Loading...
The URL can be used to link to this page
Your browser does not support the video tag.
Evaluation of Photo Red Light Enforcement Program
Lots - 3csLg 9 Final Report Evaluation of Photo Red Light Enforcement Program in the City of Miami Beach Prepared for: X11 , X'V., c it r • a lit Prepared by: Priyanka Alluri, Ph.D., P.E.,Assistant Professor Angela Kitali,Graduate Research Assistant Fabio Soto, Graduate Research Assistant Florida International University Dept.of Civil&Environmental Engg. 10555 West Flagler Street, EC 3628 Miami,FL 33174 FtvFLORIDA UNIVERSITY December 2018 DISCLAIMER The opinions, findings, and conclusions expressed in this publication are those of the authors and not necessarily those of the City of Miami Beach. ACKNOWLEDGMENTS This research was funded by the City of Miami Beach. The authors are grateful to Dr. Leslie Rosenfeld, Chief Learning and Development Officer, Organization Development Performance Initiatives, City of Miami Beach for her guidance and support throughout the project. A special thanks is due to the City of Miami Beach Police Department and the City of Miami Beach Transportation Department for providing the required data. The authors are also thankful to Mr. Hector Vargas and Ms. Liana Roque, undergraduate research assistants at Florida International University for assisting with the data collection. iii TABLE OF CONTENTS DISCLAIMER 't ACKNOWLEDGMENTS iii LIST OF FIGURES v LIST OF TABLES vi LIST OF ACRONYMS/ABBREVIATIONS vii I. INTRODUCTION 1 2.RESEARCH OBJECTIVE 2 3. LITERATURE REVIEW 2 4. STUDY DATA 4 4.1 Treatment and Non-treatment Intersections 4 4.2 Crash Data 6 4.3 Traffic Volume 10 4.4 Intersection Characteristics 10 5. METHODOLOGY 11 5.1 Descriptive Statistics 11 5.1.1 Washington Ave and 17th Street 11 5.1.241"St.and Prairie Ave 13 5.1.3 Alton Rd and Chase Ave 15 5.1.4 Indian Creek Dr and W 63rd Street 17 5.1.5 Indian Creek Dr and 71"Street 19 5.1.6 Washington Ave and Dade Blvd 21 5.1.7 Pine Tree Blvd and 23r6 Street 23 5.1.8 Indian Creek Dr and Abbott Ave 25 5.2 Statistical Models 27 5.2.1 Data Variables Considered 27 5.2.2 Poisson-gamma Model 29 5.2.3 Model Results and Discussion 30 6. REFLECTION 34 6.1 Target Crash Type and Target Crash Severity 34 6.2 General Crash Trend 34 6.3 Safety Performance of Non-treatment Intersections 36 7. SUMMARY 40 7.1 Simple Before-and-after Crash Data Analysis 40 7.1.1 Four-legged Intersections 40 7.1.2 Three-legged Intersections 40 7.1.3 Safety Performance of Intersections with No RLCs 41 7.2 FB Before-and-after Analysis 41 REFERENCES 42 iv LIST OF FIGURES Figure I: Map of Treatment and Non-Treatment Intersections 5 Figure 2: Aerial View of Washington Ave and 17th Si. Intersection 12 Figure 3: Overview of Target Crashes at Washington Ave and 17t St Intersection 12 Figure 4: Washington Ave and 17'h St Intersection-Overview of Target Crashes at Approaches with RLCs 12 Figure 5:Aerial View of W 41st St and Prairie Ave Intersection 14 Figure 6: Overview of Target Crashes at W 41" St and Prairie Ave Intersection 14 Figure 7: W 41st St and Prairie Ave Intersection-Overview of Target Crashes at Approach with RLC 14 Figure 8:Aerial View of Alton Rd and Chase Ave Intersection 16 Figure 9: Overview of Target Crashes at Alton Rd and Chase Ave Intersection 16 Figure 10: Alton Rd and Chase Ave Intersection-Overview of Target Crashes at Approach with RLC 16 Figure 11: Aerial View of Indian Creek Dr and W 63ta St. Intersection 18 Figure 12: Overview of Target Crashes at Indian Creek Dr and W 63ttl St. Intersection 18 Figure 13:Indian Creek Dr and W 63'd St.Intersection-Overview of Target Crashes at Approach with RLC 18 Figure 14: Aerial View of Indian Creek Dr and 71't St. Intersection 20 Figure 15: Overview of Target Crashes at Indian Creek Dr and 71"St.Intersection 20 Figure 16: Indian Creek Dr and 71" St. Intersection-Overview of Target Crashes at Approach with RLC 20 Figure 17: Aerial View of Washington Ave and Dade Blvd Intersection 22 Figure 18: Overview of Target Crashes at Washington Ave and Dade Blvd. Intersection 22 Figure 19: Washington Ave and Dade Blvd. Intersection - Overview of Target Crashes at Approach with RLC 22 Figure 20: Aerial View of Pine Tree Blvd and 23ra St.Intersection 24 Figure 21: Overview of Target Crashes at Pine Tree Blvd and 23r° St. Intersection 24 Figure 22: Pine Tree Blvd and 23ttl St. Intersection - Overview of Target Crashes at Approach with RLC 24 Figure 23: Aerial View of Indian Creek Dr and Abbott Ave Intersection 26 Figure 24: Overview of Target Crashes at Indian Creek Dr and Abbott Ave Intersection 26 Figure 25: Indian Creek Dr and Abbott Ave Intersection - Overview of Target Crashes at Approach with RLC 26 Figure 26: Percent of Fatal and Severe Injury Crashes by Target Crash Type 34 Figure 27: Annual Crash Trend 35 Figure 28: Target Crashes by Crash Type at Comparison Intersections 37 Figure 29: Target Crashes at Intersections Far Away from the Treatment Sites 39 v LIST OF TABLES Table 1: Summary of Recent Literature on the Safety Impacts of RLCs 3 Table 2:Treatment and Non-treatment Intersections 6 Table 3: Descriptive Statistics of Crashes at Treatment and Non-treatment Intersections 9 Table 4: Descriptive Statistics of the Data Used in the full Bayes Models 29 Table 5:Model Results for Different Target Crash Types 32 Table 6: Model Results for Different Target Crash Severities 33 vi LIST OF ACRONYMS/ABBREVIATIONS AADT Annual Average Daily Traffic BCI Bayesian Credible Interval DHSMV Department of Highway Safety and Motor Vehicles (Florida) EB East Bound FB Full Bayes FDOT Florida Department of Transportation NB North Bound PDO Property Damage Only PRL Photo Red Light RLC Red Light Camera RLR Red-light Running RTM Regression-to-the-mean SB South Bound WB West Bound vii EVALUATION OF PHOTO RED LIGHT ENFORCEMENT PROGRAM This report summarizes the key findings of the safety evaluation of the Photo Red Light (PRL) Enforcement Program in the City of Miami Beach, Florida. The report is structured as follows. • Section 1 presents a brief introduction on the usage of red light cameras(RLCs)to improve safety at signalized intersections. • Section 2 presents the study objective. • Section 3 shows the results of the most recent studies on the safety effectiveness of RLCs. • Section 4 discusses the data used in this study and the data collection efforts. • Section 5 presents the methodology adopted to evaluate the safety performance of the PRL enforcement program in the City of Miami Beach. • Section 6 reflects on the main findings from the study. • Section 7 summarizes this research effort. 1. INTRODUCTION Intersection-related crashes represent approximately 40% of all crashes (Decina et al., 2007). As such,intersection safety is a serious problem in the United States.Many of the crashes at signalized intersections can be attributed to red-light running (RLR), which "involves a driver entering an intersection after the traffic signal has turned red" (City of Fort Lauderdale, n.d.). The following crash types are commonly attributed to RLR: angle, left-turn, right-turn, and head-on crashes. These crashes are often severe and result in fatalities.For example,in 2015,on average,two people are killed every day due to RLR. From 2011 to 2015, 719 people died each year on an average from RLR crashes. In 2014, about 126,000 people were injured in RLR crashes. The estimated economic losses exceed $4 Billion annually(ATSOL,2018). Interestingly, in RLR crashes, it was not the drivers who run red lights who sustain fatal injuries, but the occupants of other vehicles, pedestrians,and bicyclists who were hit by drivers who nm red lights(IIHS-HLDI,2016). In 2010, 61 people were killed in RLR crashes in Florida, making it the third deadliest state in the nation for RLR crashes (City of Miami Springs, n.d.). Over the last decade, photo red light (PRL) cameras have been increasingly deployed to reduce the occurrence of RLR crashes. This automated enforcement method is used to discourage red light runners and decrease intersection crashes. Red light cameras (RLCs)are automated systems that photograph vehicles entering intersections after the traffic signals have turned red. The photographic evidence captured by the cameras allow camera operators to determine whether or not a ticket should be issued to the violating vehicle. In Florida, RLR tickets usually cost $158, but drivers could pay up to$262 if they fail to pay for the offence after the first notification(Florida Online Traffic School,2018). 