SIGMA : System of Icing Geographic identification in Meteorology for Aviation

Icing is a very dangerous phenomenon for aeronautics, and is very difficult to predict for meteorological forecasters. In order to help them in this task, Météo-France has developed a system for the identification of icing areas, named SIGMA (System of Icing Geographic identification in Meteorology for Aviation). This system is based on the combination of three different sources of data. Firstly, it uses the icing risk index calculated from Météo-France’s numerical weather prediction model. This index determines icing areas by using the humidity and temperature forecast values calculated at each grid point. However, as it is known that the numerical models have a tendency to overestimate the humidity and areas of icing, an additional information is used with the observation of cloudy areas from satellite imagery. This way, the initial results obtained with the icing index is filtered with infra-red data from the geostationary Meteosat spacecraft. These data re indicate the position of the cloudy areas with cloud top temperatures comprised between 0°C and –20°C. This is the first category of icing risk. Cloudy areas with temperatures below –20°C are accounted for as there is a risk of icing under the colder cloud layer as seen by the satellite. This is a second category of icing risk. Thirdly, Météo-France’s operational centimetric radar network is used for the amount of water contained in the area selected by the previous algorithms. Thus, icing risk can be identified in terms of rain in cloud, which corresponds with echoes of rain viewed by the radar network, but not necessarily viewed by surface observation systems. Some verifications have been done. An icing PIREP collection campaign was organized with French military aircraft during one winter. The PIREPs were compared with the output of the SIGMA and showed encouraging detection rates. During one winter season, the forecasters of Météo- France compared the results of the Sigma picture with the observations from the 23 points of operational rawinsounds distributed on the domain of Sigma. This study shows good results for Sigma and helps to improve the behaviour of the system.