Available online at www.sciencedirect.com

ScienceDirect

Procedía

Social and Behavioral Sciences

ELSEVIER

Procedía Social and Behavioral Sciences 20 (2011) 100-109

14th EWGT & 26th MEC & 1st RH

A management system of territorial planning and mobility:

a case study

Federica Croccoa, Carmen Forcinitia*, Domenico Walter Edvige Mongellia

aUniversité della Calabria, Ponte Bucci, Rende 87036, Italy

Abstract

This paper proposes the design of a management system of territorial planning and mobility, analyzing the relationships between the urban structure and mobility data from ISTAT statistics and field surveys.

The preliminary step involves the calibration of a matrix of data mobility through an optimization model. The next step develops a system implemented in a GIS environment aimed to shape the spatial analysis of territorial features with determined mobility data. Using GIS, thematic maps can be prepared to display the mobility flows that take place in the area. The study is mainly aimed at identifying relationships between territorial and mobility variables.

The expected results may help to define the strategies required for the planning of urban public transport and for location of suburban interchange nodes.

© 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Organizing Committee. Keywords: transit oriented development; GIS; mobility flows; O-D matrix estimate; mobility demand; network model

1. Introduction

The knowledge of mobility demand is an unavoidable prerequisite for effectively tackling the critical issues related to the strategic choices for territorial planning. The aims of the proposed study are to estimate the Origin-Destination matrices in different times of the day in the town of Mottola, to develop procedures that can be standardized and systematically applied in medium-sized municipalities, and to analyse the relationships between territorial features and mobility flows using GIS. The procedure involves the analysis of a wide range of available data concerning the mobility and it is able to acquire new ad hoc data and to process them in the last stage using the modeling of transportation systems. A broad bibliographic study allowed the acquisition of historical data from multiple sources related to mobility (O-D matrices ISTAT census, flow rates by Isfort surveys, O-D surveys, etc...) and to socio-economic context (ISTAT Statistics, municipality, etc.).

O-D matrices were subsequently obtained by the ISTAT statistics on commuting and population and housing census. However, the available data from various sources are inadequate to build a comprehensive framework so an

* Corresponding author. Tel.:+039-0984-496754; fax: +039-0984-496754. E-mail address: carmen.forciniti@unical.it.

1877-0428 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Organizing Committee doi:10.1016/j.sbspro.2011.08.015

extensive sample survey, specifically designed for the analysis of commuters, was carried out. Particularly, a questionnaire survey was designed to describe mobility in a day of reference. The resulting sample Origin-Destination matrices were then projected to the statistic universe and elaborated according to the statistics techniques and to the demand models.

The proposed algorithms were designed to the minimization of the differences between the measured trips and the corresponding trips developed by ISTAT. The algorithms were employed with the introduction of appropriate constraints on the variations of individual reports. Indices of correlation between trips recognized by survey and trips recorded by ISTAT, which are variables between 0.8 and 0.9, were obtained at the end of the procedure. The outcome of this procedure was the evaluation of a daily Origin-Destination matrix for different purposes (work and study). The next step of the proposed study was to design and to apply GIS for analyzing spatial relationships between the mobility data and the territorial features. Over the last few years, the adoption of Geographic Information Systems (GIS) has supported urban analysis. A GIS is a computer system capable of capturing, managing, integrating, manipulating, analyzing, and displaying geographically referenced information. GIS deals with spatial information which uses location within a coordinate system as its reference base (Saleh and Sadoun, 2006). Furthermore, this tool can generate new information from those stored by the automatic overlapping of geographic layers, the generation of areas of influence, the analysis of networks. A GIS allows to study analytically the spatial relationships among the variables that characterize the phenomenon under study, because it integrate common tasks performed on the database, such as statistical analysis, with the advantages of graphical representation of data and geographic analysis offered by maps. Using GIS, researchers can manipulate a large amount of data and visualize urban affairs (Okunuki, 2001). The outcome of this step was the elaboration of thematic maps.