1 2. RESEARCH OBJECTIVE The main objective of this study was to evaluate the safety effectiveness of the PRL Enforcement Program in the City of Miami Beach, Florida. The safety effectiveness of the RLCs was measured using the simple before-and-after analysis and the full Bayes before-and-after evaluation approach. The full Bayes method was implemented to take into account the regression-to-the-mean (RTM) phenomenon, which is observed due to the natural variability of crash data, to also consider the changes in traffic volume, geometric characteristics and driver behavior over time. 3. LITERATURE REVIEW Table I summarizes recent studies that focused on the safety effectiveness of RLCs. The literature review revealed that most studies have compared data for the "before installation" and "after installation" periods for treatment and non-treatment intersections to quantify the safety effectiveness of RLCs. While most studies determined that signalized intersections equipped with RLCs experienced a considerably large reduction in RLR crash types and injuries, some studies determined that RLCs contribute only to a small reduction in crashes. Additionally, these studies found that RLCs have some negative safety impacts,i.e.,they tend to increase specific crash types, such as rear-end and sideswipe collisions, by both number and severity. Understanding the methodology and findings of previous studies allowed the research team to adopt a methodology to overcome the past flaws and difficulties, such as, the RTM effect,and identify the right type of crashes to be evaluated. 2 b z F" 0w - 6; vi o A bo y o ac0- x 0 ., OU U NO M O •3 N N N O O 0 N N O I I I I I I 00 '0 00 O S V 7N N ON C\ �N 2 0 0 a O l0 t. �n 00 y. C.) A4j R N 3 ��, O y. U 1+ cg o coo °=' 0 4. o c 3 3 O y' °) o ti try c� .� a caa n i° ai .o .� as a. �. O 0 a N c0 a 0 a c) a 4) •- � c y a. s c `° D a 4. c, `° cCa co. `al o < Dao a") pa Cq o ca °) O ��]] O W 0 w 011 U OQ �v 13 U U 4� W y tx b O Egr v; 0 C vi 0 �i o „o a ..; 'LJ o V) -cs• v, y yr. a004) bdaoA a ..so U N •Ci c 74y .7 • ag o Vf S . 2 c0 •a 0U C 'zt 3 0 N .... O • v M C .� c0N. u "' o •E c`") ,. b o ai w ai 3 2 .D a� 'd :8 ;ti o m 10. .. °o cd c = _y y V y b 0 c0 o y V1 1 3 0 y o f > es E a e0 2 „, yy , a bo C • o ro • •� V N it ,00 0 '..., V R>t� O U ti W >> V .g o U O 0 b y u,., i. ° •G •y a0 •0 ccf 6)i 0 tv °� 0 ` 1 •V :1^~• y ,p aN E 2 !y a o a .= N .= y U V •.+ v > NV V y 4- 0 p C +a y, y. c) d \ . ey w .z 4) ,8 p �- 1 . •5 , ; 4 ceI11e e � y0ua a. > ed c a fi . '3- NOE N � a 0 - Q 74 Q i - P. I. -, . U ¢ w R isQ OG 03 co VZ U v) tv .r y,. 7-., • • • • • • • • •• •• •• •• •• •• o O '.) f+ , N ca L00 O 't ' M 'n V '-' N as .ai d• v; N k• N A N y N y M y V w o Cl) y.. o o .0 of 0 0 W Z 7 - Ag•a.a ' � ¢ w ` = g asE of DEQ E as E ° ` E� UZ E U E= U Et. E- U E- UZ 4) • • • • • • • • • • • • c0 y E _ -- v �_E _ 7 v N G c' N q ` 0 Q OO p N 0. od vA~ r i .b , C Q ^ N+ y ^ 144 `' p c0 .t .0 c0 0 • Mt z ,a U � QQ v; 3 � a0 0 z 4. STUDY DATA The study area included a total of 28 signalized intersections in the City of Miami Beach, Florida. The intersections were divided into two categories: • signalized intersections with RLCs,termed as"treatment intersections"; and • signalized intersections without RLCs,termed as"non-treatment intersections". As the name implies,treatment intersections are locations where RLCs were installed.On the other hand,non-treatment intersections,also known as comparison intersections,are those intersections that have similar traffic volume,roadway geometrics,and other site characteristics as the treatment intersections,but without the RLCs. These sites were manually selected after an extensive review of the City of Miami Beach's roadway network. Figures 1(a)and 1(b)show the map of the treatment and non-treatment intersections,respectively. A total of nine signalized intersections were identified as"treatment intersections". Of these nine, six are four-legged and the remaining three are three-legged intersections.A total of 19 signalized intersections were identified as "non-treatment intersections", 14 of which are four-legged, and the remaining five are three-legged signalized intersections. The following data were included in the analysis: crash data, traffic volume, roadway geometric characteristics, and traffic control features. Note that these data were required for both treatment and non-treatment intersections. The following subsections discuss these data in detail. 4.1 Treatment and Non-treatment Intersections A total of ten red-light cameras are operational at nine signalized intersections in the City of Miami Beach,as listed in Table 2 and shown in Figure 1(a). For all the intersections,Table 2 includes the number of intersection legs(i.e.,three-legged or four-legged),and the major road and minor road names. For the treatment intersections,the table provides information on the RLC installation date and the approach with the RLC. Due to the long period of construction activity, the intersection where Alton Road meets 17th Street was not included in the study. Temporary traffic controls, such as concrete barriers and traffic cones, tend to affect traffic patterns, traffic volumes, and the occurrence of crashes. Therefore, this intersection is expected to have an unusual crash frequency that cannot be associated with the operation of RLCs. The rest of the intersections(i.e., both treatment and non- treatment intersections) were manually reviewed to identify any major construction activity that might have taken place during the study period.However,no significant construction activity was found at the rest of the intersections. The review was also done to make sure that there were no other countermeasures during the study period other than the RLCs. 4 NORMANDY P6�rat SHORES Pa �c':;/ up °ATLANTIC 'enc ° % NORMANDY HEIGHTS �9.o S' ISLES nln OO co 1, �LA GORCE m Ali #W Or° aA �io0 %ry Qn j <e� ��o 0� oc 'A r- c �, D 01y 47th, ... LA GORGE I NAUTILUS 5 c a '3 _ ' a U 0a� a t alsis, MI�BEACH �e Q- n_m — f:ii: ilifji, y�a� ~was = sr fen - 0 sl MEM iii Tsem NAUTILUS. " c Nit i U ll 'il1] ' .' lip MID BEACH EtLE li : { I 4O A H m. Z7 die isl,'--3 ;.- v.•..i,iiaci-4 froQa �+ 1'k :frit aro 9 ifs16 ` ,��° >♦\tt �' r CITY CEOR �� 9C Ole ar.aay Q 7. 41D ` i 4.1." a Miami Beach �.", BAYSHORE o ,snsr II .--T _ ° O*. WEST AVENUE aCI 1' J,ae� co SOUTH BEACH ACI/0 FLAMINGO / . 0 CITY CATER LUMMUS 4- a Miami Beach: R n fID ,.,t 5th St Q Man data t701R Gnoala Terms 7 CCO n i , Map data©2018 Google Terms 2.000 ft, 1 (a)Treatment Intersections (b) Non-treatment Intersections Figure 1: Map of Treatment and Non-Treatment Intersections 5 Table 2: Treatment and Non-treatment Intersections Intersectl(N1•l• hc Item \lajor Read \-liner head Date of Operation ,Ahhrodch VA ith RI.(' 1 Washington Ave 17th St. 4/1/2010 SB 2 Washington Ave 17th St. 4/1/2010 EB ami 3* Alton Road 17th St. 9/1/2015 WB a4 co 4 4151 St. Prairie Ave 4/1/2010 NB E . v 5 Alton Road Chase Ave 4/1/2010 NB 174 °' 6 Indian Creek Dr W 63`d St. 4/1/2010 SB F- 7 71st St. Indian Creek Dr 4/1/2010 NB u 8 Washington Ave Dade Blvd. 4/1/2010 EB oa at) 9 Pinetree Blvd 23`d St. 4/1/2010 WB o r, 10 Indian Creek Dr Abbott Ave 4/1/2010 SB 11 Dade Blvd Prairie Ave 12 5th St. Alton Road 13 Arthur Godfrey Rd. Meridian Ave 14 Alton Road I6'h St. 15 Alton Road 11th St. 16 Alton Road 8`h St. oa 17 17th St. James Ave °= 18 Alton Road W 47th St. 0 4 '3 19 Washington Ave 16'h St. B 20 4151 St. Indian Creek Dr Not Applicable Not Applicable o c 21 5`h St. Collins Ave Z ~ 22 Collins Ave 16th St. 23 Collins Ave 23'St. 24 63'St. Pine Tree Dr 25 West Ave 17th St. ao 26 Pine Tree Dr Sheridan Ave ;; 27 Washington 15th St. r, 28 Pine Tree Dr W 47th St. 29 West Avenue 11t St. *This location experienced major construction during the study period.Hence,it is not included in the analysis. 4.2 Crash Data The analysis was based on five years of crash data. Since the RLCs were installed in April 2010, crash data from two years before the RLC installation (i.e., 2008-2009), and three years after the RLC installation (i.e., 2011-2013) were included in the analysis. Signal Four Analytics, a statewide interactive web-based geospatial crash analytical tool developed by and hosted at the University of Florida, GeoPlan Center, was used to identify and extract crash data. All crashes that occurred within 300 ft from the center of the intersection were identified, and the police crash reports of all these crashes were downloaded and reviewed. As is evident from the literature, RLCs impact only specific (and not all) crash types, commonly known as target crashes. The following crash types were considered to be associated with RLR and the operation of RLCs. 6 • Angle(side impact) • Right-turn(vehicle turning) • • Left-turn(vehicle turning) • Rear-end • Sideswipe The police reports of all crashes that occurred within 300 ft of the intersections were manually reviewed to identify intersection-related crashes and target crashes. The number of crashes extracted from Signal Four Analytics was 2,419. Of this total, only 1,080 target crashes were included in the analysis. A total of 389 target crashes were found to have occurred at treatment sites, and 691 target crashes were found to have occurred at non-treatment sites, during both the before and the after periods. Table 3 provides a summary of all observed crash frequencies for treatment and non-treatment intersections. In addition to crash type, crash severity was also considered in the analysis. Crashes