Content uploaded by Tim Holthaus
Author content
All content in this area was uploaded by Tim Holthaus on Jun 16, 2020
Content may be subject to copyright.
Evaluation of driven speed on German
motorways without speed limits
- a new approach -
Tim Holthaus M.Sc., Claus Goebels M.Sc., Prof. Dr.-Ing. Bert Leerkamp
5 March 2020
Abstract
The introduction of a general speed limit on German motorways is currently heavily discussed.
This paper analyses a Floating Car Data (FCD) set from April 2017 generated on German
motorways without speed limits. The data set is compared to average speeds measured at
automatic traffic counting points. Therefore, the validity of the reported speed in the present
FCD data set is analysed by comparision to the reported speed at continuous counting stations.
The results show that the average driven speed on motorways in metropolitan areas is lower
than on long distance highways and motorways in rural areas. The frequency scales offer
insights into expected effects of a general speed limit. Recommended speed (130 kph) is rarely
exceeded during a trip but a large number of trips can be observed where maximum speed is
higher than 130 kph on short parts of a motorway route.
1 Delimination of the investigation topic
The introduction of a general speed limit of 130 kph on German motorways has been frequently discussed.
As mandatory climate protection targets for transport sector were introduced by the German Federal
Ministry for the Environment, Nature Conservation and Nuclear Safety
1
in October 2019, this discussion
actually received great public attention. Prescribing the transport sector must reduce its Greenhouse
Gas (GHG)-emissions by 40 - 42 % until 2030, the discussion turns out to be very emotional in Germany
because it is one of a few countries whithout any general speed limit. In addition to the postulated effects
such as traffic safety, GHG-emission savings up to 14 % are predicted.
2
Whereas approximately 70 %
of motorways did not have any speed limit in Germany in 2015, dynamic speed limits count for approx.
6 % of total lentgh of motorway network. Accordingly, approx 25 % of the federal motorways have speed
limits.3
This article examines the average driven speed on German motorway segments without speed limits.
Comparable evaluations have already been carried out by TomTom on behalf of ZEIT Online.
4
These
results will be reproduced, compared and evaluated in order to provide a technical basis for further
discussions of general speed limits on German motorways.
2 Methodology
2.1 Research object
The Chair for Freight Transport Planning and Transport Logistics at Wuppertal University has access to
historical Floating Car Data (FCD)
5
(here in after these data are called
F CDB U W
) covering Europe and
1Cf. [Bundesregierung Deutschland, 2019].
2
Based on speed distribution of 2016 and mileage of 2018 on German motorways. Saving depends on speed limit and is
between 5 % (130 kph) and 14 % (100 kph). Cf. [Lange, M., 2020].
3Cf. [Kollmus, B. et al., 2017].
4Cf. [Biermann, K. et al., 2019].
5FCD provided by the ADAC.
1
is engaged in several research projects based on FCD data sets from Germany and neighboured European
countries. The data sets allow to reconstruct the trajectories of cars and trucks in an anonymized form.
These
F CDB U W
are generated from position data transmitted by navigation and fleet-software. Each
position point contains information about the exact time, the currently driven speed as well as the direction
and a unique vehicle ID that assignes individual measuring points to exactly one vehicle.
F CDB U W
used for speed analysis only contain passenger cars. Similar studies were carried out by
the authors with truck-generated FCD. Before starting the analysis all data points were assigned to a
motorway section and only
F CDB U W
of motorway sections without speed limits were extracted. The
F CDB U W
reporting year (2017) and state of the network model are the same. Due to low impact of
weather conditions and vacation traffic April 2017
6
is chosen for the analysis.
F CDB U W
being influenced
by construction sites with temporary speed limits are not removed from the data set, as there was no
exact information available about position and duration of constrution sites in April 2017. The network
elements are cut into edges with a maximum length of 100 m. In total this network includes 185.799
edges
7
. The map matching indentifies 629.252 vehicles on this network in April 2017. This analysis
focusses on the following two perspectives on the data:
1. Driving speed on the 100 meter network edges (average, percentiles).
a)
The
average speed
is the harmonic mean of the reported speed of all vehicles which generate
data on an edge (maximum length 100 m) of motorways without speed limit. Generally, each
vehicle is only represented once on an edge. If more than one data point per vehicle exists, the
average speed of these points is considered.
b) Percentiles
are generated from car-related speed reports, similar to the harmonic mean.
