Spatial analysis of road traffic accidents: Identifying hotspots for improved road safety in Addis Ababa, Ethiopia

Abstract Safety on road networks is the utmost important factor to consider for public well-being and transportation efficiency. This study introduces a new approach that combines Getis-Ord spatial statistics and crash rate analysis to identify significant road traffic accidents (RTAs) characterized by hotspots on road segments in Addis Ababa’s road network. The study’s results visually portray the crash locations, associating them with the underlying road network, which demonstrates a notable concentration of accident hotspots, between the years from 2014 to 2019 on Addis Ababa’s roads network. The RTAs spatial analysis resulted in the identification of hotspots on 33 road segments, 3 intersections, and 10 roundabouts/squares. Among the identified hotspots, the road segment recognized as Djibouti Street, extending from Bole Edna Mall to the 22 “mazoriya” roundabouts, stands out as the most significant accident hotspot. It exhibits an average of 37.5 crashes per kilometer per year, encompassing a road segment length of 1141 m. Using both methods in this study is crucial for validating findings by identifying high-crash segments and enhancing their reliability and hotspot accuracy. This unique validation method links each traffic accident’s spatial data with the road network using both crash rate and spatial statistical analysis, effectively pinpointing accident hotspots. Given the limited resources, this approach enhances awareness of accident-prone locations, enabling the prioritization of safety measures to improve road safety. It effectively addresses spatial analysis gaps related to RTAs in Ethiopia and holds practical significance by identifying and prioritizing safety measures aimed at reducing accidents within Addis Ababa’s road network.


Background
Ensuring safety on road networks is of utmost importance due to its significant impact on public well-being and transportation efficiency.Nevertheless, the transportation system has unintended negative consequences in the form of traffic accidents causing physical harm, loss of life, and economic costs.Road traffic accidents are now the eighth leading cause of death globally, with 1.35 million fatalities reported every year.Children and young adults aged 5-29 years are most at risk, and low-income countries experience the most significant numbers of fatalities resulting from road traffic accidents (World Health Organization WHO, 2018).In Ethiopia, there has been an increase in the number of reported road traffic accidents over the past 5 years.There were a total of 157,326 reported road traffic accidents across Ethiopia in the 2014/2015 fiscal year, and this number increased to 183,472 in the 2018/2019 fiscal year (Ethiopian Federal Police Commission EFPC, 2019).This represents an increase of approximately 16.6% over the five-year period.In response to such problems in many countries, governments around the world declared a Second Decade of Action for Road Safety 2021-2030 with the explicit target to reduce road deaths and injuries by at least 50% during that period (World Health Organization WHO, 2021).
Cities around the world are befoul in transportation, traffic congestion, air pollution, noise pollution, RTAs, and so on all have a negative impact on the aesthetics of urban environments (Odeleye & Oni, 2008).Urbanism has an extensively more effect on traffic safety as summarized by Tărîţă Cîmpeanu and Burlacu (2012), and consequently urbanism and spatial planning are geared for reaching synergy among requirements, opportunities, and road features as a way to attain sustainable development in a certain area.The emergence of traffic and subsequent traffic congestion in urban road networks are increasing worldwide with the growing number of vehicles, which results in excess delays, and reduced safety (Mazloh et al., 2016).Traffic safety is defined by the Highway Safety Manual (HSM) as "the crash frequency or crash severity, or both, and collision type for a specific time period, a given location, and a given set of geometric and operational conditions" (American Association of State Highway and Transportation Officials AASHTO, 2010, p. 3-1).Road accidents, which may be referred to as road traffic crashes or collisions, occur on public roads and may involve a single vehicle or a collision between a vehicle and one or more vehicles, pedestrians, animals, or fixed objects.This definition has been used in various literatures, such as Hoel et al. (2006) and Wang (2010).The term "public road" includes pathways for pedestrians, while "road vehicles" include both motorized automobiles and bicycles.RTAs that occur could result in harm to a person, which may range from slight injuries to fatal or serious ones, or solely property damage (Bell et al., 2006).With the development of society and improvement of living standards, the number of vehicles increases rapidly, followed by frequent RTAs that can not only bring casualties and property losses, but interfere with traffic operation and cause traffic congestion or even traffic break (He, 2013).Due to these traffic impacts, all the huge losses on the social (fatalities and injuries), economic, and environmental levels are affecting most developed and developing countries' economies.
The number of reported traffic accidents in Addis Ababa increased by approximately 37.9% from 4,262 in the 2014/2015 fiscal year to 5,874 in the 2018/2019 fiscal year.On a daily basis, road traffic crashes cause an average of 13 fatalities and 37 injuries in Ethiopia (Addis Ababa City Administration Traffic Management Agency (AATMA), 2019; Addis Ababa City Road Traffic Management Agency (AARTMA), 2019), with Addis Ababa accounting for 10% of the deaths and 26% of the injuries despite having 56% of the registered vehicles in the country (Addis Ababa City Administration Transport Bureau AATB, 2019).
Various studies have been conducted on the topic of traffic safety analysis for reduction of the impact of RTAs.Chen (2012) found that most traffic data analysis is limited to basic statistical methods which do not adequately reveal hidden patterns in traffic accident data.Statistical analysis lacks the ability to map spatial distributions and explore the relationship between accidents and road network elements.The multitude of advantages offered by GIS make it an appealing option for addressing the increasing challenges related to traffic safety (Mohan Rao, 2014).Spatial analysis will undoubtedly contribute to a reduction in traffic accidents, but the correctness, dependability, and thoroughness of the traffic accident reports are crucial to these analyses' success.In order to improve traffic safety analysis, it is crucial that traffic accident reports are accurate and thorough when entering data and doing geographical analysis (Demirel & Akgungor, 2002).
Poor traffic management and inadequate control over the existing road network have led to an increase in traffic congestion and accidents in Addis Ababa city.A situational analysis conducted in three selected regional administrations in Ethiopia under the data design research for road safety improvement in developing countries revealed that accident analysis is generally at an immature level, hindering the discovery of relevant knowledge for decision-making from the accumulated data (Beshah et al., 2012).The majority of road safety analysis techniques utilized in Ethiopia by transportation agencies and safety experts are traditional descriptive methods that concentrate on quantifying and summarizing crash data (Tola et al., 2021).As Habte (2017) mentioned, even though different researches have been done, none of these researches were keen to use GIS analysis to reduce the traffic problem in Addis Ababa.Increased awareness of accident hotspots and timing helps drivers and pedestrians avoid accidents through defensive strategies (Y.Berhanu et al., 2023).Identifying high-risk areas on a road network can help with efforts to reduce traffic accidents, and they also emphasized that the first stage in the study to improve traffic safety is the identification of hotspot spots (Harirforoush & Bellalite, 2016).Wang (2012) described a clustered area as a hotspot with a higher chance that an accident will happen based on spatial dependency and past experience.There is no clear-cut description of what defines an accident hotspot.In addition, Sajed et al. (2019) defined in their study that accident hotspots are typically areas with a high likelihood of high risk or accident incidence.These are locations where there is an unacceptably high risk of accidents.Generally, the lack of standardization and multiple dimensions characterizing hazardous road locations makes it difficult to determine the "best" definition (Elvik, 2008).Identifying hazardous road locations is one of the primary objectives of spatial analysis of road accidents, as noted by Loo and Anderson (2015).Enhancing road safety and fostering a secure driving environment revolves around identifying hotspots on road segments demonstrating higher crash densities than the surrounding network.With a scarcity of localized research on spatial approaches to identify accident hotspots, our study introduces a pioneering strategy using spatial analysis to enhance road safety.This study combines the Getis-Ord statistic, crash rate analysis, and road traffic accident spatial data in Addis Ababa's road network to identify hotspot segments across the major road network.The results are visually displayed using hotspot maps and density geometry heatmaps.This innovative approach emphasizes the importance of spatial statistical and crash rate analyses in enhancing road safety planning, and decision-making, while nurturing spatial literacy.These findings contribute to the discourse on road safety by seamlessly integrating spatial analysis techniques to address traffic-related challenges in Addis Ababa, Ethiopia.
This study is organized into several sections.Section 2 presents a review of previous research on road traffic accident safety and its spatial analysis techniques.In Section 3, the authors discuss the data used in the study, including how it was made compatible with GIS tools and the use of opensource geocoding and Open Street Maps data.The study's methodology is also outlined in Section 3, including the use of crash rate analysis and spatial statistical analysis (Hotspot analysis) to identify accident densities and hotspots.Section 5 presents the analysis and results of the findings of the spatial analysis.Finally, Sections 6 and 7 provide discussions and conclusions, recommendations, and suggestions for decision-makers and stakeholders.

