Traffic Demand Forecast of Road in Kigali, Rwanda

: Accurate forecasting of traffic demand is one of the most important issues in the feasibility study on highway projects. The existing traffic forecasting models, to some extent, have the problem of limited accuracy. In this study, two widely used methods which are Elastic Coefficient Method (ECM) and Motorized Travel Frequency Method (MTFM) were comprehensively applied to forecast the traffic volume in Kigali, Rwanda. And Comparative analysis was made between the forecasting result and the actual survey traffic result in the project’s future years. Compared with the actual survey result, the predicted result of ECM is larger and relative error is 10.49%. The result of MTFM is smaller and relative error is -7.11%. While the weighted average of above methods is closer to the actual result with a relative error in the interval of -5.00 to 5.00%. The research has shown that the combined forecast method proposed in this study, which can make up the defects in accuracy of single model, is easy to operate and owns more accuracy in traffic prediction. This study has suggested that proper combination of several methods would be an advisable trend for the traffic demand forecasting


INTRODUCTION
Traffic forecast, an important stage in feasibility study of highway, is the foundation for construction scale and standard.Only on the premise of accurate prediction on the future traffic volume of highway project, can we make sense of economic analysis and social benefit evaluation of the project.Especially with diversification of highway construction financing, returning on investment becomes the biggest concern of the investors, hence it is very important to forecast traffic volume in the preliminary stage of highway construction (Hong, 2011;Huang and Lu, 2010).
Intensive studies have been carried out in the field of traffic forecast since 1960s and several of methods have been explored whose structures vary in the degree of sophistication and data requirement.These methodologies can be classified into several types, like, regression models (Brian et al., 2002;Kamarianakis and Prastacos, 2005), traffic time series models (Stathopoulos and Karlaftis, 2003;Skamris and Flyvbjerg, 1997), gravity models, Kalman state space filtering models (Hong et al., 2011;Jae et al., 2005), grey system models and rolling models (Liu and Ren, 1994;Dia, 2001) etc. Elasticity Coefficient Method (ECM) and Motorized Travel Frequency Method (MTFM) are typical in regression models.The economy and population factors, which are directly related to traffic volume, are taken into consideration in both methods.Due to their simple principle, they are widely used in the world (Schreckenberg et al., 2001).By analyzing relationship between transport index and national economic indicators and travel frequency, this study combined ECM and MTFM to forecast traffic demand of reconstruction project in Kigali, Rwanda and proposed a more effective and convenient method for traffic forecasting (Jae et al., 2005).
Traffic demand is a descendent and traffic growth rate is related to some trend-type indicators, such as the growth rates of industrial and agricultural production, gross social product and population, in the direct impacted areas affected by economic condition and social activities (Chrobok et al., 2006;Li and Lin, 2006).Therefore, we can seize the developing regular of traffic demand by analyzing the relationship between economy and social activities, with multiple mathematic methods, for instance, elasticity coefficient model, regression analysis and econometric analysis.ECM takes full account of correlation between economic growth level and growth of traffic demand and seizes traffic development trend at the macro (Zeng, 2006).Generally speaking, the more economically developed, the lower of requirements for traffic is, hence the smaller elasticity coefficient is; while the more economically backward, the greater potential requirement for traffic, so the bigger elastic coefficient is.For one certain area, there is an urgent need for traffic in the recent period, elastic coefficient is big and gradually decreases in the succeeding period (Wang et al., 2010;Bell, 1983).Elasticity coefficient e can be calculated by following formula: The determination of the elasticity coefficient value is the sticking point of ECM.Generally, the elasticity coefficient which is related with socio-economic development level as well as regional characteristics and development strategy can be determined synthetically by analyzing history, status quo and development trend of region.For feasibility study of highway project, it is too hard to get the statistics of different regions traffic travel, so regression analysis and expert evaluating method can be used to get elasticity coefficient.The main factors affecting elasticity coefficient are socio-economic indicators and transportation related indicators.So elasticity coefficient of vehicle travel is reckoned based on section traffic of