1、Special Project AIAA SP-137-2012 Status of Inflight Icing Forecasting Products and Plans for Future Development AIAA standards are copyrighted by the American Institute of Aeronautics and Astronautics (AIAA), 1801 Alexander Bell Drive, Reston, VA 20191-4344 USA. All rights reserved. AIAA grants you
2、a license as follows: The right to download an electronic file of this AIAA standard for storage on one computer for purposes of viewing, and/or printing one copy of the AIAA standard for individual use. Neither the electronic file nor the hard copy print may be reproduced in any way. In addition, t
3、he electronic file may not be distributed elsewhere over computer networks or otherwise. The hard copy print may only be distributed to other employees for their internal use within your organization. AIAA SP-137-2012 Special Project Report Status of Inflight Icing Forecasting Products and Plans for
4、 Future Development Marcia K. Politovich, Editor Sponsored by American Institute of Aeronautics and Astronautics Approved 5 December 2012 Abstract The three papers in this Special Project Report were presented at the AIAA Atmospheric and Space Environments Conference in August 2010. They provide the
5、 current status of automated inflight icing diagnosis and forecast algorithms, and describe steps for improvement: new data inputs, improved logic, development of human-over-the-loop production methods, and expansion of the domain to cover the globe. AIAA SP-137-2012 ii Published by American Institu
6、te of Aeronautics and Astronautics 1801 Alexander Bell Drive, Reston, VA 20191 Copyright 2012 American Institute of Aeronautics and Astronautics All rights reserved No part of this publication may be reproduced in any form, in an electronic retrieval system or otherwise, without prior written permis
7、sion of the publisher. Printed in the United States of America AIAA SP-137-2012 iii Contents Foreword iv Potential Upgrades to the Current and Forecast Icing Algorithms . 1 The Global Forecast Icing Product . 11 Using Icing Algorithm Output to Create AIRMETs 19 AIAA SP-137-2012 iv Foreword Inflight
8、icing has been a strong component of the Atmospheric and Space Environment Technical Committee of AIAA. For the most part, inflight icing studies presented at AIAA conferences tend to focus on the effects of the atmospheric environment on the performance of aircraft. However, descriptions of the ici
9、ng environment, and of forecasting or diagnosing icing conditions, have also had a place in the presentations. This intermingling of related disciplines with a common goalreducing icing-related accidentshas stimulated discussions and encouraged collaborations that otherwise would not likely have com
10、e to pass. Three papers were presented at the 2010 Atmospheric and Space Environments Conference, held in Toronto, Ontario, Canada describing state-of-the-art automated forecasts and paths to future versions. At the time of the 2010 Conference, products available for inflight icing forecasting inclu
11、ded the following: Airmens Meteorological Bulletin (AIRMET): An advisory for widespread moderate or greater structural icing covering a 6-h forecast period, which may be amended. Significant Meteorological Information (SIGMET): A weather advisory for severe icing over a 3000-mi2or 7800 km2area. Curr
12、ent Icing Product (CIP): An hourly diagnosis of inflight icing environmental conditions over the continental United States (CONUS). The product includes probability of encountering icing in any of the 20-km/1000-ft grid boxes, expected severity, and likelihood of supercooled large drop (SLD; drops w
13、ith diameters exceeding 50 microns, which is outside of the certification conditions). The CIP algorithm combines numerical weather prediction (NWP) model output with observations such as geostationary satellite imagery, NexRad radar reflectivity, surface weather observations, and the national light
14、ning network. Forecast Icing Product (FIP): An output updated hourly for each hour up to 12 hours forward in time. FIP is similar to CIP but it uses NWP model surrogates for the observations ingested by CIP. The automated CIP and FIP do very well at what they are called to do: provide a medium-scale
15、 resolution product with a broad-brushed icing severity estimation over the CONUS. The intended user is an aviation meteorologist, dispatcher, or pilot looking for strategic information for flight planning. Graphical depiction, both format and content, is extremely important to these users. Consider
16、 a future air transportation system where aviation weather products are fully integrated into a seamless weather-to-aircraft process. The products will incorporate various components including weather observations, NWP models, algorithms to interpret and combine information, human-over-the-loop meth
17、ods, communications protocols, and flight planning and control systems. User needs for displays will be taken into account, but for the most part the forecaster, dispatcher, or pilot is not the end-user so much as automated aircraft and ground-based systems that plan for and monitor the many aircraf