Literature review
This chapter presents an overview of traffic safety studies, including an examination of traffic accident analysis techniques to identify hotspots.Such studies are motivated by the rise in road traffic accidents and traffic congestion, as well as negative impacts on people and the environment due to a lack of spatial approach.

Road traffic safety
Safety concerns in transportation planning are fundamental, as evidenced by a comprehensive examination of plans across 35 metropolitan areas globally (Peng & Wang, 2011).The critical importance of ensuring safety within road networks reverberates due to its extensive influence on various dimensions of people's lives.The safety performance of road transport systems hinges on a multitude of factors, encompassing network functionality, predictability, consistency, road environment, and traffic dynamics (Hafen et al., 2005).Rising concerns about the safety associated with automobiles and motorcycles are capturing the attention of road users and safety experts worldwide, with a pronounced focus on developing nations (Tavakoli Kashani & Arefkhani, 2018).Within this context, urban planners and traffic engineers are particularly attuned to the effects of diverse road network layouts on traffic safety, presenting a key arena for impactful interventions (Zhang et al., 2014).The sphere of urbanism and spatial planning significantly shapes traffic safety strategies, pursuing a harmonious equilibrium between requirements, opportunities, and road characteristics (Tărîţă Cîmpeanu & Burlacu, 2012).
An integral metric for evaluating road safety emerges through the analysis of societal costs linked to accidents and injuries (American Association of State Highway and Transportation Officials AASHTO, 2010;Hafen et al., 2005;Herbel et al., 2010;Millot, 2004).Notably, while crashes remain relatively infrequent, they represent a modest proportion of incidents in the broader transportation landscape (Oregon Department of Transportation (ODoT) Traffic-Roadway Section, (2021) ( 2021)).The severity of crashes is predominantly dictated by the most severely injured participant, prompting a structured approach for officials to prioritize safety improvements on specific road segments (Elvik, 2008;Golembiewski & Chandler, 2011;Ma et al., 2021;Schneider & Savolainen, 2011).As a pivotal initial step toward advancing traffic safety, the identification of clustered risk areas for traffic accidents assumes priority (Zhang & Shi, 2019).Within this framework, hotspot methodologies emerge as consequential tools, facilitating the orchestration of costeffective road safety strategies to mitigate traffic accidents (Xu & Tao, 2018).

