ML20149E757

Applicant Exhibit A-22,consisting of Undated Article, Relationship Between Road Accidents & Hourly Traffic Flow-I Analyses & Interpretation

ML20149E757
Person / Time
Site:	Seabrook
Issue date:	11/05/1987
From:	Ceder A, Livneh M TECHNION-ISRAEL INSTITUTE OF TECHNOLOGY, HAIFA, ISRAE
To:
References
OL-A-022, OL-A-22, NUDOCS 8802110264
Download: ML20149E757 (16)

Text

{{#Wiki_filter:50-t}y'3/yy'4 l E - (<' 3/g y a m iar.n m s.,,n.m a wommm re em,w . in cr-re t!3 2 7-- 00LKETED ?U 'k.c USNRC f RELATIONSillPS 13ETWEEN ROAD ACCIDENTS AND 110URLY TRAFFIC FLOW-1 "d8 FEB -2 A9 :28 4 p}N ANALYSES AND INTERPRETATION OFFICE W 5ECilit,ri / 3 00CKEllNG 6 HWlE i.A AvisHAi CwER and MOSHE I.,1vNEH b Transportat;on Research Institute Technion-IsraelInstitute of Technology,Itaifa, Israel (Receised l$ Ap.1l980) s Abstract-1his recarch extends the investigation of the relationships beineen measures of accidents and l trame now, and considers the hourly now instead of the average daily frame (ADT). w hich has n! ready been 7 #f;i reported. 'the Sndings of this study serve as a basis for further clari6 cation of the interactions between t.Q various levels of trame how and road accidents. Eight four.larie road sections mere studied during an 8-year .N period. providing adequate data basd on carefully predc5ncJ criteria. Power functions are fitted and iljj classi6ed according to: (1) time-sequence analysis for each roads ny section; and (2) cross sectional ar.a!) sis

dm on a one year basis. The results are presented, separately for multi and single vehicle accidents, in a 3 9 matrit fortnat. A linear dependency mas observed between the power and the logarithm of the multiple i.M constant. 'ihis mas done in a sirnilar fashion to the previously reported study of the relationshirs between rnad accidents and AD r. The results for each type of analysis and type of accident are discussed,and three

,]b examples of a practical application are given. n:i

1. INTRODUCTION This research was conducted as part of the establishment of the safety evaluation procedures h.]

for the analysis and interpretation of road accidents in Israel. The format of the entire research, %D based on gathering nationwide data, is shown in Fig.1. The basis for such tesearch is the hd availability of an adequate data bank consisting of effective reporting, storage, retriesat and compilation systems. The study was conducted in four major phases, as indicated in Fig. I: (1) di@ phase I investigates the relationships (power functions) between two measures of total ac. f)$ cidents (density and rate), and average daily trame (ADT) on four. lane and two. lane ir. erurban y,M road sections; (2) phase 11 extends the investigation of phase I for the four. lane sections yy through separate consideration of single and multi. vehicle accidents; 0) phase 111 examines Mh deterministic relationships between two wei hted accident measures and the hourly trame flow ![ P for each t>pe of accident; anJ (4) phase IV attempts to go thoroughly into the relationships between measures of accidents and hourly traffic dow by separating the traffic stream into free.now and congested-dow modes, and by interpreting the results in a probabilistic manner. h Phases I and 11 were previously reported by Ceder and Livneh [1978). Phases 111 and IV are Mi t described sequentially in this (Part 1) and in the following (Part II) papers. In the past, some aspect of road geometry has been identified as a dominant factor in accident causation at a given location. Thereafter, an attempt was made to interpret the .y frequency and number of accidents using the ADT value as additionalinformation to the road B geometry. Neverlheless, when eliminating considerations of geometry, the ADT by itself cannot y 6 be used to explain the oserallinteraction between trame flow characteristics and accidents. For .M that purpose, one should approach the actual trame Dow observed at the time of the accident. In '[.f addition, the level of risk associated with trame flow can be determined only on the basis of smaller time intervals than daily periods. This work attempts to clarify and improve the j% s 6/ understanding of the relatiorahips between measures of accidents and hourly traffic flow, which f is more fundamental than the accident /ADT relationship. The analysis and interpretation are i performed in a similar fashion to that outlined in phases I and !! of the study, ,{ 11 .G

2. SOME PREVIOUS STUDIES WN The common measure of accidents considered in relation to _ hourly traffic flow, q, is the i

"accident rate" A,. This rate is usually based on the number of accidents per year refmillion ! y M (or 10') motor vehicle. kilometers (or miles); that is, the annual number of accidents based on ] the annual amount of exposure. Several forms of rela'ionship between A, and q have been i W h y [b 8802110264 071105 / PDR ADOCK 05000443 f e g PDR y e.s

/3, i3 n 20 Armt4 CEDra and MosHE LJYmi {f p J, ] fouh 3CCM$; E:A linea"4 varitU f b. } singi)N,S k l 1

• * * *I kg.

1 i 1 ] Usf;Tjf valu.jp a v Bakf$k II _ jIl 'l Il i Il i $j 8 l 8 - -l E

pertff,

-I-) F 2 $a j e s gj i t ? 't - ij ju ~ - !l3g1 ~ } ' M S-a e 8CCH Q ~ }U j

upwQt,

J j

[1 'I-11 do*H } j(qat [ Aut9g W 8ttrilft h i - f{ singf(( genc ) ~ sim f, [ ?-{ c c ~ a! BrildO i-I - gi I fr

- j

'~ i Thisb 3,pd v a h

• "'Y.

? i r emph.3 e r wy n h,' ~ .) g y _ i 1: tt m ji i ? j@,{I.i jj [d i i' s II f f 3 Ihe 1.;a i{')l,'l .1 d "CCIE'f m 8CCK 5 4h ~~ day.f T; lWBU {c h} g., '4 ',i i l ?I -- exCQ y,2!'. ,I - fj b t t i 'i extei, j } 1.;

