{"id":{"repo_id":"vt","oai_identifier":"oai:vtechworks.lib.vt.edu:10919/141304"},"canonical_url":"https://search.dev.ndltd.org/etd/vt/oai:vtechworks.lib.vt.edu:10919/141304","repository":{"repo_id":"vt","name":"Virginia Tech","base_url":"https://vtechworks.lib.vt.edu/oai/request"},"display":{"title":"Quantifying the Impact of Pavement Surface Properties on Road Safety: A Data-Driven Approach","abstract":"The Highway Safety Manual's (HSM) safety performance functions (SPFs) are widely used for network screening and project prioritization; however, most formulations focus on volumetric exposure and geometry and omit pavement surface characteristics that govern tire–road interaction. Despite the growing availability of network level surface data such as Skid Friction Number at 40 mph (SFN40), Macrotexture via Mean Profile Depth (MPD), and pavement age and classification, there incremental value within SPFs remain under quantified. This gap leaves agencies uncertain about surface measures materially improving prediction and how large their effects are in practice. This study addresses the established gap by quantifying how adding SFN40, MPD, and Age can affect a model's crash prediction across HSM Functional classifications. A network level dataset of 0.1-mile roadway segments was assembled across selected HSM functional classes, N = 12,474; 14 classes, linking reported crashes to exposure (lnAADT), roadway geometry, and pavement surface measurements. For each class, Negative Binomial SPFs with a log link were estimated: a base specification (lnAADT, Curvature, Cross-Slope, Grade) and an Extended specification that adds SFN40, MPD, and age. Model performance was evaluated using AIC, log-likelihood, RMSE, and Dispersion. Effects are reported as incidence rate ratios (IRR) with 95% confidence intervals, and residual structure was screened using cumulative residual (CURE) plots alongside a simple multicollinearity check. Across 14 functional classes, the extended model outperformed the base in 12 classes, indicated by the difference in AIC median = -2.38, median difference in RMSE% = -2.7% (improved in all 14; range -26.4 % to -0.1%) and significantly LRT in 8 classes. Difference in AIC favored the Extended model in 8 classes (equal in 2 classes, higher in 4). Routine friction/texture measures and pavement age provide measurable predictive gains and should be incorporated into SPF calibration and network screening, with class specific effect sizes guiding surface focused maintenance prioritization. Adding friction/texture to SPFs improved fit where maneuver demand concentrates. Urban arterial intersections (ΔAIC −115.9; RMSE −5.5%) and rural multilane intersections (ΔAIC −61.9; RMSE −2.3%) showed the largest gains; freeway tangents improved modestly (ΔAIC −36.7; RMSE −0.34%). Curves and rural tangents saw negligible benefits, supporting parsimony. Coefficients are reported as IRRs; SFN40 is consistently protective (≈7%, 5%, and 1% lower expected crashes per +1 point at urban intersections, rural multilane intersections, and freeway tangents, conditional on exposure and geometry). These results support selective inclusion of surface variables in agency SPFs.","abstract_html":"The Highway Safety Manual's (HSM) safety performance functions (SPFs) are widely used for network screening and project prioritization; however, most formulations focus on volumetric exposure and geometry and omit pavement surface characteristics that govern tire–road interaction. Despite the growing availability of network level surface data such as Skid Friction Number at 40 mph (SFN40), Macrotexture via Mean Profile Depth (MPD), and pavement age and classification, there incremental value within SPFs remain under quantified. This gap leaves agencies uncertain about surface measures materially improving prediction and how large their effects are in practice. This study addresses the established gap by quantifying how adding SFN40, MPD, and Age can affect a model's crash prediction across HSM Functional classifications. A network level dataset of 0.1-mile roadway segments was assembled across selected HSM functional classes, N = 12,474; 14 classes, linking reported crashes to exposure (lnAADT), roadway geometry, and pavement surface measurements. For each class, Negative Binomial SPFs with a log link were estimated: a base specification (lnAADT, Curvature, Cross-Slope, Grade) and an Extended specification that adds SFN40, MPD, and age. Model performance was evaluated using AIC, log-likelihood, RMSE, and