Author ORCID Identifier

Year of Publication


Degree Name

Master of Science in Civil Engineering (MSCE)

Document Type

Master's Thesis




Civil Engineering

First Advisor

Dr. Reginald Souleyrette


In the context of transportation safety engineering, network screening is a method of identifying and prioritizing high-risk locations for potential safety investment. Since its release, the Highway Safety Manual (HSM) has facilitated the adoption of Safety Performance Functions (SPF) to predict the number of crashes for the network screening of any facility type. The predictive model becomes more reliable when developed from crash data with homogeneous roadway segments and this homogeneity can be attained by applying specific geometric attributes to the dataset. The caveat to this method is the requirement of adjustment factors (AFs) to adjust the predicted estimate for the segments which have different geometric characteristics compared to the base attributes. Though AFs are available from several sources, particularly the HSM and CMF Clearinghouse, there are still many attributes for various roadways for which the AFs have not been estimated yet. The absence of appropriate AFs limits the use of such crash prediction models for network screening. In that case, a generic SPF can be developed from the entire network without applying any base conditions and, the reliability of the model is compromised. The goal of this study is to evaluate the trade-offs between a more reliable SPF (that requires more AFs) and a relatively less reliable SPF (that requires fewer AFs). This leads to the following question this research attempts to answer: “Are the benefits of AFs for network screening worth the cost of developing them?”

Recommended by the HSM, this study uses “Excess Expected Crashes (EEC)”, a metric derived from the SPF and historical crash data for ranking potential sites for improvement. The study analyses found that segment rank is nearly insensitive to the choice of the SPF and developing AFs may not justify the cost of network screening. On the other hand, an SPF developed from the entire roadway data might not work as well for project-level analysis (a combination of several segments) or estimating the benefit-cost ratio for a site. This is because the magnitudes of the EEC are crucial for such cases and the generic SPF overestimates the EEC compared to SPFs developed from specific sets of attributes. for most of the segments. Therefore, the major finding of the thesis is that a generic SPF is sufficient when sites are needed to be ranked, but specific SPFs perform better when a benefit-cost analysis is required.

Digital Object Identifier (DOI)