PIARC 2026: Digitalisation of ASFINAG Winter Maintenance
- Mar 19
- 11 min read
Vienna, March 2026:
Predictive Winter Maintenance of ASFINAG:
„Predictive winter maintenance aims to prevent critical road conditions by optimising timing, clearing and dosage of deicer before, during, and after events based on forecasts, significantly reducing salt usage and increasing road safety for all users"
PIARC 2026 - World Winter Maintenance Congress in Chambery, France:
The World Congress on Road Winter Service, Resilience and Decarbonisation is a world-class event of the World Road Organisation PIARC and is being held every four years. The 17th Congress was held in Chambery, France, from 10-13 March 2026 and was an excellent opportunity to showcase our achievements in Winter Maintenance Research and Digitalisation on Highways in Austria with ASFINAG. Summarising 15 years of R&D in Winter Maintenance, we successfully submitted 5 papers and had 5 Lectures and 5 Poster Presentations highlighting key results.
Paper #1: Predictive Winter Maintenance
The paper presents a comprehensive approach to modernising and optimising winter road maintenance in Austria. It outlines the development, implementation, and improvements of predictive maintenance strategies, integrating advanced weather forecasting, road condition modelling, and real-time data analysis to enhance safety, reduce costs, and respond proactively to changing climatic conditions. Winter road maintenance is vital for traffic safety and accessibility during adverse weather conditions, accounting for climate change. In Austria, even short disruptions due to snowfall can cause economic losses exceeding the costs of WM in an entire winter season. Compared to these developments, traditional reactive approaches are increasingly inefficient, as shown in the research program WinterFIT.
At the heart of Austria’s predictive maintenance system is a holistic decision support platform called Weather 2.0. It covers the entire 2,250 km ASFINAG-managed highway network, using high-resolution nowcasting models (+12 hours) with hourly updates and accurate treatment recommendations. The core innovation of the system lies in its integration of physical models, such as thawing, freezing, residual salt behaviour, and road runoff with traffic data and road-specific parameters like curvature, texture, and rutting. This allows for accurate predictions of skid resistance over time and helps determine the most effective maintenance strategies in terms of safety and cost. All treatment effects are simulated in hourly increments, considering direction-specific variations of maintenance runs.
A major technical achievement is the system’s ability to accurately predict road surface temperatures and humidity levels, using ensemble modelling techniques. By leveraging “thermal fingerprints” of roads and accounting for local topography (e.g., bridges and tunnels), forecast precision has improved significantly compared to the state of the art, with a standard error of just 1.1°C for air temperature and 1.45°C for road surface temperature at the +12-hour horizon. Real-time monitoring of maintenance runs is another critical feature. As trucks transmit data on speed, salting dosage, brine share, and clearing activity, being used to validate forecasts, avoid redundant coverage, and support benchmarking (Figure 1).
Paper #2: The Importance of KPIs in Winter Maintenance
Efficient winter maintenance (WM) balances available resources to achieve high accessibility and road safety with the least possible costs and environmental impact. Regardless of the training efforts and provided guidelines, the winter maintenance staff has the last decision in adapting general strategies to the observed situation on the road, leading to substantial deviations in strategies and results. However, for any systematic improvement, an objective basis for a comparison of the winter maintenance performance is needed that accounts for regional and seasonal changes in winter weather, as well as other possible effects like climate change. Consistent benchmarking with key performance indicators (KPIs) allows road operators to compare strategies and results over time, normalising for weather events and climate trends. As WM is influenced by many factors, there is no single KPI that covers all relevant aspects, but rather a comprehensive framework of KPIs covering different aspects for specific improvement.
With the implementation of a holistic winter maintenance model with a validated real-time prediction of the road condition into the MDSS of ASFINAG called Weather 2.0 for the first time, such a consistent basis is given. The nowcasts in this system allow an optimisation and comparison for any strategy at any given time for any event type, as well as an assessment of their costs and resulting road safety. The paper provides an overview of the developed benchmarks and selected KPI’s for an efficient winter maintenance on highways in Austria. The developed benchmarking system and KPIs will have the data of two to three previous seasons for comparison and will go live in Winter 2025/26 with daily updating.
The first field of the benchmarks is related to the accuracy of the sensors and weather forecasts. The second field is related to the classification of weather events like snowfall, hoarfrost, freezing rain, rain, and dry periods, with the number of events as well as their duration and intensity. The third field provides insights into the actual winter maintenance normalised by event type, with salt and brine usage as well as the number of runs per event and event hour. The fourth field provides insights into efficiency and costs with a focus on winter maintenance efficiency, comparing no winter maintenance with actual and optimised strategies as well as their safety gains. All benchmarks are calculated automatically and are available for any route, time frame, maintenance depot and region.