The paper is organized as follows: in the next section a procedure for the estimation of mobility is presented; in section 3 the methodology for estimating is described; section 4 presents the procedure to correct O-D matrices; section 5 regards the analysis of the relationships between mobility flows and urban structure; section 6 refers to the analysis of the bus station service area; finally, concluding remarks are contained in section 7.

2. A procedure for the estimation of mobility

The reconstruction of mobility demand was conducted to value an Origin-Destination matrix, reliable and updated at present, obtained by integrating and processing various types of information. Available initially data can be classified in the following way:

• Historic Origin-Destination Matrix from census data.

• Sample surveys in origin and destination of trip.

• Indicators, socio-economic data, and trip data disaggregated at different levels: provincial, municipal, census parcels.

Each source provides a partial and often outdated background, in many cases statistics were compiled on the basis of surveys conducted on a much larger scale than the specific context of study. The simple integration of available data is not sufficient and, however, problematic because they were obtained with different objectives and investigation techniques related to different time and spatial scales. An assessment procedure, articulated in successive steps, was developed to make congruent data from various sources, to acquire and to integrate new data gradually.

Each step is functional in an overview and then in this context must be considered. Figure 1 shows the different stages of the procedure, pointing out that there is sequentiality and often mutual dependency. The main steps can be summarized as follows:

• Acquisition of available current and historical data, both about mobility and socio-economic context, and at different levels of aggregation.

• Planning and progress of sample surveys on commuters demand and systematic demand.

• Statistical analysis of results and extraction of Origin Destination sample matrices.

• Estimation of O-D matrices using statistical and modeling tools from the projection of sample matrices.

Figure 1 Flowchart of the procedure.

The procedure and in particular the combination of these instruments is not codified but it is necessary to play it from time to time in accordance with criteria of choice that takes into account several factors first and foremost the significance of the data available. Correction of the O-D matrix obtained from the demand models: the correction is performed on the basis of the measured matrix and the ISTAT O-D matrix. With appropriate constraints, algorithms for minimizing the deviation between the measured displacements and the corresponding displacements obtained from the ISTAT statistics are used.

3. The methodology for estimating

The classic formulation of the problem of the transport demand evaluation by counting trips from sample surveys at the origins and destinations, as explained above, concerns an improvement of the O-D matrix estimated and the possibility to update the O-D matrix continually. Although considering, as per usual practice, that the allocation model and the surveys represent the unbiased estimators of real flows, it is not usually possible to estimate the vector of demand flows with the only sample surveys.

Denoting with d* the estimate of the O-D matrix, the problem of estimating the O-D matrix can be expressed as:

d * = arg min (x, d) + z2 (v(x), f)] (1)

where:

• d* represents the matrix of unknown O-D trips;

• z1(x,d) represents a measure of the distance of the unknown O-D demand vector x by aprioristic estimation d*;

• z2(v(x),f) is a measure of the distance of demand flows vector v(x) obtained assigning x to the network from sample demand vector of movements detected f.

The measure of distance z(-) can be realized as "weighted Euclidean metric", defined by the expression (2), (3) and the distance is measured by the standard deviations weighted in inverse proportion to the quality of the information contained in the matrix O-D measured by sample survey f:

Z1 (x, d ) =

Z2 (v(x), f ) = where:

• var[sj indicates the variance of deviations between matrix obtained from ISTAT census data Vi(x) and matrix

measured by sample survey f;

• var[^od] represents the variance of the sampling error.

In this case, the equation provides the Generalized Least Squares Estimation (GLS).

The (2) and (3) allow to determine demand matrix that minimizes the sum of the deviations respect flows observed with ISTAT matrix. The standard deviations are also weighted in inverse proportion to the variance of the error Sj. For the case examined this term can be omitted, having assumed the correctness of the observed values.

