MIDA[1].qxd

370 Evaluating MIDAS
Evaluating the
benefits of MIDAS
automatic queue
protection
Steve Tucker, Highways Agency and Ian Summersgill, John Fletcher, David Mustard, TRL
This paper reports on a study of the operation of
arrive at an estimation of safety benefits. In order
These benefits were costed and compared with the
MIDAS (Motorway Incident Detection and
to estimate congestion relief benefits, a
current cost of installing MIDAS and demonstrate
Automatic Signalling) over a ten year period and
relationship was developed between the delay
that the system continues to provide good value for
covering a route 602 kilometres in length. The
attributed to non-recurrent congestion and AADT
money delivering safety and congestion benefits of
study analysed accident injury records in order to
flow.
over £60m per year.
INTRODUCTION
the effects of changes in these variables in order to distin-
guish them from the effect of introducing MIDAS. The vari-
ables that were taken into account in the analysis were:
MIDAS (Motorway Incident Detection and Automatic Sig-
• Time trend
nalling) automatic queue protection is currently in operation
• Annual average daily traffic flow
on 1015 kilometres of motorway route in England. MIDAS
• Length of the link
uses induction loops spaced at 500 metre intervals in the
• Season (by quarter)
road to detect slow moving, queueing or stationary traffic
• Number of motorway lanes
arising from an accident, incident or other cause. The system
• Road lighting
sets upstream signs automatically to show to drivers 40mph
• Intensity of roadwork activity
and 60mph advisory speed limits together with supplemen-
• Effect of other operational systems, for example, the
tary text messages such as ‘QUEUE AHEAD’ and ‘QUEUE
Controlled Motorway
CAUTION’. This protects the back of the queue from sec-
ondary collisions. The signs are mounted either on can-
The accident records that were used in the analysis were
tilevers or take the form of individual lane signs mounted on
obtained from the Department for Transport (DfT) STATS19
portal gantries.
injury accident database and these were allocated to the mo-
This study of 10 years of operations has shown a safety im-
torway links using TRL’s Microcomputer Accident Analysis
provement of 13% and significant congestion relief benefits.
Package (MAAP) software. Estimates of annual average daily
traffic flows (AADT) for each motorway link were obtained
SAFETY BENEFITS
from DfT sources, whilst the link lengths were measured
from 1:25,000 scale Ordnance Survey maps.
The accident records also provided information about the
Approach
existence of road lighting and whether roadworks were pre-
The safety benefits were estimated by analysing injury acci-
sent when an accident occurred. The proportion of accidents
dent records covering a ten year period from January 1993.
occurring during roadworks was used as a means of estimat-
They covered 602 kilometres of route. Only those sections of
ing the intensity of roadwork activity.
motorway where MIDAS had been operating for at least six
The number of lanes on each motorway link and any
months were analysed. The analysis was based on the time
changes that occurred during the study period were obtained
series of the numbers of accidents occurring in each quarter
mainly from DfT and Highways Agency (HA) sources. Infor-
of a year on each link (section of motorway between individ-
mation about the time periods and links on which the fol-
ual junctions).
lowing systems were operated were obtained from HA
Accident numbers are influenced by many variables other
sources: MIDAS operating alone; MIDAS operating in con-
than MIDAS operation. It is necessary, therefore, to estimate
junction with the Controlled Motorway; preliminary MIDAS
tec OCTOBER 2006

371
Figure 1:
MIDAS systems
studied
operation; Controlled Motorway operating alone; and Auto-
compared with the number of injury accidents that would
matic Incident Detection (an earlier form of MIDAS) operat-
have occurred if MIDAS had not been operated. This latter
ing alone.
number is not known and can only be estimated. Thus, the
Additional links without MIDAS were selected to act as sta-
benefits of MIDAS in terms of accident reduction are esti-
Table 1:
tistical controls and these included 592 kilometres of motor-
mates and there is some uncertainty in the estimated values.
‘Multiplier’ by
way route. Analysis of the statistical control sections showed
The estimates are obtained by assembling all of the avail-
accident severity
that the relationship between accident numbers and traffic
able information to develop models which can take into ac-
for MIDAS
flow was substantially different to that for the MIDAS sec-
count the effect of the variables (including MIDAS operation)
operation
tions and, in addition, the accident severity was somewhat
on accident numbers.
different. Hence the control sections are not included in the
analysis presented here.