Dispersion. Effects are reported as incidence rate ratios (IRR) with 95% confidence intervals, and residual structure was screened using cumulative residual (CURE) plots alongside a simple multicollinearity check. Across 14 functional classes, the extended model outperformed the base in 12 classes, indicated by the difference in AIC median = -2.38, median difference in RMSE% = -2.7% (improved in all 14; range -26.4 % to -0.1%) and significantly LRT in 8 classes. Difference in AIC favored the Extended model in 8 classes (equal in 2 classes, higher in 4). Routine friction/texture measures and pavement age provide measurable predictive gains and should be incorporated into SPF calibration and network screening, with class specific effect sizes guiding surface focused maintenance prioritization. Adding friction/texture to SPFs improved fit where maneuver demand concentrates. Urban arterial intersections (ΔAIC −115.9; RMSE −5.5%) and rural multilane intersections (ΔAIC −61.9; RMSE −2.3%) showed the largest gains; freeway tangents improved modestly (ΔAIC −36.7; RMSE −0.34%). Curves and rural tangents saw negligible benefits, supporting parsimony. Coefficients are reported as IRRs; SFN40 is consistently protective (≈7%, 5%, and 1% lower expected crashes per +1 point at urban intersections, rural multilane intersections, and freeway tangents, conditional on exposure and geometry). These results support selective inclusion of surface variables in agency SPFs.","abstract_has_math":false,"creators":["Molato, Nacer"],"institution":"Virginia Tech","degree_name":"Master of Science","degree_level":"masters","degree_discipline":"Civil Engineering","degree_department":"Civil and Environmental Engineering","school":null,"contributors":[],"advisors":[],"committee_chairs":["Flintsch, Gerardo W."],"committee_members":["Abbas, Montasir Mahgoub","Trani, Antonio A."],"year":2026,"date_issued":"2026-02-18","date_published":"2026-02-18","updated_at":"2026-07-22T22:19:29Z","subjects":["Friction","Macrotexture","Transportation","Crash","Grip","Interaction","Pavement"],"languages":["en"],"rights":["In Copyright"],"rights_urls":["http://rightsstatements.org/vocab/InC/1.0/"],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:45567"],"render_values":[{"text":"vt_gsexam:45567","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10919/141304","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.committeechair","label":"Committee Chair","values":["Flintsch, Gerardo W."]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Abbas, Montasir Mahgoub","Trani, Antonio A."]},{"key":"dc:contributor.department","label":"Department","values":["Civil and Environmental Engineering"]},{"key":"dc:creator","label":"Author","values":["Molato, Nacer"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-02-19T09:00:08Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-02-19T09:00:08Z"]},{"key":"dc:date.issued","label":"Date","values":["2026-02-18"]},{"key":"dc:publisher","label":"Institution","values":["Virginia Tech"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Civil Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Virginia Polytechnic Institute and State University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Friction","Macrotexture","Transportation","Crash","Grip","Interaction","Pavement"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["In Copyright"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://rightsstatements.org/vocab/InC/1.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:45567"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10919/141304"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The Highway Safety Manual's (HSM) safety performance functions (SPFs) are widely used for network screening and project prioritization; however, most formulations focus on volumetric exposure and geometry and omit pavement surface characteristics that govern tire–road interaction. Despite the growing availability of network level surface data such as Skid Friction Number at 40 mph (SFN40), Macrotexture via Mean Profile Depth (MPD), and pavement age and classification, there incremental value within SPFs remain under quantified. This gap leaves agencies uncertain about surface measures materially improving prediction and how large their effects are in practice. This study addresses the established gap by quantifying how adding SFN40, MPD, and Age can affect a model's crash prediction across HSM Functional classifications. A network level dataset of 0.1-mile