4. Procedure to correct O-D matrices

The O-D matrices estimated using analytical tools, essentially statistics and modeling of demand mobility, are to be compared with the flows data actually measured on a daily basis and in accordance with these to be corrected and amended. The optimization procedure of the matrices was carried out using iterative optimization software based on the algorithm of Monte Carlo. In the correction procedure, the need to calibrate and apply constraints to the procedure was highlighted. Particular attention was paid to the changes of the maximum number of trips of each report, regarding that the iterative minimization algorithm follows a mathematical criterion, as the minimization of the deviations, applied to the detected demand flows and the assigned demand flows.

The first constraint was applied to individual O-D relations in proportion to the size of the flows resulting from the procedure in this intermediate stage. The highest percentage increase compared to starting value was defined as equal to 200%. Weighting each pair of O-D survey in the function of minimizing the deviations was another trick to explicit the need to take into account the specific reliability of each average input demand data. Therefore, the weight of O-D pairs, whose distribution of demand values has a width of the confidence interval equal to or exceeding one fifth of the average application time, was reduced by 20%. Highly variable average values, such that leads to a particularly wide confidence interval, represent a reference that should be less binding in the process of correcting the O-D matrix. This variability is attributable to the defects of the measures and to the actual variability of demand. Finally, given the sample information, the possibility that certain reports were not represented because of a lack of sampling was taken into account. Therefore, upstream of the correction procedure, a symbolic Origin-Destination matrix, still greater than zero and destined to be developed anyway in the process of correction, was introduced.

Since this is an iterative process of correcting the O-D matrix, it is necessary to define the number of iterations at which to stop the procedure. This value, assumed constant for the various bands, was determined by the performance of the main statistical indicators, as the correlation coefficient between the O-D demand matrix and the measured O-D matrix. The maximum number of iterations was contained so as to control the effects of procedure and not to distort the initial O-D matrices. In figure 2(a) and 2(b) the results before and after correction are reported with the results of the sample survey and the percentage of variation for the different trip purposes.

\Xod - dod F varkd ]

if.i - v, (x))2

var[i?i ]

0 20 40 60

Ante-Optimaztion

0 20 40 60

Ante-Optimization

Figure 2 (a) Trips before and after optimization for work purpose; (b) Trips before and after optimization for study purpose.

Table 1 Percentage change in demand before and after optimization for work and study purposes.

Work purpose

Study purpose

Ante Post % Variation Ante Post % Variation

1 16 12 -25% 0 7 -

2 22 16 -27% 5 9 80%

3 78 46 -41% 8 19 138%

4 0 2 - 0 3 -

5 54 33 -39% 33 98 197%

6 0 5 - 0 14 -

7 22 15 -32% 8 21 163%

8 48 31 -35% 32 26 -19%

9 16 10 -38% 28 16 -43%

10 32 22 -31% 24 21 -13%

11 32 23 -28% 28 24 -14%

12 22 16 -27% 8 11 38%

13 38 23 -39% 17 16 -6%

15 91 57 -37% 20 33 65%

17 16 20 25% 31 42 35%

18 48 31 -35% 24 21 -13%

19 32 28 -13% 46 51 11%

20 48 33 -31% 25 23 -8%

21 64 37 -42% 20 22 10%

22 16 13 -19% 5 11 120%

23 16 14 -13% 12 36 200%

24 16 13 -19% 13 12 -8%

26 38 27 -29% 8 12 50%

27 16 12 -25% 8 10 25%

39 16 13 -19% 8 13 63%

5. Analysis of the relationships between mobility flows and urban structure

An urban spatial structure is a spatial arrangement of a city in which it is a result of the interaction between land markets, topography, infrastructure, taxation, regulations and urban policy over time (Bertaud and Stephen, 2003). Many urban processes are intrinsically spatial and space dependent (Paez and Scott, 2004). Railways, road networks, civil and industrial building and other constructions built on territory fit for people's needs, therefore their

location is not random but it is the outcome of processes concerned with all the elements of the urban systems.

