OPTIMIZATION OF ROAD MAINTENANCE AND REHABILITATION ON SERBIAN TOLL ROADS

Paper Nº ICMP123 8th International Conference on Managing Pavement Assets OPTIMIZATION OF ROAD MAINTENANCE AND REHABILITATION ON SERBIAN TOLL ROADS Goran Mladenovic 1*, Jelena Cirilovic 2 and Cesar Queiroz 3 1 University of Belgrade, Faculty of Civil Engineering, Assistant Professor, Serbia 2 Institute IMS, Serbia 3 University of Belgrade, Faculty of Civil Engineering, Visiting Professor, Serbia * Corresponding Author s Email: emladen@imk.grf.bg.ac.rs ABSTRACT The paper presents the application of World Bank s model RONET to a strategic network level analysis of the Serbian toll road network. Tolls have been collected on these roads since the 1980s. Despite the toll revenues, the condition of this network deteriorated considerably during the 1990s. In recent years a more substantial part of the toll revenues has been applied to the toll roads, thus gradually improving the condition of such roads. The goals of the study are to obtain the optimum maintenance and rehabilitation (M&R) strategy and related budget, estimate the impact of different funding levels on the future quality, and estimate the economic consequences of budget constraints for maintenance and rehabilitation of the toll road network. The analysis shows that a substantial part of the toll revenues can be allocated to the non-tolled part of the Serbian road network without detrimental impact on the condition of the tolled network. Application of the RONET model to the prevailing conditions on the Serbian toll road network led to an optimal M&R strategy with a good balance between rehabilitation, periodic and recurrent maintenance. Implementation of the Optimal M&R strategy would keep up the relatively good current condition of the toll road network. In other words, the current level of maintenance is close to the optimal. Implementation of higher than optimal M&R standards would lead to substantially higher road agency costs and consequently lower net benefits, while the implementation of lower than optimal M&R standards would lead to considerably worse network condition and higher vehicle operating costs. KEY WORDS: toll road network, maintenance and rehabilitation (M&R) strategy, budget constraints. INTRODUCTION Serbia is located in South East Europe (Figure 1) and extends for 77,500 square km with a population of 7.5 million inhabitants. The total length of the road network in Serbia is about 38,600 km, including 634 km of motorways and half/motorways, 4481 km of main roads, 10407 km of regional roads and 23,084 km of local roads (1). Serbia s road network density is comparable to that of other Eastern European countries. Serbia's geographic position puts it at the crossroads of the two important Pan-European corridors (Figure 1). One is connecting Western Europe and the Near East, the other is linking Greece to Central Europe. This network is known as Corridor X. All existing toll motorways in Serbia are constructed along the Corridor X. However, despite their importance, nationally and

Mladenovic, Cirilovic and Queiroz internationally, currently not all sections of Corridor X in Serbia are of Motorway Standard and the remaining sections on north to the Hungarian border, on south toward Macedonia and on east toward Bulgaria are currently under construction and expected to be completed by the end of 2012. FIGURE 1 Location of Serbia and country s toll road network. The state road network, including motorways, was until beginning of 2009 managed by the Public Enterprise Roads of Serbia (former Road Directorate). In May 2009 the new company Corridors of Serbia was founded with the primary role to complete the construction of motorway network along the Corridor X. However, the PE Roads of Serbia remained responsible for the maintenance of the existing motorways. The construction of new sections is financed primarily by loans from International Financial Institutions (IFI), while the road network maintenance is financed from tolls, state budget, excise tax on fuel and other sources. Traffic levels on the state road network have varied substantially in the last 20 years due to economic and political changes in the region and in the country itself. Currently, AADT on the motorways range between 1000 and 130,000 vehicles per day (vpd) (2). The condition of the existing motorways deteriorated considerably during the 1990s due to under-financing of operations and maintenance. In recent years, as the financial situation in Serbia has improved, financing for the road sector has gradually increased, focusing on the most hazardous and highly trafficked parts of the road network. The result was that most of the motorway network has been rehabilitated since 2000. However, due to inadequately selected maintenance treatments and/or relatively poor quality control, some sections deteriorated faster than expected and will in next few years require additional corrective maintenance treatments. The objectives of the present study are to:

(i) Obtain the optimum maintenance and rehabilitation (M&R) strategy and the related budget for the motorway road network; (ii) Estimate the impact of different funding levels on the future quality of the country s motorway network; and (iii) Estimate the economic consequences of budget constraints for maintenance and rehabilitation of the motorway network. The analysis will also show how much of the toll revenues can be allocated to the non-tolled part of the Serbian road network without detrimental impact on the condition of the tolled network. The World Bank s model RONET (Road Network Evaluation Tools), designed to assess the current characteristics of road networks and their future performance depending on different levels of interventions (and budgets) to the networks, was selected for this study. CURRENT CONDITION OF THE MOTORWAY NETWORK A condition survey of the entire state road network in Serbia has been recently completed and has provided valuable data source for this study (3). The survey was quite comprehensive and included pavement and other road assets inventory and condition data, as well as traffic levels at a relatively high level of detail. The following road parameters were measured among others: Geometric parameters of roads Pavement roughness Skid resistance Pavement strength Surface distress The data are stored in a relational database in such a format that they can be conveniently used in models such as RONET and HDM-4. The overall traffic on motorway network is still relatively low, with 90 percent of the network having AADT lower than 10000 vpd (2). It was decided to implement a uniform rather than logarithmic split between traffic bins since it provided more equal (reasonable) split. 22 percent of network have AADT lower than 5000 vpd, 15 percent has AADT between 5000 vpd and 6000 vpd, 29 percent has AADT between 6000 vpd and 7000 vpd, while 24 % has AADT between 7000 vpd and 10000 vpd. Overall, the Serbian motorway network is currently in relatively good condition with an average IRI = 2.1 m/km. Of the total motorway length, 48 percent is in very good condition (IRI < 2.0 m/km); 27 percent in good condition (IRI between 2 m/km and 2.5 m/km); and 25 percent in fair, poor or very poor condition (IRI > 2.5 m/km). Figure 2 presents the overall toll road network condition split by traffic category. The part of the network with highest AADT is currently in the poorest condition. However, these sections of motorway that go through Belgrade are currently under reconstruction, expected to be finished by the end of 2011. Other sections that are in condition worse than average are typically sections with relatively low AADT.

Mladenovic, Cirilovic and Queiroz 350.0 5 Half- motorways Lenght (km) 300.0 250.0 200.0 150.0 100.0 50.0 4.5 4 3.5 3 2.5 2 1.5 Average IRI (m/km) Very Poor Poor Fair Good Very Good average IRI 0.0 1 AADT (vehicles/ day) FIGURE 2 Overall motorway network split by AADT range and condition. OVERVIEW OF REVENUES AND EXPENDITURES IN THE PERIOD 2006-2010 The overview of overall revenues of PE Roads of Serbia is presented in Table 1 (4). The revenues are grouped in three categories: state budget, excise taxes and other resources, toll revenues and loans from International Financial Institutions (IFI) that have been extensively used in recent period for financing construction of new sections and rehabilitation of the most trafficked roads, including toll motorways. TABLE 1 Revenues of PE Roads of Serbia in period 2005-2010 (mil. ) Revenues Year 2005 2006 2007 2008 2009 2010 Budget, excise tax and other resources 151.59 119.50 162.70 252.52 193.63 212.06 Toll revenues 116.68 152.63 183.66 202.62 146.08 154.05 Loans from IFI 70.02 60.40 50.72 41.18 111.07 99.02 Total (mil. ) 338.29 332.53 397.08 496.32 450.78 465.13 The toll revenues represented between 32.4 percent and 45.9 percent of total revenues of PE Roads of Serbia. The overview of main expenditures is presented in Table 2. TABLE 2 Expenditures of PE Roads of Serbia in period 2005-2010 (mil. ) Year Expenditures 1 2005 2006 2007 2008 2009 2010 2