) 2f cach((& (. kh s i b c i.1 o Relatbnt Iuccn road acck!cnis and hourly tdc Amt 2t k' ' a fotnd in the literature. The variety of A,-q dependency is probably due to different types of h l 3 accidents, ranges of Dows in the analysis, and road designs, j f! Delmont [1953), found for two lane sections that A,(during daylight only) increases almost ! pf linearly with q, whereas Smeed [1955), has shown that A, for total accidents has a small 6 sariation for different ant.ual q values. Nonetheless, Smeed pointed out that A, values for n single ve sicle accidents hane a tendency to decrease with the increase of q, and multiochicle i 3 accidents show the opposite tendency. l Leutzbach [1966) and Gwynn [1967), hase concluded for four lanc divided sections that a U shaped dependency exists between A, (for total accidents) and q where the minimum A, ' Iq values are obtained for q salues between approx. 600 to 1300 vehlbr per two lanes. Thereaf ter, 9 13aker and Gwynn [1968), noted that A, (total accidents) increases rapidly below q = 550 vehlhr f l ) p per two lanes, but has litt e variation beyond this now value. y Pfundt [1969l, has compared three types of day and night accidents: rear end collisions due 'k j to blocked lane (s); rear end collisions due to slow and disabled vehicles; and single vehicle ,p g accidents due to loss of control. For the first type of accident, A, tends to have a convex o 'q a g upward curse with q (particularly at night); for the remaining two types, the curves are convex downward, in a different study, Leutzbach et al. [1970), have shown that on a four. lane i Autobahn section in Germany, the resultant U-shaped curve in the A,-q plane is mainly h fh attributed to rear end accidents, whereas q values have little effect upon the A, salues of k single vehicle accidents. Chapman [1971), analyzed accident and now data from England, and "[ generally r. gree with Pfundt's findings. j Recently, Drilon [1976], in a study of 8 four lane sections on German Autobahns, found g similar results (U shaped curves) to those reported by Leutzbach et al. [1970). In addition, qg s llrilon hypothesizes that the minimum A, value is obtained for the most frequent range of q. f .l j }4 j 1his hypothesis is examined, among other analyses, in a following section. It should be s emphasired that all the above mentioned studies consider the relationship between A, and q gg 3 E only on the basis of roadway classification (cross sectional anslysis). f,y Jr j

3. DATA CLASSlflCATION AND ANALYSIS ORIENTATION f[.

3 g The data were selected, based on carefully defined criteria, with the aid of the data bank of } the Israel Central Bureau of Statistics. As shown in Fig.1, the data include (i) fatal and injury G g accidents on 4 lane interurban road sections, including the type and hour of day of the accidents; and (ii) hourly trathe flow: a daily prdtile (accomplished by fixed counters), by hour of ' { p

$ }:

day, for each road section. the overall data were gathered for an 8-year period (1967-75), t el excluding the war year 1973. In order to eliminate any undesirable and/or unknown influence of g:[i external parameters, four major criteria were imposed as follows: {li (1) 1he roadway sections exclude geometric design elements which disturb the 'raffic flow p ll (steep grades, curves, roadside obstacles, etc.), and are isolated from interactions, entries and b u ctits (to eliminate the influence of cross traine); .7 (2) No changes were made in the roadway section characteristics of section length, pave-l.! ment width, and shoulder width during the period 1967-75 (to ensure a comparable base for the data); ' lHi O) There were difierent daily pro 0!cs for q: two daily peaks, one peak or no peak now (to h establish generality); and [ ! ) (4) There were similar daily profiles for q, excluding weekends, for each roadway section ).[ Mring the period of 1967-75 00 ensure steady travel characteristics), t In order to clarify criterion (4), an example of aserage daily q profiles of a roadway section @/ is exhibited in Fig. 2. The lef t illustration in Fig. 2 shows that the general tendency of each d,' I Q' q year profile is eminently presened, llowever, there are changes in the levels of each year-t profile as the result of increasing ADT salues with time.1he latter cFeet on daily q profiles can a be climinated, to some extent, by the use of normalized q values. That is, taking p = glq., y instead of q, where 0< p s 1.0 and q, is the maximum q salue, demonstrates the similarity d' betw een the average daily q value,as show n on he right i!!ustration in Fig. 2. It is w orth noting that j criterion (4) has also been applied to an examination of four seasonal average daily q profiles for l each year, to ensure the steady characteristics of the trasel pattern on each selected roadway y l i 1 i ' %V UCMf,,'j W N#DN-im IN Nhh h b)kk bhbbb bN 3 m p us htt: tad Mosur 1. rwa b f *.f 22 A where aq i r intervals of M l} t' M', /) O The mead ,9 / o, f,L l*g /, ['y.., os , ; if

I particular y

' h *$ ll 'No lf {p\\ sectional an o{ ~;e / $. ( *. *. 5-, \\a S or I;j( Q y O, n,

l,

\\\\ ll 1 3\\

,.}

'2% ll, v gi s-hl I \\ h,'\\" I 3, iX !l ,,%, "0 ), ' ^ h' ' I o. I -k nooo ~ where Ln =i i. denotes a sc ', k,f ka'o ( lf, f \\ = th?n L = Lhg \\ 6 i xy' h - o4 ll t ,, :)I e./ i. e n,o '\\ el l \\ i type) whichh a i 'i Another D \\, lf f os !! l !! l ~ d-f interval. HeQ ', dI 's ~ ot 5 fQ \\ ', ',5 f ) 'l71 /l f~, on -%', gg r t..L (I l, o /p i.% ~ where 1,(q)h I2 e se ti 24 o s s S ,e 34 o s s * (cxcluding v,.'), w e e, w.e e, n exaQ Ils 1. An etample tRoad sy 13.ser; ion 6) of daily profdes of g and p (normalized sa!ues) foe an 89 ear time sequen.' reriod. cross $cctiok' ,) -

7. section. Ily the above described procedure,a total of 8 four lane divided roadway sections were i

As the cRp $) research: selected for study. i g The accident and q data collected from the 8 roadway sections during the 8 year period were analyzed on a separate basis ( described in~ detail by Ceder and Livneh,1978), as follows: I Z. (i) time sequence analysis of data for specific roadway section over an extended period of time pl.f (8 years);(ii) cross sectional analysis of data for a given period of time (one year) for a group y of roadway sections (8 sections) having the same roadway classification. 6- 'this approach enable. in essence, the consideration of dynamic and environmental erects (3 where A,, ;g -inbr. rent in the relationship between measures of accidents and measures of traffic flow, in the (acc/10'km next two paragraphs the diRerences between a consideraion of ADT and q values with respect to the time sequence and cross sectional analyses are explained. accident ratQ In the time sequence analysis (for a specific roadway section), each ADT value refers to a U diGerent measure of accidents,due to their possible mutual changes over the considered period fi of time; hence, they constitute, say, eight data points for the 8 year period. The time sequence n.j analysis with q is based for each year on a number of data points (dependent on the onsidered $5 q range and intervals), whereas only one data point is presented when cons' ering the !;M accident /ADT relationship. The ADT value increases, usually, with time, and this increase is reflected in the time sequence analysis. The q values, however, could have only an upper i bound, and therefore,it is perhaps only possible to notice the changes with time for high q M values. Nevertheless, the time dependency,in the time sequence analysis, for the relationship p',p between measures of accidents and q, is relatively negligible, due to overlap among the q

fe ranges of the considered years.