major road, vehicle population, passenger, freight traffic and GDP (Chang et al., 2011;Zhang et al., 2007), as shown in following formula: where, Q = Traffic volume α = Adjustment factor in influenced area i x = Influence factors, e.g., travel cost, population density, Per capita income, etc., MTFM predicts future traffic volume from two direct factors, namely, growth rate of population and vehicle travel (Hung and Hong, 2009).The method presumes traffic growth is in proportion to the growth rates of these two factors.It has three steps.Firstly, it forecasts the future population in Kigali; secondly, it considers development tendency of both all-trafficmode travel frequency and proportion of trips using motor vehicle to all-traffic-mode trips to forecast the growth rate of motorized travel frequency based on survey or experience; thirdly, the traffic demand growth rate is calculated by the following formula: where, This project includes 9 sections with most of them located in the downtown of Kigali city.The improvement of current roads should be able to meet the future traffic demand.As can be foreseen with the rapid economic development in Kigali as well as in Rwanda, the urbanization and motorization in Kigali will speed up simultaneously.
Since traffic demand is basically originated from the economic and social activities, the traffic demand forecast is based on the estimation for future economic development, population, city expansion etc.Two population growth scenarios and two GDP growth scenarios were analyzed using different growth rates under different assumptions.
Correspondingly, the traffic forecast also included 2 scenarios, those were, "Lower Traffic Scenario" and "Upper Traffic Scenario".For the former scenario, the social-economic data of Lower Population Scenario and Lower GDP Scenario were adopted while for the latter one those of Upper Population Scenario and Upper GDP Scenario were adopted (Table 1 and 2).
The population-generated traffic demand was forecasted by two methodologies, i.e., Generation Rate Method and Traffic Intensity Method.The result of traffic demand forecast was achieved by aggregating both population-generated traffic demand and portinduced traffic demand.
The flowchart in Fig. 1 summarizes the framework for the traffic demand forecasting.
In this study, two widely used methods which are Elastic Coefficient Method (ECM) and Motorized Travel Frequency Method (MTFM) were comprehensively applied to forecast the traffic volume in Kigali, Rwanda.And Comparative analysis was made between the forecasting result and the actual The growth of population of Kigali will slow down under certain control by the government.It is assumed that the population of Kigali by 2025 accounts for 11% of Rwanda.Upper population scenario: The population of Kigali will grow in a relatively high speed.It is assumed that the population of Kigali by 2025 accounts for 13% of Rwanda.
Table 2: GDP growth scenarios Lower GDP scenario: It is assumed that the GDP per capita will grow slowly with a rate of 2.3 to 2.9% during 2007 to 2025 and the population of Kigali will increase following the Lower Population Scenario.Upper GDP scenario: It is assumed that the GDP per capita will grow rapidly with a rate of 3.3 to 3.8% during 2007 to 2025 and the population of Kigali will increase following the Lower Population Scenario.Compared with the actual survey result, the predicted result of ECM is larger and relative error is 10.49%.The result of MTFM is smaller and relative error is -7.11%.While the weighted average of above methods is closer to the actual result with a relative error in the interval of -5.00 to 5.00%.The research has shown that the combined forecast method proposed in this study, which can make up the defects in accuracy of single model, is easy to operate and owns more accuracy in traffic prediction.This study has suggested that proper combination of several methods would be an advisable trend for the traffic demand forecasting.

SOCIO-ECONOMIC FRAMEWORK
The socio-economic profiles of the areas were discussed in earlier sections of this study with detailed information included in Appendix.Based on the analysis on population growth trend, the population of Kigali was expected to grow at an average rate of 5.6% (Lower Population Scenario) or 6.5% (Upper Population Scenario) per annum during the period 2007 to 2010 and expected to slow down in the succeeding years (Table 3).The GDP of Kigali was expected to grow rapidly and continuously, at an average rate of 8.0% (Lower GDP Scenario) or 10.0% (Upper GDP Scenario) per annum The GDP per capita of Kigali was expected to increase continuously, at an average rate of 0.9% (Lower GDP Scenario) or 2.8% (Upper GDP Scenario) annum during the period 2007 to 2010 and expected to increase further to 2.9% (Lower GDP Scenario) or 4.8% (Upper GDP Scenario) by 2024 with the economic efficiency being improved gradually in the future (Table 4).
Detailed data of population and GDP forecast are shown in Table 1a and 2a in Appendix.

