In depth 1 First Quarter 1 from the Observatory MINISTERIO DEL INTERIOR Observatorio Nacional de Seguridad Vial www.dgt.es Measurement of free-flow speed on the spanish road network. Introduction This study was conducted by the University Institute of Automobile Research, part of the Higher School of Industrial Engineering. Technical University of Madrid. (UPM- INSIA) in the framework of the research project financed by the Directorate- General for Traffic in 9. It presents the results of applying the recommendations of the European project SafetyNet on the estimate of SPIs (Safety Performance Indicators) for free-flow speed (FFS), that is, travelling along sections with road surface and visibility in perfect condition and moderate volume of traffic, so that the only restriction that the driver has is the speed limit laid down by law in force, with a view to have a common framework for indicators (in this case, for FFS) in the European countries. To this end, in Spain a stratified sampling has been designed by regions and road types (motorways, dual carriageways, stateowned conventional roads and regional roads owned by the Autonomous regions with two types of speed limit: 9 and km/h.), dividing the territory into 8 regions: Andalusia (R-1); Central Plain excluding Madrid, that is, both Castillas and Extremadura (R-); Madrid (R-3); Valencia and Murcia (R-4); Catalonia (R-5); Galicia and Asturias (R-6); Basque Country, Navarre, Cantabria and La Rioja (R-7) and Aragón (R-8). observation locations were chosen at random where very precise devices for measuring speed have been installed. The sections of the sample were subjected to two stages of verification of compliance with the SafetyNet recommendations for producing FFS indicators: a first desktop stage using IT tools and a second in situ stage, where minimal changes in the selection and replacement of the sections chosen in the desktop stage have been made. The integrated distribution of FFS of the 8 regions has been estimated in the four road types and three vehicle types (light vehicles, motorcycles and heavy vehicles). 1. Safetynet conditions for producing a SPI of FFS The purpose of this Spanish pilot project is the estimate of Indicators (SPI Safety Performance Indicators) of the distribution of free-flow speed, in the framework of the SafetyNet Project recommendations. In order to assure reliability and to allow comparison of the measures at European level, the SafetyNet manual recommends that the locations where the speed measurements are carried out are such that the vehicle speed is not altered by factors unrelated to the drivers. These locations must: be in a straight and uniform road section be in a section where exceeding the speed limit is possible be in a section with slope under 5% in the 5 preceding meters be at least 5 meters away from intersections be at least 5 meters away from speed reduction devices be at least 5 meters away from road works be at least 5 meters away from pedestrian crossings be at least meters away from speed limit changing points or signs be away from working or parking areas or other circumstances close to the road, be in good pavement and surface conditions be at least 5 meters away from gradient changes Based on these conditions with the most restrictive assumptions, the code developed in MATLAB language pre-selects the road sections among those measuring at least 5 km in length. In every station and its observations the data quality has been monitored and a selection has been made by days of the week (restricted to Tuesdays, Wednesdays and Thursdays, that are neither a public holiday nor the day before or after a public holiday) and by time slots (with the aim of avoiding traffic jams and peak hours): from 1: to 1:59 and from 15: to 15:59 for daytime From : to :59 for night-time The additional SafetyNet recommendation on separation distance of at least 5 seconds between vehicles has also been taken into account.. Desktop work and fieldwork.1 Desktop work. Stratified sampling. Software preparation. Validation of the data The sampling is stratified by regions and road types. The 8 regions defined confirm the hypothesis of an homogeneous behaviour in each stratum. The final distribution by 4 road types: Motorways, Dual carriageways, Conventional and Regional roads with a speed limit of km/h, and Conventional and Regional roads with a speed limit of 9 km/h, is shown in Figure 1.