Traffic accidents spatial analysis
Road Traffic Accidents (RTAs) are unfortunate incidents that occur unexpectedly and unintentionally.The analysis of road accident data is essential for safety analysis and identifying areas with high accident frequencies of severe injuries and fatalities (Lee et al., 2005).The goal of traffic accident safety analysis is to prevent further accidents by identifying the cause of an accident or series of accidents (Bhalla et al., 2014).Studies have been conducted using spatial units and road networks to identify elements that impact traffic safety.Analyzing individual road segments is crucial for identifying factors that increase the risk of accidents and despite technical challenges, the use of road networks in traffic safety studies has increased (Á.Briz-Redón et al., 2019).
The study by Ziakopoulos and Yannis (2020) suggests that analyzing road safety indicators across spatial units of analysis is essential in the field of road safety.Spatial analysis of road crashes and their locations is important because road transport is distance-based.The use of Geographic Information Systems for Transportation (GIS-T) (Goodchild, 2000;Loo & Anderson, 2015;Shaw, 2000) and GIS Safety Analysis Tools have facilitated the analysis of traffic safety, including the prediction of areas where accidents are likely to occur (Mehta et al., 2015;Pulugurtha & Pasupuleti, 2013).Many spatial statistical tools are now available in GIS software to build complex models in a simple process and its ability to detect spatial relationships in accidents that cannot be detected by traditional methods of statistical data.Such spatial tools provide a geographical component to crash data, which helps in understanding the crashes in relation to the environment and road.The use of GIS in road management for traffic safety purposes allows for a variety of spatial analyses and graphical illustrations of the results for interpretation (Mohan Rao, 2014;Polat & Durduran, 2011).
The use of GIS technology has been extensively employed in road safety research to identify areas where accidents are most likely to occur.The Kernel Density Estimation (KDE) algorithm in ArcGIS has been used to analyze spatial data to determine the distribution of accident risk, as demonstrated by Budiharto and Saido (2012).Ouni and Belloumi (2019) suggest using spatial autocorrelation, specifically Moran's I and Getis-Ord statistics, as a powerful technique for spatial clustering to identify accident spatial patterns and hotspots.Moran's I measures spatial dependence to group areas with similar spatial patterns, while the Getis-Ord statistic provides a singular measure of spatial correlation and quantifies the extent of clustering in the spatial arrangement (Satria & Castro, 2016).Colak et al. (2018) also supports the use of GIS technology, specifically KDE, for identifying accident-prone highways.

Road accidents hotspot analysis
The first stage in improving traffic safety is identifying high-risk areas called hotspot spots on a road network, which is important for reducing traffic accidents (Harirforoush & Bellalite, 2016).The concept of accident proneness emerged in the early twentieth century, identifying people with a high crash rate who could be banned from driving for increased safety (Evans, 2004).The study by Chainey and Ratcliffe (2013) emphasized the importance of identifying hotspots or road segments with a higher crash density compared to other parts of the network.These hotspots are characterized by a greater concentration of traffic accidents, and they are important in improving road safety and creating a secure driving environment.Due to time and funding constraints, sites with the highest accident risk are given priority in many safety assessments.Many studies focus on implementing structural actions for traffic safety, rather than identifying hotspot detection methods Erdogan et al. (2015).However, finding road accident hotspots can help researchers understand the causes of accidents and prioritize safety enhancements (Colak et al., 2018).Tola et al. (2021) suggested that to reduce crashes through efficient safety mitigation, road traffic accident hotspot analysis aims to identify and prioritize road segments in need of immediate safety enhancements.Also, Harirforoush and Bellalite (2016) in there study stressed that to substantially mitigate the incidence of severe traffic injuries and fatal accidents, we should specifically focus on identifying high-risk segments within road networks.
Recent advances in spatial statistics and GIS have made traffic safety hotspot research possible by allowing analysts to perform advanced statistical analysis of road accident with available GPS crash location data (Atumo et al., 2021).GIS has been utilized as a management system for accident analysis to identify hotspots on highways, such as intersections and roundabouts (Erdogan et al., 2008).One commonly used method to identify hotspots is Kernel Density Estimation (KDE), but it has limitations, as it does not involve any formal statistical inference and there is currently no established density threshold for confidently identifying a hotspot (Erdogan et al., 2015).GIS has allowed analysts to perform an advanced statistical analysis of crash data (Miller, 2000).The Getis-Ord Statistic, particularly the Getis-Ord-Gi* statistic, is a spatial analysis of GIS tool used in various disciplines, including crash data, to identify high-risk or highoccurrence locations through distances and spatial correlation, revealing local clusters of spatial dependence not visible with global statistics (Getis & Ord, 1992).To locate accident hotspots, the Getis-Ord statistic is used, which computes a high index value for a collection of hot spots and a low index value for a group of low index values (Ziakopoulos & Yannis, 2020).However, combining crash rate analysis and spatial statistical Getis Ord analysis has been found to be effective in identifying high-risk areas for road traffic accidents.Several studies have used this approach, including Haghshenas et al. (2018) and Chen et al. (2020), who identified hotspots for accidents in Iran and China, respectively.
When assessing accident hotspot identification, three primary methods are often employed: Kernel Density Estimation (KDE), Getis-Ord-Gi* spatial statistics, and crash rate analysis.KDE utilizes a density estimation approach to visualize clusters, while Getis-Ord-Gi* assesses spatial autocorrelation to identify significant hotspots.Meanwhile, crash rate analysis considers the frequency of accidents in relation to exposure measures.These methodologies offer complementary insights into the spatial patterns of accidents, each with its own strengths and limitations (Chainey & Ratcliffe, 2013;Getis & Ord, 1992).The authors came to the conclusion that using Getis-Ord-Gi* spatial statistics, and crash rate analysis techniques help in identifying spatial clusters of accidents that are not evident using crash rates alone can offer a more thorough and complete view of accident risk, make it possible to validate the findings through cross-checking the results, and thus enable targeted interventions to increase road safety.
Generally, the safety of road transport systems is critical as it affects many aspects of people's lives, and various factors such as network functionality, consistency, predictability, road environment, and traffic determine the safety performance of road transport systems.Urban planners and traffic engineers have been concerned with different road network layouts and their effects on traffic safety.The use of GIS-T and GIS Safety Analysis Tools has facilitated the analysis of traffic safety, including the prediction of areas where accidents are likely to occur.Spatial statistical tools such as Kernel Density Estimation (KDE), Moran's I, and Getis-Ord statistics are powerful techniques for spatial clustering to identify accident spatial patterns and hotspots.Nevertheless, combining crash rate analysis and spatial statistical Getis Ord analysis can identify high-risk areas for road traffic accidents, including spatial clusters of accidents that are not evident using crash rates alone, leading to targeted interventions to improve road safety.