were categorized into property damage only (PDO), and fatal/injury crashes.Note that all injury severity levels(i.e.,incapacitating,non-incapacitating,and possible injury)were grouped in the fatal/injury crash category. The following information was collected while reviewing the police reports: • The approach the crash occurred o Major approach o Minor approach o Middle of the intersection o Not sure • The crash occurred on the approach with a red light camera? (applicable only for treatment intersections) o Yes o No o Not sure • Manner of collision o Rear-end o Angle/left-turn/right-turn o Head-on/front-to-front o Sideswipe o Backed into o Fixed object o Other o Not sure • 1s`vehicle maneuver action o Going straight o Making a left-turn o Making a U-turn o Making a right-turn 7 o Not sure o Other • 2"d vehicle maneuver action o Going straight o Making a left-turn o Making a U-turn o Making a right-turn o Not sure o Other • Crash involving driveway/on-street parking o Yes o No o Not sure • Crash involving a pedestrian/bicyclists o Yes o No o Not sure • Crash involving a distracted driver o Yes o No o Not sure To evaluate the direct impact of RLCs on crash frequency and severity, target crashes were excluded when one of the following conditions were present: • Adverse weather condition • Distraction • Driving under the influence(DUI) • Crashes related to emergency vehicles responding in emergency circumstances • Sickness • Sleep deprivation/fatigue Table 3 provides the descriptive statistics of crashes at each of the treatment and non-treatment intersections. 8 O O l- O l� O l- M M M l- M 0 0 0 0 [-- O O 1, M O M O l-- Vl t` r r vi 00 00 N TiV en 1 00 O n - vi -- 0\ v1 V) O b Ti - Q` vl \O Ol V1 0l ei Ti • v6 Tr U _. -- Vl - -- N .- -- .4,1N -- N N N - - N -- 4: U �s • U fj M M - 4O r- et M Tr N No M r- N V n vo N N 01 00 M V N - M I� vo so M y A= r4 n - n N M l- V1 0i [� en M I- M M Vl N --� -- -+ oa G - N en - � 0 -1 T `V M M O O M O M O M M [- O M M M l� O O 1� M M I- M O M M M M O 0 .� = - - N - -- M - O O O vO - -- N - O O - O O O - O O O O O N ' 9 = P .Eo vl O J U • +i - 0 .'' n' _ ' M O b 0.4 V1 V •O M - N N M - 00 Vl N M N 10 5 M N TF - fV - - - .-. Ot. CD w V) Q - M M n O O r- M n M M M f` O O r- O O M M 1- M co co r- M M M O r- , C I..4. _., - N N oo O - N M O - - O C. O M - M N O M M N N N 0 0 0 0 0 0 E ~ l- S l- 1- n n M co co co. co O O M O Mr- O M O l- l- l- O I-- r- O M M 0 er - - Ti Ti Ti M 00 - O - -- NO vl M M vi O --. M Ti O O - co - O - - - O•. M :2 - - N Eeci v = M M l� M O O t� t� l� M O O n M O M 0 [- M M M n O M M l- M M coO g o = - M en (-: - O - N N O O N N O M en vi - O - 0 0 0 N O 0 0 0 0 - 6 7, 0 -1' _ - 6) .6 L G F - z _ C :3 t` N 0` vo un 41 O N M 4O 41 Vl 41 C. VI Vl 41 Vl 7 Vl Vl O Vl 41 vl h Vl t �; N N d' N - 01 M - -" vi - 00 vi h ' N N v6 Ol N O 06 O V1 - N Vl - N N M N - - Cl U cl Cl L.) .b C 4. - j V1 V1 vl h vl O Ol Vl et V1 V- vl Vl Vl (- Cl vl v1 vl l- l- V• v1 N V7 VI 41 vl a VU Q c; i,= -• -- N 7 Vi - V 0\ '- - M N 'C M h - N - cY _ E >. G J M M vl V1 vl - v1 v1 O - V1 l0 - - V1 M v1 O vl - V1 - vl O - O V1 V1 v1 U F4 0 0 N6 O O -- N O O O O O ,J C. O J C R r"" .0 " i„, ..I.'. -ij .0 U H O. -' ^ N N 10 V 41 Cl Vl 41 et <t 41 h h Vl 10 N V1 d- vl M M N vl VI N - 41 7 'DCC Z y0 - C 3 • CC :� _ _ .4.1 V1 00 h 1/l kr, V1 V) V1 _. 00 -. V1 N 41 O 41 r- N 41 N - 41 h O O O 41 C .^ UO • M - - 00 O - O - O •C = -0VI41 C1 h M h N -- ken l v1 v1 l- et M -vl v1 oo vl N N -- -- v1 41 Vl --. 41 td `C' COf. T 6e y - -' en - N M '0 -- 'ct N - M 0 0 �r % . H t V •- CL 00 tN l WI 7 v1 v1 v) Vl VI O h V1 V1 O - vl O G+r -c41 M O -- M - - - - R N O - O M -- O - O .7>:i' C_ er U aa - UC U • •>.. n~- n~ .C•N Ir i. c R > > v� 0 b > > c Q > • °_' > O Q o CJ cOG 6,1 05' � a � cmcin � °= Qs • UQ cw A ig xa •o -� 4 � -�-o 0 '2 sc •@ o d .o s x E 't :5 .- . 10 a� � d g v .r ao Oil N. ^ = 0. � 3Sc4N ¢ a ¢ '� � 00ti3 � Sc.d � Nri � v = 3 - G ca a) .b 0 6) O F el QE Ca d > A ) d d ' 2 0 OGQ , • -0 'CI b b e U 6cC A.q O 0 0 a 0 Co000QQ 0 0 L 0 O V4 8 ' > 0 0 0 Vo 6) y c° y 6) 0 Q .' Ia A Of0 ec0; - 0U • c .. 0 a0: n; a: ' 0; c - ti H • Q I- sQ c a) y c c g .o Fg c c bcl) c 's cA .. .G F F •CH ., O ; s o c o s o y .9. 'C � 0vUtico 0 o -0N 0 C b .= ) 4 C �33 ¢ v ¢ , � 3a 4in41 : .. R = ¢ 37 .�1US' 3a36. 3 . . N E - N * „ vl �o ,, 00 0` O -- N M -- vl .o 5 00 01 O N M - Vl O n 00 C. 0 0 �'"• Cl N N N N N N N N N •- 0 CI iii paSSal q pa�3a! pa2231-q pagOaI-£ 11 juaugeau, 1uatu eatl-uoN . o * z 4.3 Traffic Volume Traffic volume data is included in traffic safety models because it is proven to be the main contributor to what is called crash exposure, i.e., as traffic volume increases there is a higher likelihood for crashes to occur. AADT were collected from the Florida Department of Transportation (FDOT) Traffic Online website, a web-based mapping application that provides traffic count site locations and historical traffic count data. It is managed by the FDOT Transportation Data and Analytics Office. AADT data were collected for every year of analysis, 2008, 2009, 2011, 2012, and 2013. However, it is important to note that traffic counts are not available for all years and all roads due to high data collection costs. As such, the research team made reasonable assumptions to estimate missing traffic counts.For the missing data,AADT was obtained from parallel roads with similar roadway geometric characteristics,and AADT for the missing years was extrapolated assuming that traffic volume increased by 3%each year. 4.4 Intersection Characteristics Roadway geometric characteristics and traffic control features often influence the occurrence and severity of crashes at signalized intersections. Including these data in the analysis helps understand the relationship, if any,between the RLR behavior and the characteristics of the intersections.The following roadway geometric characteristics and traffic control features were collected for each of the study intersection: • Roadway Geometric Characteristics o Number of through lanes o Approaches with left-turn lanes o Approaches with right-turn lanes o Length of the sidewalk • Traffic Control Features o Yellow and all-red times o Presence of pedestrian phase o Type of left-turn signal phasing(i.e.,protected,protected-permitted,or permitted) o Right-turn on red restriction 10 5. METHODOLOGY 5.1 Descriptive Statistics The following 9 signalized intersections were installed with RLCs in the City of Miami Beach: 1. Washington Ave and 17t St. 2. Alton Road and 17th St. 3. 41'St. and Prairie Ave 4. Alton Road and Chase Ave 5. Indian Creek Dr and W 63'd St. 6. Indian Creek Dr and 71st St. 7. Washington Ave and Dade Blvd. 8. Pinetree Blvd and 23`'St. 9. Indian Creek Dr and Abbott Ave Note that the intersection Alton Road and 17t St. experienced major construction during the study period, and hence excluded from the analysis. The following subsection discuss the safety performance of each of the remaining eight intersections. 5.1.1 Washington Ave and 17t Street This intersection is a four-legged signalized intersection. Figure 2 shows the aerial view of this intersection. The RLCs on the SB and EB approaches were installed on April 1, 2010. Figure 3 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 4 shows the number of target crashes and total crashes at the approaches with RLCs. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 108 crashes during the study period (i.e., 2008- 2009 and 2011-2013). The before-period had 27 crashes per year, and the after-period had 18 crashes per year. • Overall, there was a reduction in target crashes after the installation of the RLCs at this intersection. • After the RLC installation,both the PDO and fatal/injury crashes decreased by half. • Both total crashes and total target crashes decreased at the approaches with RLCs. Angle/left-turn/right-turn crashes decreased, rear-end crashes increased, while sideswipe crashes didn't change after the installation of RLCs. 11 A. it`, r RED LGM • PNWO • -n ENFORCED t6111161W ityf. 1 . ,.:.._ 4, .., 9. RED -�) I1- I I 1 RN0T0 iNfORCFC �i� -N L -- Wit �•./1 I "K ^. a I '', I , -_ , ' •4 ; i Figure 2: Aerial View of Washington Ave and 17th St. Intersection 16 15 14 12 14 12 12 10 10 9 87'3 8 6.0 6 612.3 4 i I 1.51.7 1.3 0 . I 0 Angle Rear-end Sideswipe Total Target PDO Fatal/Injury ■Before After Crashes ■Before After (a) Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 3: Overview of Target Crashes at Washington Ave and 17th St Intersection 8 8 8 7 7 6 15.7 6 5 5 4 3.5 3 4 3.5 3 2 2 I 3 2.7 2 1 1 0.3 0.50.7 2 0 ■ I L. 1 0 0.3 Angle Rear-end Sideswipe Total Total 0 Target Crashes PDO Fatal/Injury is Before After Crashes ■Before After (a) Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 4: Washington Ave and 17th St Intersection - Overview of Target Crashes at Approaches with RLCs 12 5.1.2 415'St. and Prairie Ave This intersection is a four-legged signalized intersection. Figure 5 shows the aerial view of this intersection. The RLC on the NB approach was installed on April 1, 2010. Figure 6 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 7 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 81 crashes during the study period(i.e.,2008-2009 and 2011-2013). The before-period had 21 crashes per year, and the after-period had 13 crashes per year. • Overall,the number of target crashes at this intersection increased after installing the RLC. Angle/left-turn/right-turn and rear-end crashes increased after the RLC installation. No sideswipe crashes were reported after the RLC installation. • Even though there was an increase in target crashes, the total crashes at the intersection decreased. • At the approach with the RLC,the before-period did not have any target crashes,however, the approach experienced three target crashes after the RLC installation. 