In depth l 1 Sample by road types 1% 17% 17% 54% Figure 1. Final distribution of the measurements by road types The software written in MATLAB language is self-contained for the whole process: selection of the sample, analysis of the data, selection of relevant data and analysis of the results of the 8 regions, combination of categories: days of the week, 3 time slots ( for daytime and 1 for night-time), 3 vehicle types (motorcycles, light vehicles and heavy vehicles) and distance of at least 5 seconds between vehicles. The procedure of verification and selection of valid data has been very strict. In addition to the predefined criteria, meteorological data have been considered to delete periods and time slots with adverse weather conditions from the database.. Fieldworks. Verification of the conditions of equipment installation On the selected road sections the SafetyNet criteria have been verified by means of an implementing protocol (check list) during the installation of the measuring equipment and in some cases visual inspection has been carried out by the investigating team. The equipment consists of direction sensing radars that capture the data with high-frequency radio signal, whose precision reaches 3% in the speed measurement. In the course of the works a small number of incidents relating to the pre-selected locations have arisen, such as: road works, speed limits, slopes, gradient changes, proximity to industrial estates, etc., leading to make some adjustments at the installation location. 3. Obtaining indicators of safety performance indicators (SPI) 3.1 Statistical methodology of the estimate process An estimate of the speed distribution involves: 1) The estimate of measures of centre, typically the average, which is used in this study as a value around which range data, and estimate of uncertainty or variability of data, in this study is measured through: The variance or its square root, standard deviation. The percentiles of the distribution:.5%, 85% and 97.5%, the interval between the percentiles.5 and 97.5 include 95% of the central data. ) The confidence intervals for a mean, variance and even for the estimated percentiles The wider the intervals, the greater the uncertainty of the estimate, which is a reflection (also influences the sample size in the amplitude of intervals) of the variability of the data. 3) In addition, it has been estimated, in a general way the percentages of vehicles exceeding the speed limit and speed limit + 1 km/h. Although the distribution of the speeds usually comes close to normal, for this analysis has been used the nonparametric technique (i.e., not dependent on hypothesis about the type of distribution) or bootstrap re-sampling. 3. Methodology of the aggregation process Different indicators have been obtained as follows: Point estimate of the mean The expression of the estimator is: µ = w1 x1 + w x + w3 x3 +.. + w 8 x 8 = 8 i= 1 w x Where W 1, W and W 8 are the relative weights of the strata and x _, x_,, x_ are the means in each stratum. The relative 1 8 weights are proportional to the populations of the strata. Confidence interval for the average The confidence interval is obtained from bootstrap or re-sampling method. In stratified sampling each stratum is re-sampled and the replicated mean is obtained with the same weighting as in the point estimate. Point estimate of variance For this study it has been adopted the following point estimator: σ st = h σ h + h= 1 h< g w w w ( µ µ ) Where σ i and μ i are the variances and means, respectively, of the strata and w i are the relative weights as defined above. h g h g i i
Confidence interval of variance As in the case of the mean, replicated variances are obtained applying the point estimator to the replicas of the samples of the strata. Point estimate of percentiles This procedure applies to the estimated percentiles of.5%, 97.5% and 85%. The first two define the central 95% of the population and the last percentile is part of the requirements of SafetyNet. To that end, samples of strata are sorted, and then each element of the sample is given a relative weight proportional to the weight of the stratum in the population. Then, the point estimator of the percentile is obtained as corresponding percentile of the weighted sample, from the set of measures realized. Confidence intervals of percentiles As in the case of the mean and variance, we obtain the replicas of the variances applying the point estimate to the replicas of the samples of the strata. Estimator of the distribution function The application of this parameter is the proportion of vehicles exceeding the speed limit or this speed + 1 km/h. The point estimator is obtained from the ordered and weighted sample, as discussed above, as the accumulated weight corresponding to the value for which is wanted to calculate the distribution function (speed limit or speed limit + 1 km/h). 4. Results To obtain SPI, it has been 83,55 the total number of measurements considered (because of the SafetyNet filtering criteria applied: date, time, weather conditions, 5-second interval), representing 1.% of total number of collected data. A summary of the indicators of average speed, percentile 85 and percentage of vehicles exceeding speed limits by road type and vehicle type, both daytime and night-time, is shown in Figures -7. MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure. SPI: average speed per road and vehicle type in daytime MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure 3. SPI: percentile 85 per road and vehicle type in the night-time
In depth l 1 MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure 4. SPI: percentage exceeding speed limit per road and vehicle type in daytime 1 14.76 13.43 119.9 116.6 11.4.3 96.68 93.84 9.4 78.47 8.98 8.49 MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure 5. SPI: average speed per road and vehicle type in the night-time To sum up may be highlighted: A large number of vehicles exceeding the speed limit in each type of road, with more than 5% in conventional roads. No significant differences in average speeds, percentile 85 and percentage exceeding the speed limit, between light vehicles and motorcycles, at night, however, it should be noted that the number of valid motorcycle records (because Safetynet calls for it) is relatively low. No significant differences are appreciated in average speeds, percentile 85 and percentage exceeding the speed limit between day and night periods measured, they are higher in the night-time. This may reveal that drivers in night hours assume greater risk in conditions of limited visibility. The average speed of heavy vehicles (including trucks and buses) is close to or exceeding the speed limit for these vehicles and that should be ensured by the limiter, especially in conventional roads with limited 9 and km/h speed.
