 |
Slipping on an icy sidewalk, bicycle path or road can be unpleasant and outright dangerous. To avoid such situations, one can salt these surfaces. Yet one cannot salt all winter long, since salting is expensive (costing up to several hundred million Danish kroner per year in Denmark), can be damaging to the environment, and because icy situations can occur as late as in April or even May. It is therefore essential to be able to forecast the risk of slippery roads and with the prediction in hand, take necessary action, e.g., salt (or not salt) roads. Note that much less salt is required if roads are salted before they freeze than after ice is formed on them.
Prediction of road conditions requires accurate forecasts of temperature, humidity and precipitation at and near the surface of the road. DMI has, since the 1994-1995 winter season, utilized a Road Conditions Model (ROCMO) forecasting system which can make such predictions.
The Data Assimilation and Forecast
The ROCMO system uses as input observations from weather stations and road stations and results from the weather model HIRLAM, (HIgh Resolution Limited Area Model), to produce five hour forecasts every hour.
The data assimilation produces a model state at the forecast initial time tf and atmospheric (HIRLAM) input data which are modified by observations. These data force the ROCMO during the forecast. The modified atmospheric data and the primary model produced data, i.e., road temperatures, road water and ice, may be considered as the complete model state since all these data contain prognostic information.
The data flow in the ROCMO system is shown in Fig. 1.
Various data are supplied during data assimilation from time td to tf and during the ROCMO forecast from time tf to time tf+L. HIRLAM data (input) are: T(k), (k<=N) is the temperature at the atmospheric model levels (°C), q(k), (k<=N) the specific humidity at the atmospheric model levels (kg/kg), T2m the temperature at 2-m height (°C), q2m the specific humidity at 2-m height (kg/kg), V10m wind speed (m/s) at 10-m height, Q precipitation intensity [kg/(m2/s)], ps surface pressure (Pa), and N number of atmospheric model levels. Synoptic data (input): Cob is the observed total fractional cloud cover and Hob the observed cloud-base height (m). Road station data (input): T2m is the observed temperature at 2-m height (°C), Td2m the observed dew point temperature at 2-m height (°C), and Ts(1) the observed road surface temperature (°C). Additional variables (output): Ts(l) (l<=M) is the road vertical temperature profile (°C), Ws the road surface water (kg/m2), Is the road surface ice (kg/m2), M the number of model levels of the ROCMO, T10m the temperature at 10-m height from T2m and T(N), q10m the specific humidity at 10-m height from q2m and q(N), and C(k) the fractional cloud cover at atmospheric model levels.
In the data assimilation, the temperature profile in the road is controlled through a careful initialization procedure. The equation of heat conduction in the road is solved during the three hour data assimilation period prior to the forecast initial time tf. Observed surface temperatures (Ts) from road sensors are imposed as an upper boundary condition during this period to produce a realistic temperature profile from observed surface temperatures and a temperature profile which is initially obtained from the previous data assimilation cycle (ROCMO state 2 of Fig. 1). The hourly HIRLAM data are interpolated linearly in the horizontal to values representing the atmospheric variables above each road station site and are then modified by means of observations.
In forecasting the correct road conditions, the basic problem is to find an accurate solution to the energy budget at the road surface. All relevant process, i.e., solar and infrared radiation, surface sensible and latent heat flux, precipitation (and the freezing and melting of it), ground heat conduction and additional heat sources, for example from traffic, are addressed.
Predictions of temperature, dew point temperature, wind, precipitation, road temperature, amounts of water and ice on the roads and cloud height and cloud cover are made.
Operational Results
Verification results of Ts forecasts are presented in Figs. 2a and 2b for the periods 12 January to 24 February 1998 and 24 February to 31 March 1998, respectively. The figures show the mean absolute error of Ts for the ROCMO forecast (brown lines) with the corresponding results of persistency forecasts (orange lines) and linear trend forecasts (green lines). The linear trend forecasts are made by linear extrapolation in time of a mean observed surface temperature tendency during the hour prior to the forecast. The results are based on the average of all 250 Danish road stations, but only in cases where Ts is between ±3°C.
Figs. 2a and 2b illustrate the predictive skill of the ROCMO system; the ROCMO forecasts are significantly better than the two other methods from the first hour on, especially from February to March (Fig. 2b).
The Consumers
The Danish Road Directorate (Vejdirektoratet), counties and municipalities throughout Denmark receive the ROCMO forecasts along with road data, which is updated every 10 minutes. With these data at hand, the road authorities have the possibility to see how road conditions will develop in the hours to come and to act accordingly.
The Future
Further improvements of the ROCMO forecasting system are planned. More accurate specification of the atmospheric state above the road station sites is the most critical point to address. The computational power of supercomputers makes it possible to run the HIRLAM model with a very high spatial resolution. New data sources such as satellite data, radar data and automatic weather stations in combination with improved analysis techniques will make it possible to analyze the small spatial scales of clouds and precipitation more accurately both in the HIRLAM model and in the ROCMO system.
|
