased by about 130,000; however, concerning the highest all-day pedestrian number at each point, in 2005 it was 35,235 at the east point in the 3rd block from the north side along Otsu Avenue, whereas in 2011 it increased to 39,915 at the west point in the 3rd block from the north side along Otsu Avenue.The larger number in 2011 could be mainly attributed to the influence of GUCCI opening in 2006, and the later renewal of a restaurant building in 2010.

CALCULATION METHOD OF SPACE SYNTAX INDICATORS
The research particularly focused on spatial configuration as a factor of hustle and bustle.Space Syntax Theory conceived by Hillier is one of the methods to analyze spatial configuration, and it seeks to establish quantitative benchmarks for accessibility in urban spaces.Our research used visibility graph analysis based on the SS Theory to calculate visible areas and integration values indicating street network integration.In addition, to clarify a relationship with the pedestrian number indicating the level of hustle and bustle on a pavement, the research calculated SS indicators by targeting the pavement area.
For calculation, DEPTHMAP, software developed by UCL, was used.A grid was laid out with one-meter intervals on a map, and any grid square where more than half of the area is occupied by a pavement was marked as a pavement.To calculate a visible area, if there is no screen and no borderline between a marked grid square and another marked grid square, it was assumed that both grid squares were visible to each other and the number of visible grid squares was counted.The visibility of all grid squares was assessed and the total number of grid squares that can be seen from a grid square containing a measurement point was considered to be a visible area at that point.In DEPTHMAP, this concept is called 'connectivity.'In addition, an indicator, Integration Value (hereinafter, IV), which represents the street network integration by applying the concept of visible area, was also prepared.The IV indicates the intensity of spatial connection in a street network, and a higher value means inadequate depth and high street network integration.To find an IV, firstly a Relative Asymmetry (hereinafter, RA) needs to be calculated (Equation 1); the RA is the relative depth (hereinafter, Depth) of each point (grid) seen from the entire area.When the RA value is higher, it is considered that many spaces must be traversed and the relevant space is positioned at a deep and complex location.Next, since the RA is affected by the total number of grids, a Real Relative Asymmetry (hereinafter, RRA) is found (Equation 3), which is a value standardized by a correction coefficient Dk determined by only k (Equation 2).The inverse number of the RRA value is an IV (Equation 4).Moreover, the midtown area was set as the range of the research; therefore, a Global Integration Value (hereinafter, GIV), which covers the entire area, was used and the GIV on a pavement measured. (1)

CALCULATION OF SPACE SYNTAX INDICATORS IN AN OPEN SYSTEM SPACE
This section describes specific procedures.Firstly, by using the GIS data, the surface of the urban area was divided into pavements, screens, and others.A pavement is considered to be: a pavement between a city block borderline and a road; a pavement around a public park; or a subway entrance.Vacant ground created due to the setback of a building, parks and paths in a park are not categorized as pavements.A screen is considered to be: a wall, hedge, fence or an entrance to an elevated road, all of which are 1.5m or more in height.When there is a screen, it is assumed that a pavement located beyond the screen cannot be seen.Others refers to parks, vacant ground, trees, fountains, and elevated roads that do not restrict visibility.For configurations of the ground, especially for any vertical drop, the following assumptions are made: • the height above sea level in the district varies between a high of 12.5m and a low of 7.5m; however, such differences in elevation are not taken into account; • outside the district, it is assumed that any area beyond a contour line of 1.5m or higher cannot be seen and a screen surface is established.Fig. 4 shows the range visible from the district.

Figure 4
The visible range from the district When a pavement is incorporated in a roadway, and they are not clearly divided, the width was measured as zero (0).

CORRELATION MATRIX BETWEEN PEDESTRIAN NUMBERS AND CANDIDATE EXPLANATORY VARIABLES
A correlation matrix between the pedestrian numbers in 2005 and 2011 and candidate factor variables was created (Table 2, Table 3).To create the matrix, the same values were used for the data of GIV, connectivity, distance from the station, and pavement width, because blocks and roads had not changed over the six years between 2005 and 2011.With regard to the correlations between the pedestrian numbers and candidate factor variables, the following point stood out for 2005: (Y1a)2005_ pedestrian numbers had a slightly high correlation with (X1a)2005_commercial use GF-space ratio, (X2a)2005_commercial use floor-space ratio, and (X6)distance from the station with a correlation coefficient of 0.639, 0.534, and -0.551 respectively.(Y1b)2011_pedestrian numbers in 2011 also had a slightly high correlation with (X1b)2011_commercial use GF-space ratio with a correlation coefficient of 0.512.
VIFs between candidate explanatory variables were calculated (Table 2, Table 3).To avoid multicollinearity, from combinations with a VIF exceeding 2, a candidate explanatory variable showing the highest correlation with the pedestrian number was selected and others were rejected.For 2005, from among combinations of (X1a)2005_commercial use GF-space ratio (the correlation with the pedestrian numbers (hereinafter, the same) was 0.639) with (X2a)2005_commercial use floor-space ratio (the same: 0.534), and with (X6)distance from the station (the same: -0.551), (X1a)2005_commercial use GF-space ratio was selected.For 2011, from among combinations of (X1b)2011_commercial use GF-space ratio (the same: 0.512) with (X2b)2011_commercial use floor-space ratio (the same: 0.390), and with (X6)distance from the station (the same: -0.408), (X1b)2011_commercial use GF-space ratio (the same: 0.512) was selected.Moreover, in combination of (X5)connectivity (the same: 0.239 in 2005, and 0.079 in 2011) and (X7)pavement width (the same: 0.463, and 0.285), (X7)pavement width was selected.