CURRENT TRAFFIC DEMAND OF THE EXISTING ROAD
A survey on the traffic volumes of the existing roads was conducted.The current traffic volume of each road section is shown in Table 5.

Traffic demand forecast:
Elasticity coefficient method: Social-economic activity is the root for traffic to occur and develop and the progress of social-economy will also be promoted by development of traffic.ECM is based on this assumption; it forecasts the future value of elasticity coefficient and determines the traffic demand in future on the basis of analyzing the development trend of future productivity and property structure.
Future traffic demand can be forecasted, adopting ECM, by formula as below: where, t r = Growth rate of traffic demand (%) e = Elastic coefficient g r = Growth rate of economy, this report adopted GDP growth rate (%) In the light of general experience in various cities and with consideration of the current traffic situation in Kigali, the values of elastic coefficient of this project are as below: Based on the forecast of GDP growth rate in Kigali, with the elasticity coefficient in Table 6, the traffic demand growth rate of this project could be calculated.The results are shown in Table 3a in Appendix.
According to current traffic volume and the formula as follow, we could forecast the traffic volume in next few years: The brief results are shown in Table 7.
In this study, average traffic demand of the project means the weighted average traffic volume calculated by the following formula: where, T = Average traffic demand of the project T j = Traffic volume of road section j L j = Length of road section j Motorized travel frequency method: According to general experience in various cities and consideration of the current income level and traffic situation in Kigali, the motorized travel frequency is taken as 0.8 at 2006 and gradually increased to 1.0 in 2024.
According to forecast result of the population growth rate in next few years, with the formula and coefficient, we could calculate the traffic demand growth rate of this project.The results are shown in Table 4a in Appendix.
Gone upon current traffic and the formula as follow, we could forecast traffic volume of every road section of this project: The brief results are shown in Table 8.

Motorized travel frequency
Lower traffic scenario ------------------------------------------------Upper traffic scenario ------------------------------------------------ economic development.The relative errors of both methods are larger.The weighted average of two methods, which takes account of influence of economic and population, is more accurate than each of them with a relative error in the internal of -5.00 to 5.00%.It is easy to operate as well as compensating shortage of a single method to combined ECM and MTFM.This points out that the proper combination of several methods should be one of the developmental trends for the traffic demand forecasting.More research is encouraged to analyze the relationship between weight distribution and economic development as well as travel frequency in follow-up study so that the results obtained in this study can be generalized further.

CONCLUSION
This study analyzes two widely used methods which are Elastic Coefficient Method (ECM) and Motorized Travel Frequency Method (MTFM) was comprehensively applied to forecast the traffic volume in Kigali, Rwanda.And Comparative analysis was made between the forecasting result and the actual survey traffic result in the project's future years.Compared with the actual survey result, the predicted result of ECM is larger and relative error is 10.49%.The result of MTFM is smaller and relative error is -7.11%.While the weighted average of above methods is closer to the actual result with a relative error in the interval of -5.00 to 5.00%.The research has shown that the combined forecast method proposed in this study, which can make up the defects in accuracy of single model, is easy to operate and owns more accuracy in traffic prediction.This study has suggested that proper combination of several methods would be an advisable trend for the traffic demand forecasting.

Fig. 1 :
Fig. 1: Framework for traffic demand forecasting growth rate in Year i i T = Traffic volume in Year i growth rate in Year i i T = Traffic volume in year i

=
Predicted traffic volume in year i ie A , im A = Results of ECM and MTFM in year i e error of forecasted traffic volume in year i (%) i B = Actual survey traffic volume in year i --------------------------Motorized travel frequency method -

Table 1 :
Population growth scenarios Lower population scenario:

Table 3 :
Estimated population growth of kigali

Table 4 :
Estimated growth rate of GDP/GDP per capita of kigali

Table 5 :
Current traffic volume of the existing road

Table 9 :
Result of traffic forecast of the project (pcu/day)

Table 3a :
Traffic growth rate by elasticity coefficient method

Table 4a :
Traffic growth rate by motorized travel frequency method