Example: For the upper 10 % percentile all reported speed measurements are sorted in
ascending order and then the speed is taken that is not exceeded by 90 % of the cars on the
edge.
2.
Reported maximum speed and average speed of the individual vehicles identified on network edges
without speed limit.
a) Maximum reported driven speed of a vehicle on all motorways without speed limit for
the whole day. For this purpose, all reported data points of a vehicle are chosen, if this vehicle
is detected at least once on an edge without speed limit. Out of this data set the highest
reporting speed of each vehicle is chosen.
b) Average of reported driven speed of a vehicle
on all motorways without speed limit for
the whole day. For this purpose, all reported data points of a vehicle are chosen, if this vehicle
is detected on an edge without speed limit. Out of this data set the average speed for each
vehicle is calculated.
2.2 Validation of FCD reported speed
Reported speed at continuous counting stations are compared to
F CDB U W
based speed in order to
ensure that speeds generated with
F CDB U W
forms a valid sample and represents real driven speed. In
2016, the German Federal Highway Research Institute (BASt) published an analysis for driven speed
on German motorways. For this purpose, speed data recorded by continuous counting stations
8
were
analyzed. The comparision with
F CDB U W
based speed distribution on edges close
9
to these continuous
counting stations (
Figure 1a
) only shows small differences. Especially in lower speed classes (1 - 70 kph)
the amount of
F CDB U W
data points is higher. On the one hand this can be explained by inaccurately
reported positions of the continuous counting stations in the available database
10
. A counting station
position near an exit or entry lane will measure lots of accelerating and breaking vehicles and therefore
lead to higher shares of slow vehicles. Assuming the real position of the continuous counting stations is
located between exit and entry, the differences in the lower speed classes could be explained. In a few
6
April is choosen to reduce the influence of weather condition and vacation. In addition the impact of congestion in April
is low due to the congestion analysis of the ADAC e. V. (cf. [ADAC e.V., 2018, p. 7]).
7
Cutting the network into edges that are no longer than 100 m can result in very short edges. In the context of the network
matching of the
F CDB U W
, it can therefore happen that individual short edges exist on which no
F CDB U W
data are
machted.
8Cf. [Löhe, U., 2016, p. 8].
9F CDB U W
are generated in intervals. Therefore they cannot be taken at a defined single point, in contrast to continuous
counting stations. Instead F C DBU W are taken from a 100 m edge on which the counting station is located.
10
Database of continuous counting stations can be downloaded at
https://www.bast.de/BASt_2017/DE/Verkehrstechnik/
Fachthemen/v2-verkehrszaehlung/Aktuell/zaehl_aktuell_node.html. Retrieved on 2 March 2020.
2
cases
F CDB U W
are assigned wrongly to the main edge although they in fact are located on a parallel
running entry lane. On the other hand the
F CDB U W
data set shows lower speeds than BASt analysis on
some edges without any exit or entry lane nearby. This difference is probably effected by construction
sites which could not be eleminated in F CDB U W data set yet.
Because of the above mentioned effects, the distribution generated from
F CDB U W
shows a significant
difference in the speed classes 71 - 90 kph. This difference can be eliminated by identifying the construction
sites. For example
Figure 1b
shows a continuous counting station on motorway A3, which was under
construction in April 2017.
11
This counting station is passed by about 17.000 vehicles (according to
F CDB U W
) which are about 10 % of all counted vehicles in April 2017, which were detected on the above
mentioned continuous counting stations.
(a)
Distribution of reported speed at continuous counting stations, with and
without counting station No. 5033 (A3 near Leverkusen)12
(b)
Distribution of reported speed on motorway A3 near conting station No.
5033 (A3 near Leverkusen) during influence of construction site
Figure 1: Distribution of reported F C DBUW speed in April 2017
11 Information about road construction on motorway A3 can be found in https://www.strassen.nrw.de/de/projekte/a3/
sanierung-zwischen- opladen-und- hilden.html. Retrieved on 2 March 2020.
12 The F CDB U W based speed at the continuous counting stations reported in [Löhe, U., 2016] are used the evaluation.
3
Due to missing information about road construction in 2017, it was not possible to eliminate all edges
which were under construction in Germany in April 2017. As a result, edges which were under construction
might have a lower speed within the results than in reality.