e;y in the cross sectional analysis (for a given year), each ADT vahie refers to a diHerent y measure of accidents, due to the various roadway sections; hence, they constitute, say, eight d data points for 8 roadway sections. Similar to the explanations for the time sequence analysis, l k: when considering q values, one obtains several data points for each road section rather than a g single data point (for the accident /ADT ietationship). 9 To summarire, for the time sequence analysis, the relationship between measures of t,h accidents and q emphasizes the uniqueness of each roadway section, while for the cros? hg sectional analysis, the uniqueness of each year (or other given period)is emphasized, hy, q [ h 4 MEASURES AND MODELS The data were gathered, basically, by matching each type of accident with the aserage q rial value at the time of the accident.1he q value is considered with respect to the interval q 2 Aq, g d i 7 - n.-,wi m,w,.,,o.. ,.n i ,,w-c,, w. er. 7 I t ] (V ,j.3 ( ) ' M'j Relatioahips L.Je2 toad accucats and hourly hallic 202-1 2) q: (h where og = 50 vehlhr per direction of travel (two lanes).1 hat is, each q range was divided into 5 intervals of 100 sch/hr on a two-lane basis. The meatures of accidents were carefully denned and selected as follows: 14 denotes either a i r, zM particular year (for the time sequence analysis) or a partici.ar roadway section (for the cross M.9 sectional analysis). {.l[. 3,(q),, 'y M9.]. f accident density (acc/km)' h L, ,for the mterval q og ,,;g t where L, = the length, in kilometres, of roadway section i; in the cross sectional analysis i l.'ld I o, ~ denotes a section and l4 is a sariable, while in the time-sequence analysis i denotes a year and j@ h I then 1 = L,i.e. has a constant value;and N,(q)= total annual number of accidents (of a given 0 h type) which occurred during the five work days of each week for the interval q ! Aq. I'lb 4 ^ Another component which should be taken into account is the exposure time of each q og /[.'.s intersal. llence: the annual exposure time of T(q) = IJg) = traffic flows within the interval q Aq ;

J l, [

where I,(q)= the daily exposure time for the interval q ! Aq at the jih day, f = l,2,...,261 , p(.[, (excluding weekends). b.g An example of T(q) distribution is shown in Fig. 3, where the upper illustration is for the NN time sequence analysis (one section, over an 8 year period), and the lower illustration is for the 'h cross sectional analysis (one year for 8 sections). Sp etions recte As the consequence of the above definitions, two accident measures were selected for this ,E;h h research: 'Q*! k (d rcar period h-i as follows: A,,(q) = (1) y iod of time I9I I g;,Q for a group ( A,(q) = A,gq).10^ (2) h'h Ital effects Ilow. in the where A,,(q) = weighted accident (acc/10'km hr), which means the accHent density ,t t hJ Qg,(M .ith respect (accll0'km) per one hour c2 pome of trame flows within the interval q aq: and A,(q) = H accident rate (acell0' veh-km). W*V M fh, refers to a cred period Moo 1 i i i i i 1 iTl i rTTT P T~ (, Considered 6000 ~ roohor ts.secton 6 ~ h.pg[

3. sequence

.ik' idering the $W,% increase is d2 1~' !.k ~ < an upper ,,b,4 for high q fd O

elationship L

9 nong the q y g a dillerent ~ 8 '* ~ s , say, eight f ce analysis, 7,. j l l ill.cr than a ,y g ["3j eg teasures of ~ 'd,1 woo r the cross ' y,'s,. o-I d. $.3 O Foo doo 6CO 800 1000 tMo 6400 160o IF00 e;- e bevhow) r y' F avelige q lig 1. An e sample of Tlehlistributi.rts fur the inne sequence (urrer rait) and crosi sectional (lon er part)

val q i Aq, analpes

,.[ j E.,il M \\ j 3$ SkNk(Spgvc,q@9eitM,lbh%$$shhM[y Pf: M t '.'.-)PC 'y

• W

, %]$$$$i,fI/fN[Mdb,,$bh8 P*lPMM DN - c ~. M@ ( j. 0 Ausuu Cuu and Moss Umn N 111e models w hich were found appropriate for the analyses made are power functions (same h I as the models in phases I and II, see Fig.1): 'r, j. A,,(q) = a3nn (3) W $j A,(q) = a.q'a (4) .: n w [ ) where a3, a and pi, p. are constant parameters which are determined bj a linear regression M yj technique. Note that the powers p3 and p determined the functional tendency (certainly for R pos;tive ai, a. salues): h p M p3,p4 > l$ convex upw ard; g'J p3,p.= l$ linear upward; gb DI { l > p3,p.> 0$concase upward: D); j p3, p. = 0$ constant linear; and ? j; 0 > p3,p.$ convex downward, M; The models represented by eqns (3) and (4) are fitted separately for single and multi vehicle kq c }y)

^ -]

accidents. The sum of these models for each category of analysis (time sequence and cross l sectional), might reveal the possible lependency between the total accidents and q, and might WG ~ 0,! also yield the earlier mentioned U shaped function. If the latter is the result, then one can arrive V.k at equations satisfying optimum conditions, i.e, dA,Jdq = 0 or dAldq = 0, which yield the Q y]q f following optimum parameters: ,.} '.4 euip;o;j)"'P'P b,7, ' pia (5) ,y y) N'* ~ (\\ pia;Q**'D @h - p;a h) /

O l

where the prime represents the parameters of eqns (3) and (4) for single vehicle accidents and W 3 i v.' the double prime for multiochi:le accidents.The substitution of qu and q,,(prosided that each l'9 h is a positive value) in eqns (3) and (4), tespectisely, gises accordingly the optimum weighted d accident density measure, A,,,, and the optimum accident rate measure A,,.

• 4.m

+ r 1fy -s d.

