Development of Trip Attraction Rates and Parking Standards for Supermarkets in Kandy Area S.Suthakaran 1, U.G.S.Dharmakeerthi 1 and I.M.S.Sathyaprasad 1 1 Department of Civil Engineering Faculty of Engineering University of Peradeniya Peradeniya, Kandy SRI LANKA E-mail: nanbansutha@gmail.com Abstract: Trip attraction rates are major inputs needed in traffic impact assessment of commercial developments. The objectives of this study are to develop models to estimate trip attraction rates for person trips and vehicle trips and to develop parking standards for three-wheelers, motorcycles, cars/vans for supermarkets in the Kandy area. The number of vehicles entering and leaving the supermarkets and the number of people visiting the supermarkets at every fifteen minute interval during peak hours will be surveyed using a smart-phone based Android application. Multiple linear regression analysis will be used to analyse data. Central limit theorem will be used to develop parking standards. For each category a sophisticated model for supermarket designers and simplified model for local authorities will be developed. The results show that the wine shop area is the governing factor of the trip attraction rate and parking standards in the Kandy area. Keywords: trip attraction rates, parking demand, supermarkets, regression models, parking durations 1. INTRODUCTION A trip is a movement of a person from one place (origin) to another (destination). Trip attraction represents a trip starting or ending in a non-residential area. Most typically trip attraction rate refers to the number of trips attracted per day per activity center. Trip attraction rates are major inputs needed in traffic impact assessment of commercial developments. It is essential for the design of transportation facilities and services, and also for planning, investment, and policy development. In order to avoid traffic congestion in the road near an activity center, it is important to estimate the parking demand of the activity center. Even though a number of trip attraction models have been developed in different countries, it is not possible to use them in all occasions as the models have not been developed based on the local conditions of the study area. This study aims to develop suitable trip attraction models for vehicles and people and parking demand models for motorcycles, threewheelers, cars and vans for the supermarkets in the Kandy area. So, these models will be beneficial for supermarket parking designers and traffic engineers in the Kandy area. A total of twelve supermarkets in the Kandy area have been identified as survey locations for this study 1.1. Objectives i. Develop models to estimate trip attraction rates for person trips & vehicle trips of supermarkets in Kandy area ii. Develop parking standards for motorcycles, three-wheelers, cars & vans for supermarkets in Kandy area. 2. LITERATURE REVIEW Since these studies deal with statistical data analysis, it is important to have an idea about the methods adopted to collect data. Mamun et al. 2014, Uddin et al. 2012, Fillone et al. 2003, Kattor, K.V. & George 2013 and Kikuchi et al. 2004 used visual observation methods to count the persons and vehicles in each fifteen minute interval during a week day three hour peak duration. As visual 198
observation method is economical just to count vehicles and people, peak hour count will be enough to estimate the demand and Highway Capacity Manual uses this fifteen minute interval as the base unit for capacity computation (Kikuchi et al. 2004). The collected data have to be statistically analysed using an appropriate method. Mamun et al. 2014, Fillone et al. 2003, Kattor, K.V. & George 2013 and Kikuchi et al. 2004 used the multiple linear regression analysis method to analyse the data and produce a model. Uddin et al. 2012 used the average and standard deviation method to analyse the 2.1. Factors Affecting Trip Attraction Rate and Parking Demand In the past studies the authors used several factors as independent variables. Since those studies were based on shopping centers, some of the factors may not be suitable for this study. Table 1 describes those factors. The factors which suit this study are underlined. Table 1 Factors Affecting Trip Attraction Rate and Parking Demand Factor Remarks Reference Gross floor area(gfa) Mamun et al. 2014, Uddin et al. 2012 & Kikuchi et al. 2004 No. of Years in As this factor increases the trip Fillone et al. 2003 & Kattor, K.V. & operation attraction rate also increases. George 2013 Parking availability No. of stores No. of floors Commercial floor area Residential floor area Restaurant availability No. of employees Maintenance fee Type of the building Type of goods sold Width of major corridor Discount availability No. of entrance Supermarket is a single store. This factor can be accommodated into GFA This is out of scope (They used to analyze shopping characteristics of an area) Restaurant is not a part of supermarkets.