ENVIRONMENTAL ENGINEERING

Transcription

1 ENVIRONMENTAL ENGINEERING May 22-23, 2008 The 7 th International Conference Faculty of Environmental Engineering Vilnius Gediminas Technical University Saulėtekio ave 11, LT Vilnius, Lithuania Phone: ; Fax.: ; ap2008@ap.vgtu.lt DEVELOPMENT OF A SYSTEM OF AUTOMATIC SPEED CONTROL ON POLAND S NATIONAL ROADS Stanislaw Gaca Cracow University of Technology, Institute of Road and Railway Engineering, Warszawska 24, Cracow, Poland, sgaca@pk.edu.pl Abstract. The paper gives an overview of a proposed system for automatic speed control on Poland s national roads. Work on the implementation first started in 2007 and has been continued as part of a programme to reduce accidents on national roads. In the first phase some 700 measurement stations will be put up. The speed cameras will be located in high risk sites that have been chosen after in-depth studies of speed and its effects on accidents. The paper presents some of the results.. Keywords: traffic safety, speed control, speed cameras, accident prediction models 1. Introduction While Poland s national roads make up a mere 6% of the entire road network, they carry 38% of traffic. In recent years 21% of all accidents and more than 37% of all road deaths occurred on national roads. Given this high number of accidents and deaths, the decision was made to introduce a special road safety programme on national roads. In developing the programme, the focus was on the following main road safety problems [1]: high accident severity with as many as 13 fatalities per 100 accidents in built-up areas and 24 deaths outside built-up areas per 100 accidents; unusually high number of pedestrian accidents (22%) compared with the actual pedestrian traffic volume. Pedestrians make up on average 29% of total road deaths on national roads, but there are strong differences from region to region. Pedal cyclists are equally at risk with an average of 6% of all injuries and 9% of all killed, while they only make up some 1 2% of traffic; a high number of fatalities in head-on collisions at 29% of all fatalities on national roads with head-on collision making up 17% of total accidents; accident risk at junctions usually involving side impact accidents at 24% of accidents; the presence of roadside lateral obstacles and frequent collisions with these on average 10% of all accidents, but with regional variations (7.2% 14.8%); speed not appropriate to the conditions being the most frequent cause of driver related accidents. On average in a year 26% of accidents involve speeding and claim 29% of injuries and 26% of all road deaths; failure to observe speed limits in non built-up areas and on roads passing through towns. The problems identified on national roads indicate clearly the effect speed has as a direct and indirect cause of accidents. This is why effective speed management is considered a top priority among road safety policies and is to be achieved through legal regulations, planning, road infrastructure, enforcement, campaigns and advanced technologies. Reducing the number of speeding drivers and introducing new local speed limits has a lot of potential and can significantly bring down road accidents and casualties. One of the measures that can reduce the actual speed is automatic speed control. But the effectiveness of this measure is frequently limited and drivers tend to change their behaviour only where the control occurs. This is why it is important to select the right place for a speed camera and link the site to a real accident risk. Below is a description of the conditions for organising automatic speed control systems on national roads. 2. Speed on Polish roads and how it affects road safety The reason behind the decision to deploy a speed control system on national roads is the studied effect of speeding on accidents and the phenomenon of moving at speeds significantly higher than considered safe for the conditions. While the effect of speed on accidents seems to be clear, it continues to be studied: 1141

2 a) experts carry out detailed analyses of why selected groups of accidents occurred; b) before and after analyses to identify the changes in accident numbers or the values of accident indicators recorded for different speeds (before and after speed reduction measures were introduced) in comparable road and traffic conditions; c) attempts to identify empirical relations between accident indicators and vehicle speed parameters. Regression models are usually used to describe these relations. Because of the design of regression models, the quantitative description of phenomena they identify covers the studied space only. The causal links identified in before and after studies are more obvious than they are in studies of randomly selected road networks (regression analyses). But the opportunities for before and after analyses are relatively limited and it takes a long time to accumulate the data. This explains perhaps the popularity of regression models; d) analytical models are built to describe vehicle movement. These models are useful for studying the effects of accidents under the assumption that the effects are proportionate to the kinetic energy of the vehicles when they crash. Analytical models are developed into