13 ..•.,t:.= ' #. A.. . ( ice - .;,\ 'ataw j l �4 w c F [Mk ii RIB ,_�^z�_'c _a s V 4 I"St -: I 1 I it: - �, �i C. ^F � 1 ► lam• 0 IFC•:t ! r,. LLB, �, i r O:CFMPoNCfC • � � I L.?,I' - — In Figure 5: Aerial View of W 41"St and Prairie Ave Intersection 6 5 6 5 5 4.7 4.5 4 4 4 3.5 ` 3 3 2 2 1.3 1 1 JO ; 1 0 0 _ Angle Rear-end Sideswipe Total Target PDO Fatal/Injury Crashes •Before After ■Before After (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 6: Overview of Target Crashes at W 41"St and Prairie Ave Intersection 2 2 2 1.5 1.3 1.5 1 1 I 1 _ 1 0.5 0.3 0.5 0.3 0 0 0 0 0 i 0 0 0 0 Angle Rear-end Sideswipe Total Target Total PDO Fatal/Injury Crashes Crashes •Before After ■Before After (a) Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 7: W 41"St and Prairie Ave Intersection -Overview of Target Crashes at Approach with RLC 14 5.1.3 Alton Rd and Chase Ave This intersection is a four-legged signalized intersection. Figure 8 shows the aerial view of this intersection. The RLC on the NB approach was installed on April 1, 2010. Figure 9 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 10 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 76 crashes during the study period(i.e.,2008-2009 and 2011-2013). The before-period had 16 crashes per year, and the after-period had 14.7 crashes per year. • The results show that while angle/left-turn/right-tum and sideswipe crashes decreased, rear-end crashes increased after the RLC installation. • Overall,this intersection experienced similar crash trend after the RLC installation. • The approach with RLC did not experience any angle/left-tum/right-turn or sideswipe crashes during the study period. However, multiple rear-end crashes were reported. Rear- end crashes doubled after the RLC installation. 15 - • — v. . . . . . ,... . _ . - ;lifi , , i ... „,. ._ . . ___ . , . .„..-„. ,- , / . al : " �4 i - Tir. :ti IL:. ' DNI PHOTO • ENFORCED Figure 8: Aerial View of Alton Rd and Chase Ave Intersection 6 5.5 5.7 6 5 4.7 5 5 4.3 4 4 3 3 3 2 1.5 ii 1 1 2IIII 1.3 0 I 0 I 1 0.5 hi .. Angle Rear-end Sideswipe Total Target 0 • Crashes PDO Fatal/Injury ■Before After ■Before After (a) Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 9: Overview of Target Crashes at Alton Rd and Chase Ave Intersection 6 5.67 6 5 5 4 4 4 3 3 3 3 2 1.5 1.5 — 2 2 1 1 1 00 I 0 01 0.50 0 0 Il Angle Rear-end Sideswipe Total Total PDO Fatal/Injury ■Before After Target Crashes ■Before After Crashes (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 10: Alton Rd and Chase Ave Intersection - Overview of Target Crashes at Approach with RLC 16 5.1.4 Indian Creek Dr and W 63rd Street This intersection is a four-legged signalized intersection. Figure 11 shows the aerial view of this intersection. The RLC on the SB approach was installed on April 1, 2010. Figure 12 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 13 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 114 crashes during the study period (i.e., 2008- 2009 and 2011-2013). The before-period had 19.5 crashes per year, and the after-period had 25 crashes per year. • This intersection experienced a reduction in all target crashes (i.e., angle/left-turn/right- turn,rear-end, and sideswipe crashes)after the RLC installation. • There was a reduction in PDO target crashes; while the target crashes resulting in injuries experienced similar crash trend after the RLC installation. • The approach with RLC experienced a positive effect; angle/left-turn/right-turn, rear-end, and sideswipe crashes decreased after the RLC installation. • The approach with RLC experienced a reduction in PDO crashes. No fatal/injury crashes were reported on this approach during the study period. 17 r� I OMI r-:''s.y..4. P PINor:; • / / : EMIORCEC I x �J. • v"� ..,� ' H611 4,*• i_ 1 Figure 11: Aerial View of Indian Creek Dr and W 63rd St. Intersection 10 10 9 10 8 I 7.3 8 6 6 5 • 6 • 4 3.7n ; > 4 2 1.5 I 0 ■ 2 1 1 Mi il Angle Rear-end Sideswipe Total Target 0 Crashes PDO Fatal/Injury •Before After •Before After (a) Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 12: Overview of Target Crashes at Indian Creek Dr and W 63rd St. Intersection 7 6.5 6.7 7 6 6 5 5 4 4 4 3 2 3 2 2 I jO•71,0.3 0 1. 11 1 U 0 0 Angle Rear-end Sideswipe Total Total PDO Fatal/Injury Target Crashes ■Before After •Before After Crashes (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 13: Indian Creek Dr and W 63rd St. Intersection -Overview of Target Crashes at Approach with RLC 18 5.1.5 Indian Creek Dr and 71'Street This intersection is a four-legged signalized intersection. Figure 14 shows the aerial view of this intersection. The RLC on the NB approach was installed on April 1, 2010. Figure 15 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 16 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 103 crashes during the study period (i.e., 2008- 2009 and 2011-2013). The before-period had 23.5 crashes per year, and the after-period had 18.7 crashes per year. • This intersection experienced a reduction in all target crashes (i.e., angle/left-turn/right- turn,rear-end,and sideswipe crashes)after the RLC installation. • There was a reduction in PDO target crashes,but target crashes resulting in injuries slightly increased. • The approach with RLC experienced a positive effect; angle/left-turn/right-turn, rear-end, and sideswipe crashes decreased after the RLC installation. The total number of crashes also decreased at this approach. • While PDO target crashes decreased by more than half, fatal/injury target crashes increased. It is important to note that no fatal/injury target crashes were reported in the before-period. 19 • '. •i1rW tet` —;_ :le— I• ra 4 Slav[ , 411i -�K r .,i _;,er•.. 1—L \ .. :ucxr '� .. -.g ` • PHOTO ' `,, - -. - E F I Figure 14: Aerial View of Indian Creek Dr and 71St St. Intersection 20 19 20 16.5 15 14.3 15 11.5 11.0 10 b 10 54 3-3 5' 2-5 0 - Angle Rear-end Sideswipe Total Target pjyj Fatal/Injury ■Before After Crashes •Before After (a) Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 15: Overview of Target Crashes at Indian Creek Dr and 71" St. Intersection 3 3 3 3 2.5 2.5 2.5 2 2 1.5 1.51.3 1.5 1.3 I 10.7 1 0.5 0.50.3 0.50 10.0 ' 1 - I 0.5 0 0.3 Angle Rear-end Sideswipe Total Total 0 Target Crashes PDO Fatal/Injury ■Before After Crashes •Before After (a) Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 16: Indian Creek Dr and 71"St. Intersection -Overview of Target Crashes at Approach with RLC 20 5.1.6 Washington Ave and Dade Blvd This intersection is a three-legged signalized intersection. Figure 17 shows the aerial view of this intersection. The RLC on the EB approach was installed on April 01, 2010. Figure 18 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 19 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 51 crashes during the study period (i.e.,2008-2009 and 2011-2013). The before-period had 10 crashes per year, and the after-period had 10.3 crashes per year. • Overall,target crashes slightly decreased at this intersection after the RLC installation. • After the RLC installation, this intersection experienced an increase in angle/left- turn/right-turn crashes but a decrease in rear-end and sideswipe crashes. • PDO target crashes decreased while fatal/injury target crashes increased. • It is important to note that the approach with RLC did not experience any angle/left- turn/right-turn or rear-end crashes.There were only two target crashes(sideswipe)reported during the before-period. • At the approach with RLC,no target crashes were reported after the RLC installation. Total crashes decreased at this approach. • The approach with RLC experienced a reduction in PDO crashes after the RLC installation. It is important to note that there were no observed fatal/injury crashes during the study period. 21 . .„Ow'. ,: _ _WM 11.74141•VOIV , pipp/ . :4,11 ' - . t'-_,1-10 11 I j ,,,-; °;,. ,•-' �i .F111111ilV1 REQ %% t IT . r =) - 0 LGRI ,1 • (r. I /� 'y- y- PHOT: / O ( i'� l . FRFOTRCPC 0. 