Materials and methods
The general methods employed are described in this section.The recognition and understanding of the potential of accident hotspot identification to improve road traffic safety is widespread within the fields of transportation engineering, urban planning, and road safety research.This study introduces an innovative strategy that combines the Getis-Ord statistic, crash rate analysis, and road traffic accident spatial data in Addis Ababa's road network.By employing spatial analysis techniques, we identify hotspot segments with higher crash densities than the surrounding network.The results are visually displayed through hotspot maps and density geometry heatmaps.This study highlights the importance of spatial statistical and crash rate analyses in enhancing road safety planning, and decision-making, while addressing the scarcity of localized research on spatial approaches.These findings contribute to the discourse on road safety by seamlessly integrating spatial analysis techniques that offer valuable insights for addressing traffic-related challenges in Addis Ababa, Ethiopia.The study's overall process is depicted in figure 3.

Study area
The study area considered for this research is Addis Ababa, the capital and largest city of Ethiopia, serving as the country's commercial, manufacturing, and cultural center, as shown in Figure 1 including its sub-cities and road network.The city is situated in central Ethiopia on a plateau crossed by numerous streams and encircled by hills (New World Encyclopedia contributors, 2021).The city rests at an elevation of approximately 2440 m (8000 ft) above sea level.It serves as the focal point for a comprehensive highway network, hosts an international airport, and marks the terminus of a railway connecting to the Gulf of Aden port in Djibouti, the capital of the neighboring state.According to United Nations population projections, the metropolitan area population of Addis Ababa in 2020 was 4,794,000, reflecting a 4.4% increase from 2019 (Macrotrends, 2023).

Accident data
In order to choose safety-improving countermeasures, collected data should be examined and reviewed to pinpoint areas with current or prospective future safety hazards.Crash historical data provide the main information on the traffic safety environment, driver behavior, and vehicle performance.The time frame of historical traffic accident data used to identify accident hot spots varies from 1 to 5 years (Elvik, 2008).The article notes that due to the random nature of traffic accidents, we should collect data over multiple years to account for the variability in accident numbers.Golembiewski and Chandler (2011) suggest that gathering data from the previous 3-5 years is generally recommended, with a minimum of 3 years.The study examined traffic accident data between 2014/15 and 2018/19, collected from the Addis Ababa Traffic Management Authority, the Addis Ababa transport bureau's annual report, and the Addis Ababa traffic police office.From over 67,000 accident records, 64, 878 data are analyzed in the spatial analysis and the remaining become incomplete data.These data include attributes such as accident ID, location, type, time, date, and driver information, and some fields have missing data like accident location positional information, which are challenges for spatial analysis.Based on the attribute data, chart 1 and Figure 2 illustrate the varying degrees of road accidents in Addis Ababa spanning from 2014 to 2019, categorized into fatal, serious injury, and slight injury incidents.Notably, the year 2016 emerged as the most hazardous, recording 700 fatalities, while 2014 experienced the fewest fatalities at 194, when considering the total number of accidents in each respective year.Intriguingly, the year 2015 witnessed the highest overall frequency of road traffic accidents of all types.Among these, slight injury accidents constituted the majority at 60.94%, followed by property damage-only (POD) accidents at 23.74%, with fatal accidents representing a minority at 3.94%.Serious injury accidents comprised 11.37% of the total.It is noteworthy that while both accidents and fatalities per accident exhibited an upward trend from 2014 to 2016, they exhibited a downward trend in 2017 and 2018.By 2019, there were 2,242 accidents, resulting in 0.122 deaths per accident.