simulation models with vehicle running tracks, vehicle deformation (damage) upon crash and the location after the crash; e) tests are carried out using real objects and physical models. The authors of the paper mainly used before and after studies and regression models [2]. The following are some general conclusions from the studies: while it is possible to describe quantitatively the effects of speed on accidents and casualties, speed must be seen as a component of a set of factors that determine the accident risk in regression models speed is only one of the variables; there are strong links between the effects of speed on safety and the technical standards of roads and junctions and roadside development. Road accessibility comes up as an important factor; in many cases higher speeds and limited accessibility do not always mean deteriorated road safety; road user behaviour, including their choice of speed, depends on the type of road and roadside development; excessive speed and a variety of speeds within a traffic stream are frequent causes of accidents; pedestrians and pedal cyclists are at particular risk when involved in accidents caused by speeding. Risk exposure variables have an important role to play in regression models of accident prediction as illustrated in this model developed using data from 62 non built-up : AC = 0,484 0,737 0,0195 V A + 0,044 p 0,0018 Q L e hv (1) where: AC number of accidents annually on a road section L, Q daily traffic volume [P/24h], L length of section [km], p hv percentage of heavy goods vehicles using traffic [%]. In relation (1) the effect of average speed on the number of accidents is as expected, just as the effect of the percentage of trucks. Variable p hv may express indirectly the effect of speed dispersion on accidents. Model (1) explains only 45% of the changeability of the number of accidents AC for reasons different from those expected based on the Poisson distribution. The match is much better (75%) in the non built-up area accident prediction model the EURO model [3]: AC = 5,663 Q 0,748 L 0,847 V exp(0,038 LJ 0,056 B j 2,492 A PV + 0,023 V 0,114 where: PV percentage of drivers exceeding the speed limit [%], LJ number of junctions on analysed section, B j carriageway width [m], V p speed limit. Where roads pass through small and medium towns, a sample of 62 has produced the following regression model for accident prediction [2]: 0,492 0,715 AC = 0,0086 Q L exp(0,137 Bj + + 0,11KL 0,12KL + 0,02KL ) A B where Q and L B j have the same meaning as in formulas (1) and (2). KL A, KL B and KL C are symbols denoting the carriageway class (KL A pavements, dense development, KL B carriageway without pavements, dense or medium density development, KL C - carriageway without pavements with dominance of through traffic, loose development or medium density development). Variables KL A, KL B and KL C take on value 1 if the carriageway qualifies for the class and 0 in the other cases. Model (3) explains 71% of the variability of the number of accidents AC for reasons different from those resulting from the Poisson distribution. The effect of speed is indirectly included in this model by taking account of the carriageway class. This is because of the strong correlation between the average speed and these variables. The examples of accident prediction models demonstrate the effects of speed and speed limits on road safety. The results can be used to argue the case for intensified speed control on national roads notorious for exceeding speed limits. This is confirmed in studies, which the author has been co-running for the National Road Safety Council since 2002 [ ]. Table 1 gives the average speeds (based on daily data) V A, the V 85 percentile and percentage of cars exceeding the speed limit PV on national roads in different groups. These values are calculated for vehicle stream data collected on 115 road (427 daily measurements) in non built-up areas and from 89 of roads passing through small towns (124 daily measurements). Table 1 gives the average values of the parameters and the minimum and maximal daily values in parentheses. The average values are calculated for C p ) (2) (3) 1142

3 different seasons. Speed values tend to be higher during seasons with good weather. Table 1. Average daily speed parameters on national roads by type of carriageway limit) are calculated for 85 of roads passing through towns. This is given in Table 3. Table. 2. Speed parameters in before and after tests in comparable samples of roads passing through towns between 5.00 a.m. and p.m. Location of section Built-up Non built-up Type of carriageway *) Speed parameters V A V 85 PV [%] Z1 ( ) ( ) ( ) Z2 ( ) ( ) ( ) M3 ( ) ( ) ( ) Z1 ( ) ( ) ( ) Z2 ( ) ( ) ( ) Z2+D ( ) ( ) ( ) Z3 ( ) ( ) ( ) *) Z1 single carriageway 7.0 m wide and bit. shoulders Z2 - single carriageway 7.0 m wide and hard shoulders Z2+D as Z2 with trees in road crown Z3 - single carriageway < 7.0 m wide and hard shoulders M3 - single carriageway m wide with pavements Apart from seasonal speed changes, there are also distinct changes in driver behaviour regarding speed during the day. Where roads pass through towns, nighttime speed of cars is significantly higher than during the day. What makes this particularly