5. RESULTS AllD FINDINGS 1he results of the time-sequence and cross sectional analyses are summarized in Tables I h) y%)fk and 2, respertisely. These regression results, by accident type, refer to egns (3)-(5), and are

'O3 n accompanied by the standard error salues (SE,,, and SE, in units of A,., and A,, respectively). g lhe results given in Tables 1 and 2 are shown in Figs. 4 and 5 for multi and single vehicle 1 Vb accidents, respectively; also an attempt is made in Fig. 6 to show the summation of the curves. b) Note that in Figs. 4-6 all the curves are within the actual q range. From these results and G E. analyses, four major findings are identified: 3 (a) Ti>r the multi tchicle accident models all p,> 0 with a convexity tendency for the /M time. sequence models (i < pi< 2), and mised functional tendency for the cross sectional -Q [E models. While A,,(q) is always increasing with q, the A,(q) is either increasing or slightly

• j,'

L decreasing with q(-0.37 < p.< 3.82). There are two interesting observations: (i) three out of 4-r Q.h four time sequence models in which A,(q) does not increase sharply with q are characterized q p by low upper bound value of the q range; and (ii) the two cross sectional models obtained [p L, y for two > ears before the energy crisis (1971-72), indicate a lower safety level than the two )1 models af ter the energy crisis (1974-75), where all four models hase similar q ranges. i, (7s (b) Ilir the single rrhicle accident models all p.< 0 indicating convex dow naard curves in the A,-q plane. On the other hand, there is a miteu functional tendency for the A,,-q g. relationship. The comparison between single-and multi-accident models demenstrates less Md cune-dispersion of the former for both the time sequence and cross sectional analyses. J h (c) For the summation of the single and multi rchicle accident models halI of the A,(q) N~ {E models are characterized by U. shaped cunes, and the remaining half by consex downward [.- cunes.1hree out of four of the A,(q) models which do not hase U shaped curves are indicated Rf (' Ls i. f; &) 9 t 5 d _ -. + wyw=, msy.-w,* m e m c.-,..%.,. ..,w,,..w.m = ? .h* Yh. ? i h ff k (- y_ i 13 Relatbnshiri Cn rosJ acchtents and hounty us5c Bow--! L' i:

lions (same

3. :

e lf 1 (3) . v ',.,i, (4)

• . c,.

[, j *.-

a; -

r regression ettainlyfor

• .. p-

.t 4 (2 R t i i i 4

j.

• c..

I. G. / A g 4 i i ' -l>Q 3 0 (r a ::

: R S :: : :

0 *. r E * .Y M llt ' i a a a 4 a i i a i aa a a ai

• y

,.h multi vehicle m , o %l$ g 9 g-1 i t i - 9 9 ~ t 7 i ?- ~ cc and cross 9.- ~ ~

• y ' g(

? ~ ~ q, and rn. ht ig ~ % j ,4 q y 'p qq , q qq g 'e p,7 p i '.ne can arrive j, e g

:: ;a

~ ich y,ield the -l1 2 e -A a* a a aA i a 'I W T J E f f 9 'I e ,f,.. i ($) -i ..t, ~ ,w.,e = g ~r ' <. i t.1 18 l 'h ?kb gil'i i g, %q f accidents and E f 1

• E IE E.! g? 35 3,

1E ided that each .l i i a i i a i a a i i a i i aa k f F. C tum ecighted h(,3 :' t ? s vr W = 8 s a f. 3 5 :i l,M M' i i 4 i a f4 i d i 4 i 4 i d u-f (: ) 8 {n:,g d. Tables 1 'n 3 T, Ij in W), and are ,i f. E j, q 'p q q i, j i i ; ; jj [i j.,u,< ~ q 1,,f..e respectn ely). ~ o ~ ~ ~ - ~ , w I, j

.:; 4 4.b:

single vehicle i j , i.', 6 the curves. I E y~ ',.n w se results and t(a f;-: .- a t e, k@i liy ~ Jency for the } ross sectional j-

M

ing or slightly jj ~ ~ h, i) three out of I ;f.i (. characterized ...f. .idels obtained t t I than the two ,I g..rge

t....._

.,i 6 ,nges. -j . (:,,} ) - ward curves in or the A,,- q l Q.p ionstrates less N

halyses, j

M il Cf the A,(q) i .t 6 ~ set doun :iard '.A s are indicated g,. %y { ! *l. .,3 ' hM. e.Ef [g ,) N *I

• U E

--Y^WD

• h D b%

3 y, i.j h. 'k' ' f t y-m / e n. d ; 4 a f s r b.%,.. ' _ g i \\ / e-26 LIUSMAI CEMR arid Mosne LIY/:f H l, .: S J.'.'o i d. V.; t 1 e.. % b. i::K h.1;; a

t i

i8 i'p y .e ):N g i. 6 8e [ j 4 hs (f e ..a, i 1:n d 4 4 4 5 t. 'M 19 ?'d? yi g. 't 9 9 o e s R R C C 3. +.!< a h ~ ~ pe e: -i 4 - 4 ! ~?g / ,. e d b [ $. $ $. k $k kk $ ". $k A. k. $k } .j ty p. . e o e o e o e . e k E. k$ E. I. k kk E. 8 k A. 1 es e-e1 s ..a n g. 3.- g / Y ']) d ,'f] 3 3 'e R : 2 2 ". : 2 2 ' - ho = k N* N. ' k. k. 5* kE k.

• k

,E ~,

v., j g

x o .. :,o. ),i T. 7 8, ',.s.3 s v',; I 7 7 5 4 e e ,e y ag 4 4 i- '4 2 [1 o a i i %w l e J . a-s r.<.~ t C ',. i '3 la I sE g g S $ 3 $ = ; 2 0 R E .' ? I 3 = ,a d 6 4 4 4 6 4 6 4 4 4 4 4 e d 4 it '/ j w j ~- s r : : : : : : : : : E s: s: l' i N.h M 0 a a ss aJ a4 4a ss a. 44 1 e y y E l 1 111.-. 1 2 5:h; "T S ?

• 7 e

o 27, i k-T 'o 'o 'o 'a 'e 2

• 2 e

1~

1. %

? 7 7 7 & y4 1 e [} a

a 3 A : 3 : 3 9

9 9

,1 J 4 4 4 4 4 4 4 e - l ;i<

-I% '!. F 4 j I ji? e ,.*vg 3 3[ . s . s . e o ( 3 y 3 6 0,.:@ tr y g g e e m 1 1* L., A 7.I b

v..

-i;: ,*e Ml 5:, f. y', 'W 4 v i eb 1 SJ' l' f I ~ f .1% ~y:(d l '3 s$ r 6 r., e ij Iit ..a n

y. 'g, C a

4 o p, s. n' 1 e s t. g .) .c ! '. 'c. g e. .w sime r or.y r...