The above mentioned validation process shows, that
F CDB U W
can be used for speed related evaluations.
Nevertheless, it must be considered, that there are still some inaccuracies in the
F CDB U W
because of a
lack of information like missing databases about construction sites. For further analyses the
F CDB U W
have to be merged with external event information about detours, construction sites, weather etc. to
enlarge the significance of the results.
3 Results
3.1 Speed aggregates on edges
The average speed on German motorways without speed limits can be divided into two major groups of
motorways: Motorways inside of metropolitan areas and those in rural areas. The density of the network
is probably an indicator for the average speed. About 75 % of all edges have an
average speed
of 130
kph or lower. Only some motorways connecting major cities and predominantly passing rural areas (e. g.
Ulm – Munich, Berlin – Rostock) show a higher average speed than 130 kph (see
Figure 2
). On these
motorways long distance cummuters dominate the normal daily traffic and they obviously choose higher
speed.
Figure 2: Speed on edges for whole April 2017 incl. weekend and night - average speed
4
The
maximum
driven speed on motorways without speed limits in rural areas is also significantly higher
than in metropolitan areas (see
Figure 3
). Drivers in metropolitan areas obviously do not utilise the
opportunity to accelerate to high speeds. Possible reasons are short distances between entry and exit
lanes or short distances travelled on motorways combined with a higher share of daily commuters.
Figure 3: Speed on edges for whole April 2017 incl. weekend and night - 10 % percentile
Figure 4a
compares the fastest 10 %, the fastest 25 % and the average driven speed per edge. Combined
with the maximum reported driven speed per vehicle (see
Figure 4b
), it can be seen, that only 1 % of all
motorway edges have an
average speed
of more than 140 kph while 55 % of the vehicles have a
maximum
speed higher than 140 kph during their trips.
In order to derive the effects of general speed limits not only the average speed on the edges or the
maximum speed of individual vehicles must be considered, but also the average speed driven of each
individual vehicle and each trip on the unlimited motorway edges. Accordingly, most of the vehicles do
not permanently drive faster than 130 kph. Only 21 % of the vehicles have an average driven speed (on
motorways without speed limit) faster than 130 kph and only 9 % faster than 140 kph.
5
(a) 100 meter edges by speed class
(b) Vehicle with maximum occuring speed by speed class
Figure 4: Cumulative frequency
3.1.1 Other data samples and analyses
TomTom
As already described in chapter 1,
F CDB U W
speed data analysis is compared to TomTom speed data
published by ZEIT Online.
13
According to TomTom, more than 61 billion traces per day are recorded
worldwide.
14
Out of this data set one week in March 2019 was taken to produce the published results for
Germany.
The comparison between
F CDB U W
and TomTom shows a significant difference (see
Figure 5
). It has
to be considered, that the data from TomTom are generated on motorways with and without speed limits.
Therefore, the speed class below 100 kph covers nearly two thirds of the TomTom speed measurements.
F CDB U W
show significantly less cars in the lowest speed class (
≤
100 kph) while the speed class from
101-110 kph is nearly congruent in both samples. The classes between 111 kph and 130 kph are nearly
doubled by TomTom compared to
F CDB U W
. The distribution of the
F CDB U W
shows a slightly left
crooked normal distribution with the average in the class of 121 kph to 130 kph. Nearly 50 % of the edges
can be found between 111 kph and 130 kph.
13 Cf. [Biermann, K. et al., 2019].
14 Cf. [TomTom, 2020].
6
The distribution of the TomTom data shows a left crooked normal distribution and a higher variance
with the average in the class of 130 kph to 140 kph. Taking into account the data shows motorways with
and without speed limits, this is a significant difference.
Figure 5: Average speed out of F C DBUW for whole
April 2017 vs. TomTom data
Figure 6: Comparision of 10 % percentile out of
F CDB U W for whole April 2017 vs. Januar 2017
These differences may result in an unknown composition of the sample in the TomTom data. For policy
decision it is necessary to use transprarent data and methods.
15
The comparision of the distribution of
reported speed at continuous counting stations (see
Figure 1a
) with
F CDB U W
shows that the provided
F CDB U W sample can reproduce the real driven speeds (see chapter 2.2).
Some of the differences occure due to different months in which the data where captured.