bad is the reduced visibility at night-time potentially leading to increased accident risks. However, both day and night-time values of the V 85 percentile are above the speed limits. In evaluating the behaviour of drivers using national roads, it is important to study driver responses to the new speed limit in built-up areas. There was introduced a change in speed limit in Poland in May 2004 from 60 km/h to 50 km/h between 5.00 a.m. and p.m. The results of some early studies of driver response are given in [7, 8]. Before and after data from 25 towns are given in Table 2. It presents the average speeds in a 50 km/h area. The conclusion from Table 2 is that samples from different sites show a statistically significant reduction in average speed V A and V 85 percentiles. However V A and V 85 continue to be seriously over the speed limit of 50 km/h. The data are derived from a relatively small sample and should be seen as preliminary results. The studies will continue. For the purpose of speed enforcement it is important to identify how much faster cars go above the limit. This is described using parameter DV which is the difference between the speed of cars over the limit and the value of the limit. The average values over the limit DV A during the day (50 km/h speed limit) and night (60 km/h speed 1143 Z1 *) carriageways Parameter V A V 85 Before After Difference Z2 carriageways Before After Difference M3 carriageways Before After Difference *) As in Table 1. Table 3. Average values over the limit DV A and standard deviation S DV within the sample of analysed sites where roads pass through towns. Type of carriageway *) Daytime: 5.00 a.m p.m DV A S DV Night-time: p.m 5.00 a.m. DV A S DV Z Z M *) As in Table 1. The results clearly show that cars exceed the speed limits both day and night. This problem should be addressed by intensifying speed enforcement. However, the sample included a significant spread of speeds over the limit for each group of in the different carriageway types. The results of the tests show that speed limits compliance is very poor on national roads, and things are particularly bad where the roads pass through towns. This presents a new argument in support of an automatic speed control system. 3. Selecting speed cameras sites Based on foreign experience and a diagnosis of accident risks including speed tests, the following criteria have been formulated for selecting speed camera locations and methods of enforcement:

4 speed enforcement and control should be ensured on sites where speeding accidents are particularly frequent. Equally important are high risk sites with speed having an indirect effect on the causes of accidents. Before selecting the locations for speed cameras, high accident density of roads must be identified where accident rates are significantly above the average; once selected, the should form a network of measurement points on roads that are interlinked within the region; that way we can increase coverage and improve detection rates; while a single speed camera can be moved between several measurement points, speed should be recorded at least once a week for 6 24 hours; apart from stationary speed cameras, mobile devices should be used for measurements on sites selected at random; fixed speed cameras should be resistant against vandalism, but the design may not pose any additional safety risks; fixed speed camera sites must be adequately marked and drivers must be able to see the cameras; existing local speed limits in or around a speed camera site must be reviewed to ensure that drivers can accept and understand them; campaigns to explain the reasons for installing cameras and the expected benefits must be conducted; high risk sites should be identified by searching accident databases for section of 1 km with steps every 100 m. The identification criterion should be related to average accident density or other accident indicators. This could be e.g. the number of accidents on a 1 km section within 5 years calculated from the Poisson distribution: λ k e λ P( k) = 0,05 k! (4) where λ is the average value of the accident indicators and k is the boundary value of the number of accidents (integer); the critical should be identified for accidents directly or indirectly caused by excessive speed; with no speed control must not be too long. This is why the boundary values of indicators for identifying high risk sites should be determined based on average accident risks on the particular of the network rather than on the entire road; Once selected, of roads must be verified on site with special focus on: road design that encourages high speeds and whether that could be changed; errors in road infrastructure that could be a possible cause of accidents; the clarity of local speed limits, if there are any, and the possibility to use additional road measures to reduce speed. While the main objective of building a network of speed cameras is to reduce road deaths, all accidents must be taken into account in identifying high risk sites. Earlier the relations between accident density and casualty rates were analysed using data from high risk sites (Fig. 1). It was found that there is a relatively strong link between accident density and injury density but a