• M.pr**. we - --- - - - -

?.. .,,w s t h. b Y TYs5$.h b $'!YYY Nb Y h! /' a 7,..;. Relat6nships ticticea read acckfe;,ts and hourly trame fba-I 27 , !43 I!': I"I I I I I I i i I I 1 1 1 I I I & I i i !N MG -0fi (rQl%. Sf(j@ Predg

• f{

k "" " 7. - a N; 4 ; bg ,; y* t' N '[ b N i} j -n .r. JA { $r l

p

i )'~ .d :~ 9). r ?=. /p g&' &m he q.e 9 hg h.1, y }. } s c..M .g . l'[.J '. k

• S g f5

'.*..f,.' y - a e, D V2 ' J, i M/ ,/ t ' f. I'S . qq / ?m . t 0 M **' ' I 1 ' ' 'W ''ii M PC SM PM fSM M l'O SM t?4 f* M 'f '.[ g f av4 /M d u ip'f

• .7 l'ig. t The resuht t.f the time-sequence and the cross sectional mode's for muki4ehkie acciJents.

f $.! g.,':.} %.1. + 5 vr, r,7 m i , i i l,,,,,,,o j ? [.{ ii 1 i, ,-. n. --,

s.,...~ -n 3

i g t.( i q,, . c.. ,f .e 9 .$ \\ \\.,o ..w..- f SR W.4 g { pn '4dI [;$ g jb_4 - \\ __ I_- 21.,.. 's g.'.[ 'i Q[ r , U,5 i - L }' _ * %a e n. f% [ .) r .3 3 3 S

a i

U g. ) ..., p. b s .4 'ed s. \\ .w e t t' a 9 '^ i \\. ' ~. e. h SI'**. ,.31 N-(( es\\ l ^' d[,t.' 8, km... ,x ,e ~ ~ me ,m ~ . e. .ww 5

• \\

lig 3 The resuhi of the time sequcNe and the cross sectiwi models for s' sle.ichkie accidents. jf m 4.11 j i4 i s ", g ') f { WT.'T " 3 ') .] \\ 0 M 0 F'T -W [ -- 6, h .. Ik5 3 h h i l O k (o) !.;i V(3 A,m. c.m.a umar us,n v

Q by urper bound on the q range below 1000vehlhr per two lanes. This might be (one of t the valu\\

reasons for this observation. Nevertheless, the optimum average q,, value for the minimum deph'h A,(q) value is 500 sch/hr for those U. shaped models. This optimum value results from opposing tendencies of mulliachicle accident models (A, increasing with 4) and single vehicle ,pp l.ly g accident models ( A, decreasing with q).1he average q,, value is below those (600-1300 vehlbr) ,,, [,E use which were determined by Leutzbach ll%6) and Gwy nn [1967), probably due to a higher upper U' bound on the q range (about 3000 vehlbr) than that indicated in Fig. 8. ,,o,d thre (d) The hypothesis made by Drilon [1976), that the minimum A, value (for total accidents)is (3,,[,, .i obtained for the most frequent range of q,is not strengthened by our data. In fact, the opposite i diffe ? r; is the caye, since out of sixteen data sets of T(q) distributions, none agrees with Drilon's Tabid h hy polhesis. Theh)y

6. MATRIX REPRESENTATION OF Tile RESULTS uqul y

Following a similar method to that outlined by Ceder and Livneh [1978),(phases I and 11 "'*fjf, 1 indicated in Fig.1), the results are plotted on the matric the y-axis is the logarithm scale of a3 '**((y.I fh or o., and the 1. axis represents the scale for pi or p, respectively, F,.1 1 :p 3 $ ..,9 ?f. and (:!.4 q a 3 ..} D_MW M*Il Crog,.5ettrrol enod*Jg, y .~, 4 we tcri ee g l !. 'g, k{ '~ 3 N$ -Q Q g> impo ;g NJ.; j, b howt c.. 'd.f, safel,.7

e. no., y

[ p ast'- 7 c-. g/" ~ M; "j, pp wa, wl

1. k 0

g.w1 [f i;'. '. 5D u.t ~ CI i* g s': -i,f J 'h s .v 'g s, i , fi U6 j$ 'J[h l I Jh.. )

a;;

+. w.a, ,i.e, :

• +

E,g. i = h \\J/ Ijt w Dn.. W g.- s .Ml Ir.4

s. ?.?

v m 1 , 1.... 1,... m r% 4 SO 'So W am e a.o So N no (% ew 9 .iveh (Nw l }.)J h r, 6. m,,,,, i_.s m m.a,. ..sww.<< ai.... a i 1 + 'T9MFWwM P san. M* *p s 7 P' WW v 7 6 m w s ' j t t t 2 f ,] (,.g Rclalbrnh@t .cn it'ad accdcra,3rd hourly DaCc flas-l kj D v t lhe log a3 values sersus p., values are illustrated in Fig. 7, and the log a. values versus p. C (0"C op the salues are illustrated in Fig. 8. From these Ogures, one can eminent;< obserse the linear l d,d the c, umum v dependency that exists between the variables. Consequently a linear regression procedure was results from app ied to each set of results.1he Cited models are shown in Figs. 7 and 8 and ci-8'"EIC'VChICIC are indicated in Iable 3 with their coemcient of determination t'. In addition, the F statistic is O'. Ll300 vehlhi) used to examine the possibility of combining the linear models of both the time. sequence and ' '/i i higher upper cross sectional analysis.1his examination leads, ined, to a common model through a 2d it acc, dents),s three-stage test:(1) between the sarianees:(2) between a;s or #,'s; and (3) between a,'s or F,'s . ;N i i (see TaNe 3). It was found that for both the A,,(q) and A,(q) models, there is no signi6 cant zith Drilon's difference between the linear models at the 95"e lesel. The common models are speci6ed in , T( I,the oppoute TaNe 3, and me shown on the left part of Figs. 7 and 8 with respect to each t>pe of accident. I'.*N The remaining parts of these 0gures i!!ustrate, for each type of accident, separate time-sequence and cross sectional models. Some of the Sndings mentioncJ in the presious section K(! ', are clearly and sy stematically demonstrated in this matrix representation. hases I and 11 Each linear rnodel shown in Tab!e 3 represents a family of curses which intersect at a W bm scale of a n unique point. For example, this point in the A,-q plane (A!, q:) is obtained by: )'E A 1 = a.(q iY*, W$ j log A! = log a. + r4 leg (q t) W ' ~. and therefore, log A; = g,,, log (qi) = - pi, and similarly, log Al = a, log (q:) = - an 1hese intersecting points are symbolized with an aucrisk in Table 3. I

