1 PROJECT OBJECTIVES AND ACCOMPLISHMENTS
The original goal of this project was to investigate and develop, if possible, improved algorithms for in-flight icing and other aviation impact variables for the Alaska Region. Early, on, it was decided that the principal focus of the project would be in-flight icing. Specific objectives were as follows:
In addition to the Principal Investigators listed above, the following additional personnel have been involved in the project:
Ben Bernstein, Scientist, NCAR/Research Applications Program (9/98-present)
Dennis Fielding, Graduate Student, University of Alaska-Fairbanks (9/98-5/99)
Debra Y. Harrington, Research Programmer, University of Alaska-Fairbanks (5/99-10/99)
Cynthia Weatherby, Research Programmer, University of Alaska-Fairbanks (11/99-present)
Also, the following individuals from the Alaska Region of the National Weather Service have been involved in the project on mostly an advisory capacity:
Kraig Gilkey, SOO, NWSFO/Fairbanks, AK
Lee Kelley, MIC, NWSFO/Fairbanks, AK
Gary Hufford, Regional Scientist, NWS/Alaska Region HQ, Anchorage, AK.
1.3 Division of Responsibilities
For this project, the NWS provided several sets of observational and modeling data, including Alaska synoptic station network data for use in the in-flight icing algorithms (as appropriate), archived pilot reports of icing (PIREPs) as verification data for the algorithms and Alaska Eta model data for preliminary algorithm development work. The NWS also provided expertise on the utility of the PIREPs and on an in-flight icing algorithm (IFA) provided to AAWU from NCAR under a separate project funded by the Federal Aviation Administration. The NWS also provided suggestions for appropriate case study experiments.
For this project, NCAR also provided observational and PIREP data sets from their archives as well as codes, information and expertise relevant to the pre-existing in-flight icing algorithms NCAR also provided considerable expertise on in-flight icing based on their observational and algorithm development work pertaining to the continental United States (CONUS). Their assistance and expertise were invaluable to this project.
For this project, UAF/GI provided access and expertise related to the Alaska version of the Penn State/NCAR MM5 mesoscale modeling system as well as expertise on Alaskan forecast problems UAF took the lead in the testing and evaluation of the pre-existing IFAs and worked closely with NCAR on preliminary work towards new algorithm development.
1.4 Description of Research/Development Accomplishments
1.4.1 Testing and Evaluation of Pre-existing in-flight icing algorithms driven by the Alaska MM5.
A number of case study events were chosen as the basis for testing and evaluating three separate in-flight icing algorithms:
Algorithms (1) and (2) are entirely model-based and thus application of MM5 data is straightforward to those algorithms. Algorithm (3) requires a matching (and thereby interpolation) between the MM5 model grid and the observation. It also differs from the other algorithms in that it can only be applied within a certain radius of an observation location, that radius being the distance for which the surface meteorological conditions can be approximated by a given surface observation. With the strong topographic relief over much of Alaska, this radius cannot be reasonably set to values larger than 50 km over any part of the region. The net result is that the coverage of algorithm (3) in Alaska is substantially less than the entire MM5 grid area. Thus intercomparisons between the algorithms must take this structural difference into consideration, and in our work we have attempted to the extent possible to do so.
The case study events examined were as follows:
All three of these events were associated with large synoptic cyclone systems containing embedded mesoscale precipitation elements. As such, MM5 simulations were conducted on 60 and 20 km grids. The simulations all used the used the so-called "Reisner 1" microphysics scheme, in which cloud droplet water, cloud ice, cloud snow and cloud rain water mixing ratios are prognostic variables. Temperature values are a standard MM5 output variable while relative humidity (RH) is derived from the water vapor mixing ratio information using standard formulae.
The MM5 simulations were first evaluated as to their quality; in some cases the observed structure of the cyclone system was reproduced extremely well save for the fact that the propagation of the system was slightly slower than observed. In these cases, the simulations were retained but it was kept in mind during the verification process that such lags existed. Given that PIREPs from a several hour period surrounding the verification time were used in the process, relatively short lags in model propagation did not adversely impact the results.
Once the simulations were complete, the three algorithms were applied to the MM5 output data for both domains at 12-hour intervals over a 48 or 72-hour simulation period (dependent on the event). The algorithms were applied both in a columnar mode (any diagnosis of icing counts as a "yes" point) and in a three dimensional mode ("yes" points only at the vertical level diagnosed). The resulting diagnoses/forecasts of icing were then compared to available PIREPs of icing conditions during a window +/- 2 hours surrounding the verification time. Algorithm performance was then determined via a contingency table approach (e.g., reference, etc.) for both the columnar mode and the three dimensional mode predictions by the algorithms.
The results of these evaluations have been discussed in part within the references and/or presentations noted in section 3. Here we will summarize the general results. Both algorithms (1) and (2) tended to overpredict icing conditions though algorithm (1) (Schultz/Politovich) had a greater degree of overprediction. On the other hand, probability of detection (POD) statistics for both these algorithms were quite good. The stove pipe algorithm (algorithm (3)) did reasonably well at predicting icing conditions near the observation locations. However, it could not diagnose icing in areas exceeding the ~50km radii from the station noted above and thus had lower POD statistics with a current lower degree of overprediction. Statistics for the algorithms in a columnar mode generally showed better performance than for the three-dimensional mode, which met our expectations. Closer examination of the algorithm results and the PIREP reports, however, showed the encouraging sign that the algorithms were often only 1-3 vertical model levels (in a sigma coordinate sense) off in their predictions, implying that higher vertical resolution of the cloud systems might produce more favorable results in the three dimensional mode.