(they are available in shopping centers) Since we can go into the supermarket and pick what we want no. of employees will not affect the study. Supermarkets of this study do not collect this fee It may affect the trip attraction rate, but difficult to convert into numerical value All supermarkets of this study sell similar types of goods. This is not applicable for supermarkets. Only seasonal discount is provided by all supermarkets of this study All supermarkets of this study have a single entrance Mamun et al. 2014, Uddin et al. 2012, Kikuchi et al. 2004, Fillone et al. 2003 & Kattor, K.V. & George 2013 Fillone et al. 2003 & Kattor, K.V. & George 2013 Fillone et al. 2003 Mamun et al. 2014 Uddin et al. 2012, Fillone et al. 2003 & Kattor, K.V. & George 2013 Fillone et al. 2003 Kattor, K.V. & George 2013 Kikuchi et al. 2004 Uddin et al. 2012 & Fillone et al. 2003 2.2. Available Parking Standards for Shopping Areas Available parking standards of several cities were referred and given in table 2 as parking space requirement per 100 m 2 of gross floor area. From these values, it is clear that the parking requirement is varies city to city. 199
Table 2 Available Parking Standards Region Country City Parking space Reference Colombo 2.00 Colombo Development Plan 2008 Kandy Development Plan for Urban Kandy 1.00 Sri Lanka Development Area n.d. Sri Jayewardenepura Parking and Building Regulations 2.00 Kotte 2008 India New Delhi 1.67 Pakistan Ahmedabad 0.65 Asia Bangladesh Dhaka 0.50 Indonesia Jakarta 1.33 Japan Tokyo 0.36 Philippines Manila 1.19 Parking Policy in Asian Cities 2011 Malaysia Kuala Lumpur 2.40 Thailand Bangkok 2.15 South Korea Seoul 0.78 China Beijing 0.35 Australia Sydney 2.83 Melbourne 1.48 Christchurch 2.14 Dunedin 1.91 Australia Hamilton 1.25 Douglass 2011 New Zealand New Plymouth 1.51 Hutt City 1.02 Tauranga 1.64 Porirua 1.26 Europe England Essex 7.14 Parking Standards Design and Good Cambridge 2.00 Practice 2009 Sheffield 2.86 North 0.40- America Washington America 0.45 Davidson & Dolnick 2002 2.3. Regression Analysis The multiple linear regression model can be represented as in Eq(1), (Ilvento & Tom 2003) Y = 0 + 1X 1 + 2X 2 nx n (1) Where Y is the dependent variable; X l, X 2... X n are independent variables; 0, 1 n are the regression coefficients, representing the parameters of the model for a specific population. The basic theory behind this analysis is minimizing the summation of square of errors. In order to use this analysis method, the following assumptions must be satisfied. Each value of independent variables X i and dependent variable Y must be observed without error. The relationships between dependent variable Y and each of the independent variables X i are linear. Distribution o 2003) & Waters 2006 and Ilvento & Tom 2003) The values of error term are independent Tom 2003) 200
The independent variables X i and Ilvento & Tom 2003) Error terms are normally distributed. (Osborne & Waters 2006 and Ilvento & Tom 2003) Error terms are independent from each of the independent variables. (Ilvento & Tom 2003) In addition to that the following requirements must be satisfied for this analysis. (Ilvento & Tom 2003) The sample size is at least one greater than the number of independent variables. (Higher the sample size higher the accuracy of results) Same number of observations for all variables. Thus if one variable is not known in a particular study location, then the entire k e used. The accuracy of the model can be measured by R 2 which is defined as in Eq(2). (Orlov 1996) R 2 = 1 (U 1 2 + U 2 2 n 2 ) / ((Y 1-2 + (Y 2-2 n- 2 ) (2) Where U 1, U 2 n are the error terms Y 1, Y 2 n are the dependent variables and is the average value of dependent variables. 