7. EXAttPLES OF A PR ACTICAL APPLICATION

.fj Knowledge of the proper relationship (and limitations insobed) betw een A,, er A, and q is i ^ important from various aspects: trame planning, design, operation and research. This section, hl-@ however, introduces examples of only one practical application. That is, the evaluation of the y f:'; safety lesel either before and after implementation of a roadway improvement project, or af ter 6 short term operation of a new roadway section.1his practical application is discussed also in yn y'a phases I and 11(see Fig.1),in siew of the relationship between measures of accidents and AD f. 1}, ] 3 Etample 1. It is assumed that in section No^. 6 of roadway No.13 a safety improsement &,'S was carried out in January,1976.1he data co!!ected after one year are' 'd ' hi 9 h.]1 + .w .4,. r o w y:3, R taires el rwa. ,We. m u' u. tMr. rndn. m(c. !5 > :

• I q

DO sch s el, = c h. ich ich ich L) $ ) g.- 0-5:0 1044 0 114 iW 0IM 016) 0 416 0 652 b1 50(Lino 261 0lm 0IN 0W 04'o 0 511 0 61) ,F', 1ML1%0 $22 0.343 0 175 10'*0 03M 0 843 0 2NI kM - IW 2310 4437

4. l Mt O l3f, 0 933 eC4 0 533 0 014 P{,, l:

a ._.,_.,._._.y. s<ce.wo trewren) g;d ( - rv,.,co kyn.,c np s 5Me - iewe

• W<*

C r.11 - SPC 'or cl A'

• \\

(;ig) f 7 s enesewe : \\ g g, s. yg_uy _,, _ f ,e p.4_ f g wewwo_ ,x 1 N N I .n s .o ' 4 s

• e.

(.h.' h t y r, te .e

• v s.

4 1 ej\\ p. g; b Y.

• '[

. ; -.m e a

t j

N, p r e, y n, 7 m,,,,,, _ u-, n,. m a a,,m,,, e %.,s, at 4, tesala b accdcr.t twc anJ the rema rie, pri-b Mh acodent tge ard 1pc et ar,alysis O w 4 Ww

.,w,w nsw. m~w wrwtwwn~~~mm.

s h. k. f. / 'l b

y.&., j 4

r ..r. .v' WX.-) W ' 9" ', 4

M. *

? E 4.. L ) r a &. ~ g r .c s:. - l -b :

W

o. N Swwge-ve'ucle muf% - we%Cle -b seve-vemck koM -M s mum-ww komrnoni-J { ,\\g /a ao* cross-wc ioret Sme-sequence - \\. l

• \\

cross -wctionoi --- l \\ tene-sequence - ,o $s. .k . sd *,,- s\\ r \\ b ;C' ? I 5" F t w %,".. Es.

• \\

A.. x v s C 1 M&. h(, O J d' r:;; -; N 2 c c' N @4 g ,,-Q so'* g L W. 3: \\ JL..* N g s .c.- g WIL w o* g \\ s g 7 s 'O ~ A. \\ 's E r r. ~?! t s. a 1 Q~, \\ ws n x s $<.: S g .1 l l s g@.,. 3 \\g l t I l l ~ l @ g,i $.s :*. }< -4 -3 -2 -t O t 2 3 4 -4 -3 -2 -t 0 -t O I 2 3 4 kT P. P. P. $?D s (27.l Fig. 5. Matris representaron of the moders parameters s. and p =here the left rart datiastasheS the .t resuhn by accxkM type.and the remasseg part-by both accalcat type and type of analysa. L$ W:. W@,i ~_ en.s. o-c " ~o. n,o n r.,, s.. v - - n,, O n 3 - n e n u se .s p, a w. n 2 :. m # $ py.:. l -,,.:..-..n :.c.,l,.,.- - ~. ~.;..- : l-{ x //f E (4-

• *...,,' Q_,k.:

3. E =.

M 3 n. ..w*-s?'..g;*3 y*f g*p t: ~m - - ~* 7 , Q j; : z. *4 '_,..j..l;.;} ; ;\\',ff;.dji;,. p<. ',a i;; % y,.* e_ >:-?l > t *. .,, >.,.. ;..;3i %,Q;.cx. i 4 ...}.. . p. 7m ) 3 Relatbaship(f ca read acciJents and hovely tianc flowl V 31 S

• ,/kk TsNe 3 Rest.its of the liner dependency bet:cca e, anJ r, and s and f.

i.q: -tu,.n.i. . i.,, i..,. u. i. ,q

• b". L.

P. '! ti. ~4. i ..wm. .. cia.i

• ~a

.. ww. mismi ~.f 0.40 0.47 0.14 0.34 -0.94 -0.76 l ' ;.

• g 3.04 2.7

.I.73 2.11 .t.37 .t.52 r e 0.H 0.99 0.99 0H 0.99 0.99 .'. 4 i.w. t 0.72 0.62 0.62 -1,71 -0.74 .l,76 ',,, F' M s..t ..s .I w t 1.89 i.7) l ss 0.34 0,e4 0,94 f.'. A 2.*1

0. n 0.57 0.46 0.14 0.l?

., ' J y dv. i y iian.is m.i, us.03 no. : vi.34 m.n 5

fG C.

8, 0.01 0,86 0.10 .Q.33 0.43 0.35 ?.i - t'. <5 TE e, ..rs .:.s3 ..n ..si .:.u lN:. t P' 5 11 e 0.99 0.99 0.99 0.94 0.97 0.98 %., ; j h. 1.,,i~... W

0. n

0. n

0. n

.i.n

3. n A

3; i. -g u s... 3.32

3. n 3.

0.9 -0.o. 0.04 N,'. n Ig

w 'l

a,. C.,s 0.4e 0.7,

0. u n.n 0.o

,g e2 y m. r. nr.n in.,3 m.n en.n n.n

)s, t

h, ;. 4 j[rh l witr'i suitable units to those indicated for the models. The question is, whether the level of 1 safety improved, and to what magnitude. According to this research approach, the results 9,'s { devised by th'e time sequence analpis can be applied to this example, llence, the evaluation g procedure is based on the results inJicated in Table I for roadway No.13, section No. 6 for R y each q range along with a confidence interval. Since the row er functions are intrinsically linear $;, )4 g1 (can be expressed by naturallogarithms,in a linear form),95Fr confidence limit can be found W3 'g-l for the new data (after the improsement), in(A,),. or In(A,.),,.. in the transformed plane d ag accurJing to: F.,i.2.- gy c f.gl g t f(n - 2,0.975) s. I4 I + S" 'I'" "A' T" 5 I{ lin (A,1, -(In A,))2 MM g4 tcp. t., gg for a two. sided 95Fc lesci t. test u.,ing the f. table with (n - 2) degrees of freedom, and where n is i p.. (W, k. P, s tlr the number of data points, s is the standard error for either in A, or in A,,, and (In A,) or T u, j (In A,,) is the mean salue of in A, or In A., respectively. Note that this confidence limit is i %E based on the assumption that the residuals in the transformed scale are normally distributed th? '. k with mean zero and constant sariance. If, for example, the confidence interval is t SE,, and k;,$ I W 2 SE, for A,.(q) and A,(q), respectisely, then t:y substituting the data in the models, the b @i following results are obtained:

p

/* A,,(g) A,(g) ,T.,., Ng. nudli-safety sinste safety rwlti-safety sing'e safety l' y.. 3' ( s eh. c hange