1 131 141 15 131 15 1 113 11 1 97 99 89 MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure 6. SPI: percentile 85 per road and vehicle type in the night-time % 7% 66.67% 67.% 63.39% % 58.67% 58.4% 56.8% 5% 51.1% 49.4% 5.36% % 34.96% 3% % 19.16% 1.% 1% % MOTORCYCLES LIGHT VEHICLES HEAVY VEHICLES Figure 7. SPI: percentage exceeding speed limit per road and vehicle type in the night-time 5. Conclusions Confidence intervals of all SPI selected have been obtained by implementing an advanced methodology (bootstrap or resampling). Having obtained fairly narrow confidence intervals, the maximum amplitude corresponds to the average speeds of motorcycles; even so they are relatively small, which confirms the reliability of the estimates. Many parameters have been estimated from the point of view of safety. Percentiles 85% and 97.5% are very interesting, since the behaviour of drivers whose speeds are further away from the mean value is relevant to safety. These parameters are closely related to the percentage of vehicles exceeding 1 km/h or more the speed limit which has been determined.6% of free flow conditions, weighting all roads as a whole. Daytime indicators for motorcycles show a more moderate behaviour than that of light vehicles, but these data should be viewed cautiously because the sample sizes of motorcycles are relatively low, even more in the case of the night. There is little difference between day and night, but night values are generally slightly higher. This piece of information is relevant, from the point of view of road safety, for the risk of low visibility and higher risk of accident. 6. Annexe 1: Comparative with other countries in europe Coinciding with the end of this work, it has been published an updated report on average speed and percentage exceeding the speed limit (SPI for two years, and 7, except the Czech Republic only 7) on motorways of 9 European countries and of light vehicles except the Czech Republic and Switzerland, including all types of vehicles. Published data have been compared with those of Spain in the year 9 obtained in this project. In this report it should be noted that Finland has motorways with different limits, in winter and 1 in summer, as in Denmark, where it is noticed that until speed limit in all motorways was 11 and that +5% of them have 13 limit since 4. However, the data provided in this report should be considered unofficial indicators obtained as mean monthly, as in Holland.
MINISTERIO DEL INTERIOR Observatorio Nacional de Seguridad Vial Observatorio Nacional de Seguridad Vial Josefa Valcárcel, 44 71 Madrid www.dgt.es D.L.: M-5617-1 Nipo: 18-1-15-6 AVERAGE SPEED LIGHT VEHICLES (except Czech Republic and Switzerland that include all vehicles). -7. Spain 9 (free-flow condi ons) 1 Finland km/h (winter) Denmark ( km/h) United Kingdom Ireland Finland Czech Denmark Switzerland Holland (summer) SPAIN France Austria Republic (13 km/h) Average speed (km/h) 1.1 119.1 11.7 16 114 114.8 114.9 1 15.5 1. 5 Average speed 7 (km/h) 13.1 117.7 11.7 11 17 115.1 114.3 1.51 15 1. 119.4 119.9 Figure 8. Comparative of average speed in other European countries % LIGHT VEHICLES EXCEEDING SPEED LIMIT (except Czech Republic and Switzerland that include all vehicles).. -7. SPAIN 9 (free-flow conditions) 7 67.8 57.44 5 51 47 38 36.6 39.9 3 31 7.9 4 1 3 1 Finland ( km/h in winter) United Kingdom Ireland Switzerland Finland (summer) SPAIN France Austria Percentage of offenders (%) Percentage of offenders (%) 7 Figure 9. Comparative of percentage of offenders in other European countries In the published data there are three groups of motorways, 1G: different limits (Finland, Denmark and United Kingdom), G: limit 1 [Switzerland, Finland, Netherlands, Spain (** 9)], 3G: limit 13 (Czech Republic, Denmark, France and Austria). In 7 all countries of the 3 rd group have average speed, about 1 km/h below the limit (13), as Ireland, which belongs to group. Czech Republic (3G) has a very low average speed because includes all vehicles, not only light vehicles, like Switzerland, in group. By contrast, the average speed of Finland M- (1G) is over the limit, whereas UK remained on the limit values. Motorways in Denmark DK- exceed by 7-9 km/h limit. Comparing with 7 measurements, there has been a significant reduction in average speed in France, Switzerland (light + heavy) where significant measures to control speed have been undertaken, whereas in Austria there was low decreasing. Ireland is the only country with increasing average speed (but there have been changes in speed limits in January 5 and new motorways). It is remarkable the stability of the average value in Denmark despite the increased limit from 11 to 13. Among all the countries in group (1), the average in Spain is very close to the limit 1 (1.51 average speed weighted light vehicles in day-night), but it is higher than the average in most countries of 3G (13). As for the percentage of vehicles exceeding the limits: the lowest are in Ireland and Switzerland (G: 1), the highest in Finland and the United Kingdom in 7, besides it has also increased the percentage of traffic offenders in Finland especially in the M- in winter. It has fallen sharply in France and Switzerland (average speed too), followed by Ireland (effect of increasing the speed limit from January 5) and Austria. In Spain, the percentage of traffic offender drivers reaches 57.4% in 9.