Figure 1 shows details of the MIDAS schemes that were in-
‘Multiplier’
cluded in the study together with those sections of motor-
Severity
Lower
Central
Upper
estimate
ways used as controls.
95%ile
95%ile
Form of analysis
All injury accidents
0.81
0.87
0.93
The number of injury accidents that occurred during the pe-
Fatal + serious injury accidents
0.76
0.90
1.07
riods when MIDAS was operated is known. This must be
Slight injury accidents
0.80
0.86
0.93
tec OCTOBER 2006

372 Evaluating MIDAS
dence interval for the fatal and serious injury accidents com-
l
a
t
a
F
i
c
c
a
t
n
e
d
s a
s
e
v d e
p r
a
e
y r
9
bined is larger than for either slight injury accidents or all in-
e
S i
r u
o s i u
j
n
y
r
i
c
c
a
t
n
e
d
s a
s
e
v d e
p r
a
e
y r
47
jury, since there are relatively few of the former (about 11 per
cent of all injury accidents).
i
l
S
t
h
g
j
n
i
y
r
u
t
n
e
d
i
c
c
a
s a
s
d
e
v
e
p r
a
e
y r
3 3
6
The proportional reductions (‘multipliers’) in accident
a
m
a
D
e
g
y
l
n
o
t
n
e
d
i
c
c
a
s a
s
e
v d e
p r
a
e
y r
3 1
,
4
8
numbers attributable to MIDAS operation have been de-
scribed. However, in order to estimate the reductions in the
Value of accident savings per km per year
£48,750
absolute numbers of accidents it is necessary to estimate the
numbers of accidents to which the proportions apply.
Value of accident savings per year on 800km
£39.0 million
The method that has been adopted is that of estimating
the average number of injury accidents per kilometre per
The methods of Generalised Linear Modelling (GLM), an
year on the schemes in 2003.
Table 2:
advanced form of regression analysis (McCullagh and Nelder,
The value was 3.50 injury accidents per km. It can there-
Estimated annual
1989), were used to relate the numbers of injury accidents to
fore be estimated that the number of injury accidents per km
benefits of
the explanatory variables using the GENSTAT statistical pack-
would have been expected to have been 1/0.87 times as great
MIDAS for
age. The methodology is explained in Maher and Summers-
if MIDAS had not been operational. This gives a value of 4.02
800km of route
gill (1996).
injury accidents per km.
Separate relationships have been developed for each of the
It would also be possible to obtain similar estimates sepa-
21 schemes included in the study. However, in order to elim-
rately for fatal and serious injury accidents and for slight in-
inate unnecessary detail, only the relationships for all the
jury accidents. However, there are relatively few of the for-
schemes combined are considered in this paper.
mer and hence the national (GB) accident statistics for mo-
Estimation of safety benefits
torways in 2003 (Department for Transport, 2004a) were
Relationships were developed for various groups of injury ac-
used to estimate the proportions of fatal, serious and slight
cidents and schemes (including those for individual
injury accidents per km per year. These were: fatal (2.10 per
schemes) during the course of the study. However, only the
cent), serious (11.23 per cent) and slight (86.67 per cent). On
more important are discussed here:
motorways there are estimated to be 7.6 damage only acci-
• All injury accidents
dents for each injury accident (Design Manual for Roads and
• Fatal and serious injury accidents
Bridges, 2004).
• Slight injury accidents
The average economic values (Department for Transport,
2004a) for the prevention of motorway accidents at 2003 val-
The model results that are presented include central (best)
ues and prices were: fatal accident (£1,630,430); serious in-
estimates of the ‘multiplier’ representing the effect of MIDAS
jury accident (£195,150); slight injury accident (£23,500);
operating alone, together with the upper and lower limits of
and damage only accident (£2,080).
the 95 per cent confidence interval. The confidence interval
Table 2 presents the estimated overall annual benefits of
is the range within which the true value may confidently be
MIDAS operation for the 800 kilometres of route on which it
expected to lie. For example, a ‘multiplier’ of 0.87 means that
operated at the time of the study.
there are estimated to be only 87 per cent of the accidents
There are two important questions that need to be assessed
with MIDAS as there would have been without it (a 13 per
to which the study has provided answers:
cent reduction in injury accidents).