About an urban area, it is interesting analysing the spatial distribution of buildings and infrastructures according to the territorial features and the distribution of population and human activity. In fact, this analysis allows understanding how urban structure develops over time and how the different urban sub-systems influence each other. To find the processes of spatial distributions, it is necessary to manipulate a large amount spatial data about urban areas using spatial analysis techniques. These techniques were recently used in the study of urban analysis.

In this paper, the interactions between transportation system and land-use are drawn. Particularly, the relationships between mobility flows and urban structure are underlined. The mobility flows, defined in the previous paragraphs, are commuting trips produced within the town of Mottola toward near towns.

The matrix of total daily trips for both studying and working was represented graphically by a thematic map in which each census parcel is associated the average number of daily trips issued (figure 3). The census parcels with the largest number of daily trips are represented with a darker shade of colour and are located at the city centre and close to public transportation station. Moreover, the figure shows that, for the most census parcels, the number of daily trips decreases with the distance from the bus station. This trend comes out from a first global analysis of the data. Indeed, the number of trips issued from a zone does not depend only on the distance from the bus station but also by other variables such as population, age, employment, socio-economic conditions, and car availability. Not all these variables, however, are considered in the analysis undertaken in this work.

Legend Total daily trips

0-10 11-20 21-30 31-40

■ 41-50

■ 51-60

■ 61-70

■ 71-80

■ 81-90

■ 91-100

■ 101 -110 ■ 111 -120

■ 121 -130

■ 131 -140 H Bus Station

Figure 3 Total daily trips.

The analysis of the urban structure was performed according to the zoning of the development plan (Piano Regolatore Generale). The area of each census parcel was divided according to the different planning zones which have certain characteristics. Among all zones, those where there are residential buildings were selected.

The zoning was overlapped on the census parcels for identifying the volume of residential building of each zone present in each census parcel. An indicator of population distribution was obtained by dividing the volume of residential building of the zone to the total volume of the census parcel. Using the indices determined in accordance with this procedure, the number of inhabitants was determined for each zone in the considered census parcel.

The figure 4 shows the comparison between the distribution of population in the central town area and the number of trips issued for census parcel. The section which has the highest number of daily trips is characterized by high concentrations of population and is about 300 meters from the bus station that can be reached by walk. Other census parcels have the same characteristics but a minor number of daily trips. Finally, the suburban placed in south respect to the center are the parcels census with the fewer daily trips.

Figure 4 Relationship between total daily trips and urban structure.

6. Analysis of the bus station service area

For defining the bus station service area, the road network of Mottola was constructed using an extension of GIS. Each link of the network was drawn in GIS, which automatically calculated their length. The roads were classified in three groups: primary, secondary and local. Each group has an average road speed depending on the road type and traffic. The bus station service area was determined in two ways: considering the travel time measured in minutes, and the road distance measured in meters along the network.

In the first case, travel time was determined according to the average road speeds; therefore, the bus station service area, showed in the figure 5, is related to motorized trips. From the performed analysis, it resulted that 29% of the resident population can reach the bus station in less than 1 minute, 56% spends from 1 to 3 minutes, 2% from 3 to 5 minutes, and the remaining 12% spends more than 5 minutes. Accordingly, about 85% of the resident population can reach the bus station by making a motorized trip less than 3 minutes long.

Figure 5 Bus station service area evaluated in travel time (minutes).

In the second case, the bus station service area, showed in the figure 6, was defined in terms of road distance to take into account pedestrian trips. In fact, knowing the path length it is possible to decide if the trip can be made walking. In general, the max distance which a person can walk is about 500 meters.

Figure 6 Bus station service area evaluated in road distance (m).

Figure 7 Bus station service area for the town centre evaluated in road distance (m).

Another thematic map was elaborated to value if the bus station can serve the resident population in the town centre who decided to walk (figure 7). This map shows the bus station service area until 2000 meters. Most of the town falls in the range between 500 and 1000 meters whereas a substantial part is over 1000 meters away from the bus station.