1.4.2 Application of Evaluation Results towards development of new icing algorithms
The results seen here were encouraging as to the potential of including MM5 model output as an integral part of an in-flight icing algorithm. Simultaneously with this project, the NCAR PI and her colleagues developed the first versions of the Integrated Icing Diagnostic/Forecast Algorithm (IIDA/IIFA) that uses satellite and radar information in addition to model output and surface observations. Developed originally for application in the CONUS, it uses the Rapid Update Cycle -2 data for the model component and has performed well.
In the context of this project the UAF and NCAR PIs, as well as NWS Alaska Region personnel and other NCAR personnel, have extensively discussed and planned for development of an IIDA/IIFA system for Alaska. Currently, a version of the IIDA for Alaska is under development at NCAR using the Alaska Eta DataStream for the model component of the Alaska IIDA. An early prototype of this IIDA using solely the Alaska Eta data has been tested with encouraging results. As an extension to this work at NCAR, UAF is now developing a version employing MM5 data based on the results reported above.
1.4.3 Operational Evaluation and Implementation of Icing Algorithms
Unfortunately, due to staffing constraints at the AAWU as well as some personal circumstances involving the AAWU PI, many of the objectives for this portion of the project have not been realized. AAWU has tested versions of the Schultz-Politovich algorithm and NCAR Algorithm (the immediate predecessor to the IIDA/IIFA) in-house and found, consistent with the results of this project, a general tendency to overpredict, particularly with the Schultz-Politovich algorithm. Otherwise, there has been no testing or implementation of an MM5-based icing algorithm at the AAWU associated with this project.
2 SUMMARY OF UNIVERSITY/NWS EXCHANGES
In contrast with previous COMET efforts, University and NWS exchanges were more limited in duration and scope with respect to this project, though the good rapport between the University and NWS/Fairbanks office, which had a minor role in this project, continued. Below the exchanges are briefly summarized.
The UAF and NWS participants took part in an Alaskan aviation weather workshop sponsored by NWS during April 1998, and also in a NOAA/CIFAR review during September 1998. During the latter event, some of the early work regarding the utility of the MM5 model system as a driver for the pre-existing IFAs was presented. Besides these presentations, the UAF PI visited the AAWU and NWS/Alaska Region headquarters twice in 1998 and once in 1999 for consultation and discussions and data exchanges regarding the project as well as other Alaskan forecast issues. Several other exchanges occurred via phone and email during the course of the project, and there was intermittent informal discussion of the project with the NWS/Fairbanks SOO and MIC.
3. PRESENTATIONS AND PUBLICATIONS
Tilley, J. S., 1998: Towards Improved Forecasting of Icing and Turbulence in the Alaskan Region. 1998 Spring Air Fair Expo, Northern Alaskan Aviation Symposium, Fairbanks, AK, 6 March.
Tilley, J.S., 1998: Application of Mesoscale Numerical Weather Prediction Models to Aviation Weather at the University of Alaska: Past, Present and Future. National Weather Service/Alaska Region Aviation Weather Workshop, Anchorage, AK, 5-10 April.
Tilley, J.S, K. G. Gilkey, D.-L. Wilkinson, M.K. Politovich, C. A. Scott, H. L. Kelley, G. Hufford and C. Dierking, 1998: Mesoscale Modeling in Support of Observational Studies in the NWS Alaska Region. NOAA Review of Cooperative Institute for Arctic Research, Fairbanks, AK, 16 September 1998.
Tilley, J. S., D.-L. Wilkinson and M.K. Politovich, 1999: Application of Mesoscale Model Data to Algorithms for In-flight Icing over the Alaska Region. Eighth Conference on Aviation, Range and Aerospace Meteorology, 10-15 January, Dallas, TX [preprint][also appears in preprints to The Ninth PSU/NCAR Mesoscale Model Users' Workshop, 23-25 June, Boulder, CO]
Tilley, J.S., J. Long and C. Weatherby, 2001: On the Performance of the AFWA version of the PSU/NCAR MM5 model for short-range forecasting in Alaska, the Western Arctic and North Pacific (in preparation for preprints to 6th AMS Conference on Polar Meteorology and Oceanography, May 2001, San Diego, CA).
Tilley, J. S., C. Weatherby and M. Politovich, 2001: MM5 Model-based icing algorithm performance over an annual cycle in high latitudes. (in preparation for preprints to 11th PSU/NCAR Mesoscale Model Users' workshop, June 2001, Boulder, CO)
4. SUMMARY OF BENEFITS AND PROBLEMS ENCOUNTERED
Prime among the University benefits was the improved understanding of Alaska aviation weather forecast problems and the challenges faced by the NWS and FAA in providing improved services with respect to aviation weather forecasting in the Alaska Region. In addition, the University has gained much through this project through the participation of NCAR in terms of improved understanding of the physics involved with in-flight icing and also in the types of approaches used in algorithm development, as well as their physical basis. NCAR has benefited from interactions with both the NWS and the University in terms of a fuller understanding of Alaskan (and high latitude in general) meteorology as well as the special circumstances (different from forecast offices in the CONUS) under which the Alaska Region offices operate. The NWS has gained an improved understanding of the advantages and limitations of mesoscale numerical weather prediction models in the Alaska Region for not only aviation forecasting, but other forecast problems as well. With respect to all three parties, the interactions associated with this project has laid the groundwork for substantial advances in aviation weather services in Alaska over the next 5 to 7 years, and this benefit is one shared with all Alaskan aviators.
4.2 Problems Encountered
The greatest problem encountered in the course of this project was the inability of the AAWU PI to really spend much time on the project outside of a few discussions and some data exchange. Staffing levels at the AAWU during 1998-2000 were such that the AAWU PI spent many hours at the forecast desk and had little time to work on this project. In addition, personal circumstances resulted in the AAWU being away from the office for extensive periods during the course of this project. The result is that the operational side of this project could not really be realized, which was a disappointing outcome.