18、t in the air. This publication, which comprises three presented papers, offers ideas for extending the existing suite of inflight icing products into new geographic domains, with richer information content and opportunities for forecasters to add additional skill and insight gained from experience a
19、nd knowledge of the atmosphere. This research, which is underway, forms a solid basis for automated aircraft icing diagnosis and forecasting, which offers the reader a glimpse into future products. AIAA SP-137-2012 1 Potential Upgrades to the Current and Forecast Icing Algorithms Marcia K. Politovic
20、h,1Cory A. Wolff2, Frank McDonough2, Julie Haggerty3National Center for Atmospheric Research, Boulder, CO 80303 and Kenneth Howard4National Severe Storms Laboratory, Norman, OK 73072 The Current and Forecast Icing Product algorithms generate icing diagnoses and forecasts across the CONUS. These have
21、 been approved for use in operational decision-making in aviation. However, there is both a desire and need for upgrades to improve accuracy through use of new or improved weather prediction models or observations. This paper describes the upgrades planned at this time and outlines the reasoning beh
22、ind choosing new candidate information for those upgrades. Two products under consideration for addition to CIP, images from the Advanced Satellite Aviation-Weather Program and grids from the 3-D NexRad radar mosaic produced by the National Severe Storms Laboratory, are described in more detail. I.
23、Introduction T is not difficult to provide an automated icing forecast. Looking for moist air in the appropriate temperature range will give a fairly good forecast of icing this was the basis behind the Schultz-Politovich algorithm1, which was based on temperature and relative humidity fields from t
24、he Nested Grid Model in the early 1990s. If the user is only looking for a broad-based forecast of where icing conditions are likely, this fills the niche very nicely. However, thats not the entire picture. As the user demands more details such as severity (especially moderate or greater), condition
25、s outside Appendix C (including SLD, supercooled large drops with diameters exceeding 50 microns), higher resolution in time and space (especially vertically) and expanded geographic coverage, the forecast process becomes increasingly more complex. There are many areas in seemingly favorable tempera
26、ture and humidity ranges that do not include icing conditions. This paper will describe planned upgrades to the Current and Forecast Icing Products, and how upgrade data and process candidates are chosen. The development and operational transfer of icing products are closely linked to NextGen schedu
27、les and requirements; however this paper will focus on technical aspects of the upgrade process. II. Algorithm Basics The Current Icing Product2(CIP) provides an hourly diagnosis of icing conditions. The CIP algorithm examines model outputs and observations, and extracts clues about the icing enviro
28、nment from these information sources. The process is straightforward and traceable. Its relatively easy to assess the impact of humidity fields or satellite-derived cloud top temperatures on the final product. The forecast version, FIP, is purely model-based and uses models outputs as surrogates for
29、 most of the observations. 1Deputy Director for Science Aviation Applications Program, Research Applications Laboratory, NCAR, PO Box 3000, Boulder, CO 80303, AIAA Senior Member 2Associate Scientist, RAL 3Project Scientist, RAL 4Meteorologist, Stormscale Hydrometeorology R FIP calculates 1, 2, 3, 6,
30、 9, and 12-h forecast windows. The processes used in CIP and FIP are discussed below. A. Gridding Data are mapped to a grid based on that of the NWP model being used. Since the observational data are available at different resolutions, they must be mapped using various techniques. For those data set
31、s with higher resolution data such as satellite or radar, the whole field of data contained within one horizontal grid footprint is utilized by finding percentiles of the distribution of parameter values in that area. Usually, 25th, 50th, and 75thpercentiles are extracted and carried through the ana
32、lysis rather than one value to characterize the entire grid. This way, the variation of those values can be used to assess the uniformity of the icing environment. For point values with coarser resolution than the underlying grid (and generally available at non-uniform spacing, such as METARs), a sy
33、stem is used by which each grid point seeks the “best” information by comparing information at various radii from the point. Depending on the parameter, the meaning of “best” changes, depending on the impact of the parameter on the icing condition for example, there is a hierarchy of precipitation t