3.2.1.1.
Geocoding.Safety analysis relies on crash data to identify hotspot areas, assisting practical accident reduction by identifying frequent severe injury and fatality areas.However, datasets are rarely complete and often require pre-processing.Hence, to apply the data in spatial analysis, geocoding techniques are used to locate more than 7000 accident sites in a dataset containing inaccuracies and missing accident location coordinate data.Geocoding is the process of converting addresses to geographic information, such as latitude and longitude, to map their locations.Methods used for geocoding include ArcGIS Pro Geocode Addresses Tools, geocoding with Python libraries such as Geopy and Geopandas, and geocoding with Google Earth and Google Maps.However, each method has its limitations, and the accuracy of the results depends on the quality of the input data.

Road networks data
The road network data, sub-city, and Addis Ababa boundary shapefiles map data were collected from the Addis Ababa City Administration (AACA) and Addis Ababa City Roads Authority (AACRA).However, due to issues related to road segmentation problem with the collected road network data, we decided to generate a new road network of Addis Ababa city using Google Open Street Map (OSM).The analysis utilized a linear network consisting of 4,584 vertices and 4,647 road segments, with a total length of 1,455 km, and included various road types.The segmentations of the road network are done by using network analysis technique which is explained in Section 4. The precision of the generated road network from OSM was verified using shapefiles data from the AACA, World Topographic Map, groundtruthing Orthophoto map, and GPS points and found fitting.Hence, we also used the AACRA road network for checking the spatial reference correctness of the generated OSM.

Data validation
After the completion of the database preparation, a data validation procedure was implemented to verify the consistency and compatibility of the road traffic accidents (RTAs) point locations, OpenStreetMap (OSM) road network, and shapefiles obtained from various sources with the actual ground conditions.This verification was achieved by utilizing Google Earth, GPS coordinates, orthophoto maps, city shapefile data, and the World Topographic Map.

Network analysis
The original road network file, which was collected from AACRA, encountered compatibility challenges in relation to road segmentations when subjected to analysis using spatial analysis tools.As a result, there was a need to adapt to a new data format.A spatial network analysis tool is utilized to segment the road networks and build a road network dataset, facilitating the identification of accident hotspots.Visualizing these hotspots on segmented maps through spatial analysis assists decision-makers in comprehending concentrated problem areas.

Building road network
Following the establishment of the road network using OSM in QGIS, these files were imported into ArcGIS Pro to facilitate network construction.This stage plays a pivotal role, serving as a crucial preliminary step before implementing spatial statistical and crash rate analysis functions.These analyses conclude in the identification of accident hotspot areas and segments, achieved by merging traffic accident location points with corresponding road network segments.The construction of road networks occurs within the Network Analysis tool, involving the reinstatement of network connectivity and attribute data within the dataset the linkage of segments node to node.Notably, post attribute edits in the source feature class, the network's connectivity is exclusively restored in the edited regions to streamline the construction procedure.

Crash rate analysis
Crash Rate Analysis is a method used to identify the frequency and locations of traffic accidents in a particular area.The Crash Rate Analysis tool is used to identify high-risk areas and develop effective strategies for reducing the frequency and severity in terms of accident types (fatal, severe, or minor injury) and the impact of road traffic accidents.The results of the analysis can be visualized using various spatial analysis tools, which can aid transportation planners and policymakers in making informed decisions regarding road safety (Esri, n.d..).Analyzing traffic safety is crucial when evaluating road networks.A significant part of the safety assessment is analyzing accidents occurring at intersections and along roads to identify areas with a high number of severe injuries and fatalities.To enhance safety, data-driven techniques and frequent traffic crash analysis are required.

Hot spot analysis (getis-ord Gi)
Spatial statistics use mathematical computations that directly incorporate space and spatial relationships to quantify spatial distributions and relationships.This helps minimize subjectivity, identify patterns and trends, and make confident decisions based on more than just visual analysis.Spatial statistics tools provide quick and valuable information to better understand data and make informed decisions (ESRI, 2019).It is crucial to consider the spatial element in the study of accident analysis because every accident location is spatially dependent (Atumo et al., 2021;Colak et al., 2018;Polat & Durduran, 2011;Ziakopoulos & Yannis, 2020).Hence, it is important to select spatial statistics method, which is a mathematical approach that incorporates spatial relationships into data analysis to identify significant hotspots of high or low values for traffic safety spatial analysis.Spatial statistical Getis Ord analysis is a more advanced technique that takes into account the spatial distribution of crashes and identifies clusters with high or low crash rates in a given area.This technique uses statistical methods to determine whether the observed distribution of crashes is significantly different from what would be expected by chance.However, the application of Getis Ord analysis provides a more comprehensive understanding of the spatial distribution of accidents and can uncover accident hotspots that may go unnoticed when using other methods.As Ouni and Belloumi (2019) discussed in their study that the identification of hotspots increased the potential to look at a given main road by verifying hazardous segments.A feature must have a high value and be surrounded by additional features with high values in order to be a statistically significant hot spot.This model creates an Output Feature Class with z-scores, p-values, and a confidence level bin (Gi*Bin) field for each feature in the Input Feature Class (ESRI, 2022).According to ESRI (2019), the Getis-Ord Gi* model identifies hot spots using three different confidence level bins, with +3 representing a 99% confidence level, +2 representing a 95% confidence level, and +1 representing a 90% confidence level.These values are used to determine areas with statistically significant clustering of high or low values based on the model's input data and parameters.The toolbox for Spatial Statistics provides a range of statistical methods for examining patterns, processes, relationships, and distributions in a spatial context.One of the key advantages of these tools is their ability to account for spatial factors directly in their calculations.This feature enables the toolbox to identify important characteristics of spatial distributions, identify clusters and anomalies, evaluate patterns of clustering or dispersion, group features based on attributes, choose an appropriate analysis scale, and investigate spatial relationships.Gi, a z-score for each feature in the dataset, can be calculated using Equation (1) (ESRI, 2014).
Calculations-the Getis-Ord local statistic is given as: Where: x j is the attribute value for feature j, w i;j is equal to the spatial weight feature i and j, n is the total number of features, � X and S are given by Equations ( 2) and (3) below.
Gi* statistic produces a z-score for each feature.Positive and significant z-scores indicate greater concentrations of high values or hot spots.Mapping these clusters displays the locations and sizes of hot and cold spots, answering the question of where spatial clusters are as follows (Fischer & Getis, 2009).