3. METHODOLOGY As the goal of this study is to estimate the demand using governing factors, it is preferable to incorporate multiple linear regression analysis method. This analysis method can be followed easily with Microsoft Excel. 3.1. Pilot Survey The purpose of pilot survey is to find the best time period to collect data. It is obvious that the trip attraction rates and parking demand of a supermarket are on their peak when the sales of the supermarket on its peak. Hence the researchers went to each selected supermarkets and investigated about the time period of peak sales. Through the pilot survey, it was found that 4 p.m. to 7 p.m. is the peak sales duration for supermarkets in the study area of this study and also it occurs only during weekdays. And also it was found that distance to the nearest supermarket, parking availability at the nearest supermarket, no. of counters in the supermarket and wine shop availability in the supermarket also may affect the trip attraction rate and parking demand of the supermarket. 3.2. Field Procedure Trip attraction and parking demand will be collected on a week day during 4 p.m. to 7 p.m. at fifteen minute intervals. As the accuracy of manual recording is less during the peak time, an Android application which developed for this purpose will be used for this study to record above mentioned data. In addition the following data will be collected and recorded manually. Gross floor area of the supermarket No. of years the supermarket has operated Parking availability at the supermarket Parking availability at the nearest supermarket Distance to the nearest supermarket No. of counters in the supermarket Initial count of motorcycles, three-wheelers, cars & vans. Area of wine shop attached Figure 1 shows the interface of the Android application. 201
Figure 1 Interface of the Android application 3.3. Data Analysis Multiple linear regression analysis method will be used to analyze the collected data. Figure 2 & 3 illustrate the methodology adapted in this study. Collect data Collect data Sample mean & variance Central limit theorem Sample mean & variance Population mean Regression model Central limit theorem Regression analysis Population mean & variance 95% confident demand Regression model Regression analysis Figure 2 Trip Attraction Rate Flowchart Figure 3 Parking Demand Flowchart The idea of regression analysis is to obtain an equation for a dependent variable in terms of independent variables. In this study, five different models will be developed. The dependent variables for each model are population mean of trip attraction rate of persons, population mean of trip attraction rate of vehicles, 95% parking demand of motorcycles, 95% parking demand of three-wheelers and 95% parking demand of cars and vans respectively. Parking demand will be developed by assuming distribution of parking accumulation as normal. Independent variables for all of those models are gross floor area of the supermarket, year of operation of the supermarket, number of counters of the supermarket, wine shop area, parking availability of the supermarket, parking availability of the nearest supermarket and distance to the nearest supermarket. Since trip attraction rate and parking demand are inter related to each other, the same independent variables will be considered for both cases. But during the analysis, some of the variables may be discarded due to their failure to satisfy the assumptions. Then sophisticated version of the model will be obtained. Then the independent variables that are less significant on dependent variable when comparing with other variables will be discarded to obtain a simplified version of the model. 202
