AN IMPORTANT NOTE TO CIP AND FIP USERS ---------------------------------------- CIP and FIP users should be aware that icing conditions sometimes change rapidly in time and space. CIP and FIP flight level and cross-section plots may not be able to capture rapid changes and/or small-scale differences in icing. They also will not capture all occurrences of icing. Users are advised to look at several of the icing plots to assess the icing situation along their route. Flight route cross-sections show icing within ~60km of a straight line route, and icing conditions may be significantly different just beyond this +/-60km window. Flight level plots show the icing that is expected at that altitude, and icing conditions may be significantly different at altitudes just above and below these heights. The concept behind the integrated icing algorithm ------------------------------------------------- A number of aircraft icing diagnostic and forecast algorithms have been developed in the past several years. These algorithms have used information from meteorological models, satellites, surface observations and radar mosaics in a variety of ways and have met with some success. Although all of these techniques have their individual merits, they also have their limitations. In brief, algorithms which are based purely on model output capture most icing pilot reports (PIREPs), but overforecast icing by indicating it in locations where clouds do not exist. Algorithms which are based primarily on data from instruments (satellite, radar, surface observations) tend to be quite accurate in the locations where they indicate icing, but they underforecast icing because none of these instruments can identify all icing locations by themselves. This has been demonstrated by Carriere et al (1997), and they suggest the use of data from instruments such as satellites, radars and surface observations in combination with model output to improve icing algorithm skill. The integrated icing diagnostic attempts to take advantage of the abilities of both the instruments and the models to try to capture the maximum number of PIREPs while indicating icing in the smallest amount of 3-D space possible. Strengths and weaknesses of algorithms which are purely model-based ------------------------------------------------------------------- Algorithms which depend solely on forecast model output rely heavily on relative humidity and temperature fields. The temperature output from these models (e.g. the Rapid Update Cycle - RUC) can be used fairly reliably to indicate locations where it is either too warm or too cold for icing to exist, and to highlight the temperature ranges where icing is most common. However, the use of the model-based relative humidity output to indicate the locations where clouds exist has significant problems. One would expect that if the model were perfect, then icing forecasters could simply apply a relative humidity threshold of 100 percent to indicate the locations of clouds. Of course, this is not the case (e.g. due to model resolution, averaging and inaccuracies), and icing algorithms have had to use lower thresholds of relative humidity to try to find clouds. This places the algorithm developer in a position of trade-off between lowering the relative humidity thresholds to capture as many of the clouds (and icing PIREPs) as possible, and raising the relative humidity thresholds to keep the impacted air space to a minimum. Typically, the result has been for the developer to choose a "reasonable" minimum relative humidity threshold of 60 to 80 percent (e.g. Schultz and Politovich, 1992). In some cases, developers have used additional information such as thermodynamic structure (Forbes et al. 1993) and derived vertical motion fields (McCann, 1997) in combination with relative humidity to get a better handle on the locations of clouds, and even extend the relative humidity range to lower values when appropriate thermodynamic and/or vertical motion structures were present. Still, none of these algorithms could absolutely identify the locations of clouds in 3-D space, and thus, they diagnose and forecast icing in cloud free locations. Overall, while purely model-based icing algorithms capture a high percentage of icing PIREPs, they overforecast the icing. Strengths and weaknesses of algorithms which are purely instrument-based ------------------------------------------------------------------------ Algorithms which rely on instruments to identify locations where icing is likely to exist are typically quite efficient. In other words, when they identify a location as having icing, it often does exist and icing PIREPs will tend to show up there. The areas identified are often relatively small compared to those from model based algorithms. This combination of small areas/volumes