Point density analysis
In this study, the determination of traffic accident spot density involves the utilization of point density analysis.The process begins by identifying every point on a road, followed by the application of the Spatial Join tool to determine the number of accidents that have taken place along each road segment.
According to Haghshenas et al. (2018), density analysis plays a crucial role in the combined crash rate analysis and spatial statistical Getis Ord analysis of hotspot methods.This entails calculating the accident density within a specific area.After locating all the points on the road network, the Spatial Join tool is used to count the number of accidents associated with each road segment.This information proves beneficial for crash rate and spatial statistical analysis, serving as one of the input parameters.

Analysis and results
It is crucial to assess and analyze the data that has been collected in order to pinpoint areas that currently provide a safety risk or that could do so in the future.Then, safety-improving solutions can be chosen.Hence, Hotspots spatial analysis, Crash Rate Analysis, and statistical analysis have been performed to identify road traffic hotspots on the collected data of Addis Ababa city roads as they are being discussed in the following sections.Combining spatial statistical analysis and crash rate analysis identifying hotspots for traffic accidents can be done using Getis Ord analysis.This method has been used in numerous studies to determine high-risk locations and enhance traffic safety.Results showed that there are locations where there is an unacceptably high risk of accidents and it can be noted that the location could be a spot, a segment of the road, or a range in the road networks.

Crash rate analysis result
This analysis allows identifying high-risk areas by highlighting locations with high crash rates.Hence, using the historical traffic accident data of Addis Ababa city, the result in Figure 4 shows the analysis of crash data and crash rate layer that defines the frequency and rate at which crashes occur along a roadway section or at an intersection.The results of the analysis indicate roads that have the highest frequency of crashes along the road segments.These roads are symbolized with red, while those with low significant crash segments are symbolized with yellow.The symbolization is based on crash frequency, which is measured as crashes per kilometer per year.This analysis can be referred to as the high injury network or roadways in Addis Ababa that represent roads most frequently involved in accidents.
The analysis and the resulting map in Figure 5 show areas with high crash frequencies.These areas are represented by yellow-red colored road segments of Addis Ababa road networks.The first class with yellow color represents areas with a low crash rate, indicating a relatively safe road segment.The second and third classes are from moderately low crash rate and with a moderate crash rate.Classes 4 and 5 indicate relatively high crash rate and the highest crash rate, highlighting areas of extreme concern called hotspots.These are locations with a significantly higher crash rate than the average and deserve immediate attention.The hotspots are primarily located in highly populated areas, areas with high traffic volume and road networks that encompass the majority of the market and commercial centers.Specifically, these areas are concentrated in the sub-cities of Addis Ketema, Arada, Lideta, and Kirkos, which are four out of the 10 sub-cities in Addis Ababa city.The results of this analysis help to assess and map the relative safety of different areas, identify high-risk locations, and make informed decisions about road safety improvements.