New construction / Reconstruction 46.39 80.57 79.21 85.39 132.70 143.68 Maintenance 270.70 340.72 333.81 314.83 265.54 194.46 Roads 262.66 329.49 317.89 308.10 251.15 178.72 Routine/Prev./Rehabilitation 240.67 312.45 295.95 264.92 202.87 141.88 Winter maintenance 21.99 17.03 21.93 43.19 48.28 36.84 Structures 8.04 11.24 15.92 6.73 14.39 15.74 Note: 1 The operation cost and annual loan repayments are not included in the table 2 Projection The road maintenance budget in last six years ranged between 140 million and 310 million. Such budget applies to the entire state road network of approximately 15000 km of main and regional roads, including 634 km of motorways and half-motorways. DESCRIPTION OF THE RONET MODEL The Road Network Evaluation Tools (RONET) is a model that can be used by decision makers to assess the current state of the road network, its relative importance to the economy (e.g. asset value as a percentage of GDP) and to compute a set of monitoring indicators to assess the performance of the road network. RONET simulates the future performance of the road network under different road maintenance standards and different levels of funding. It determines, for example, the minimum cost for sustaining the road network in its current condition and estimates the savings or the costs to the economy for maintaining the network at different levels of service. RONET determines the optimal maintenance standard for each road and determines the funding gap, defined as the difference between current maintenance spending and required maintenance spending. The Road User Revenue Module estimates the level of road user charges required (e.g., toll and fuel levy) to meet road maintenance expenditures under different budget scenarios. This could be used by road agencies to prepare a business case to negotiate and revise road tariffs on a sound basis. RONET is publicly available from the World Bank website at: http://go.worldbank.org/a2qqyznfm0. CONFIGURATION OF RONET FOR LOCAL CONDITIONS To ensure that the RONET model simulates the conditions that realistically reflect the Serbian toll road network and provides reasonable outputs, it was considered important to: Adjust the model s decision matrix so it can accommodate more realistic assumptions for motorways Calibrate RONET pavement deterioration models and road works effects models (5) Develop a set of road user costs models for the local vehicle fleet Configure RONET analysis parameters with local conditions and maintenance practice The RONET model uses a simplified road deterioration model for paved roads, developed by Archondo-Callao (6): where: ( K ( ) ) gm m t 5 diri = K α e 1+ SNC α YE4 + α t + K m IRI gp o 1 2 gm a (1)

Mladenovic, Cirilovic and Queiroz diri = annual roughness increment (m/km), m = environmental coefficient, t = pavement age since last overlay, reconstruction or new construction (years), SNC = modified structural number, SNC = 0.0394 ai Hi + SNSG a i = layer coefficients (see Table C2.7 in 3), H i = layer thicknesses (mm), SNSG = 3.51 log CBR 0.85 ( log CBR ) 2 1.43 CBR = subgrade California bearing ratio (%), YE4 = annual number of equivalent standard axles (million ESA/lane-year), IRI a = pavement longitudinal roughness at start of the year, expressed in terms of the International Roughness Index (m/km), α 0, α 1, α 2 = model coefficients, and K gp,k gm = calibration factors (7). The model coefficients and calibration factors, as well as the environmental coefficient, can be specified by the user. The default value of coefficient α 0 is 134, the same value as recommended in the HDM-4 (see Table 2.32 in 8) model. Coefficient α 1 reflects the reduction of strength of the pavement due to presence of cracking and its default value is 0.7947. Coefficient α 2 reflects the increase in roughness progression of the pavement due to presence of cracking, rutting, and potholes, and its default value is 0.054. Calibration factor K gp based on comparison with already calibrated HDM-4 deterioration models (5) and the value of 1.2 was used in the analysis. The appropriate value of the environmental coefficient m, for a country such as Serbia, is between 0.035 and 0.060 ( Temperate-cool to Temperate-freeze temperature classification and Semi-arid to Sub-humid moisture classification) (see Table C2.31 in 8). Values of 0.035 and 1.0 are used for the environmental coefficient m and the environmental calibration factor K gm respectively. A set of road user costs models for typical local vehicle categories was developed using the RUCKS model (9) that is based on HDM-4, version 1.3 road user effects equations. The models have the following shape: 2 3 o 1 2 3 URUC($ / vehicle km) = a + a IRI + a IRI + a IRI (3) where: URUC = unit road users cost ( /vehicle-km), IRI = pavement longitudinal roughness (m/km), and a 0, a 1, a 2, a 3 = model coefficients that are input to RONET. RONET defines five default treatments that should be applied depending on the road condition, which means that one maintenance treatment is applicable only to the road in a certain condition. The intensity and cost of maintenance treatment can be adjusted in the program. Maintenance treatments used in the study of toll road network in Serbia are summarized in Table 3. TABLE 3 Maintenance Treatments with Associated Unit Costs