and decent probability of detection (up to 60 percent) results in relatively high efficiencies (POD per unit area or volume; see Brown et al. 1997 for statistical results on some algorithms discussed here). The shortcoming of these instrument-based algorithms is that they miss a large percentage of the icing due to limitations of the instruments themselves or the techniques applied to them. For example, satellite-based icing algorithms (e.g. Thompson et al 1997; Ellrod and Nelson 1997) use a combination of data from several satellite channels (visible and infrared) to identify clouds which are likely to have liquid tops at temperatures below freezing. While this technique is quite successful when single-layer, warm clouds are the only clouds which exist in a certain location, it fails to identify icing clouds when 1) a multi-layered clouds structure exists (i.e. when a cold, upper cloud shield masks a warm, liquid cloud at lower altitude) and 2) a deep, classic freezing rain cloud exists (a deep, cold cloud masks a layer of freezing precipitation below). Radar-based icing algorithms use information from the national radar mosaic and model data to identify subfreezing locations in and around radar-indicated precipitation as those which are likely to have icing. While icing often exists in these locations, a great deal of icing occurs in locations hundreds of kilometers from identifiable echo in radar mosaics. This is typically because the drops which cause the icing are too small to be seen by the radar, except at very close ranges. The exception to this is freezing rain, which can be readily seen by radars, even at long ranges. How the integrated icing algorithm pulls it all together -------------------------------------------------------- The idea behind the integrated icing diagnostic is to take advantage of the abilities and minimize the shortcomings of both the model-based and instrument- based approaches, to try to capture the maximum number of PIREPs while impacting the smallest amount of area/volume possible. In general, the algorithm first integrates information from the GOES-8 satellite, surface observations and RUC model to identify the three-dimensional extent of cloud, then uses information from these resources plus the national radar mosaic to identify the locations and likelihood of both conventional and supercooled large drop (SLD) icing across the United States and Canada. A situational approach is used which applies information from the different data sources in different ways, depending upon the physics expected to be at work at each location within the domain. The integrated algorithm uses information from all four data sources (satellite, surface observations, radar mosaic, RUC model). By using information from several of them in a situational approach, it is possible to minimize the impact of bad data in any one field. Images of the resultant icing and SLD fields, as well as the ingredients from which they are derived are displayed on this page. Basic descriptions of how the various ingredients are derived and used in the algorithm are also available. REFERENCES ---------- Brown, B.G., B.C. Bernstein, T. Kane and R. Bullock, 1997: Diagnostic and comparative verification of algorithms for the detection and forecasting of in-flight icing. Preprints, 7th Conf. on Aviation, Range and Aerospace Meteorology, Long Beach, CA, 2-7 February. Amer. Meteor. Soc., Boston, 88-93. Carriere, J., S. Alquier, C. LeBot and E. Moulin, 1997: Some results of a statistical evaluation of forecast icing risk algorithms. Preprints, 7th Conf. on Aviation, Range and Aerospace Meteorology, Long Beach CA, Amer. Met. Soc., Boston, 106-111. Ellrod, G.P. and J.P. Nelson, 1997: An experimental GOES image product to identify conditions favorable for aircraft icing. Preprints, 7th Conf. on Aviation, Range and Aerospace Meteorology, Long Beach, CA, 2-7 February. Amer. Meteor. Soc., Boston, 112-115. Forbes, G.S., Y. Hu, B.G. Brown, B.C. Bernstein and M.K. Politovich, 1993: Examination of soundings in the proximity of pilot reports of aircraft icing during STORM-FEST. Preprints, 5th Conf. on Aviation Weather Systems, Vienna, VA, 2-6 August. Amer. Meteor. Soc., Boston, 282-286. McCann, D.W., 1997: Five ways to produce supercooled drizzle drops. Preprints, 7th Conf. on Aviation, Range and Aerospace Meteorology, Long Beach, CA, 2-7 February. Amer. Meteor. Soc., Boston, 94-99. Schultz, P. and M.K. Politovich, 1992: Toward the improvement of aircraft icing forecasts for the continental United States. Wea. and Forecasting, 7, 492-500. Thompson, G., T.F. Lee and J. Vivekanandan, 1997: Comparisons of satellite- based aircraft icing diagnoses. Preprints, 7th Conf. on Aviation, Range and Aerospace Meteorology, Long Beach, CA, 2-7 February. Amer. Meteor. Soc., Boston, 132-137.