34、ypes that infer increasingly hazardous conditions aloft. Also during the gridding procedure, locations without cloud or precipitation are flagged as no-icing and further calculations are not applied which saves considerable computing time. B. Decision Tree Vertical columns of NWP model output are us
35、ed to determine the type of cloud present which, in turn, influences how data are interpreted and combined downstream. A shallow, non-precipitating stratiform cloud has different physical processes at work than does a deep convective storm. Once a decision is made on a cloud type it is carried throu
36、gh that column in the future, this may also be “fuzzified” (see below) to account for uncertainties in or combinations of conditions. C. Fuzzy Logic Figure 1. Examples of CIP severity, probability and SLD fields. Figures are composites showing the highest value in a vertical column. AIAA SP-137-2012
37、 3 The fuzzy logic techniques used in CIP/FIP are simple but powerful. Information from diverse sources is mapped to a scale of -1 (completely negates) to +1 (completely supports) depending on how “interested” the algorithm is in using that information to determine icing and its characteristics. Zer
38、o means the information has no effect on consideration of icing conditions. Interest maps are created from all relevant parameters, then weighted according to their significance, given a confidence value depending on source and time of day (for example, the satellite information is less reliable nea
39、r the terminator), and a final number is produced for each grid point. Separate fuzzy logic calculations are performed for icing probability, severity, and SLD potential. III. Upgrades Without introducing a discussion of “how good is good” we always strive for improvement in the algorithms. Our bigg
40、est problem at this time is overforecasting of icing conditions and we believe that is mainly due to an inability to diagnose glaciated cloud conditions. Some upgrades are forced on us, for example, when the National Center for Environmental Prediction (NCEP) makes changes in NWP models or when agin
41、g satellites are replaced and the detected radiation wavelengths differ slightly with the new instruments. Other upgrades represent changes in algorithm logic as we learn more about atmospheric processes and how to use observations more intelligently. Additional upgrades are pursued as new instrumen
42、ts are developed with a potential for becoming operational platforms. Our guidelines for considering upgrade candidates are these: New NWP models and instruments No choice (model changes made at NCEP, new GOES, etc.) They add new information not available elsewhere They confirm other information and
43、/or serve as backups in case of data loss They form a basis for the future (i.e., at current temporal and spatial resolution they may not seem to add much but they could make a potentially huge difference in the future) Logistical Increased domain Alaska, global Increased spatial resolution high res
44、olution NWP models, radar and satellite data enhancing coarser model output More frequent update limitations in computational speed, human comprehension. What is too fast an update rate? What can a human handle, and what makes sense in terms of how fast atmospheric processes are changing? What are t
45、he relevant scales? Algorithmic Improved microphysics in models weights change in the algorithm Improved logic for extracting and combining information Changes in interest maps to accommodate model changes Planned upgrades for CIP/FIP are scheduled for early 2011, coincident with the transition from
46、 RUC to the Weather Research and Forecast Rapid-Refresh model (WRF-RR), and for 2013. Beyond that time, we anticipate major changes in input data sets, both from observational platforms and NWP models. Minor upgrades due to model and instrument changes, or for bug-fixes, can be implemented as needed
47、. Prior to implementation, upgrades are incorporated into test versions of FIP and CIP at NCAR and run for several weeks to allow scientists to gage their performance. Statistical comparisons with baseline versions of the algorithms, and with pilot reports of icing, are used to quantify improvements
48、. When the scientists are satisfied that the upgrade provides improvement (or, at the least, does not degrade the product), and has no unexpected downstream effects, engineers code the upgrades into the algorithm for further testing, verification, and implementation. The evaluation and verification
49、processes to approve operational versions of the products is rigorous and is undergoing changes to comply with NextGen requirements; they will not be discussed here. For the 2011 upgrade (Table 1), most of the work consists of bug-fixes, slight changes to fuzzy logic interest maps to accommodate changes in the model output, and some upgrades to bring the CIP up to date with FIP. (The current and forecast algorithms had been on different update cycles and some of the logic in FIP had been upgraded with no concurrent change in CIP.) Th