• Which sections of motorway have provided the greatest
Attempts were made to obtain relationships between the
accident reductions?
effect of MIDAS on accidents (the ‘multiplier’) and AADT
• Which sections would be likely to provide the greatest
flow, both for all injury accidents and separately for fatal and
benefits from future MIDAS provision?
serious injury accidents. However, the relationships varied
with the form of model used to determine them and there-
In order to answer these questions, it is first necessary to
fore they cannot be regarded as reliable. In the remainder of
determine the relationship between the number of injury ac-
this article it is assumed that the MIDAS ‘multiplier’ does not
cidents and AADT traffic flow. The relationships are pre-
vary with AADT flow.
sented in Table 3.
Table 1 (previous page) shows the ‘multiplier’ by accident
The monetary value of accident reductions according to
severity for MIDAS operated alone. For all injury accidents
the AADT flow can be estimated based on the assumptions
combined, the ‘multiplier’ has a central (best) estimate of
set out previously. These are presented in Table 4.
0.87 with a 95 per cent confidence interval ranging from
Table 4 shows that the value of accident savings attribut-
0.81 to 0.93. For fatal and serious injury accidents combined,
able to MIDAS varied from about £38,000 per km per year for
the ‘multiplier’ has a central (best) estimate of 0.90 with a 95
a section of motorway carrying an AADT flow of 80,000 vehi-
per cent confidence interval ranging from 0.76 to 1.07. For
cles per day to about £74,000 per km per year for a section of
slight injury accidents, the ‘multiplier’ has a central (best) es-
motorway carrying an AADT flow of 180,000 per day.
timate of 0.86 with a 95 per cent confidence interval ranging
from 0.80 to 0.93. It can be seen that the 95 per cent confi-
CONGESTION RELIEF BENEFITS
AADT flow
Number of injury accidents
(veh per day)
per km per year
Approach
An estimate has been made of the congestion relief benefits
No system
MIDAS
of MIDAS automatic queue protection. The approach used
80,000
3.11
2.71
has been to develop a relationship between the delay attrib-
100,000
3.75
3.26
utable to non-recurrent congestion and AADT flow. Non-re-
Table 3:
120,000
4.36
3.79
current congestion is defined to be that which is not due to
Number of injury
140,000
4.95
4.31
the volume of traffic, but to incidents and roadworks.
accidents per
160,000
5.54
4.82
The non-recurrent delays were calculated using journey
kilometre per year
180,000
6.11
5.32
times derived from MIDAS loops. Journey times from this
tec OCTOBER 2006

373
Table 4:
source are available for 2003 and 2004 separately for each
Value of accident
Value of accident
carriageway of each link. On average, non-recurrent delay
Two-way AADT
savings (£ per year
made up 25 per cent of all delay. Relationships between non-
(vehicles per day)
savings of MIDAS
per km of route)
recurrent delay and AADT flow were derived from this data
per km per year
using regression analysis.
80,000
£38,000
by daily traffic
The proportion of non-recurrent delay that is attributable
100,000
£45,000
flow
to accidents is estimated to be 60 per cent based on the re-
120,000
£53,000
sults of other unpublished studies. Delays caused by acci-
dents are therefore estimated to form 15 per cent of all delays
140,000
£60,000
on motorways.
160,000
£67,000
It has been assumed that the secondary accidents that
180,000
£74,000
MIDAS automatic queue protection prevents would generate
the same delay per accident as primary accidents produce.
These various components were combined to estimate the
1000km of motorway route delivering safety and congestion
relationship between the congestion relief benefits of MIDAS
benefits of over £60m per year.
and AADT flow.