Spatial Statistic Analysis (Hotspot Analysis) result
In this analysis, road traffic accidents occurring within a minimum and optimal fixed distance band of 0.4 km are assigned weightage to identify hotspots.Various researchers have employed distinct distance band lengths due to the absence of a specific predetermined range for the fixed distance band in prior research endeavors focused on identifying traffic accident patterns and hotspot areas (Atumo et al., 2021).In prior research, arbitrary fixed distances have been employed to partition the road network for spatial analysis, without considering variations in traffic speeds (Zahran et al., 2017).Results depicted the association of each crash with the underlying road segments and looked for crash hot spot locations.The final outcome of the analysis is displayed in Figure 6 using various colors to signify the degree of accident propensity in various areas.The study analyzed hot spots in Addis Ababa using Spatial Statistic Hotspot Analysis, where red areas represent high values or point densities and blue areas represent cold spots.The darkest red features indicate strong clustering of high values with 99 percent confidence that the clustering is not a random chance.The confidence levels in Spatial Statistic Analysis provide a measure of the strength of evidence for identifying hotspots.The higher the confidence level, the greater the statistical certainty that the observed clustering of accidents is not a result of random variation, making the hotspots more reliable and actionable for road safety.The analysis was used to identify accident hotspots by taking into account both the weight of each accident and the spatial weights of the road network.
After processing data using hotspot analysis, it resulted with the identification of 33 road segments, 3 intersections, and 10 roundabouts/squares hotspots in terms of their vulnerability to highly frequent road traffic accidents as shown in Figure 5. Compared to all other hotspots identified, the road segment called Djibouti Street, which starts from Bole Edna Mall roundabout to 22 "mazoriya" roundabouts, was found to be the most accident-prone segment in the Addis Ababa road network.This segment had 212 accident counts and an average of 37.15 crashes per km per year over a length of 1141 m.
Overall, hotspot analysis of road traffic accidents has proved useful in identifying vulnerable road segments, intersections, and roundabouts/squares in terms of highly frequent road traffic accidents.The findings from the researches demonstrated that driving behavior, ability, and demographic factors (such as age, driving license length, employment, education, gender, and the season in which accidents occur) and driving skills have a direct impact on road traffic accidents (Shirmohammadi et al., 2019).The age group of drivers is a key variable influencing the frequency of traffic injuries, as revealed by research conducted by Tewolde (2007) in Addis Ababa.Hence, the findings of this study can inform policymakers, city planners, and safety engineers on where to focus their efforts to reduce the occurrence of accidents and improve road safety.In addition, the spatial analysis of traffic accidents in Addis Ababa's road networks, as shown in Tables 1 and 2, tells road segments, intersections, and roundabouts/squares with a high traffic accident rate, referred to as hotspot locations, along with their corresponding length and pavement type.
The resulting map in the following figure, along with Tables 1 and 2, will assist residents in the community and road users in identifying streets, roundabouts, and intersections along their daily commute that carry a higher risk of traffic crashes in Addis Ababa's road network.This information will enable them to take proactive safety measures.
In contrast to the pointmap, a heatmap has the ability to offer greater clarity in regions where multiple points are clustered together (Zahtila & Knura, 2022).A density heat map is a type of thematic map that uses color shading to represent the density of point features in a given area.The density heat map of road traffic hotspots provides valuable information regarding the concentration of road traffic accidents in a specific area.Figure 6 displays a visual representation of accident density across the road network, where darker shades of color indicate areas with high density, while lighter shades represent areas with low density.Within the identified accident hotspots, point features indicate the precise locations of these accidents.The density heat map highlights that the central part of the city, yellowish color, is more susceptible to road traffic accidents compared to the outer areas, light green, as shown in Figure 6.This validates the precision of the spatial analysis of hotspots resulting from the background shown in the density heat map.This information can be utilized to identify accident hotspots and guide decision makers in prioritizing road safety improvements and other necessary interventions.

Discussions
This study employed crash rate analysis and spatial statistical analysis hotspot identification to assess the road traffic accident (RTA) patterns within Addis Ababa's road network.By crash rate analysis Class 4 (29.581280-60.582760) and Class 5 (60.582761-1511.380220)have been identified as areas falling into the higher classes, signifying potential hotspots.The highest priority should be given to areas in Class 5 because they have the highest crash rates and represent the most critical hotspots.Class 4 areas should also be a focus for safety improvements, aiming to reduce crash rates and enhance road safety.It is also important to continuously monitor and evaluate the identified hotspots.In spatial statistical analysis, confidence levels indicate the reliability of identifying RTA hotspots.Results in Figure 5 show that 90% confidence level suggests reasonably significant clustering, with a 10% chance of randomness.At 95% confidence, hotspots have a higher reliability, with only a 5% chance of being random.The most robust findings occur at 99% confidence, where hotspots are highly unlikely to be random (only a 1% chance).Hotspots, areas with statistically significant clustering of accidents, are of utmost importance.The 99% confidence hotspots represent the strongest findings, demanding immediate attention and resource allocation.The figure underscores the significance of targeted safety interventions and continuous monitoring to effectively reduce accidents and improve road safety.
Crash rate analysis and hotspot analysis (Getis-Ord Gi*) are distinct methods in spatial analysis.While they operate independently, they can complement each other.Crash rates from the former can be incorporated into the latter to identify significant accident higher-risk areas.Conversely, Hot Spot Analysis results can guide the focus of a more detailed crash rate analysis as shown in the following figure.These methods were applied independently but produced overlapping results as shown on Figure 7.The common high-crash segments represent areas of agreement and validation between the analyses indicated by arrows in Figure 7. High-crash road segments from crash rate analysis can help identify areas with frequent accidents, and hotspot analysis can identify areas with statistically significant clustering (hotspots) of accidents.This combination and iterative analysis offer the reliability of the findings and the accuracy of hotspot identification, and a comprehensive view of traffic accident patterns, targeted safety interventions, and improved decision-making insights.The spatial analysis results, shown in Figure 4, have identified clusters and hotspots of road traffic accidents (RTAs).Table 1 presents a detailed highlight of streets and road segments within Addis Ababa city, offering insights into hotspot segments, road classifications, pavement types, and segment lengths.This data serves as a valuable resource for stakeholders such as city planners and transportation officials, aiding in the identification of areas requiring maintenance, road surface improvements, traffic management enhancements, and prioritization of construction projects with a specific focus on high-traffic hotspots road segments.
Furthermore, Table 2 presents crucial details on hotspot segment lengths at various roundabouts and intersections.Notably, the study identified 33 road segments, 3 intersections, and 10 roundabouts/squares as hotspots with a significant frequency of accidents.Some of the identified hotspots agree with a previous study by Aychew (2020) in Yeka Sub city, but not much with Habte's (2017) study which used the KDE method in Gulele Sub city for 3 years of road traffic accident data.In contrast, this study used 5 years of data from all Addis Ababa sub-cities and conducted a spatial analysis using Getis-Ord Gi* statistic.However, four hotspots were located in Yeka sub-city, consistent with a previous study by Aychew (2020) focused on that area.Djibouti Street emerged as the most accident-prone segment, with 212 accident counts and 204 crashes per mile per year.Leveraging these hotspot findings, road managers and administrators can prioritize targeted interventions for improved road safety.The density heat map in Figure 6 offers multifaceted advantages in visualizing hotspot outcomes, encompassing intensity, high-and low-density zones, overlapping hotspots, subtle spatial patterns, and dynamic, interactive visualization capabilities.Simultaneously, the segment length data in Table 2 empower traffic planners and policymakers to allocate resources efficiently, with Mexico Square, Gofa Gebriel Square, and German Square Area Residential roundabout exhibiting the longest hotspot segment lengths.
The study's crash rate map (Figure 4) and hotspot analysis map (Figure 5) harmoniously identified high-risk regions, verified by their alignment with high-crash rate segments and hotspot clusters, as vividly portrayed in Figure 4's density heat map.This density map suitably conveys the concentration of road traffic accidents, distinguishing high-density zones in darker hues and lower-density areas in lighter shades.The combined results of these two methods provide a comprehensive approach to road traffic accident analysis.Spatial statistical analysis identifies accident hotspots at different confidence levels (90%, 95%, and 99%), emphasizing the statistical significance of road segments, while spatial crash rate analysis categorizes road segments into five classes based on crash rates, allowing for the prioritization of safety measures in higher-risk areas.This holistic approach ensures that safety interventions are not only targeted but also statistically validated, eventually leading to more effective accident reduction and improved road safety.Overall, this study furnishes valuable insights for traffic planners, policymakers, drivers, and pedestrians.Despite limitations stemming from traffic volume data and other factors, this study underscores the validity of identifying accident hotspots through a combination of spatial statistical methodologies, crash rate analysis, and hotspots and heat map visualization.This approach carries significant implications for future research, highlighting the potential benefits of combining spatial statistical methods, crash rate analysis associating traffic accident spatial data to their corresponding road segments, hotspots, and heat map visualization for enhanced result interpretation in improving road safety.