roadway segments was assembled across selected HSM functional classes, N = 12,474; 14 classes, linking reported crashes to exposure (lnAADT), roadway geometry, and pavement surface measurements. For each class, Negative Binomial SPFs with a log link were estimated: a base specification (lnAADT, Curvature, Cross-Slope, Grade) and an Extended specification that adds SFN40, MPD, and age. Model performance was evaluated using AIC, log-likelihood, RMSE, and Dispersion. Effects are reported as incidence rate ratios (IRR) with 95% confidence intervals, and residual structure was screened using cumulative residual (CURE) plots alongside a simple multicollinearity check. Across 14 functional classes, the extended model outperformed the base in 12 classes, indicated by the difference in AIC median = -2.38, median difference in RMSE% = -2.7% (improved in all 14; range -26.4 % to -0.1%) and significantly LRT in 8 classes. Difference in AIC favored the Extended model in 8 classes (equal in 2 classes, higher in 4). Routine friction/texture measures and pavement age provide measurable predictive gains and should be incorporated into SPF calibration and network screening, with class specific effect sizes guiding surface focused maintenance prioritization. Adding friction/texture to SPFs improved fit where maneuver demand concentrates. Urban arterial intersections (ΔAIC −115.9; RMSE −5.5%) and rural multilane intersections (ΔAIC −61.9; RMSE −2.3%) showed the largest gains; freeway tangents improved modestly (ΔAIC −36.7; RMSE −0.34%). Curves and rural tangents saw negligible benefits, supporting parsimony. Coefficients are reported as IRRs; SFN40 is consistently protective (≈7%, 5%, and 1% lower expected crashes per +1 point at urban intersections, rural multilane intersections, and freeway tangents, conditional on exposure and geometry). These results support selective inclusion of surface variables in agency SPFs."]},{"key":"dc:description.abstractgeneral","label":"General Abstract","values":["Vehicular accidents are partly random due to human interaction, but they also follow patterns we can measure. This thesis uses statistical models to test whether pavement surface conditions, how much a tire grips the road (friction) and how rough the surface is (macrotexture), help explain and predict where crashes happen beyond traffic and road design alone. Using over 12,000 0.1-mile segments in Virginia, a base model using traffic and geometry and an extended model which adds on roadway surface measurements collected by modern friction testing. Results show that adding surface information improves prediction in most places with heavy braking and turning especially in urban and rural multilane intersections and provides smaller but measurable gains on high-speed freeway segments. In a small low sample curve category, simpler models remained the better choice, underscoring that more variables aren't always helpful. The practical takeaway is straightforward: agencies can use routinely collected surface data to sharpen safety screening and target treatments (like resurfacing or friction restoration) where they will prevent the most crashes. Where the gains are small, agencies can keep a simpler approach. By identifying when surface data adds value, this work helps direct limited safety and maintenance funds to locations where they can do the best."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Master of Science"]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["ETD"]},{"key":"dc:title","label":"Title","values":["Quantifying the Impact of Pavement Surface Properties on Road Safety: A Data-Driven Approach"]}]}],"canonical_facts":{"dc:contributor.committeechair":["Flintsch, Gerardo W."],"dc:contributor.committeemember":["Abbas, Montasir Mahgoub","Trani, Antonio A."],"dc:contributor.department":["Civil and Environmental Engineering"],"dc:creator":["Molato, Nacer"],"dc:date.accessioned":["2026-02-19T09:00:08Z"],"dc:date.available":["2026-02-19T09:00:08Z"],"dc:date.issued":["2026-02-18"],"dc:description.abstract":["The Highway Safety Manual's (HSM) safety performance functions (SPFs) are widely used for network screening and project prioritization; however, most formulations focus on volumetric exposure and geometry and omit pavement surface characteristics that govern tire–road interaction. Despite the growing availability of network level surface data such as Skid Friction Number at 40 mph (SFN40), Macrotexture via Mean Profile Depth (MPD), and pavement age and