significantly weaker relation between accident density and fatality density. Despite that it was agreed that given the random and insufficient fatality accident data as well as data inaccuracy, speed camera sites cannot be selected based on fatality accidents only. Casualty density [number/1km/5yrs] Injured: y = 0,025x 2 + 0,942x + 2,28 R 2 = 0,71 Fatalities: y = 0,142x + 0,645 R 2 = 0, Accident density [acc./1km/5yrs] Fig 1. Relation between accident density and casualty density. In the first stage of automatic speed control implementation on national roads, the roads selected were the core roads in the network (roads numbered 1 to 12). Selection started in 2006/2007, basing on the information on the accidents count in years Table 4 presents the numbers of casualties involved in speeding accidents on these roads. The data refers to all and critical with increased number of accidents. Table 4. Casualty numbers involved in speeding accidents between on all and critical of roads All Critical Percent of critical Length [km] % Injuries % Fatalities % The data provide an accurate picture of how much can be achieved to improve safety through better speed enforcement. On of critical roads 1 12 the plan is to put up 711 measurement stations (speed camera boxes). The process of automatic speed enforcement will include the typical before and after studies to compare speed parameters and analyse the changes in the number and structure of accidents. 1144

5 4. Conclusions To address the high number of accidents occurring on Poland s national roads, including those caused by excessive speed, speed management measures must be undertaken. Implementation of lower speed limits in 2004, from 60 to 50 km/h in build-up areas, was not fully accepted by the drivers and did not change their behaviour in a significant way. Average reduction in the mean speed on analysed roads was between 1,7 and 3,3 km/h. Empirical studies have confirmed drivers disrespect for speed limits, also in rural areas, which calls for more restrictive speed enforcement measures. To that end creation of a system of automatic speed control is under way in Poland based on the rules presented in the paper. Speed cameras add to the other road safety measures which are introduced during road improvements. Early experience from the implementation of the automatic speed control system has highlighted the following problems: - the main objective of building a network of speed cameras is to reduce road deaths. However, given the random nature and small number of fatality accidents there is little room for using statistical methods for selecting speed camera sites based on the fatality density criterion only; - because accident databases are not entirely reliable, speed camera sites cannot be selected on this basis alone and additional on sites checks had to be conducted; - in introducing an automatic speed control system it is important to obtain public acceptance by informing the public about the location of the sites and reporting back on the results; - in introducing an automatic speed control system it is important to carry out studies of the effect of speed cameras on speed, changes in accident numbers and structure and public support. The results can be helpful in planning the location of new or existing speed cameras. Such studies are already under way. References 1. Jamroz K. et al.: Traffic Road Safety Program for National Roads in Poland GAMBIT-DK Diagnosis (Program Bezpieczeństwa Ruchu Drogowego dla Dróg Krajowych w Polsce GAMBIT-DK diagnoza stanu bezpieczeństwa ruchu drogowego na drogach krajowych). FRIL- PG-PK, Gdańsk/Kraków, 2007, p. 70 (working report in Polish) 2. Gaca S.: Vehicle Speed Research and the Effect of Speed on Road Safety (Badania prędkości pojazdów i jej wpływu na bezpieczeństwo ruchu drogowego). Zeszyty Naukowe PK, Nr 75, Kraków, 2002, p. 196 (in Polish) 3. Baruya A.: Speed-accident relationship on European roads. Proceedings of the Conference Traffic Safety in Europe. VTI konferens 10A, Part 10. Swedish National Road and Transport Research Institute, 1998, p Gaca S. Jamroz K., Ząbczyk K. et al.: Analysis of selected aspects of road users behavior. Final Report. Signalco Trafik Heusch Boesefeldt. Secretariat of the National Road Safety Council, Warszawa, 2003, p Gaca S. Jamroz K., Ząbczyk K. et al.: The Nationwide Study of the Measurements of Vehicles Speed and Safety Belt Use (Ogólnokrajowe studium pomiarów prędkości pojazdów i wykorzystania pasów bezpieczeństwa) Periodical Reports. Signalco Trafik BIT. Secretariat of the National Road Safety Council, Warszawa, 2007, p. 350 (in Polish) 6. Gaca S.: Speed as a Reason of Heavy Accidents and Possible Improvement Measures (Prędkość jako okoliczność ciężkich wypadków i możliwe środki poprawy). Drogownictwo 7-8/2006, p (in Polish) 7. Gaca S., Kieć M.: Research of drivers reaction on speed limit change at urban area (Badania reakcji kierujących pojazdami na zmianę ograniczenia prędkości na terenach zabudowy). Transport Miejski i Regionalny 12/2005, p (in Polish) 8. Gaca S., Kieć M.: The influence of decreasing the speed limit to 50 kmph in urban areas on drivers behavior. Regional Development and Transport Logistics 5 th International Conference Żilina, 6 th 8 th December 2006, p