Road surface type Asphalt Road condition IRI range (m/km) Maintenance treatment Periodic maintenance treatment cost ( /km) Routine maintenance cost ( /km) Very Good < 2.0 Routine maintenance (RM) 5,954 Good 2.0 3.0 RM+Preventive maintenance 25,000 6,750 Fair 2.5 3.5 RM+Mill & Overlay 50 mm 108,946 5,329 Poor 3.0 4.0 RM+Mill & Overlay 120 mm 225,773 13,528 Very Poor >3.5 RM+Reconstruction 325,911 17,700 The unit costs presented in Table 1 are applicable for half-motorway. ROAD MAINTENANCE STANDARDS RONET defines seven default maintenance standards. The standards are based on different trigger levels for each maintenance treatment. The Very High standard represents a withoutbudget-constraints scenario with a high level of periodic maintenance and rehabilitation works. The High, Medium, Low and Very Low standards represent scenarios of reduced levels of road works expenditures. The Do Minimum standard represents a scenario where only reconstruction is applied at a relatively high road roughness, while no capital road works are applied over the evaluation period. In addition, it is possible to define custom standards. The maintenance standards for toll roads applied in this study are presented in Table 4. TABLE 4 Evaluated Maintenance Standards for Paved Roads Maintenance Standard IRI 3.0 Preventive Roughness Range and Required Maintenance Treatment 2.5 < IRI 3.5 3.0 < IRI 4.0 Mill & Overlay Mill & Overlay 50 mm 120 mm 3.5 < IRI 4.5 Reconstruction Code Name Roughness Threshold, IRI (m/km) A Very High Standard 2.00 2.50 3.00 3.50 B High Standard 2.25 2.75 3.25 3.75 C Medium Standard 2.50 3.00 3.50 4.00 D Low Standard 2.75 3.25 3.75 4.25 E Very Low Standard 3.00 3.50 4.00 4.50 F Do Minimum - - - 5.00 G Do Nothing - - - - Note: IRI units are m/km The routine maintenance costs presented in Table 3 are adopted for the Very High Standard. These values were gradually reduced for lower maintenance standards, down to zero for the Do Nothing standard. NETWORK-LEVEL MAINTENANCE STRATEGIES For each road class, defined by road condition and traffic level, RONET was used to evaluate several maintenance standards, including the Optimal one, defined as the standard that leads to the lowest total society costs and highest Net Present Value (NPV). Total society costs are the sum of road agency and road users costs. The Optimal strategy at network level consists of optimal strategies for each road class. The program also evaluates two higher ( Optimal+1 and

Mladenovic, Cirilovic and Queiroz Optimal+2 ) and three lower ( Optimal-1, Optimal-2 and Optimal-3 ) strategies that consists of application of higher and lower maintenance standards compared to the optimal one for each road class. The base maintenance strategy, used for the calculation of benefits, is the Do minimum strategy that consists of the application of pavement reconstruction at relatively high roughness level. In addition, the program evaluates the Do nothing strategy that consists of no maintenance works. Table 5 presents the total agency and users costs and net benefits at a discount rate of 8 percent, using an analysis period of 20 years. For the Serbian toll road network, the total road agency costs range between 0 and 247 million; the corresponding discounted annual budgets for maintenance range between 0 and 21.8 million. The economic internal rate of return for the optimal strategy is 18.5 percent, as calculated by RONET. The Optimal strategy consists typically of the application of Very low or Do minimum maintenance standards as a result of relatively low traffic level on most of the toll road network and its relatively good current condition. The required annual maintenance and rehabilitation budget for the Optimal scenario is about 15.4 million per year, which represents annual requirements of about 25,000 per kilometer. TABLE 5 Total Discounted Costs and Benefits and other parameters for All Maintenance Strategies Total costs in years 1 to 20 Internal Average Network Net Road Road Total Rate of Roughness maintenance Benefits Agency Users Society Return IRI strategy mil. mil. mil. mil. % m/km Optimal -1 187 17866 18052 0-2.38 Optimal 169 17806 17975 77.2 18.5 2.16 Optimal +1 189 17819 18008 44.0 12.8 2.01 Optimal +2 247 17840 18086-34.0 4.2 1.87 The two strategies that include higher maintenance standards ( Optimal+1 and Optimal+2 ) would result in lower net benefits for much higher Agency costs. However, these strategy would also result in slightly improved network overall condition. Strategy Optimal -1 is essentially equal to Do minimum strategy and provides higher user and agency costs resulting in network deterioration over analysis period. Figure 3 presents the distribution of the total road works requirements during the analysis period between rehabilitation, periodic and recurrent maintenance for different M&R strategies.