Figure 6
shows the same analysis for January and April. The maximum speed is significantly higher in April than
in January. This can be explained by two main factors:
•weather conditions and
•winter tires.
Due to the coincidence of weather, further analyses of enriched data is required. The winter tires often
reduce the maximum speed to 160 kph. This explains the massive break within class 161 kph to 170 kph
in January.
These results show the importance of a continous analysis over serveral months and the need for further
data (e. g. weather information).
German Enviroment Agency
The German Environment Agency (UBA) has recently published a study on expected effects of a general
speed limit on German motorways. The effects on speed distributions were estimated by transforming
given speed distributions on speed restricted motorway segments (120 kph) on segments without speed
limits.
16
As motorways with speed limits are mostly located in metropolitan areas, it has to be assumed,
that the speed distributions are unlikely to come into being in rural areas. Subject to the condition that
the speed limit will be enforced by intensive speed controls, the speed distribution will most likely show a
left crooked distribution with a high share of speeds in the range of 110 kph to 130 kph (for speed limit
120 kph). Therefore the predicted effects on GHG-emission savings are giving a maximum estimation and
tend to be overestimated.
15
Assuming that every person worldwide would have a vehicle and record an average of 3.1 trips per day (average travel
frequency from Mobilität in Deutschland (MiD) for all modes (Cf. [infas et al., 2018, p. 3]), this corresponds to only
approx, 25 billion traces a day. This is only nearly a third of the daily produced data by TomTom.
16 Cf. [Lange, M., 2020, p. 15 ff.].
7
3.2 Influence of day time
The influence of day time is shown in
Figure 7 and 8
. Regularily, the amount of traffic in the morning
is higher due to overlapping activities (morning peak). Accordingly, the average driven speed is lower.
This has to be considered when using trajectory data to parametrize or calibrate models in transport
planning - especially if the speed is not manually caped at 130 kph and the real driven speed is used.
One application purpose is the useage of extracted speeds from
F CDB U W
in network design. Following
the German Guidelines for Integrated Network Design (GIND) the parametrized network should represent
an empty state for initial assignment of relations to network and a stressed but not overstressed state for
traveltime evaluation.17
In planning, usually edges without speed limit will be considered with a speed of 130 kph. To consider
or to evaluate real travel times no manual modification of speed limits is done.
Figure 7 and 8
show
two exemplary network states for a stressed network following GIND.
Figure 7
represents the network
state for the morning hours while
Figure 8
represents the network for the afternoon. Compared to the
monthly average (Figure 2) these two states are slightly slower.
Figure 7: Average Speed on edges for Mo-Fr in April 2017, 7 a.m to 11 a.m.
Moreover, the afternoon state is slightly faster than the morning state which can be explained with
more likely overcrowded motorways in the morning peak. The spatial distribution of slower and faster
edges is more or less the same. In the morning state velocity on 20.6 % of the edges is higher than 130 kph
17 Cf. [Forschungsgesellschaft für Straßen- und Verkehrswesen (FGSV) - Arbeitsgruppe Verkehrsplanung, 2008].
8
in average while in the afternoon state approximately 25.5 % of the edges show an average speed higher
than 130 kph.
Figure 8: Average Speed on edges for Mo-Fr in April 2017, 2 p.m to 7 p.m.
In addition to the speed differences between both data samples (see
Figure 5
) it is questionable which
data sample reflects reality best. In order to answer this, the raw data of both data samples must be
compared with each other and with other data such as the continuous counting stations. This is the only
way to create a solid database for spatial and transport planning.
4 Conclusion
For a reliable discussion different data sources have to be compared. All of this different data have to proof
representativity. Therefore, it is important not only to rely on cumulative results but also on individual
data evaluation. Furthermore, also the origin of raw data and the representativeness of the data is of
great importance.
This article provides results from different data sets with a different sample and thus offers an expanded
basis for a discussion about general speed limit. The majority of the edges have an average speed of
130 kph or below. In relation to the maximum reported speed of the vehicles, however, it can be seen that
less than 30 % of the vehicles on the edges are below the target speed (130 kph). The average reported
speed on motorways without speed limit for each vehicle is nearly 80 % lower or equal 130 kph. This
should be considered when calculating GHG reductions by transferring speed distributions of limited
motorways to unlimited motorways.