classification, there incremental value within SPFs remain under quantified. This gap leaves agencies uncertain about surface measures materially improving prediction and how large their effects are in practice. This study addresses the established gap by quantifying how adding SFN40, MPD, and Age can affect a model's crash prediction across HSM Functional classifications. A network level dataset of 0.1-mile roadway segments was assembled across selected HSM functional classes, N = 12,474; 14 classes, linking reported crashes to exposure (lnAADT), roadway geometry, and pavement surface measurements. For each class, Negative Binomial SPFs with a log link were estimated: a base specification (lnAADT, Curvature, Cross-Slope, Grade) and an Extended specification that adds SFN40, MPD, and age. Model performance was evaluated using AIC, log-likelihood, RMSE, and Dispersion. Effects are reported as incidence rate ratios (IRR) with 95% confidence intervals, and residual structure was screened using cumulative residual (CURE) plots alongside a simple multicollinearity check. Across 14 functional classes, the extended model outperformed the base in 12 classes, indicated by the difference in AIC median = -2.38, median difference in RMSE% = -2.7% (improved in all 14; range -26.4 % to -0.1%) and significantly LRT in 8 classes. Difference in AIC favored the Extended model in 8 classes (equal in 2 classes, higher in 4). Routine friction/texture measures and pavement age provide measurable predictive gains and should be incorporated into SPF calibration and network screening, with class specific effect sizes guiding surface focused maintenance prioritization. Adding friction/texture to SPFs improved fit where maneuver demand concentrates. Urban arterial intersections (ΔAIC −115.9; RMSE −5.5%) and rural multilane intersections (ΔAIC −61.9; RMSE −2.3%) showed the largest gains; freeway tangents improved modestly (ΔAIC −36.7; RMSE −0.34%). Curves and rural tangents saw negligible benefits, supporting parsimony. Coefficients are reported as IRRs; SFN40 is consistently protective (≈7%, 5%, and 1% lower expected crashes per +1 point at urban intersections, rural multilane intersections, and freeway tangents, conditional on exposure and geometry). These results support selective inclusion of surface variables in agency SPFs."],"dc:description.abstractgeneral":["Vehicular accidents are partly random due to human interaction, but they also follow patterns we can measure. This thesis uses statistical models to test whether pavement surface conditions, how much a tire grips the road (friction) and how rough the surface is (macrotexture), help explain and predict where crashes happen beyond traffic and road design alone. Using over 12,000 0.1-mile segments in Virginia, a base model using traffic and geometry and an extended model which adds on roadway surface measurements collected by modern friction testing. Results show that adding surface information improves prediction in most places with heavy braking and turning especially in urban and rural multilane intersections and provides smaller but measurable gains on high-speed freeway segments. In a small low sample curve category, simpler models remained the better choice, underscoring that more variables aren't always helpful. The practical takeaway is straightforward: agencies can use routinely collected surface data to sharpen safety screening and target treatments (like resurfacing or friction restoration) where they will prevent the most crashes. Where the gains are small, agencies can keep a simpler approach. By identifying when surface data adds value, this work helps direct limited safety and maintenance funds to locations where they can do the best."],"dc:description.degree":["Master of Science"],"dc:format.medium":["ETD"],"dc:identifier.other":["vt_gsexam:45567"],"dc:identifier.uri":["https://hdl.handle.net/10919/141304"],"dc:language.iso":["en"],"dc:publisher":["Virginia Tech"],"dc:rights":["In Copyright"],"dc:rights.uri":["http://rightsstatements.org/vocab/InC/1.0/"],"dc:subject":["Friction","Macrotexture","Transportation","Crash","Grip","Interaction","Pavement"],"dc:title":["Quantifying the Impact of Pavement Surface Properties on Road Safety: A Data-Driven Approach"],"dc:type":["Thesis"],"thesis:degree_discipline":["Civil Engineering"],"thesis:degree_level":["masters"],"thesis:degree_name":["Master of Science"],"thesis:institution_name":["Virginia Polytechnic Institute and State University"]},"updated_at":"2026-07-22T22:19:29Z"}