• ch change seh change ve h.

c hange Mf',d N('e" 29 0 140 01D 0 044 0 500 .o Qf k:7 no on O co:.... o.343 Oow .p' ' v 12 % IN2 n 0 077 un 1 394 n 000 m 4 W.7, luo 16:2 o cv s 32n OM .$ 4 ni'K ant imprm eine nt ' } {'N T .an imptotement, but not s;gnifwant . [e ma deterbralbn. but ret sign.ficant ""ognirKant dettibraboa i V [1 !\\. 't r;.

w.. m r4,erm,w 7 my, pr-mew -rvew=+;

WI 4 ( p g r 32 y/ Amnu Crt<a and Mnus 1 nm V j The asterisks attempt to interpret the results with respect to the conndence interval. It is h worth noting that the data could also be analyzed by the A,(ADT) and A,(ADT) models W p}Q l specified for the considered toadway section in Table 3 of Ceder and Uvneh l1978). In the tJ latter case, the results indicate significant improvement for multi. vehicle accidents and l signincant deterioration for single. vehicle accidents.Certainly,the consideration of q instead of h C[ ADT determines, more speciGcally, the relative changes in the safety level after the improve-4 p ment, Furthermore, the knowledge of the exposure time for each q range might lead to isolation d,Y of the problematic daily hours from a safety standpoint. g Eiample 2. If the data from e.rample I are associated with a new fuur lane section (also k p{ after one year of operation), then it is only possible to select an appropriate cross sectional 6 i model. That is, the time-sequence model cannot be applied due to lack of comparable basis n g (.3 and/or information. The selection of a cross sectional model might be based on ttvo criteria:(i) Q gr h ?,$ that it includes the upper bound q range of the considered data; and (ii) that it reflects the [j$ ), /.. environmental characteristics of the considered new roadway section (usually exhibited by the y l latest > car model which satisfies criterion (i)). Consequently, the model selected is that of 1973 [ and, in a similar way to example 1, the results are obtained by substituting the data in the 'S, i models indicated for 1975 in Table 2 (based on SE,. and SE,): ,2;fH y 4 H .;n sq A.w A.w U q muhL Safety $irig'c. Safety muki-Safety Sinate-Safel) h' ( s e h. change vt h. tharge ve h, charge ic h. thange .',T 2% 0.102 m O t19 0394 m 0 507 m 7', p !+ 7s0 0349 m 0 074 **** 04)3 m 00 6 me wherd f;9y 12s0 0 61: m 0 0$$ me 0 436 m 0 007 m L 1730 04o m 0 041 u 0 467 m 000) ou 'E j ' sign;fgant improvement mith reyect ?o similar roadway net-Chan N ' $.* p, "an im[rottment.but not nigmfgant tions thKh hast the same charac. impr;3 Q ma deterWahon, but not pgmfgant g %y after;y,7 I }2 :.e. l gl m*sigmfgant deters rathsn .E.)A', the rq)y 3

• pf Perhaps the major Gnding is the signincant deterioration in single vehicle accidents at the p'v Q'D mid.q range. When considering the A,(ADf) and A,(ADT) models in phase I, the results are a vf,

.~ D l significant improvement in the safety level of the rew roadway section for multi-vehicle poss'g';.8 ni accidents and an improvement, but not significant, for single-vehicle accidents. In essence,in ,M this example, the ignorance of q produces a situation in which the relative safety deterioration interg. W 7 at the mid g range cannot be detected. Etampft 3. An alternative means of estimating the A,.(q) and A,(q) models is use of the hour] f interrelationship between pi or p. and log a3 or log a., respectively. In fact, for any given 4, type. basid. T(q) and the number of accidents on a four larse section,one can obtain a crude estimation of PI j such models. For example, on a four lane section, the hourly flow, during specine daily hours, sing'. s / increased from 500 to 800 vehlbr for two lanes (due to either a closure of a parallel road or by each ) byg means of traffic direction).1he question is w hether the levelof safety has been changed and to w hat typei' np magnitude.The As, and A, values for q ~ 500 schlbr and this q salue are then substituted in stud; P 4 the common linear modeh indicated in Table 3. This substitution determines the appropriate l '*,h rnodels to be used for the "af ter change"(q = 800 vehlhr) data. 't hat is, d,f q) for multi vehicle ADE y [II, accidents is obtained through determination of a. and p., based on the observed value inte ) i.gj "r A, = 1.10 acc/10' veh.km: The4 addh g' 1.1 = a. 500'*

a. = 370 cidey; t.

h p% log a. = - 0.1 - 2.85p, ' p = - 0.936 flo *. ,k ultir? 1 L and similarly, 5,(q) for singleachicle accidents and 2,.(q) for multi and single vehicle folld atso i, accidents can be calculated. The complete "before and af ter" data and results are as follow:: r 7 q h e' ( t. k, o l, I .m... . ~.7,y7 = ~ ~, - .y rh h, h I$l . fl ( Relaimhip ; see, cead auiJer,ts aW bowly haic ik%l 3) 'Q ,[]' interval. It ' ADT) models Au.ie w ws. mrue Arier 1978). In the w averne e 2 o i j,,, iccidents and pr4 j" g g ,ggn t 3 q instead of the imptwe- "lh r, 0 633 h; 1; asolation g oy f. O