9
An evaluation of each individual trip is of great importance. Such an evaluation is time-consuming
but also essential for the evaluation of a general speed limit. In contrast to the analyses done in this
paper, it is recommended to analyse each trip separately and not cumulatively in further examinations.
As a result this approach can be used to make statements about spatial structure differences, travel
distance-related speed profiles and thus an exact statement about a possible reduction of GHG-emissions.
References
ADAC e.V. Staubilanz 2017. ADAC e.V.. Ressort für Verkehr - Verkehrspolitik, 2018. https://www.adac.de/_mmm/pdf/statistik_
staubilanz_231552.pdf. Retrieved on 2 March 2020.
Biermann, K., Blickle, P., Loos, A., and Venohr, S. Wo Deutschland rast. Zeit Online, 2019. https://www.zeit.de/mobilitaet/2019-
02/autobahnen-geschwindigkeit- tempo-schnelligkeit- raser-verkehr. Retrieved on 2 March 2020.
Bundesregierung Deutschland. Klimaschutzprogramm 2030. Bundesregierung Deutschland, 2019. https://www.bundesregierung.
de/resource/blob/975226/1679914/e01d6bd855f09bf05cf7498e06d0a3ff/2019-10- 09-klima- massnahmen-data.pdf?download=1. Re-
trieved on 2 March 2020.
Forschungsgesellschaft für Straßen- und Verkehrswesen (FGSV) - Arbeitsgruppe Verkehrsplanung. Richtlinien für integrierte
Netzgestaltung - RIN. FGSV-Verlag, Köln, 2008. ISBN 978-3-939-71579-5.
infas, DLR, IVT, and infas 360. Mobilität in Deutschland - MID. Bundesministerium für Verkehr und digitale Infrastruktur, Bonn,
2018. http://www.mobilitaet-in- deutschland.de/pdf/MiD2017_Ergebnisbericht.pdf. Retrieved on 2 March 2020.
Kollmus, B., Treichel, H., and Quast, F. Tempolimits auf Bundesautobahnen 2015. Schlussbericht zum Arbeitsprogramm-
Projekt F1100.6110020. Bundesanstalt für Straßenwesen, 2017. https://www.bast.de/BASt_2017/DE/Publikationen/
Fachveroeffentlichungen/Verkehrstechnik/Downloads/V1-BAB- Tempolimit-2015.pdf?__blob=publicationFile&v=6. Retrieved on 2
March 2020.
Lange, M. Klimaschutz durch Tempolimit. Umweltbundesamt, 2020. https://www.umweltbundesamt.de/sites/default/files/medien/
1410/publikationen/2020-03- 02_texte_38-2020_wirkung- tempolimit.pdf. Retrieved on 2 March 2020.
Löhe, U. Geschwindigkeiten auf Bundesautobahnen in den Jahren 2010 bis 2014. Bundesantsalt für Straßenwesen, 2016.
https://www.bast.de/BASt_2017/DE/Publikationen/Fachveroeffentlichungen/Verkehrstechnik/Downloads/Geschwindigkeiten-BAB-
2010-2014.pdf?__blob=publicationFile&v=3. Retrieved on 2 March 2020.
TomTom. Location Intelligence. 2020. https://www.tomtom.com/industries/gis-location- analytics/. Retrieved on 2 March 2020.
About the authors
Bert Leerkamp
owns the Chair for
Freight Transport Planning and Trans-
port Logistics since 2009. His main
research topics are providing new data
for transport modelling, the analy-
sis of trip generation (especially in
freight transport), network design and
focusses on integration and sustain-
ability of freight transport in trans-
port planning.
Mail: leerkamp@uni-wuppertal.de.
Tim Holthaus
works since 2015 at
the Chair of Prof. Dr.-Ing. Bert
Leerkamp. His main research top-
ics are network design, rechability-
analyses and the development of
regional/city-logistic concepts. He is
a member of the Working Committee
1.3 of the Road and Transportation
Research Association (FGSV).
Mail: holthaus@uni-wuppertal.de.
Claus Goebels
works since 2016 at
the Chair of Prof. Dr.-Ing. Bert
Leerkamp. His main research topics
are trip generation, data extrapolation
and the development of new statistic
methods for transport planning. He
is a member of the Working Group
3.10.6 of the Road and Transporta-
tion Research Association (FGSV).
Mail: goebels@uni-wuppertal.de.
10