Creation of an All-weather Road Impact Prediction System (ARIPS) [Summary]

Details

• Creators:
  Reeves, Heather D. ; Rosenow, Andrew A. ; Duarte, Jorge
• Corporate Creators:
  Cooperative Institute for Severe and High-Impact Weather Research and Operations (CIWRO)
• Corporate Contributors:
  Missouri. Department of Transportation. Construction and Materials Division ; United States. Department of Transportation. Federal Highway Administration
• Subject/TRT Terms:
  Flash Flooding Floods Machine Learning Rainfall Street Closure
• Resource Type:
  Brief
• Geographical Coverage:
  United States
• Edition:
  Research Summary
• Corporate Publisher:
  Missouri. Department of Transportation
• Abstract:
  This project is designed to examine the utility of using the database of flood-induced road closures collected by MoDOT to develop an Artificial Intelligence/Machine Learning (AI/ML) model. This was accomplished by assembling a database of predictors. These predictors include quantitative precipitation estimates, hydrologic model output, and a series of static fields relevant to flooding, such as ground permeability, elevation, and slope. These disparate sources of data were assembled into an AI/ML ready dataset. The second component of the project was the evaluation of the utility of this dataset for the task of training an AI/ML model. In general, the bulk evaluation of the dataset showed promise for this task. Two important shortcomings were identified as part of this process. First, the closure database records only the day of the closure, which is a challenge for flash floods, which occur on the time frame of hours. Secondly, the MoDOT dataset represents only events where flooding occurred - for a model to predict a probability of an event, it must have a representation of times where the event did not happen to learn the difference between events and non-events. Thus, a database of non-flooding rainfall is needed alongside the MoDOT closure database. Both of these identified issues can be rectified in future research efforts using the database assembled in this initial project.
• Format:
  PDF
• Funding:
  TR202513
• Collection(s):
  US Transportation Collection
• Main Document Checksum:
  urn:sha-512:1010421de8dc820b46c42e1f736849dd4ce4e78d3ef9ba453d2670df66951a1bce30887582f382da559d363c3049008f7119cf74803080b193d39aa79f3d4956
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/89401/dot_89401_DS1.pdf
• File Type:
  [PDF - 387.04 KB ]