section (also A,,

02o om ross sectional , ; f,. g. 0 024 Iparable basis while is 0.M l l r criteri: (i) E 0 23 /!-ld} D redects the hibited by the A, Llo Li2 l.% b that of 1975 muk 6 M "M ra - 0" W D data in the }by A, 031 ,;yy. 62.53 [9{.$) A, n 40 0 25 .chkw r, -o sl3 9@ F, I 0 27 ) I f %v, /! j where A, and As, (The expected safety rneasures if the safety level rem tins the samtQ are the 4 j results of substituting q = S00 in the models. The comparison between A,, and A,, and [M! between A, and I for q = 800 rescals that: (i) sing!c vehicle accidents base atmost not shanged, though the absolute A, value afice the change decreased by 4G!; and,(ii)idatise improsement is obsersed for the multiochicle accidents though the absolute As< s alue 4 after the change increased by 6G! For such applications,it is also adsisable to determine that }, h,,y the result nt r, and ra salues are within or close to the indicated range in Table 3. h t.j ( hl I' l'1f ycidents at the $ results are a CONCLUDING REM ARKS cultiachicle 'lhis research attempts to find quantitatise models (pa*er functions) to represent the p, In essence,in possible dependency between two measures of accidents and the hourly now for eight h, y p r'eterioration interurban road sections during an 8 year period. jy;g[ A.; ;,f from these attempts to search for proper relationships betw een measures of accidents and the f.(.63 b b use of the hourly Dow,it apparent that the technical;.rocedure involves a combination of two primary $ any givea q. t)res of analy.sts: time-sequence (for each roadw ay section) and cross sectional (on a one par ) estimation of basis). For each type of anal) sis, the total accidents are primarily separated into multi-and (.7 h daily hourt, singleachicle accidents. The latter separation enables one to:(1) distinguish accident ecsts for pl*" p ([.Y P' illel road or by each ts pe of accident and for each q range;(2) find the differential effects of traffic flow on each 'M' ged and lo c hat t)pe of accident; and 0) perform a more reliable safety evaluation for "hefore and af ter" 3 substituted in studies. k.g/j ac appropriate Phases I and 11 of the oserall stuJy, described in Fig. I, are concerned with the induence of $ multia chide ADT on the rneasures of accidents. Ilowever, this consideration by itself cannot explain the V!;}. S @M3 >bsersed v:lue interactions between road accidents and trame flow, since it is only based on a daily average. i, The consideration of houily fMws prosiJes a better understanding of these interactions, in y'p;( addition, it is possible to mose one step forward in order to further unJerstand the ac. (fh. cidentiltrame fbw dependency by separating the hourly now into free Dow and congested. 00*. This separation into components of both 13 pes tif trafhe Saw and type of accident will i't$(g j$# ultimately lead toward more accurate accident prediction based on the trde now. The 9 singlea chide follo*ing paper ICeder,1982), describes this further ana!) sis (phase IV in Fig.1), and attempts $c as follocs: also to determine and compare the probabihties for each accidentiflow type cornponent M g,y ..m a w .w q Y _w 7.w s.w ,ypmwww..mw, w,4 . mum m.m.,v.% d. [ 14 Amul Cir4a and Maws Lupi w i r, e. s. I R E F L R(!NCE$ I* Baker W. T. nM 0 3nn D. W., Retarbnsh:rs of accWent r A ii+;th hourly trame solames. LLision of Research & ,hy Evdatien. New Jeney Derartmer.1 of Transrettatba,1%f. i 6.- g tkisent D. M. rfte:t of nietage speeJ arJ wlme en a, ore v>kle arcWents on too lane targents llRD troc. 32, 4 181J95,105). W} Brince W, Unf atigeschehen uW VerkehspNauf Tersc Anas Strasrenhas a9J StrassesserteArstec Aa,1,llef t 201,1976. f.vl 7 CeJet A. anJ thneh M. Fwiher eval.iatka of the relationships serleeen ro:J acciJents anJ aierage Jai!y trais. Acc. Anal. A free. 10. 95-10r,.1978. h' i Ceder A, Relatinshirs Wi*een road accWests aW lourly tran: N+: II. Probataistic steroach $sbmitteJ to AccM. 6:^ T Asel & Tree, 14,35-44. 3,1 3 Chrman R. A, Some rela'ionships between road accMeat frequenc es ad estatures of enrosure to risk. Fb D. y i i 'f d,ssertation, Unhenity College l oWos. h41971. q Gm3na D. W., Relationship of accuent tales ar4 accWer. inichem.ents mith Louity volurr.e. Trafic Quart.111,3 July I'p' 1%L k i Leutzbach W., Trame accuenti and trame kw on the FrankfurtNannheim motorosy-laiestigation of the relationship g N =ces trame accidents ned t+aie few (in German). llenisches MiaWierium fue %tschaft und Verkehr, March 1%6.

,y L4uubach W., Siegener W. srJ Wedemann R.0a the connecthe be

seen trs%c accidents and tra5e volume on a section

-j of a German autobahn. AccW. Anal. A fret. 2,14kl44 (in German),1970. l3 h Pfundt K.1hree d5culties in the comparison of accileal rate. AccM. Anst & Pree.1,25k259,1%9 Ni ,w Smeed R. J., Accident rates. lat. LaJ Sa/ct; & frsfc Res..K2),44),1955. .t,-. w b I M( '1 C; u l' g W. R a un f,$$ This p accid y:t{1,; ba5eg The t I7.,hel, Mg- [477 comp'M ae: geste dCter ...;. c; thii,;# dp aCCidyg $1% W,l', f %' f, Afti'l'.;l vestil N.. ) J theseQ Lo Of m? know (.j,jtM theory gg, neate da Q;6 molic 'i h'M rules,3 4!1h stream di c@ a pii are try 3 dire ip,;! a folti

fp thesg

' fu upon M I-h sing 1(.i .ht multifD s, =. 'd < ip( i M !i only (d j-flow ,kf mdejj u .i o.; t .s - 3 i [4 5 b j y _ _., -, m -,..,.. v,. 1 1 R$. 4 .~ LAf:. (j y _ . 3. s _ )Q y-*' p: 3g y + p ,8 Rf04 ),h g 4 ' NUCLE R REGU ORY COMMISSION ~~ WASHINGTON, D. C. 20555 6-f S ? .... 9 )a 1, N

s q,

s ( ?.i fiay-26, 1987

j 3.

a )- MEMORANDUM FOR: John Milligan Technassociates i-l Emile L. Julian eting Chief FROM: l Docketing and Service Branch

SUBJECT:

SE A BR40M, EXHIBITS s\\ _r' Any documents fi?Gd on the open record in the SEABR,oeK pro-ceeding and made a part of the official hearing record as an exhibit is considered exempt from the provisions of the United States Copyright Act, unless it was originally filed under seal with the court expressly because of copyright concerns. ^ + - All of the documents 'D,,nt to TI for processing f all within the Q j exempt cl a s sificcicitin.. ' s s h'4( e s 7; h ns-h 4, 3 . b 's 4 4e A. [ ,g 4 i ^ t; N 4 e i s* % 'N% t f p 1 s s 1, w 4 s -}}