4. RESULTS AND DISCUSSION In this paper, Y vehicle,y people,y 3w,Y bike,y car, X G, X W, X D, X P, X Pn and X N are representing vehicle trip attraction rate (Vehicles/hour), people trip attraction rate (People/hour), 95% parking demand for three-wheelers, 95% parking demand for motor cycles, 95% parking demand for car/van, gross floor area of the supermarket (m 2 ), wine shop area of the supermarket (m 2 ), distance to the nearest supermarket (km) and parking availability in the supermarket, parking availability in the nearest supermarket and no. of counters in the supermarket respectively. For trip attraction rate, fifteen minute interval data was converted into one hour data and population mean found as shown in figure 2 for one hour data. For parking demand, initial count of each vehicle category was known. Then using the Android application data which contains entry and exit of each vehicle with time parking accumulation was found in each five minute interval. Then mean and standard deviation of parking accumulation was found. By assuming the distribution of parking accumulation as normal the data for 95% confident parking demand was estimated. 4.1. Correlation Test A correlation test was carried out to check the independency level of the independent variables. It was resulted that the number of counters and parking spaces are related to gross floor area. Hence this study will consider gross floor area and omit the number of counters and parking spaces. 4.2. Regression Models Two set of regression models were developed; a sophisticated version and a simplified version. Sophisticated version is more accurate than simplified version as it relates more independent variables than that of simplified version. It can be possible to collect the values of all independent variables for sophisticated version when a supermarket designer wants to estimate trip attraction rate and parking demand. But when a public authority wants to issue some standards, it is difficult to include all independent variables in sophisticated version because some of them are not the feature of a particular supermarket. Hence a simplified version is developed with only the features of a particular supermarket. Sophisticated models are given in Eq(3-7). From twelve data sets, nine were used to develop these models. Y vehicle = - 0.799 + 0.136X G + 1.546X W + 0.145X D 0.399X P (R 2 = 0.938) (3) Y people = 49.48 + 0.192X G + 4.85X W + 28.6X D 0.539X P (R 2 = 0.949) (4) Y 3w = 5.586 + 0.010X G + 0.319X W + 1.62X D 0.115X P (R 2 = 0.880) (5) Y bike = 11.097 + 0.015X G + 0.137X W + 3.3X D 0.259X P (R 2 = 0.605) (6) Y car = -15.61 + 0.082X G + 3.33X D 0.423X P (R 2 = 0.899) (7) Except for the motorcycle parking demand, others show a better R 2 value. As expected parking spaces of the nearest supermarket show negative coefficient as it reduce the dependent variable and others show positive coefficient as they increase the dependent variable. Also for motorcycle parking demand there may be some other governing factors like rain, socio economic factor etc. which may affect. do not give priority to wine shops. Also for car parking demand the intercept is a significantly large negative value which may due to car/van parked illegally near to the supermarket which to capture during the survey. Simplified models are given in Eq(8-12). Y vehicle = -4.64 + 0.131X G + 1.64X W (R 2 = 0.925) (8) Y people = 18.336 + 0.183X G + 5.269X W (R 2 = 0.932) (9) Y 3w = 3.292 + 0.008X G + 0.363X W (R 2 = 0.843) (10) Y bike = 6.257 + 0.009X G + 0.042X W (R 2 = 0.240) (11) Y car = -15.822 + 0.073X G (R 2 = 0.865) (12) In the simplified version, also except for the motorcycle parking demand, others show a better R 2 value thus the models are good. In the simplified version also the reasons for lower R 2 value for 95% motorcycle parking demand and significantly large negative intercept value for 95% car parking 203
demand may same as that of sophisticated version. 4.3. Validation of Regression Models Among twelve data sets, the remaining three sets were used to check the validity of the models. It was resulted that among ten developed models, seven models are estimating the dependent variables with acceptable error. But simplified version of motorcycle parking model showed unacceptable error as its R 2 value is less. Both models of car parking demand estimating demand as negative for gross floor area less than around 220 m 2 as they have significantly large negative intercept value. 4.4. Parking Standards Parking standard derived from the simplified versions of regression model. When there is no wine shop, 1 standard parking bay per 15 m 2 of G.F.A. and 1 motor cycle parking bay per 60 m 2 of G.F.A. is required. On the other hand, when there is a wine shop, 1 standard parking bay per 13 m 2 of G.F.A. and 1 motor cycle parking bay per 50 m 2 of G.F.A. is required. 