Operational bias correction of RWIS pavement temperature forecasts

The forecasting of pavement temperatures for winter road maintenance operations is becoming increasingly important in the effort to increase operational efficiency through better use of equipment and freeze-point depressant chemicals during winter weather events. Road Weather Information Systems (RWIS) facilitates this through use of road weather observations, including pavement and subsurface temperature sensing, coupled with weather forecasts and a heat balance model to predict pavement temperature over the next day or two. Currently, one heat balance model in vogue is METRo developed by Environment Canada which explicitly forecasts pavement temperatures and road conditions assuming only crude plowing activity takes place. As METRo tends to exhibit diurnal biases in its forecast pavement temperatures, an operational correction scheme was implemented to improve forecast accuracy on an hourly basis. The use of weighted and blended METRo model data to generate hourly bias corrections, based on the recent performance of respective model runs as verified by the actual in situ values at RWIS environmental sensor stations, is explored. By applying such corrections using recent performance the general forecast result was improved while stations with unique exposures, such as those in recurring shadow for part of the day due to being in rock cuts or mountainous terrain, are implicitly accounted for; an important consideration considering the high impact solar radiation has on pavement temperature. One challenge was that the METRo forecasted pavement conditions, assuming nothing more than routine plowing, and these are largely dependent on moisture levels on the road and whether or not the concurrent pavement temperature is above, at, or below freezing. When bias corrections changed the sign of the pavement temperature the forecast road conditions needed to be altered as well in order to be logically consistent.