1 ••• P. � C. , f I ` y ''civ; Figure 17: Aerial View of Washington Ave and Dade Blvd Intersection s 5 5.0 4.s 4.3 q 4.0 3.3 3 2.7 3.0 2 2 1.5 1.5 2.0 1 1 1 0.7 II 1.0 0.5 0 0.0 - Angle Rear-end Sideswipe Total Target PDO Fatal/Injury •Before After Crashes •Before After (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 18: Overview of Target Crashes at Washington Ave and Dade Blvd. Intersection 3 3 3 2.5 2.5 2 2 1.5 1.5 1 1 I 1 1 I 0.5 0 0 0 0 0 0 0.5 0 I 0 0 0 Angle Rear-end Sideswipe Total Total 0 Target Crashes PDO Fatal/Injury ■Before After Crashes ■Before ■After (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 19: Washington Ave and Dade Blvd. Intersection -Overview of Target Crashes at Approach with RLC 22 5.1.7 Pine Tree Blvd and 23'd Street This intersection is a three-legged signalized intersection. Figure 20 shows the aerial view of this intersection. The RLC on the WB approach was installed on April 01,2010. Figure 21 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 22 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 46 crashes during the study period(i.e.,2008-2009 and 2011-2013). The before-period had 12 crashes per year, and the after-period had 7.3 crashes per year. • Overall,target crashes decreased by half after the RLC installation. • This intersection experienced a reduction in angle/left-turn/right-turn, rear-end, and sideswipe crashes after the RLC installation. • After the RLC installation,there was a reduction in PDO target crashes at the intersection, but there was a slight increase in fatal/injury target crashes. • It is important to note that the approach with RLC did not experience any angle/left- turn/right-turn or rear-end crashes during the study period. Similarly, there were no observed fatal/injury crashes during the study period. • PDO target crashes decreased at the approach with RLC after its installation. 23 ',r-'s , %' .: /\• Ric ri- .it-.,, P S'�/ 0,:„,;'! i N( f-:, ' j " t ';''.. 4 \ .,; - • A, • 4'...2 s,', , . 1. 1 .0 y i`' 40" - Z C' Figure 20: Aerial View of Pine Tree Blvd and 23rd St. Intersection 4 4 4 4 3 3 2 2 1.5 1.5 1.3 2 1.7 I r 1 1 0.7 - 1 0 1 0 0 0.3 Angle Rear-end Sideswipe Total Target 0 Crashes PDO Fatal/Injury ■Before After ■Before After (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 21: Overview of Target Crashes at Pine Tree Blvd and 23rd St. Intersection 1.0 1.0 1.0 0.5 0.5 0.5 0.5 0.5 0.3 0.00.0 0.00.0 0.0 0.0 I 0.0 0.0 0.0 0.0 0.0 Angle Rear-end Sideswipe Total Total PDO Fatal/Injury Target Crashes •Before After Crashes •Before r After (a)Target Crashes by Crash Type (b)Target Crashes by Crash Severity Figure 22: Pine Tree Blvd and 23rd St. Intersection -Overview of Target Crashes at Approach with RLC 24 5.1.8 Indian Creek Dr and Abbott Ave This intersection is a three-legged signalized intersection. Figure 23 shows the aerial view of this intersection. The RLC on the SB approach was installed on April 1, 2010. Figure 24 shows the number of target crashes by crash type and crash severity at this intersection before and after the RLC installation. Figure 25 shows the number of target crashes and total crashes at the approach with RLC. Note that target crashes include angle/left-turn/right-turn, rear-end, and sideswipe crashes. Some of the key findings include: • This intersection experienced a total of 60 crashes during the study period(i.e.,2008-2009 and 2011-2013). The before-period had 13 crashes per year, and the after-period had 11.3 crashes per year. • This intersection experienced a reduction in all target crashes (i.e., angle/left-turn/right- turn,rear-end, and sideswipe crashes)after the RLC installation. • Similarly,both PDO and fatal/injury target crashes decreased after the RLC installation. • The approach with RLC did not experience any angle/left-turn/right-turn, rear-end or sideswipe crashes after the RLC installation.There were no observed target crashes on this approach after the RLC installation. • Even though the approach with RLC didn't experience any target crashes after the RLC installation, the total number of crashes at the approach with RLC increased slightly. 25 ►E0 h lL ` i L. PMOiO � MI FRTOE OC ' �/. • . terliaill . - Figure 23: Aerial View of Indian Creek Dr and Abbott Ave Intersection 5 5 4 4 4 3 2.7 I 2.3 I2 1.5 1.3 I I o -I ll -0.3 1 MI Angle Rear-end Sideswipe Total Target 0 Crashes PDO Fatal/Injury ■Before After ■Before After (a)Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 24: Overview of Target Crashes at Indian Creek Dr and Abbott Ave Intersection 2 2 1.7 1.6 1.5 1.6 1.2 1.- 1.2 0.8 - 0.8 0.5 0.5 0.5 0.5 11 00 I 0 I 0 0 0 0 0.4 . 0 II o 0 Angle Rear-end Sideswipe Total Total PDO Fatal/Injury Target Crashes •Before After Crashes ■Before •After (a)Target Crashes by Crash Type (b) Target Crashes by Crash Severity Figure 25: Indian Creek Dr and Abbott Ave Intersection -Overview of Target Crashes at Approach with RLC 26 5.2 Statistical Models Traditionally, naïve statistical methods have been used to quantify the safety performance of roadway countermeasures. In these approaches,the long-term safety effectiveness of the roadway countermeasure—in this case intersections with RLC—is quantified based on estimated crash rate. The estimated crash rates in this case are obtained by averaging observed crash rates over a few years on sites where the countermeasure have been installed. Despite being the most used technique for evaluating safety effectiveness of roadway countermeasures,this approach however, suffers from methodological and statistical limitations, including but not limited to: • Failure to estimate reliable estimates when there is scarcity of sample size. The estimated long-term crash rate obtained by averaging observed crash rates over a few years can be unduly influenced by a single year with an unusually high(or low)number of crashes. • Inability to account for the uncertainty of the crash data. • Failure to account for the regression-to-the-mean bias, which is often predominant when the treatment sites are identified based on prior crash experience. These limitations can be potentially addressed by employing a more reliable approach such as a full Bayes (FB) method in lieu of a crash rate approach. The FB method uses the data from the intersections of interest in combination with information from similar sites to complement the limited data. A FB approach has the ability to account for most of the uncertainties in the dataset and model parameters(Park et al.,2016).This method is also independent of sample size,yielding robust results even when used with small sample size (Li et al., 2013). Additionally, the FB approach divides the periods into time intervals(yearly in this case)and models each time interval as a separate data point to account for time variations, unlike the crash rate methodology which averages the data into a single data point. Detailed discussion about the FB method is presented in Carriquiry&Pawlovich(2004). This study evaluated the safety effectiveness of RLCs at signalized intersections. The study employed the FB methodology to evaluate the overall safety effectiveness in terms of the effectiveness for target crashes by crash type(angle/left-turn/right-turn, rear-end, and sideswipe), and crash severity(PDO and fatal/injury crashes). Note that the models were developed only for four-legged signalized intersections. Only three treatment intersections are three-legged; the sample size is too small to yield reliable results from the FB statistical analysis. 5.2.1 Data Variables Considered In this study, the FB before-and-after with comparison sites approach was used to assess the effectiveness of RLCs at signalized intersections. The RLCs were installed in 2010. The before- period included two years(2008 and 2009)before the installation of the RLCs;and the after-period included three years after the installation of RLCs (2011-2013). As mentioned earlier, treatment intersections are those where RLCs were installed.Comparison(i.e.,non-treatment)intersections, on the other hand, include intersections similar to treatment intersections but without RLCs. Independent variables included in this study are listed below(see Table 4): 27 • Treatment indicator. It represents the countermeasure effectiveness (i.e., RLCs in this case). It measures the difference in crash count between treatment and comparison intersections. o comparison intersections (code 0) o treatment intersections (code 1) • Time indicator. It accounts for the changes in crash frequency due to the intervention. o before period(code 0) o after period(code 1) • Treatment by time indicator. It accounts for different time trends across treatment and comparison intersections. • Jump parameter. It accounts for a sudden drop or increase in crash frequency upon installation of RLCs. • Average annual daily traffic(AADT) on the major street(Major AADT). • AADT on the minor street(Minor AADT). • Speed limit on the major approach (Major speed) o <30 mph (code 0) o > 30 mph(code 1) • Speed limit on the minor approach (Minor AADT) o 5 30 mph (code 0) o > 30 mph (code 1) • Yellow time o <4 seconds(code 0) o >4 seconds(code 1) • All red time o <2 seconds (code 0) o >2 seconds(code 1) • Length of the pedestrian crosswalk • Number