In summary, the developed holistic winter maintenance model and decision support system, Weather 2.0 provide a sound basis for systematic benchmarking in all relevant areas. Nonetheless, the implementation of the benchmarking system with automatic calculation of KPIs on a daily basis is the key to moving from a functioning system to a framework that allows constant improvement of both models and winter maintenance practice, both for specific events and in general. The coming winter seasons, along with the benchmarking results, will reveal the specific areas of improvement of WM on highways in Austria and help to realise it (Figure 2).
Paper #3: Optimisation of winter maintenance strategies with real-time prediction:
Common winter maintenance relies heavily on the experience and dedication of its staff in the adaptation and application of its strategies into practice under ever-changing conditions. With the increasing availability of sensor data and more accurate local weather forecasts, the decision-making has shifted from a practical, reactive approach towards a predictive one. Nevertheless, the sensor data as well as the weather forecasts still have to be interpreted on a case-by-case basis, leading to a large deviation of strategies for similar situations. In this situation, without an objective basis for comparison, the development of innovative strategies is problematic. Furthermore, without such an objective basis, it is very hard to convince the winter maintenance crew to adopt innovative approaches and findings, as they will always fall back to their own approaches from experience.
The goal of the research program WinterFIT was to change this situation, providing a basis for an objective comparison and optimisation of all possible strategies, relying on a real-time prediction of weather parameters and road conditions with and without winter maintenance. In this paper, the focus is on the optimisation of winter maintenance strategies for any given event. For optimisation, several criteria and target functions are being presented, considering both the improved skid resistance for a given strategy compared to doing nothing, as well as the costs of these strategies. The strategy generation for optimisation allows the combination of any number of consecutive runs for a given cycle time, share of brine and application rate for +12 hours in one-hour time increments.
In the paper, the optimisation of winter maintenance strategies for a severe snowfall event and a hoarfrost event is provided in detail, showing exactly under which conditions which application rate, brine share, time of application or cycle time are optimal and why. Nonetheless, the winter maintenance model is capable of optimising strategies for any given constraints and event type, being a reliable basis for comparison and improvement. Beyond the optimisation of winter maintenance strategies based on nowcasts, the winter maintenance model also allows back-casting, using data from actual events. With this, it is possible to calculate benchmarks, compare to implemented strategies and reconstruct road conditions in case of severe accidents.
With the implementation of the holistic winter maintenance model on the entire highway network in Austria since 2023/24, it is now possible to predict the road condition and optimise the winter maintenance in 250 weather sections and on 136 routes. The prediction and optimisation are repeated every hour, taking into account previous treatment runs as well as changes from the sensor data and the weather predictions. The model runs in the background, providing nowcasts on all sensor system locations and road weather sections as well as the treatment recommendations with timing, salting rate, and share of brine into the MDSS of ASFINAG called Weather 2.0. Despite the extensive efforts in the model validation, there will always be deviations between prediction and reality, which can be addressed by an adaptation of the treatment recommendations by trained staff. Future avenues for research will be a further refinement of the winter maintenance model and the strategy optimisation, the implementation of a real-time benchmarking system, and exploring the possibility of using real-time data from winter maintenance vehicles and floating car data to update and validate the predicted road condition and skid resistance in real-time (Figure 3).
Paper #4: Hoarfrost in Winter Maintenance - The Underestimated Danger:
Winter maintenance, as a key task of road operators, ensures accessibility and safety during the winter. As part of the research program WinterFIT of ASFINAG headed by Hoffmann Consult now been running for more than 15 years, the paper covers the topic of hoarfrost from the physical mechanisms to measurement and prediction, as well as treatment optimisation and integration. With 40 to 60 days of hoarfrost in Austria, the first challenge was to better understand the hoarfrost mechanism and develop approaches for a systematic, continuous measurement of events. With the developed sensor system and data from standardised weather stations in several different regions over several winter periods, both an analysis of events and predictions of hoarfrost events have become possible. The statistical analysis has shown that the formation of hoarfrost is a continuous process starting between 22:00 and 01:00 in the evening and reaching a peak between 7:00 and 9:00 in the morning. The typical Weather conditions are decreasing temperatures between -1°C and -6°C at high relative humidity levels above 88%, resulting in an ice mass of 50 to 150 g/m2. For the predictive winter maintenance of ASFINAG, several models (logistic regression, ANN) have been developed, allowing a prediction of future events with 88 to 92% accuracy.
However, for these results to have any practical meaning, it was furthermore necessary to develop highly accurate local nowcasts of air and road surface temperature as well as relative humidity prediction models combining road sensor and RWIS station data with Weather forecasts. Based on these predictions, it was now possible to develop standard strategies for typical hoarfrost events in the form of guidelines that are being used on all road levels in Austria. Nonetheless, the research did not stop there, as the weather and road conditions during winter maintenance are constantly changing, and a more robust approach accounting for all types of weather events and recent treatment runs with specific recommendations for each road section and maintenance route was needed.