In conclusion, the analysis of the bus station service area indicates that the bus station is accessible enough for the population who travels by car whereas it is not very accessible by walk because it is localized at the edge of the downtown.

7. Conclusion

The study was designed and developed not as an occasional act to estimate the demand of commuters within the town of Mottola, but in perspective of a systematic application both in different territorial context and in the same territorial context regularly over time. The periodic and regular update makes it possible to monitor the status of mobility and its evolutions allowing to evaluate the effects of policies and introduced structural interventions that, directly or indirectly, can affect the mobility and territorial planning system.

The study was conducted to estimate the Origin-Destination matrices, to develop procedures that can be standardized and systematically applied in medium-sized municipalities, and, finally, to investigate the travel behavior of the resident population. The experience has provided many ideas to improve the efficiency and effectiveness in future investigations. These proposals for:

• increase the representativeness of the sample,

• reduce costs and lead times,

• improve the quality of the data,

• analysing the spatial distribution of urban structure and mobility variables.

References

Ashok, K., Ben-Akiva, M. (1993). Dynamic Origin - Destination Matrix Estimation end Prediction for Real Time Traffic Management System, C.F. Daganzo Ed., International Symposium on Transportation and Traffic Theory, Elsevier Science, 465-484.

Bell, M., (1991). The estimation of origin-destination matrices by costrained generalized least squares. Transportation Research Part B, 25, 1322.

Bertaud, A., Stephen M. (2003). The Spatial Distribution of Population in 48 World Cities: Implications for Economies in Transition. Online. The Centre for Urban Land economic research. University of Wisconsin. http://alainbertaud.com/ AB_Files/ Spatia_ Distribution_ of_ Pop_50_Cities.pdf

Camara, G., Carvalho, M.S., (2005). A Tutorial on Spatial Analysis of Areas. http://edugi.uji.es/Camara/spatial_analysis_areas.pdf

Cascetta, E. (1991). Estimation of trips matrices from traffic counts and survey data: a generalized least squares estimator. Transportation Research Part B, 18, 289-299.

Cascetta, E. (2006). Modelli per i sistemi di trasporto. Teoria e applicazioni. Franco Angeli, Milano.

Cascetta, E., Montella, B. (1998). Metodologie per la redazione e la gestione dei piani urbani del traffico e della mobilita. Franco Angeli, Milano.

Cascetta, E., Postorino, M.N. (2001). Fixed point models for the estimation of O-D matrices using traffic counts on congested networks, Transportation Science, 35, 134-147.

Chen, Y., Florian, M. (1995). A coordinate descent method for the bi-level O-D matriz adjustament problem. International Transaction in operational Research, 2, 165-177.

De Luca, M. (2000). Manuale di pianificazione dei trasporti. Franco Angeli, Milano.

Di Gangi, M. (1988). Una valutazione delle prestazioni statistiche degli estimatori della matrice O/D che combinano I risultati di indagini e/o modelli con I conteggi di flussi di traffico. Ricerca Operativa, 51, 23-59.

ISTAT (2001). 14thCensimento popolazione abitazione.

Noresah, M.S., Ruslan, R. (2009). Modelling urban spatial structure using Geographically Weighted Regression. 18th World IMACS / MODSIM Congress, Cairns, Australia 13-17 July 2009.

Okunuki, K. (2001). Urban analysis with GIS. Geo Journal, 52, 181-188.

Paez, A., Scott, D.M. (2004). Spatial statistic for urban analysis: A review of techniques with examples. Geo Journal, 61, 53-67.

Piano Regolatore generale (2004). Delibera Giunta Regionale n. 483 del 31/03/2005. BURP, n.61 del 22/04/2005.

Saleh, B., Sadoun, B. (2006). Design and implementation of a GIS system for planning. International Journal on Digital Libraries, 6(2), 210218.

Yang, H. (1995). Heuristic algorithms for the bi-level Origin/destination matrix estimation problem. Transportation Research Part B, 29, 231242.