of through lanes on the major approach (Major through lanes) o <2 lanes (code 0) o >2 lanes (code 1) • Number of lanes on the minor approach (Minor through lanes) o < 1 lanes (code 0) o > 1 lanes (code 1) • Number of driveways within 250 feet from the center of the intersection (Number of driveways) o <3 (code 0) o > 3 (code 1) 28 Table 4: Descri►tive Statistics of the Data Used in the full Ba es Models ;triable l nits Mean Std. DeN. Minimum NIcdian Vatinn►n► Major AADT Vehicles per day 31,721 13,674.06 4,900 35,000 83,500 Minor AADT Vehicles per day 9,581 8,802.09 2,700 5,500 39,500 Major speed Miles per hour 32.13 2.48 30 30 35 Minor speed Miles per hour 30.38 2.35 20 30 35 Yellow time Seconds 3.94 0.20 3.06 4 4.3 All red time Seconds 2.36 0.55 2 2 4 Length of pedestrian crosswalk Feet 72.27 18.7 48 67.3 120 Major through lanes Count 2.15 0.53 1 2 4 Minor through lanes Count 1.225 0.47 0 1 2 Number of driveways Count , 3.325 1.53 0 3 8 5.2.2 Poisson-gamma Model The Poisson-gamma statistical model was considered to assess the safety effectiveness of RLCs. In all cases, Kt in Equation 1 denotes the crash count observed at intersection i(i = 1,2,3,...,n) during year t (t = 1,2,3, ...,5) and can be modeled with a Poisson distribution with mean and variance equal to bit. Yit lett ^' Poisson (eit), (1) The Poisson mean Sit can be written as shown in Equation 2. 1n(eit) = ln(iii.t) + £i (2) Using the Poisson-gamma, the random effect Ei is assumed to follow a Gamma distribution as presented in Equation 3. E (c p,Gamma( 1)� where �P Gamma(1,1) (3) The log-gamma model for crash density is described as a piecewise linear function (Equation 4) of predictor variables, such that the function is continuous at the change point toi. toi represent the interventions year (2010) for the i`h treatment intersections. The piecewise linear function is defined by at least two equations, each of which applies to a different part of the domain, i.e., before-and-after installation of the RLCs in this case. The linear-intervention model allows for different slopes of crash frequency on time before and after the installation of the RLCs and also across the treatment and comparison intersections. ln(eic) = ao + a1Ti + a2lt>to; + a3Tit + a4Tilt>toi + 131X1it + 32X2it + ...+ R10X10 (4) 29 5.2.3 Model Results and Discussion To obtain the FB estimates of the unknown parameters, it is required to specify prior distributions for the hyper-parameters. The most commonly used priors are vague normal distributions (with zero mean and large variance)for the regression parameters.The posterior estimates of the model's parameters for the FB methods were obtained via four independent chains with 25,000 iterations whereby the first 25,000 were used as a burn-in sample.The posterior means and the 80th percentile Bayesian Credible Intervals (BCIs) of the posterior distributions for each crash type and crash severity are shown in Tables 5 and 6, respectively. The predictor variable is considered to be significant at 80%BCI if the values of the 10%and 90%percentiles do not include zero (0), i.e., they are both negative or positive.The findings from the analyses are summarized below by target crash type and target crash severity. Total Target Crashes For both treatment and non-treatment four-legged signalized intersections, • In general,major and minor approach AADT, speed limit of over 30 mph, amber time z 4 seconds,length of pedestrian crossing,and number of driveways z 3 within the intersection influence area increase the target crash probability. • Longer all red interval has a positive impact in reducing the target crash probability; however, it is not significant at 80%BCI. • Major approach with more than two through lanes and minor approach with more than one lane result in reduced target crash probabilities. For four-legged signalized intersections with RLC, • Compared to the before-period, the after-period experienced fewer target crashes. • Fewer target crashes are expected at the treatment intersections compared to the non- treatment intersections;however, it is not significant at 80%interval. • There is a significant drop in target crashes immediately after the installation of RLCs. • The target crashes dropped immediately after the installation of RLCs, and then continued to increase,but they are still lower than the target crashes in the before-period. Angle/Left-Turn/Right-Turn Crashes • Compared to the before-period, the after-period experienced fewer angle/left-turn/right- turn crashes. This observation is significant at 70%BCI. • Fewer angle/left-turn/right-turn crashes are expected at the treatment intersections compared to the non-treatment intersections;however,it is not significant at 80%interval. • There is a significant drop in angle/left-turn/right-turn crashes immediately after the installation of RLCs. • The angle/left-turn/right-turn crashes dropped immediately after the installation of RLCs, and then continued to increase,but they are still lower than the before-period. 30 Rear-end Crashes • Compared to the before-period,the after-period experienced fewer rear-end crashes. • More rear-end crashes are expected at the treatment intersections compared to the non- treatment intersections. This observation is significant at 70%BCI. • There is a drop in rear-end crashes immediately after the installation of RLCs; however, this drop is not significant at 80%BCI. • The rear-end crashes dropped immediately after the installation of RLCs, and they continued to increase at a steeper rate. Sideswipe Crashes • Compared to the before-period,the after-period experienced fewer sideswipe crashes. • Fewer sideswipe crashes are expected at the treatment intersections compared to the non- treatment intersections. • There is a significant drop in sideswipe crashes immediately after the installation of RLCs. • The sideswipe crashes dropped immediately after the installation of RLCs, and then continued to increase,but they are still lower than the before-period. Target PDO Crashes • Compared to the before-period,the after-period experienced fewer target PDO crashes. • Fewer target PDO crashes are expected at the treatment intersections compared to the non- treatment intersections; however, it is not significant at 80%BCI. • There is a significant drop in target PDO crashes immediately after the installation of RLCs. • The target PDO crashes dropped immediately after the installation of RLCs, and then continued to increase,but they are still lower than the before-period. Target Fatal/Injury Crashes • Compared to the before-period, the after-period experienced fewer target fatal/injury crashes; but it is not significant at 80%BCI. • Fewer target fatal/injury crashes are expected at the treatment intersections compared to the non-treatment intersections; however, it is also not significant at 80%BCI. • There is a significant drop in target fatal/injury crashes immediately after the installation of RLCs. •. The target fatal/injury crashes dropped immediately after the installation of RLCs,and they continued to increase,but they are still lower than the before-period. 31 o R N M V) 'et d' 00 ' 00 N N M 00 a 00 O M O I ) 00 V) .. et N V) M M Ir) O) O N C .. O O — .-. C C(,) .. d 1 o eg .- 0, N -I 00 VD M ...1L o N N C V) CO 000 0 O 00 b I-- M M N )D O Tr - IA O 0 O 00 7 O 00 .-. Ir) R )D .• .. I i 1 I I I 3 1 r> O) N O) b er )D .-. N Ir) Tr I ) )DN C) J - M N N 00 O) O V) V) M O 00 N :.., V) O N .. O N 00 O O, N N ems) O O M V) O O O C .. O O O O O I .)I O i U• ' I CO .. C 1 u) 00 is IT O) C 00 a eh o) :° � °Ov ..1. 0 r) 00 In In In C CN Cr, O O I^ I-. ON N N on 00 C C O O C � O C Ic C C 0 4, I 'C ea v 0 cno M )- ) -,. 0 In )D 00 M M 'C * M •,-. h •L Ir) O` .. N O N '4 ON •• N •• )C 'V .� O M M N O O O IA O O 00 O b O In .. • '- N o O C O O O O M O ''R O O LM) :J 1 IH O C C • = )O 00 1r) 0, M O' N N I!) N V) •K N Ir) N M ccI. O) N O M O en ^ 0 WI N C L NC is O) V) O O I . M--- N O N .. O O O O O NI I O O N O .. o I U w~ e G )O N N )D O) M V T O) 3 N M C G o G. O 00 N N M en O 00 O ^ .. I� • L O .-. N .. . O R N et .. en O N y = C' Ir) O C O O .• O O O O I.. O O 1 O a: I O ca y CL a) 40 IX o e N N M cn M M M N el 00 N N O tel V) 0000 ,,.I N O M i s O V) C C "" O Ir) I^ O d: M N b •er •! 0 i t �i C C C I I O O O C C O "iO 1 I I - w v C)v • 00 N .--. N of N C' In Ir) Ir) •I I/) N N ;o er O 0' O O 00 O) O )O NN 00 00 N N a „ bL :s V) O C N •• O) - O d• N M 00 O 'C j_ • 00 O C O O OI C O C O '^ C O O 4> -= .t I I I I I C I 1 I C CA <C L v Li NI! ON V) )D V) )D N 0 ON )D R V) 'V 00 )D 00 .0 • •. M N O 10 ON )O . . N O )O N vt r-- 4. 4. O r OOI. •• 'O O O N M N N N N .dM a> O O O o O O OO o O C .I s ' s, s '7 :C rcco E... L • , cd O) A Ir) O) 00 N 0,,,,?, V) M )D en N N 00 0 C 15 C N V) try C N O O� )D N N N V) (0 O .. o O N N O r r .. R M 00 .. Ia Ce. iL '- V) O O O O O C O fa O 7. O O O O O (.. I O _ 00 A ^ O) N N N Ir) v N N In ON N -- N M FY p - ... 00 00 .. 00 en — ef• 00 VIeh (.., V M N O .