The Maintenance Decision Support System (MDSS) was developed over the last 10-plus years to aid winter maintenance stake-holders when making decisions regarding snow removal activities. The system uses Numerical Weather Prediction (NWP) model data, observations, a road temperature model and AI post-processing techniques to produce optimized weather and road temperature forecasts at specific locations out to 72 hours. These forecasts are then used to recommend snow removal treatments as needed. Data from the system is made available to a Java application which displays the information over the client's domain in an operational setting with automatic forecast updates.

The weather forecasts used to drive the system are generated by the Road Weather Forecast System (RWFS) which is based on the Dynamic Integrated foreCAST system (DICast). RWFS forecasts are generated by first applying a Dynamic MOS (DMOS) linear regression technique to NWP model output and observations, then combining multiple DMOS inputs together as a weighted sum. Regression equations are updated weekly, while input component weights are updated daily. The result is an optimized, automated forecast which adapts to changes in model performance and weather regimes.

Over the years the system has been extended in many ways, but experimental domains have been limited to state Departments Of Transportation (DOT), airports, or other domestic road authorities. In 2010-2011, the system was deployed over Madrid, Spain, using new input data sets, including WRF model data supplied by the University of Cantabria. This paper examines some of the results of this effort.