Filter
Select the crash category, analysis window, and data-quality standard.
Spatial traffic-safety prioritization · Washington, D.C.
Infrastructure-aware crash prioritization.
CrashMap 360 integrates crash severity, spatial clustering, traffic exposure, and infrastructure proximity to rank candidate locations for engineering review.
i Configurable for different cities, infrastructure types, and agency priorities.
Top 10 severity clusters30 m
Night crashes beyond the assumed reach of the nearest mapped streetlight.
Actual cluster centroids · severity ranking
Analytical contribution
Infrastructure proximity is integrated into cluster selection and ranking.
Each crash is evaluated against the nearest related infrastructure before clustering. Candidate clusters are then compared using severity, capped severity, crash frequency, and exposure-normalized risk.
Analysis pipeline
Six stages from filtering through engineering review.
The logic remains consistent across modules while the infrastructure type and defensible parameters change.
Measure
Calculate each crash’s distance to the nearest relevant infrastructure.
Screen
Retain crashes occurring beyond a documented infrastructure threshold.
Cluster
Group repeated crash patterns into compact candidate locations.
Rank
Compare frequency, severity, capped severity, and exposure-normalized risk.
Review
Use the rankings to focus field and engineering assessment.
Crash-specific analyses
Night, speeding, and cyclist crash prioritization.
Select a crash type and ranking method. The map and table use the final outputs reported in the study.
Top 10 severity clusters30 m
Night crashes beyond the assumed reach of the nearest mapped streetlight.
Actual cluster centroids · severity ranking
Selected analysis
Night / Streetlights
Night crashes beyond the assumed reach of the nearest mapped streetlight.
1,424night crashes beyond 30 m
- Screen
- Nearest relevant mapped infrastructure
- Cluster
- Complete-linkage spatial grouping
- Rank
- Severity, frequency, capped severity, and exposure
Severity rankCrashesSeverity sumLatitudeLongitude
1 Night / Streetlights31238.828147-77.017848
2 Night / Streetlights31138.889910-76.973359
3 Night / Streetlights41138.852248-76.963688
4 Night / Streetlights3738.906000-77.011737
5 Night / Streetlights4638.881656-77.011924
6 Night / Streetlights3538.820787-77.018431
7 Night / Streetlights3538.920742-76.957523
8 Night / Streetlights4538.900194-76.980346
9 Night / Streetlights3438.869424-76.959220
10 Night / Streetlights3438.881660-77.011502
Configurable parameters
Parameters for local implementation.
Agencies can configure infrastructure types, distance thresholds, clustering rules, severity weights, exposure measures, and analysis windows. The Washington, D.C. study uses the specifications shown here.
Sensitivity analysis
Alternative thresholds are compared to measure how parameter changes affect cluster matches and ranking stability.
ParameterD.C. specificationTested / adaptableBasis
Night / streetlights30 m22.5 m · 37.5 mAssumed lighting reach
Speeding / devices59.4 m44.55 m · 74.25 m · 200 mMedian D.C. block length
Cyclist / bike lanes100 m75 m · 125 mInfrastructure context radius
Severity weights7 · 4 · 1ConfigurableFatal · major · minor injury
Data qualityMAR ≥ 100ConfigurableMPD location-confidence screen
Research and stress tests
Comparing methods and ranking stability.
CrashMap 360 tests how candidate priorities change when the method, threshold, or denominator changes.
Alternative rankings
Severity, capped severity, crash frequency, average severity, severe-crash percentage, and exposure-normalized risk.
Robustness checks
Spearman rank correlation and top-N overlap compare how candidate ordering responds to alternative thresholds.
External comparisons
Complete-linkage results are compared with DBSCAN and DDOT’s corridor-based High Injury Network.
Interpretability
Logistic regression examines characteristics associated with high-severity clusters; it is not presented as causal proof.
View code repository ↗Published paper and citation link will be added after publication.
Scope
Interpretation requirements.
01
Infrastructure proximity identifies candidates for review; field conditions determine the appropriate response.
02
Mapped infrastructure, crash coordinates, and AADT exposure introduce measurement uncertainty.
03
Future work should test whether this prioritization improves engineering decisions or safety outcomes.
Researcher
Sachin Kundra
Sachin Kundra is a student researcher interested in applying technology, spatial analysis, and data science to real-world problems.
He developed CrashMap 360 to examine how crash severity, traffic exposure, spatial clustering, and infrastructure proximity can be combined into a location-prioritization framework for engineering review.
The project and its potential agency applications have been discussed with staff from DDOT and MCDOT.
Project origin
A question prompted by a familiar intersection.
CrashMap 360 began with a question Sachin had carried for years about a fatal school-bus crash at an intersection on his family’s regular commute. The victim was nine years old—the same age as his younger sister at the time. Later, after hearing a discussion about the rise in pedestrian deaths and the concentration of crashes at night, he began asking whether public crash data could reveal patterns that conventional hotspot lists might overlook. That question developed into an infrastructure-specific prioritization framework for Washington, D.C.