• 0 O N Ir) N M 0 Ir) N 5 a. G e�) O O O O O O O •• O O C C O O v I I 1 I Vi �i 4. a > a) o o C-. O O •Er 2 0 /E— FI 0 4 .4 .4 .� N n c9 n <1.)› to, y y w cn4 L o E 0 yl y N 0cn 000 "0 tn G N C = 4. > F. 4. ao 0,o ,E E 4. R. 00 a comma C O� ,o b ^ C .0 as ,o C C 'C I!) g s 40 r v. o O 0 A ca A 3 .d .Ly .� ,4 z 0 -0 0 -0 N N o q (I.) c .b a V N N az • 0 ob o c " C H • a o 3 E •E .y, ca ca a.2 - ti C :3 _ •'03 v .. N 0 ° = ov 0 o •MI o C o o alai w 6) E = °: w g o F i C. 4 4 5 0- >- ( Q ti .J 8 2 2 `5 z 3 .5 E• .b E—I L. E� E—I '.. Z .. N M -. V) --- of '.D .-• S 00 M a M N V) .•• M 00 N N 00 o 00 M 00 N el O V O N N O N M V) •-. O C O C .-. C C 0 O C C O O v. ' u i • .,0 00 M 00 VI V) 0 VO N0 Vi 00 'O 'D 00 00 e. -. 00 V• N V) N N 1!1 N 000 'C a en O 00 -- •--( O O O N O O N O �t t N • i. — O O O O O O 0 N - CF" ,,,,0O O C+ M 00 0 e! N M in '.D 7 N N O% 00 N M ., 0 0 0 M 00 1!1 s O O ti M O •-•• •-r VI O O 00 IN -. N O+ et j vi O O O C O C O O O O O O O O N 0000 000 V) 47 M N N s.D en N s oo et N 0 O'. O C'. en N 0 •--- .0 0 0 00 N N N N N N 0 0 O O O O O O O O o O O O .-r 1 A VI V) O. en N .-- Os N N 00 0 et M ' '.D 7, -C. in V) NO 0 0 N 00 0 0000 N IN 0 kr)N 00 .-. ..r O' O ..r 1n N N •--. .-. O O O O O O v '- O O O O O I I , , r C O+ - N .1' - V) et R en N d' .0 •--. ' 0 O'. N N 'O .- N N — V) 0 N C. O 0 N 0 O O O M Vi N M O et N 7 e+i 0 0 0 O o O . O 0 O OO O en 0 .-. = U L "er O. V) N V) I) N Q\ .-1 N 0 V) N CO N 00 -- : O O '~ .--' `D co O N M N N N M O O O O O O O O 0 O O O coni i 1 W1 v. --. U e OS ON V) Q\ 00 00 00 V) M 0 en N --- N 00 4.. 0 en N VI en N N en O O, .0 N N — N V) 'v y O O -. v O 0 N N 0 N N •-. et en 00 c.., V) O O O O O O o — O O O O O O U L.� CL 7' i+ CO CO L _ _ .. ON N N N V) et N N V) O% N --- N M CO 6>• . N M O N .-• 00 00 .. 00 en .--- e! 00 V) eh 0 M N �D 0 •.--� O 0 N IF) N en O I) N O -Coo� 7 M O O O O O O O •-� O O O O O Ooo A L. W al U 4. N w Z VI E Fko 17C C 00 0a -0 x .� 4: y .. a. O. ii 3 .c R ° ° `N° •p F- F- .4 s .s N a .0 .0 , ,- 0 u «~� F o 0 = I < < 0 E CO E A„ n 4) CO o •v — a� o cw _ .. )., 1- 0.M C M '- O i N ,_ N b O 49 ' 'a 45 N 0 tcp 7, m • o mm1 o 0 o c E y a cc 6 . E E •alaaioy . c °' v °' YE T - L ° 7.) ~ o 40 a c 0 . .. - a 66. O F i _ 2 m 2 ;) >- Q a t 2 2 Z 3 .E [F b F- .-',. ..E-. E-. Z 6.REFLECTION The main objective of this study was to evaluate the safety effectiveness of the PRL Enforcement Program in the City of Miami Beach, Florida. The simple before-and-after analysis and the full Bayes before-and-after evaluation approach were used to quantify the safety effectiveness of the RLCs.The analysis was based on target crash type which includes angle/left-turn/right-turn,rear- end,and sideswipe crashes,and target crash severity which includes PDO and fatal/injury crashes. In general,three main inferences could be drawn from the analysis results. 6.1 Target Crash Type and Target Crash Severity The presence of RLCs at the study intersections tend to increase rear-end crashes and decrease angle/left-turn/right-turn and sideswipe crashes. This observation is consistent with several other previous studies that have focused on analyzing the effectiveness of RLCs at intersections (Hoye, 2013;Ahmed and Abdel-Aty,2015; Llau et al.,2015; Claros et al.,2017). Rear-end crashes are often less severe compared to angle crashes. For example,Figures 26(a)and (b)show the proportion of fatal and severe injury crashes by these crash types in the City of Miami Beach and statewide,respectively.Note that these figures are for the year 2013 and for all crashes that occurred on non-limited access facilities. It can be inferred from Figure 26 that in the City of Miami Beach, about 2.4% of all angle/left-turn/right-turn crashes result in fatal or incapacitating injury, while a relatively lower 0.8% of all rear-end crashes are fatal and incapacitating injury crashes. A similar trend,nonetheless at a higher magnitude,is observed in Florida(Figure 26(b)). 3.0 2.4 6.0 5.5 2.5 2.1 § 5.0 3.9 rIx i, y 2.0 � y 4.0 al 1.5 at i 3.0 2.1 U1.0 0.8 1 2.0 1.0 o '' 0.5 I 0.0 o 1.0 a "a 0.0 6 ..e 0.0 6. a.. li.- 4�a�a •,�5��4 t° a qS°O 4�te1 `�5�,4 So (a)City of Miami Beach (b) State of Florida Figure 26: Percent of Fatal and Severe Injury Crashes by Target Crash Type 6.2 General Crash Trend From the full Bayes model results, it can be inferred that crashes at the treatment intersections were on an increasing trend. However, this trend may not be attributed to the presence of RLCs. In general,recent years have seen an increasing trend in crashes.At the city,state,and the national level, crashes are generally on an increasing trend, especially since 2011. Figure 27 gives the annual crash trends from 2008-2016 in the City of Miami Beach, the State of Florida, and the United States. 34 7,000 6,000 5,000 ••"" — ••. 4,000 • U 3,000 2,000 1,000 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year (a)Annual Crash Trend in the City of Miami Beach(Source: Signal Four Analytics, 2018) 800,000 700,000 H 600,000 500,000 •••• co U 400,000 0 300,000 I 200,000 100,000 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year (b)Annual Crash Trend in Florida(Source: Signal Four Analytics,2018) 8,000,000 7,000,000 6,000,000 ..... ti 5000000 4,000,000 75 3,000,000 2,000,000 1,000,000 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year (c)Annual Crash Trend in the U.S. (Source: NHTSA,2018) Figure 27: Annual Crash Trend 35 6.3 Safety Performance of Non-treatment Intersections The simple before-and-after study revealed that most of the treatment intersections experienced fewer target crashes after the installation of RLCs. The average number of target crashes at the non-treatment (i.e., comparison) intersections have also reduced in the after-period (i.e., from 2011-2013). Drivers,when they encounter a RLC at an intersection,may anticipate RLCs at other intersections within the region, and as a result drive more cautiously. This behavior may result in a reduction in target crashes at non-treatment intersections as well. Moreover,jurisdiction-wide publicity of RLCs and the general public's lack of knowledge on the exact installation locations of RLCs may result in fewer target crashes within the region.Figure 28 shows the average number of target crashes at non-treatment intersections during the before- and after- periods. Note that target crashes include angle/left-turn/right-turn,rear-end, and sideswipe crashes. A total of five signalized intersections that are far away from the intersections with RLCs were identified. The average number of target crashes at these far-away intersections was found to be higher during 2011-2013 (i.e., after RLC installation) compared to 2008-2009 (i.e., before RLC installation). In general, angle/left-turn/right-turn and sideswipe crashes increased in 2011-2013, while rear-end crashes reduced. In other words, the intersections in the vicinity of RLCs, in general,were found to experience fewer angle/left-turn/right-turn and sideswipe crashes,while the intersections far away from the treatment sites were found to experience an increase in angle/left- turn/right-turn and sideswipe crashes in 2011-2013 compared to 2008-2009. 36 10 9.5 16.0 15 8 7 14.0 12.0 9.7 6 -5 5.7 10.0 4.08.06.5 4 6.0 4.5 3.3 4 3.3 1111 2 2 ,0.7 : 02.0 0 0.0 i Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After ■Before After Arthur Godfrey Rd. and Meridian Ave Alton Rd.and 16'h St. Intersection 9 8 7.7 20 19 IL 7 15 -Ii 13.7 6 5 10 8.5 3 3 3 3 5.3 6.5 5.3 2 2 2 1.7 5 4 3 Il IlL O ' Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes to Before After ■Before After Alton Rd. and 11'h St. Alton Rd. and 8'h St. 18.0 16.5 9.0 8 16.0 8.011 7 14.0 7.0 12.0 6.0 4.5 10.0 8 7 8 5.0 4.0 4 3.7 8.0 6.0 3.0 2 4.0 3.3 2.0 2.0 1.5 1.3 I 3 1.0 0.5 0.3 0.0 0.0 El,. Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After ■Before After Washington Ave and 16'h St. 41s'St.and Indian Creek Dr Figure 28: Target Crashes by Crash Type at Comparison Intersections 37 4.0 3.5 3.7 8.0 7 3.5 7.0 3.0 6.0 2.5 2.5 5.0 2.0 3.3 1.5 3.0 2 2.3 1 2 4.0 3.5 i 2.0 1.5 1.0 0.7 1.0 10.3 10.7 , 0.5 0 0.0 0.0 Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After ■Before After 17'h St. and James Ave 5th St. and Collins Ave 2.5 2 2 1.7 2.0 1.5 1.5 12 1.5 1 1 1 0 1.0 0 7 0.7 10.7 0.5 0.5 1 0.3 0.5 '0.3 I 0 0 0 1110.0I Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After ■Before After 631"d St. and Pine Tree Dr Pine Tree Dr and Sheridan Ave(3-legged) 3 4 4 2.5 3.5 3 4 3 3 2 1.7 3 1.5 I 2 2 1 1 2 1.3 1 1 I I0.3 0.3 1 0.5 0.7 0 0 0 0 . Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After •Before After Washington Ave and 15th St. (3-legged) West Ave and 11'h St. (3-legged) Figure 28 (cont'd): Target Crashes by Crash Type at Comparison Intersections 38 I 1.8 1.7 12 10.7 1.6 10 1.4 1.2 8 7 1.0 6 a 0.8 0.7 0.7 4.5 0.6 0.5 liii 4 3.5 0.4 0.3 2 1 1.7 2 0.2 p 0 0 I El 0.0 0 Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes •Before After ■Before After Meridian Ave and 8111 St. Washington Ave and 1 l'h St. 2 1.7 3 2 2 1 2 1 1 2 1.3 1 1 0.7 1 0.5 0.5 1 0.7 0.5 0.5 0 ] 0 0 I 0 0 I 0 0 1 0 111 0 0 Angle Rear-end Sideswipe Total Target Angle Rear-end Sideswipe Total Target Crashes Crashes ■Before After ■Before After Meridian Ave and 11'h St. 5`11 St. and Ocean Dr(3-legged) 8 7.3 7 6 5 4 3.7 3 2.3 2 1.3 1 1.5 1 0ll 0.5 I 0 Angle Rear-end Sideswipe Total Target Crashes •Before After Washington Ave and 6`h St. (3-legged) Figure 29: Target Crashes at Intersections Far Away from the Treatment Sites 39 7.SUMMARY The main objective of this study was to evaluate the safety effectiveness of the PRL Enforcement Program in the City of Miami Beach, Florida. The simple before-and-after analysis and the full Bayes before-and-after evaluation approach were used to quantify the safety effectiveness of the RLCs.The analysis was based on target crash type which includes angle/left-turn/right-turn,rear- end,and sideswipe crashes,and target crash severity which includes PDO and fatal/injury crashes. A total of ten RLCs are operational at nine signalized intersections in the City of Miami Beach. Due to the long period of construction activity,the intersection where Alton Road meets 17th Street was not included in the study. The simple before-and-after crash data analysis was conducted for the remaining eight signalized intersections. The advanced full Bayes before-and-after analysis was conducted only for the five four-legged signalized intersections. Only three treatment intersections are three-legged; the sample size is too small to yield reliable results from the FB statistical analysis. 7.1 Simple Before-and-after Crash Data Analysis 7.1.1 Four-legged Intersections At four-legged intersections with RLCs, • In general,there was a reduction in target crashes after the installation of RLCs. • Among the target crash types, angle/left-turn/right-turn and sideswipe crashes usually decreased while rear-end crashes usually increased. In other words, the presence of RLCs at the study intersections were found to increase rear-end crashes and decrease angle/left- tum/right-turn and sideswipe crashes. However, rear-end crashes are often considered to be less severe compared to angle/left-turn/right-turn crashes. • In general, there was a reduction in PDO target crashes, but target crashes resulting in injuries slightly increased. At approaches with RLCs, • At three of the five intersections,target crashes reduced after the installation of RLCs. • At two of the five intersections,total crashes reduced after the installation of RLCs. • At three of the five intersections,rear-end crashes increased after the installation of RLCs. 7.1.2 Three-legged Intersections • Overall, the intersections with RLCs experienced a reduction in target crashes after the installation of RLCs. • At all the three intersections,rear-end and sideswipe crashes reduced after the installation of RLCs. • Angle crashes reduced at two of the three intersections. • In general, there was a reduction in PDO target crashes, but target crashes resulting in injuries slightly increased. • At approaches with RLCs,there were no target crashes after the installation of RLCs. 40 7.1.3 Safety Performance of Intersections with No RLCs The average number of target crashes at the non-treatment intersections (i.e., signalized intersections with no RLCs)that are in the vicinity of treatment intersections have also reduced in the after-period (i.e., from 2011-2013). Drivers, when they encounter a RLC at an intersection, may anticipate RLCs at other intersections within the region,and as a result drive more cautiously. This behavior may result in a reduction in target crashes at intersections with no RLCs as well. Moreover,jurisdiction-wide publicity of RLCs and the general public's lack of knowledge on the exact installation locations of RLCs may result in fewer target crashes within the region. The average number of target crashes at five signalized intersections that are far away from the intersections with RLCs was found to be higher during 2011-2013 (i.e., after RLC installation) compared to 2008-2009 (i.e.,before RLC installation). The intersections in the vicinity of RLCs, in general,were found to experience fewer angle/left-turn/right-turn and sideswipe crashes, while the intersections far away from the treatment sites were found to experience an increase in angle/left-turn/right-turn and sideswipe crashes in 2011-2013 compared to 2008-2009. 7.2 FB Before-and-after Analysis • In general, crashes at the treatment intersections were on an increasing trend. However, this trend may not be attributed to the presence of RLCs. Recent years have seen an increasing trend in crashes. At the city, state, and the national level, crashes are generally on an increasing trend,especially since 2011. • There is a significant sudden drop in all types of target crashes immediately after the installation of RLCs. • Compared to the before-period, the after-period experienced: • Fewer target crashes • Fewer angle/left-turn/right-turn crashes • Fewer sideswipe crashes • More rear-end crashes • The sideswipe and angle/left-turn/right-turn crashes dropped immediately after the installation of RLCs,and then continued to increase,but they are still lower than the before- period. • The rear-end crashes dropped immediately after the installation of RLCs, and then continued to increase,but they increased at a steeper rate. 41 REFERENCES Ahmed, M. M., & Abdel-Aty, M. (2015). Evaluation and spatial analysis of automated red-light running enforcement cameras. Transportation Research Part C. Emerging Technologies,50, 130- 140. American Traffic Solutions (ATSOL). (2018). Red-light running dangers in the United States. Retrieved from https://www.atsol.com/wp-content/uploads/2017/06/ATS-RLR-Dangers-Cutsheet-2017.pdf Carriquiry, A., & Pawlovich, M. (2004). From empirical Bayes to full Bayes: methods for analyzing traffic safety data. White Paper, Iowa State University. City of Fort Lauderdale. (n.d.). Red light safety camera program. Retrieved from https://www.fortlauderdale.gov/departments/transportation-and-mobility/red-light-safety- camera-program City of Miami Springs. (n.d.). Get the facts—red-light cameras save lives. Retrieved from https://www.miamisprings- fl.•ov/sites/default/files/fileattachments/s olice/.a•e/19539/•et the facts - red light cameras save lives.pdf Claros, B., Sun, C., & Edara, P. (2017). Safety effectiveness and crash cost benefit of red light cameras in Missouri. Traffic Injury Prevention, 18(1), 70-76. Decina, L. E., Thomas, L., Srinivasan, R., & Staplin, L. (2007). Automated enforcement: A compendium of worldwide evaluations of results. Report prepared for National Highway Traffic Safety Administration(NHTSA), Kulpsville, PA. Retrieved from https://www.google.com/search?q=HS810763.pdf&rlz=1C1GGRV enUS752US753&oq=HS81 0763.pdf&aqs=chrome..69i57.10366j0j8&sourceid=chrome&ie=UTF-8 Florida Online Traffic School. (2018). Red-light camera citations in Florida. Retrieved from https://www.floridaonlinetrafficschool.com/articles/red-light-camera-tickets.aspx Hoye, A. (2013). Still red for red light cameras? An update. Accident Analysis & Prevention, 55, 77-89. Insurance Institute for Highway Safety – Highway Loss Data Institute (IIHS-HLDI). (2016). Turning off red light cameras costs lives, new research shows. Retrieved from https://www.iihs.org/iihs/news/desktopnews/turning-off-red-light-cameras-costs-lives-new- research-shows Ko,M.,Geedipally, S. R.,Walden,T. D.,&Wunderlich,R. C. (2017). Effects of red light running camera systems installation and then deactivation on intersection safety. Journal of safety research, 62, 117-126. 42 • Li, W., Carriquiry, A., Pawlovich, M., & Welch, T. (2008). The choice of statistical models in road safety countermeasure effectiveness studies in Iowa.Accident Analysis&Prevention,40(4), 1531-1542. Llau,A.F.,Ahmed,N.U.,Khan,H.M.,Cevallos,F. G.,&Pekovic,V. (2015).The impact of red light cameras on crashes Within Miami—Dade County, Florida. Traffic Injury Prevention, 16(8), 773-780. National Highway Traffic Safety Administration(NHTSA). (2018). Traffic safety facts 2016 data (DOT HS 812580).NHTSA's National Center for Statistics and Analysis, Washington D.C. Park,J.,Abdel-Aty,M.,&Lee,J.(2016).Use of empirical and full Bayes before—after approaches to estimate the safety effects of roadside barriers with different crash conditions.Journal of Safety Research,58, 31-40. Pulugurtha, S. S., & Otturu, R. (2014). Effectiveness of red light running camera enforcement program in reducing crashes:Evaluation using"before the installation","after the installation",and "after the termination"data.Accident Analysis &Prevention, 64,9-17. Shin,K.,&Washington,S. (2007). The impact of red light cameras on safety in Arizona.Accident Analysis&Prevention,39(6), 1212-1221. The Geoplan Center, Department of Urban and Regional Planning,University of Florida. (2018). Signal Four Analytics. The Geoplan Center, Department of Urban and Regional Planning, University of Florida, Gainesville, Florida. https://s4.geoplan.ufl.edu/ 43