4.5. Hypothesis Testing on Parking Duration Parking Duration(minutes) Table 3 Parking duration Three-Wheelers Motorcycles Car/Van Wine shop availability Wine shop availability Wine shop availability Yes No Yes No Yes No Median 3 9 4 12 5 9 85 th percentile 7 22 12 24 16 21 95 th percentile 18 46 23 43 23 26 As show in table 4, Three-wheelers and motorcycles showed a comparatively low parking duration when wine shop available. In order to statistically verify this behavior a Hypothesis testing was done. It was resulted that for three-wheelers and motorcycles, data is convincing enough to say that the average parking duration of three-wheelers and motorcycles in supermarkets is significantly differs with wine shop availability. But for car/van data is not convincing enough to say that the average parking duration of car/van in supermarkets is significantly differs with wine shop availability. 5. CONCLUSIONS AND RECOMMENDATIONS The conclusions are: i. Models developed for trip attraction rates and parking demand were shown. ii. Wine shop is the governing factor for trip attraction and parking requirement of three-wheelers and motorcycles. iii. Separate parking for wine shop can be recommended to reduce parking uncertainties in the supermarkets. The recommendations are: i. These models are valid only for the supermarkets in the Kandy area. Further studies can be done for other areas in a similar way. ii. Supermarkets which have shared parking spaces were omitted. Some further studies can be done to find the parking share of the supermarket. iii. For motorcycle parking demand model accuracy can be improved by studying further governing factors. 204
iv. It is recommended to count the roadside parking near to supermarket when the parking is full. v. It can be recommended to increase the sample size for better results. 6. ACKNOWLEDGEMENTS The writers would like to thank Tokyo cement P.L.C. for their financial assistance to this study, supermarket staffs for their help during pilot survey and data collection, Eng. Titus Nanda Kumara for preparing the Android application to this survey. Finally they would also like to thank, their friends who helped to collect data and their reviewers for the support and encouragement they gave at various occasions. 7. REFERENCES 'City of colombo development plan' 2008, Amendment. Davidson, M & Dolnick, F 2002, 'Planning advisory services', Parking solutions. Douglass, A 2011, 'Trips and parking related to land use', Transport Agency research report, 453. Fillone, AM, Tecson, MR, Sia, R & Viray, P 2003, 'Trip attraction of mixed-use development in manila', Proceedings of the Eastern Asia Society for Transportation Studies, Vol.4, Manila. George, P, Kattor, GJ & K.V, AM 2013, 'Prediction of trip attraction based on commercial land use characteristics', International Journal of Innovative Research in Science, Engineering and TechnologyVolume 2, Special Issue 1, pp. 352-359. Ilvento & Tom 2003, 'Multiple Regression Analysis', University of Delaware, College of Agriculture and Natural Resources,Food and Resource Economics, Delaware. 'Kandy development plan for urban development area', Regulation 31, Kandy. Kikuchi, S, Felsen, M, Mangalpally, S & Gupta, A 2004, 'Trip Attraction Rates of Shopping Centers in Northern New Castle County, Delaware', Delaware Center for Transportation, Newark, Delaware. Mamun, MS, Rahman, SMR, Rahman, MM, Aziz, YB & Raihan, MA 2014, 'Determination of trip attraction rates of shopping centers in dhaka city', 2nd International Conference on Advances in Civil Engineering, CUET, Chittagong. Orlov, ML 1996, 'Multiple linear regression analysis using microsoft excel', Chemistry Department, Oregon State University, Oregon. Osborne, JW & Waters, E 2006, 'Four Assumptions Of Multiple Regression That Researchers Should Always Test', North Carolina State University and University of Oklahoma, Oklahoma. 'Parking and building regulations' 2008, Sri Jayawardenapura Kotte Municipal area 'Parking policy in asian cities' 2011, ISBN 978-92-9092-352-7. Parking standards design and good practice 2009, viewed 20 January 2016, <www.essex.gov.uk>. development, Coras Iompair Eireann, Dublin. ns of the linear regression model', research and Uddin, MM, Hasan, MR, Ahmed, I, Das, P, Uddin, MA & Hasan, DT 2012, 'A Comprehensive Study on Trip Attraction Rates of Shopping Centers in Dhanmondi Area', International Journal of Civil & Environmental Engineering IJCEE-IJENS Vol:12 No:04, pp. 12-16. 205