Conclusions
Studies have shown that hotspot identification techniques can enhance traffic safety.The field of traffic safety analysis has advanced with the emergence of modern statistical techniques for linear networks.This study utilizes such techniques to analyze a geocoded dataset of road traffic accidents that occurred in Addis Ababa from 2014 to 2019, utilizing a roads network for analysis.This study introduces a new approach that combines the Getis-Ord statistic with crash rate analysis, linking spatial data pertaining to road traffic accidents with the respective road segments, along with their corresponding accident counts.This approach efficiently identifies and visually maps hotspots in Addis Ababa's major road network, emphasizing their role in improving road safety.It also enhances the road safety conversation by seamlessly integrating spatial analysis to tackle traffic-related challenges in Addis Ababa, Ethiopia.Hence, the traffic accident spatial analyses resulted in the identification of a total of 46 hotspot locations on the Addis Ababa road network.Among these, 33 are situated on different sections of the road network, while 13 are located at roundabouts/intersections. Following a comparison of the number of accidents occurring in each segment of the major road network in Addis Ababa, it was determined that Djibouti Street, which extends from the Bole Edna Mall roundabout to the "22 Mazoriya" roundabouts, is the most accident-prone hotspot segment on the Addis Ababa road network, recording a total of 212 accident counts.The significance of employing both methods in this study becomes evident through the overlap of shared high-crash segments, which symbolize areas where both methods concur and validate one another.This strengthens the reliability of the study's outcomes and the precision of hotspot identification.Combining crash rate and spatial statistical analyses enhances our understanding of road traffic accident patterns, pinpointing areas for targeted safety interventions.Spatial statistical analysis identifies hotspots with confidence, while crash rate analysis categorizes road segments by risk, enabling precise and effective accident reduction measures for improved road safety.To facilitate enhanced interpretation, the study employed a hotspot and heat map visualization strategy.
This study identifies accident hotspots on Addis Ababa roads, aiding safety prioritization for transportation authorities and road users.Challenges include integrating spatial methods and ensuring data accuracy.To enhance road safety, we propose using GPS and GIS systems for real-time data collection.Additionally, we suggest establishing government-run health centers near hotspots for rapid first-aid assistance and reducing fatalities.Future research can explore low and non-significant values of Getis Ord statistical analysis and other parameters in accident causes and compare spatial analysis methods for accuracy at different scales.
Figure 2. Severity of RTA by year.

Figure 5 .
Figure 5. Map of hotspots location on Addis Ababa road networks.

Figure
Figure 6.Density Heat map of Hotspots location on Addis Ababa Road Networks.

Figure 7 .
Figure 7.Comparison of hotspots and crash rate maps Analysis.

Table 1 . Hotspots identified junction, and segments
As conditions change or safety measures are implemented, it is essential to assess the effectiveness of interventions and modify strategies accordingly.