The development of the holistic winter maintenance model during WinterFIT made this possible, allowing to simulate and optimisation of maintenance strategies with timing, application rates, and share of brine for each treatment run for any Weather event. In the paper, we have shown both the mechanism of preventive or late treatments for hoarfrost events as well as the optimisation of strategies and sensor-based validation afterwards. In the case of hoarfrost, the results prove that preventive treatment with high shares of brine is superior compared to a late treatment in case of a missed event in the morning. However, if the preventive treatment run is too early, it is suboptimal, as most of the salt is lost before the hoarfrost event
The entire winter maintenance model with hoarfrost prediction and optimisation of treatment strategies has been successfully implemented in the maintenance decision-support software (MDSS) of ASFINAG called Weather 2.0 since 2022/23, providing specific treatment recommendations in case of hoarfrost on 250 road weather sections as well as 136 winter maintenance routes with specific timing, application rate, and shares of brine. Both the offline fallback strategies and the Weather 2.0 system have been received well by the maintenance crews. However, as the best models cannot account for everything and the drivers have to adjust the recommendations to the current situation, periodic training and a motivated crew are key to a successful winter maintenance. In this regard, the validation of events and results in the automated winter maintenance benchmarking of ASFINAG is key to ensuring continuous improvement of models, strategies, and training. The goal of ongoing research is therefore to automatically detect sensor errors, improve prediction accuracy and further refine treatment recommendations and validation of results (Figure 4).
Paper #5: Winter maintenance – why do we have limits?
In the old design guidelines of Austria, a longitudinal incline of up to 4% for highways and 6% for motorways was still allowed, with the critical sections on the A21 and the A13 being the only highway or motorway sections with an incline of 6% and periodic closures during severe snowfall events. Since 2014, the new design guidelines generally limit the maximum incline for motorways and highways to 4% to ensure homogeneous traffic flow with high truck traffic and avoid problems during severe snowfall events. This paper provides an overview of the methods and results of a case study, analysing the limits of winter maintenance on two sections with maximum longitudinal slopes of 6%. Despite intensive winter maintenance with significantly decreased cycle times, these sections experience periodic closures during severe snowfall events every 1-2 years in the last decade, prompting a detailed investigation into whether and how temporary highway closures can be avoided.
The study combines vehicle dynamics with a holistic winter maintenance simulation model from the research program WinterFIT of ASFINAG to identify conditions under which trucks lose traction on inclines. Based on analysis of weight in motion data (WIM), it was found that especially semi-trailers with an unfavourable relation of driving axle to total weight are most prone to losing traction on steep inclines. The results of the driving dynamics calculation revealed a critical skid resistance threshold of µ ≈ 0.2 being necessary to maintain uphill motion on 6% slopes. Below this threshold, stalling of trucks becomes very likely, with stuck trucks blocking both traffic and winter maintenance vehicles.
For assessing the impact of different winter maintenance cycle times and strategies, the developed holistic winter maintenance model of WinterFIT was adjusted, including only snowfall events and a refined time scale to cover cycle times below one hour. The validated simulation results show that at snowfall rates exceeding 1.0–1.5 cm/h, even very short clearing and salting cycles (e.g., <30 minutes) often fail to restore sufficient skid resistance above the critical threshold. The main reasons for this are the limited thawing capacity of de-icers, given the short cycle times with large shares being lost due to traffic, runoff, and frequent clearing. Thus, physical limits are largely reached with additional resources or even shorter cycles being theoretically feasible, but have almost no effects despite their costs.
The case study also shows that while flatter slopes not exceeding 4% incline can be largely managed with current cycle times of 90 to 120 minutes, 6% slopes remain vulnerable due to traffic dynamics and the physical and economic limits of winter maintenance. Operational alternatives like rerouting, block clearance, or enhanced driver preparation are discussed as well in the case study. Block handling of traffic during severe events behind winter maintenance and rerouting is currently deemed most practical for the A21, also requiring timely preparation and signalling. On the A13, this is not reasonably possible, but given the better preparation of truck drivers on the Brenner pass route, the closure times remain limited and will most likely be less frequent in the future due to climate change (Figure 5).
Acknowledgements:
Our special thanks go to ASFINAG Service GmbH and the Winter Maintenance Personnel in the Maintenance Depots for the trust and continuous support during research, development and implementation for many years. Furthermore, we would like to thank our various collaboration partners from the regional road departments in Austria and the Technische Universität Wien - Institute of Transport Sciences, who also contributed in various stages of the research.
Comments