Spatial Clustering of Road Traffic Crash Locations on IntercityCorridors Using National Road Safety Records

Abstract

A crash hotspot map is only as good as the coordinates it is built from, and in most of theworld those coordinates record where a responder attended rather than where the crash occurred. We ask what that costs a corridor prioritization decision. Using two intercity corridors,the 54.0 km Ilorin to Ogbomoso trunk in Kwara State and the 92.9 km Osogbo to Ibadan route inOsun State, with segment geometry and junction structure taken from OpenStreetMap and crash totals calibrated to sixteen quarters of published national returns, we reconstruct areference crash intensity for each corridor and score against it the record that severity-dependentunder-reporting and displacement toward response landmarks leave behind, at rates taken frompublished comparators and swept across their plausible ranges. Three results follow. At the calibrated reporting rate and displacement scale, under-reportingcosts more of the ranking than displacement does: with 28 percent of crashes entering the record,the observed ranking recovers 65 percent of the true top decile on one corridor and 77 percent onthe other before displacement acts at all. The two corridors then part company under displacement, Osun losing a further0.124 of its top-decile overlap where Kwara loses 0.003, so the damage is a property of thecorridor rather than of the reporting process. And displacement resisted detection from the observed records alone. Each of five detectors withuseful power rejected undisplaced records at 12 to 49 percent against a nominal 5 percent, because the truerisk process concentrates at the same junctions and settlements that displacement pulls toward.Detection becomes reliable only when calibrated against known true locations, which is preciselywhat an analyst working from official records does not have. We conclude that positionalprovenance should be published alongside crash counts, because no amount of spatial statisticsrecovers it after the fact.