Multihazard Exposure Assessment on the Valjevo City Road Network

Abstract Valjevo City is one of the regional centers in western Serbia. Its road network is important for the local economy, which is based on light industry, mining, and agriculture. Over the years, its road network was exposed to various natural hazards. Due to the hilly/mountainous terrain configuration, it occasionally suffers from landslides and flash floods, but it especially suffered great damage in the 2006 and 2014 floods. In the present work, we have attempted to allocate critical sections along the Valjevo road network for the following three natural hazards: landslides, riverine floods, and flash floods. The area was firstly mapped for the corresponding hazards, based on heuristic and deterministic methodology, wherein various spatial factors were included and scored for their influence. The hazard exposure criteria for each hazard type were determined by field investigations, including estimations of average runout distances for landslides, and distance buffers along the potentially flooding streams and rivers. When compared against the available inventories, the flood and landslide hazard model had substantial accuracy in mapping very high hazard zones (accuracy>80%), while the flash flood model did not perform very well, (accuracy>30%). However, its reference inventory is incomplete, which might explain the poor performance. The calculation of the exposure required overlaying of all three hazard raster models by the road network vector, which included only the official state roads, with 43 links, 300 junctions, and was 265 km long in total. Road links were then split into smaller 500-m segments for a more detailed overview, which opened a nonstandard segmentation and raster-to-vector overlaying problems, solved externally in R, and then reintroduced back into the geographic information system (GIS). The exposure value was summed across each of these segments and visualized in typical green-to-red color coding to highlight the most exposed parts. This was repeated for all three types of hazards and, finally, the multihazard exposure map of the road network was obtained by weighted averaging of these individual hazard exposures. It is successfully demonstrated that the segmented map gives a better insight into realistic road exposure to these hazards, as it resulted in about 16% or about 40 km of very high exposure, whereas the link map overestimated the exposure to 39% or about 100 km. Both maps, however, can find their practical applications in appropriate road planning, mitigation, and mainstreaming emergency resilience.