Article

Road Surface Temperature and Bridge Prediction Using Random Forest and DNN Regression

Authors:
To read the full-text of this research, you can request a copy directly from the author.

Abstract

Road weather conditions have a major impact on the safety of Oklahoma state Roads. The cold fronts in winter can bring sleet and snow. Due to the temperature sudden drop the ice melts and becomes hazardous as it reduces the vehicles traction which leads to accidents and fatalities. The Oklahoma department of transportation (ODOT) treats the roads before the cold front hits the states without having enough information about the future conditions of the Road Surface Temperature (RST) and the Bridge Surface Temperature (BST). Therefore, we used the machine learning modeling with the meteorological historical data to come up with a 24 hours prediction model for the RST and BST on Oklahoma state roads. In this Study, we are validating the performance of Random Forest and Neural network regression models on road and Bridge surface temperature for 24 hours future prediction; we are using the data from Road Weather Information System (RWIS) deployed by ODOT and the online future data predicted by the Global Forecast system (GFS). The model data features are based on meteorological data collected by RWIS stations along the I-35 Highway of Oklahoma State. Test results showed that the predicted temperatures matched well with the observed Surface temperatures collected from the Stations along the highway within an average of ±4°F. The model will help ODOT to know the road conditions in the next 24 hours and take precautions.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the author.

... Statistical models based on historical data are more commonly used and easier to implement [11,12]. The most common statistical models encountered in the literature use the regression method and facilitate relatively high-quality forecasts [13,14]. ...
Article
Full-text available
This work presents the results of the research on the thermal state of the road surface measured by means of a mobile road condition sensor. A 15 km route circumnavigating the city centre and used by urban traffic was taken as the research area. Sixteen test runs were performed under summer and winter conditions. An analysis of the locations at which the rate of change of surface temperature had extreme values between consecutive test runs (which lasted around 25 min) was carried out. Based on that, critical sections of the test route were identified, and thus points characterised by readings deviating significantly from the mean values. Based on the thermal mapping methodology, an analysis of the rate of surface cooling was carried out. It was shown to be characterised by temporal (as well as spatial) non-uniformity. The rate of change of surface and air temperatures was calculated as a function of time, which was equal to −2.22 and −1.97°C/h, respectively. During the analysis, it came to light that hourly intervals represent the optimal frequency for thermal monitoring of a road network. Exceptions to the above are the aforementioned sections selected on the basis of the mobile measurements, in which permanent monitoring of road conditions via ESS stationary systems are recommended.
Article
Full-text available
Forecasting of road surface and traffic conditions is an important aspect of traffic safety and winter road maintenance, especially in the harsh northern climate. The weather conditions can change quickly, for example, with the onset of snowfall or during rapid temperature variations. A prior knowledge of road weather is important from a public road safety standpoint. Proper consideration of upcoming weather events also helps the road maintenance authorities to attend the roads in an effective and economical manner. In Finland, the Finnish Meteorological Institute (FMI) is duty bound to issue warnings of hazardous traffic conditions to the general public. To strengthen these services towards more efficient estimation of rapidly varying conditions of the road surface at a national scale, a simulation model, RoadSurf, has been developed. As input, the model employs numerical weather forecasts, either directly or after modifications made by meteorologists, as well as observations from synoptic or road weather stations and radar precipitation measurement network. As output, the model produces not only road surface temperature, but also road surface condition classification and a traffic index describing the driving conditions in more general terms, as well as road surface friction. The model has been in operational use since 2000. In addition to the original goal of providing road weather forecasts for the national road network, the model has been used in several other applications, for example, in predicting pedestrian sidewalk conditions and in numerous intelligent traffic applications. The present study describes the road weather model RoadSurf and its main applications.
Article
Full-text available
The main objective of this study is to validate a performance forecasting model for road surface temperatures ranging from 3 to 24 hours in the future using real-time weather forecasts as provided by the Korea Meteorological Administration (KMA). The model used here has been developed by researchers at the Korea Institute of Construction and Technology. The core concept of the model is based on heat transfer between road surfaces and the atmosphere. Preliminary tests were performed through comparing historically observed road surface temperatures obtained from road sensors with predicted temperatures from model runs at a specific site on the CheongWon-SangJoo Expressway. After testing, the model was partially run on the CheongWon-SangJoo Expressway at the Bo-Eun Branch of the Korea Highway Corporation to predict road surface temperatures 24 hours in advance. Test results showed that predicted road surface temperatures matched well with the observed temperatures. In addition, the average (either hourly or daily) difference between the two temperatures lies within ±2°C, indicating that the model performance is reasonable. For a real-time model run, road surface temperatures were predicted only at night as there were no instruments to measure solar radiation in real time along the roads. Nonetheless, a 12 hour forecast performed well, showing the average temperature differences between observed and predicted temperatures of about ±0.5°C.
Article
In 1992/93 the Deutscher Wetterdienst (DWD) introduced SWIS (Straßenzustands- und Wetterinformationssystem) as a nationwide road weather information system. In order to forecast road conditions and road-surface temperature over a time period of 27 hours, an energy balance model is used. The course of the energy fluxes is estimated by the meteorological parameters. It is thus very important that the forecaster provides an excellent synoptic input for the energy balance model. The energy balance model is able to account for five different road site characteristics; i.e. unobstructed road sections with regular and little traffic, respectively, shaded road sections, bridges and road sections within urban influences. To evaluate the model's quality, the technique of hindcasting was employed by running the model for a day in the past (running time 0900 UTC of a given day to 1200 UTC the following day), using measured values of a specific road weather monitoring station complemented by observations from a nearby synoptic station. Preliminary results show that from 1500 UTC to 1800 UTC the mean errors are 1 to 2 K, during the remaining forecast time the mean errors are often less than 1 K; results are even better during clear sky conditions.
Road pavement temperatures and their impact on travel during snow storms
  • Donald M Moore
  • Nws
  • R J Billings
  • Miller
Donald M. Moore, NOAA/NWS, Billings, MT; and R. J. Miller. 2007. "Road pavement temperatures and their impact on travel during snow storms". American Meterological Society.
21-Degree Drop in 4 Minutes: The Biggest Temperature Swings from Winter Storm Oliver
  • Jonathan Belles
Belles, Jonathan. 2018. "21-Degree Drop in 4 Minutes: The Biggest Temperature Swings from Winter Storm Oliver." The Weather Channel. The Weather Channel.