Spatial Adjustment The CRIS data received was geolocated. For all crashes not geocoded to TxDOT on-system roads, their locations were used as they were received. For crashes that were geocoded to TxDOT on- system roadways, post-processing was performed to adjust their location. This was necessary because CRIS snaps all on-system crashes to the combined centerline of the roadway mainline. This means that there is not a way to spatially differentiate crashes that occurred on different sides of a divided highway, nor crashes that occurred on a frontage road from the mainline highway. This effect was more pronounced on one-way paired couplet highways, where the actual directional centerlines were two physically different roads but the combined centerline was the geographic median and ran through the middle of the block between the two roads. To correct this, attributes from the CRIS crash data, as well as the attributes and spatial properties of the TxDOT roadway inventory, were used in combination. Crashes were subset using the on-system flag and identified as either mainline or frontage roads using the roadway part ID. To adjust the crash locations, roadways were subset from the overall TxDOT roadway inventory as well as identified as mainline or frontage based on record type. The cardinal direction of the roadway segment was also calculated, with the direction of travel being determined by the segment’s roadbed identifier. Crashes were then associated in a one-to-many relationship with these roadway segments based on highway number (attributes present in both datasets) and as part of a search tolerance based off the roadway’s width, except for one-way paired couplets, where the distance was based off the distance between the two couplets. The CRIS data dictionary lists cardinal roadway direction as an attribute within the crash level dataset, however, that attribute was not present within the dataset acquired from the CRIS website. Instead, CRIS used vehicle direction of travel, although this information was not always present within the vehicle level data, and some multi-vehicle crashes had different vehicle directions. For any crashes not matched based on vehicle direction, it was attempted to interpret roadway direction based on directional elements within the street name reported in the crash, i.e., ‘North’, ‘Northbound’, ‘NB’, ‘NBND’, etc. Any crash (still matched on the previously described attribute and spatial criterion) with a difference between interpreted direction and roadway azimuth less than 45° for vehicle-based direction, and 90° for name-based direction (vehicle used a lower tolerance because vehicle directions were provided in 45° increments, i.e. northwest, southeast, etc., while name based directions were assumed in 90° increments, i.e. north, south,
52
Made with FlippingBook - Share PDF online