REFERENCES
Estimation of congestion relief benefits
It has been estimated that MIDAS automatic queue protec-
tion reduces accidents by 13 per cent. If it is assumed that
MCCULLAGH P AND NELDER J A (1989). Generalised Linear
each accident prevented by MIDAS (secondary accident)
Models. Chapman and Hall, London.
would on average have contributed the same amount of
MAHER M J AND SUMMERSGILL I (1996). A comprehensive
delay as a primary accident, then the estimated congestion
methodology for the fitting of predictive accident models.
relief benefit of MIDAS is 13 per cent of 60 per cent of non-re-
Accident Analysis and Prevention, Vol 28, No 3, pp281-296.
current delay (7.8 per cent). Similarly, the estimated conges-
Pergamon Press, London.
tion relief benefit of MIDAS is 13 per cent of 15 per cent of
DEPARTMENT FOR TRANSPORT (2004a). Road Casualties
total delay (2.0 per cent).
Great Britain 2003. Annual Report. The Stationery Office,
The average cost of a vehicle-hour of delay was obtained
London.
from the DfT (Department for Transport, 2004b) and was
DESIGN MANUAL FOR ROADS AND BRIDGES (2004).
Table 5: (below)
£11.28 at 2002 market prices. This was used to convert the re-
ECONOMIC ASSESSMENT OF ROAD SCHEMES: THE COBA
Estimated annual
ductions in delays to monetary values.
MANUAL. Vol 13, Section 1. The Stationery Office, London.
congestion relief
Table 5 shows the estimated overall annual congestion re-
DEPARTMENT FOR TRANSPORT (2004b). Values of Time and
benefits of
lief benefits of MIDAS, for the 800km of route on which it
Operating Costs: TAG Unit 3.5.6. Transport Analysis Guidance
MIDAS for
operated at the time of the study. The estimate of the conges-
(TAG). Available from: www.webtag.org.uk.
800km of route
tion relief benefit per km per year (veh-hr) was obtained
using the relationships between congestion relief benefit and
This paper has been produced by the authors and is based on
AADT flow. This was converted to a monetary value using
work carried out by TRL Limited under contract for the Highways
Table 6: (bottom)
the figure of £11.28 per veh-hr. The estimates for 800km of
Agency. Any views expressed in it are not necessarily those of the
Estimated
route were obtained by multiplying the benefits per km by
aforementioned organisations.
congestion relief
800.
benefits of
Table 6 summarises the estimated congestion relief bene-
The authors
MIDAS per km
fits per year per kilometre of motorway in relation to AADT
The correspondent author,Steve Tucker, Highways Agency, can be
per year by daily
flow.
contacted on +44 (0) 117 372 8225 or by email. Ian
traffic flow
Summersgill, John Fletcher and David Mustard are all with TRL .
CURRENT POSITION
Congestion relief benefit per km per year
1000 veh-hr
The study has estimated that, on average, the safety benefit
of MIDAS automatic queue protection had a value of £49,000
Congestion relief benefit per year on 800km
800,000 veh-hr
per km per year of motorway route. The associated conges-
tion relief benefit had a value of £11,000 per km per year at
the time when the study was conducted. The corresponding
Value of congestion relief benefit per km per year
£11,300
value of accident savings on the 800 km on which MIDAS
was then operated was £39 million per year and the associ-
Value of congestion relief benefit per year on 800km
£9 million
ated delay savings were £9 million per year.
The benefits were dependent on the traffic flow, so that on
a busy motorway carrying 180,000 vehicles per day, the
Congestion relief benefits
safety benefits were estimated to be £74,000 per km per year
Two-way AADT
(veh-hr per year per
(£per year per km
together with congestion relief benefits of £30,000 per km
(vehicles per day)
km of route)
of route)
per year.
MIDAS studied here was provided as part of a package of
60,000
390 veh-hr
£4,400
improvements, such as the provision of new message signs
80,000
590 veh-hr
£6,700
and new fibre cables, over a number of years. As these works
100,000
850 veh-hr
£9,600
have varied between schemes it is not possible to give a pre-
120,000
1,170 veh-hr
£13,200
cise cost for just MIDAS. However recent installations indi-
140,000
1,550 veh-hr
£17,500
cate that MIDAS costs are in the order of £210,000 per km,
which demonstrates that MIDAS continues to provide good
160,000
2,020 veh-hr
£22,800
value for money. MIDAS is currently operated on over
180,000
2,600 veh-hr
£29,300
tec OCTOBER 2006