Total Road Agency Costs (million euro) 500 450 400 350 300 250 200 150 100 50 0 Optimal+2 Optimal+1 Optimal Optimal-1 Rehabilitation Periodic Maintenance Recurrent Maintenance FIGURE 3 Distribution of the total road works requirements. The periodic maintenance costs represent a major part of the total M&R costs, between 216 million and 281 million. The recurrent maintenance costs vary between 47 million and 127 million. The rehabilitation costs are relatively substantial only for the Optimal -1 strategy. Figure 4 presents toll network performance under different maintenance strategies over the 20- year evaluation period. IRI (m/km) 10 9 8 7 6 5 4 3 2 1 0 0 5 10 15 20 Year Do nothing Optimal-1 Optimal Optimal+1 Optimal+2 FIGURE 4 Roughness over the analysis period for different maintenance standards. It was found that the Optimal strategy keeps up the relatively good current condition of the toll road network. The Optimal+1 and Optimal+2 strategies maintain the network continuously at a slightly higher quality level. The Optimal-1 strategy provides significantly lower IRI in the first 12 years, after which the condition significantly improves. Figure 5 illustrates the predicted network condition at the end of the 20-year analysis period.

Mladenovic, Cirilovic and Queiroz 100% Percentage of toll road network length 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Optimal+2 Optimal+1 Optimal Optimal-1 Very Poor Poor Fair Good Very Good M&R Strategy FIGURE 5 Impact of funding level on network condition at the end of the 20-year analysis period. The Optimal strategy would lead to a considerable improvement in the condition of the road network, throughout almost the entire evaluation period. The percentage of roads in very good and good condition slightly increases compared to the current condition. Current annual toll revenues are about 154 million Euros. The annual needs for pavement maintenance and rehabilitation are about 17 million Euros for the optimal alternative (equivalent to a discounted total of 169 million Euros, over the 20-year analysis period). The annual operating costs (e.g., toll collection and users support) and annual debt service (e.g., for loans used for rehabilitation and construction of new sections in recent years) can be estimated respectively at 17 million Euros and 50 million Euros. Consequently, it appears that an annual amount of about 70 million Euros of toll revenues can be used for maintenance and rehabilitation of other parts of the network, or for construction of new motorway sections, without detrimental impact on the condition of the tolled network. While the RONET model has been used in other countries, reportedly with satisfactory results, it would be convenient to validate its application and methodology for Serbian conditions. This is proposed as further research, and will be accomplished by applying the methodology described in this paper in future years, when more historical data will be available on the condition of the Serbian toll road network. CONCLUSIONS The paper presented the application of World Bank s model RONET to a strategic network level analysis of the Serbian toll road network. The current state of the network is good, with large variations in traffic range. Application of the RONET model to the prevailing conditions on the Serbian toll road network led to an optimal M&R strategy with a good balance between rehabilitation, periodic and recurrent maintenance. The corresponding ( optimal ) budget for the Serbian toll road network was found to be 169 million over the 20-year analysis period. The implementation of the Optimal M&R strategy would keep up the relatively good current condition of the toll road network. Implementation of higher M&R standards would lead to substantially higher road

agency costs and consequently lower net benefits, while the implementation of lower M&R standards would lead to considerably worse network condition for approximately the same or even higher agency costs. The analysis also shows that a substantial portion of toll revenues can be used for maintenance and rehabilitation of other parts of the road network, or for construction of new motorway sections, without detrimental impact on the condition of the tolled network. REFERENCES 1. Road network in the Republic of Serbia, PE Roads of Serbia, http://www.putevisrbije.rs/en-gb/putna-mreza/road-network-in-the-republic-of-serbia, Accessed Feb. 22, 2011. 2. Traffic counting on the Main Road Network in the Republic of Serbia, Annual Average Daily Traffic in 2009, PE Roads of Serbia, 2010. 3. Transport Rehabilitation Project, Road Database, Final Report, PE Roads of Serbia, 2009. 4. Strategic Study with Proposal of Medium-term Program of Development, Reconstruction, Maintenance and Protection of State Road Network in Period 2010-2015, PE Roads of Serbia, 2010. 5. Transport Rehabilitation Project, Road Database, Final Report, Appendix G, HDM-4 Pavement Deterioration Model Calibration to Local Conditions in Serbia, PE Roads of Serbia, 2009. 6. Archondo-Callao, R. Road Network Evaluation Tools (RONET), Version 2.00, User s Guide, The World Bank, 2009. 7. Bennett, C.R. and W.D.O. Paterson. Highway Development and Management Series, Volume 5, A Guide to Calibration and Adaptation, PIARC/AIPCR and World Bank, 2000. 8. Odoki, J.B. and H.G.R. Kerali. Highway Development and Management Series, Volume 4, Analytical Framework and Model Descriptions, PIARC/AIPCR and World Bank, 2000. 9. Archondo-Callao, R. HDM-4 Road Use Cost Model Documentation, Version 1.20, User s Guide, The World Bank, 2007.