Table 2
The correlation matrix and VIFs between candidate explanatory variables ( 2005) Table 3 The correlation matrix and VIFs between candidate explanatory variables ( 2011)

EXAMINATION OF MODELS OF THE MULTIPLE LINEAR REGRESSION ANALYSIS RESULTS
Multiple linear regression analysis was conducted to derive multiple regression equations, and their applicability was examined.From among the four explanatory variables, (X1a)2005_commercial use GF-space ratio, (X3a)2005_office use floor-space ratio, (X4)GIV, and (X7)pavement width for 2005, and (X1b)2011_commercial use GF-space ratio,  (X3b)2011_office use floor-space ratio, (X4)GIV, and (X7)pavement width for 2011, explanatory variables satisfying significant at a 5% level in a t-test were extracted for each year.As a result, for 2005, two variables, (X1a)2005_commercial use GF-space ratio and (X4)GIV, were selected, giving a standard partial regression coefficient of 0.585 and 0.310 respectively, and the multiple correlation coefficient of the regression equation was 0.708, and the coefficient of determination was 0.501 (Table 4).For 2011 as well, the two variables, (X1b)2011_commercial use GF-space ratio and (X4)GIV, were selected, giving a standard partial regression coefficient of 0.468 and 0.240 respectively, and the multiple correlation coefficient of the regression equation was 0.562, and the coefficient of determination was 0.316 (Table 5).

Table 5
Multiple linear regression analysis ( 2011) The above results suggest that both in 2005 and 2011, the larger the ratio of the GF commercial use of a building to the relevant district area, and the higher the street network integration, the higher the pedestrian number will be.The absolute values of the standard partial regression coefficients enable an interpretation that the influence of the factor is strong in the order of the commercial use GF-space ratio followed by GIV.Next, focus was given to a change in the pedestrian numbers (Fig. 6).The total number of all 61 points showed a 14% decrease, and from the fact that the number of daily passengers using Sakae Station, the main actor of the relevant district, dropped from 110,000 in 2005 to 105,000 in 2011, it can be assumed that the pedestrian number decreased across the entire district.When the rate of change at each point was examined, the maximum fall was -60%, and the maximum increase was +218%.At the points along a main street or at which a large store is located, the rate simply decreased; on the other hand, the rate in alleys tended to increase.
In 2005 and 2011, although such changes to the pedestrian numbers can be seen, a SS indicator GIV satisfied a significant difference at a 5% level in a t-test for the two points of time, which allows an interpretation that even though a point of time changes, the indicator has a certain degree of influence as a factor of the pedestrian number.the maximum fall was -60%, and the maximum increase was +218%.At the points along a main street or at which a large store is located, the rate simply decreased; on the other hand, the rate in alleys tended to increase.In both 2005 and 2011, the GIV satisfied a significant difference at a 5% level in a t-test for the two points of time, which allows an interpretation that even though a point of time changes, the indicator has a certain degree of influence as a factor of the pedestrian number.
As tasks to be tackled in the future, further examination of points of time along with understanding the likely influence of SS indicators if other factors significantly change will clarify the validity of the SS indicators.
DISTRICT: CASE STUDY SUBJECT Nagoya City with a population of about 2.2 million has two main commercial and office clusters in the heart of the city; the areas around Nagoya Station and Sakae Station (Fig. 1).The Sakae-South District, the subject of the case study, is located to the southwest of Sakae Station and comprises 25 ha of commercial properties, surrounded by Hirokoji Street, Hisaya Main Street and Wakamiya Main Street.In the area sandwiched in between Otsu Avenue and Hisaya Main Street and running from Hirokoji Street to Wakamiya Main Street, a row of eight large-store buildings can be found.In this area, there are two subway stations, Sakae Station with a daily passenger throughput of 105,000, and Yabacho Station with 27,000 passengers (as of 2011).

Fig. 5
Fig.5shows the spatial distribution of the SS indicators calculated based on the visible range shown in Fig.4.Regarding connectivity, any street with a wider pavement or roadway had a larger visible area.This is possibly because a visible area depends upon the area of a pavement, and an unobstructed view that gives depth; a wider pavement often has a greater area; and a street with a wider roadway tends to have good visibility.This is particularly noticeable on Wakamiya Main Street, which with the widest pavement in the district measuring 10m on each side had the highest visible area.It was also clarified that the areas along the three main thoroughfares, Otsu Avenue, Hirokoji Street, and Hisaya Main Street with pavement widths of 4m, 6m, and 6m respectively had high visible areas.For the GIV, any street with a wider pavement or roadway had relatively high street integration.When comparing GIV with connectivity, the connectivity of Wakamiya Main Street differed greatly from Otsu Avenue, Hirokoji Street, and Hisaya Main Street, whereas Wakamiya Main Street and Hirokoji Street similarly had high GIV values, followed by Hisaya Main Street and Otsu Avenue; the relative value for each street was slightly different.

4. 3
CONSIDERATION OF COMPARISON OF MODELS BETWEEN TWO POINTS OF TIMEFor the 2005 and 2011 models resulting from the multiple linear regression analysis, these two points of time were compared and the applicability of the models and intensity of factors were examined.The multiple correlation coefficients of the models decreased from 0.701 in 2005 to 0.562 in 2011.For the standard regression coefficients as well, the commercial use GF-space ratio decreased from 0.585 to 0.468, and the GIV from 0.310 to 0.240.As the factors for such decreases, focus was given to changes to the commercial use GF-space ratio and GIV, both of which are candidate explanatory variables.The GIV showed no change, and the commercial use GF-space ratio had a rate of change at about 5% or less at all 61 points; this would indicate that no significant change occurred.

Figure 6
Figure 6The change in the pedestrian numbers