Segments_PCR_2020_2024

This segmental (line) dataset was developed to support safety planning efforts at the Iowa Depatment of Transportation. The data used in this analysis were for crashes that occurred in the time period between 2020 and 2024. There are previous analyses that were done that analyze data during different timeframes. Initially data was developed for the primary road network only but recent (and this current) analysis does include segment data for secondary and municipal routes. A safety performance function (SPF) is an equation used to predict the average number of crashes per year at a location as a function of exposure and, in some cases, roadway or intersection characteristics. SPFs were developed for 8 categories of primary road segments. Divided High Speed Divided Low Speed Freeway High Speed Freeway Low Speed Undivided High Speed Undivided Low Speed Undivided Multilane High Speed Undivided Multilane Low Speed Generally, SPFs more realistically demonstrate the relationship between crashes and traffic volume. SPFs account for the regression to the mean by using an Empirical Bayes statistical method. The advantages of using this method are more accurately calculating the potential for safety improvement and acknowledging the complex, non-linear relationship between crash frequency and volume. SPFs are regression equations that estimate crash frequency as a function of traffic volume. The predicted number of crashes for a given traffic volume is found on the SPF curve (1). But for any specific intersection, the observed number of crashes (2) is likely to be above or below the predicted number calculated by the SPF. The observed crash count is corrected using the EB method resulting in the expected number of crashes (3) at that location. The difference between the expected number and the predicted number gives the potential for safety improvement (4) also known as potential crash reduction. A high PCR indicates a poorly performing segment of road In general, the primary road segment categories were developed based on the posted speed, if it was access controlled, and whether the route was divided or undivided. Speed was used instead of rural vs. urban since the two categories are historically difficult to define. "High speed" was defined as greater than 45 mph and "low speed" was less than or equal to 45 mph. The statewide road and crash databases were used to determine the locations, traffic volumes, and crash histories. Segment crashes were not included as Intersection crashes in the Intersection SPF/PCR development. Segments with extremely short (< 0.1 miles as suggested by HSM) and long lengths have been excluded from the model development analysis to exclude the outliers and develop a more robust model to better fit the data. The PCR for excluded segments then were calculated utilizing the model equation developed for each category. A segment was placed into one of the 8 categories based on the segment characteristics. An SPF model (negative binomial regression) was developed for each category to calculate the predicted number of crashes (see example equation below): Predicted Crashes/ 5-year = L * exp^(α) (AADT)^β The segment length and weighted average AADTs (from the middle year of the analysis period) were used in the equation to determine the predicted number of crashes on the segment. The Empirical Bayes method was applied to the observed crashes to determine the expected number of crashes at a site. The potential for crash reduction (PCR) was calculated by subtracting the predicted number of crashes from the expected number of crashes along the segment. Each Primary Road (Primary Highway) segment has four PCR rankings which are based on PCR per Year (PCR/Year) values and four based on PCR per Mile per Year (PCR/Mile/Year) values. Due to the variability in segment lengths and to have a normalized comparison among segments, the PCR/mile/year value is typically used: 1. Statewide Primary Road segment ranking for All crashes (KABCO) 2. Category Primary Road segment ranking for All crashes (KABCO) 3. Statewide Primary Road segment ranking for Severe crashes (KAB) 4. Category Primary Road segment ranking for Severe crashes (KAB)本分段（线路）数据集旨在为爱荷华州交通部（Iowa Department of Transportation）的安全规划工作提供支撑。本次分析所用数据取自2020年至2024年间发生的交通事故。此前已有针对不同时间跨度数据开展的相关分析研究。最初本数据集仅针对主干道网络构建，但本次及近期分析均纳入了次要道路及市政道路的分段数据。 安全性能函数（Safety Performance Function, SPF）是一类用于根据暴露量（以及部分场景下的道路或交叉口特性）预测某一地点年平均交通事故数的方程式。我们针对8类主干道分段构建了安全性能函数，分别为：分隔式高速路段、分隔式低速路段、高速公路高速路段、高速公路低速路段、非分隔式高速路段、非分隔式低速路段、非分隔式多车道高速路段以及非分隔式多车道低速路段。 总体而言，安全性能函数能更真实地反映交通事故数与交通流量之间的关联关系。安全性能函数通过采用经验贝叶斯（Empirical Bayes）统计方法，对均值回归效应进行了修正。该方法的优势在于能够更精准地计算安全改善潜力，同时兼顾事故发生频率与交通流量之间复杂的非线性关联。安全性能函数属于回归方程式，可基于交通流量估算事故发生频率。针对特定交通流量的预测事故数可通过安全性能函数曲线（1）获取。但针对任意特定交叉口，实际观测到的事故数（2）往往会高于或低于安全性能函数计算得到的预测值。通过经验贝叶斯方法对实际观测到的事故数进行修正后，可得到该地点的预期事故数（3）。预期事故数与预测事故数的差值即为安全改善潜力（4），也称为事故削减潜力（Potential Crash Reduction, PCR）。事故削减潜力（PCR）越高，代表对应路段的安全表现越差。 总体而言，主干道分段的分类依据为路段限速、是否为出入受控路段，以及道路是否采用分隔式设计。相较于乡村/城市道路分类，我们采用限速作为分类依据，原因在于传统的乡村/城市道路分类往往难以清晰界定。本次研究中，"高速"定义为时速高于45英里，"低速"定义为时速小于或等于45英里。 研究采用全州范围的道路与事故数据库，以获取路段位置、交通流量及事故历史数据。在构建交叉口安全性能函数与事故削减潜力模型时，分段路段的事故数据不会被纳入交叉口事故统计范畴。为剔除异常值并构建更稳健的模型以更好适配数据，本模型构建分析过程中排除了极短（小于0.1英里，符合《公路安全手册》（Highway Safety Manual, HSM）建议）及极长的分段路段。随后，我们将利用各分类对应的模型方程式，计算被排除路段的事故削减潜力。将根据路段特性，将其归入上述8个分类中的一类。 我们为每个分类构建了安全性能函数模型（负二项回归模型），用于计算预测事故数（示例方程式如下）： 5年预测事故数 = L × exp(α) × (平均年日交通量（Annual Average Daily Traffic, AADT）)^β 方程式中将采用路段长度与分析周期的中间年份的加权平均年日交通量（AADT），以计算该路段的预测事故数。通过经验贝叶斯方法处理观测事故数据，可得到某一路段的预期事故数。事故削减潜力（PCR）的计算方式为路段预期事故数减去预测事故数。 每条主干道（干线公路）分段均拥有4项基于年事故削减潜力（PCR/Year）的排名，以及4项基于单位英里年事故削减潜力（PCR/Mile/Year）的排名。由于路段长度存在差异，为实现路段间的标准化对比，通常采用单位英里年事故削减潜力作为排名依据： 1. 全州主干道分段所有事故（KABCO）排名 2. 分类主干道分段所有事故（KABCO）排名 3. 全州主干道分段严重事故（KAB）排名 4. 分类主干道分段严重事故（KAB）排名