J. Ceiling and Visibility

[Background] [Ceiling and Visibility Expert Algorithm]


1. Background

The prediction of cloud ceiling and visibility (C&V) continues to be a formidable challenge. Aviation safety and operational efficiency depend heavily upon accurate and timely forecasts of these atmospheric conditions. Led by K. Petty, RAP scientists are working towards the goal of improving the accuracy of short-term (0-12 hrs.) C&V forecasting and extending this forecasting capability to the continguous U.S. (CONUS) and Alaska domains. This work is being conducted in coordination with the Federal Aviation Administration's (FAA) National Ceiling and Visibility Product Development Team (NCVPDT). This team is made up of government institutions, private organizations and universities.

The development of the C&V forecast system (Petty et al., 2000) is built around recognition that no single technique will succeed above all others across a range of forecast periods, meteorological conditions or geographic environments. The expert algorithm produces C&V analyses and forecasts through use of a data fusion process that adaptively weights and combines a changeable set of observations and forecast products derived via statistical and numerical modeling techniques.

The following section outlines the primary scientific activities that took place during 2000 in support of C&V expert algorithm development.

2. Ceiling and Visibility Expert Algorithm

The program of research and development focuses effort on the improvement of critical links in the logical chain that applies scientific understanding to practical C&V diagnosis and forecast problems.  These critical links and areas of scientific development are 1) expert algorithm enhancements, 2) improved remote sensing techniques, 3) improved translation algorithms and numerical models, 4) creation of statistical C&V forecast methodologies 5) organization and execution of a winter weather field program, and 6) the development of verification techniques.

RAP scientists investigated methods for data fusion and outlined steps for implementing and testing various data fusion techniques. The main focus is to understand how to optimize the adaptive weighting process used in the expert system. In addition, RAP has investigated and validated other data sources and forecasting techniques that may be potentially beneficial to the analysis and forecast of ceiling and visibility. The algorithm development uses a phased-based approach in which the influence of a single input source is determined, analyzed, and compared to other inputs, prior to its final incorporation into the algorithm. Currently, Geostationary Operational Environmental Satellite (GOES) data, surface observations (METAR), model data (RUC-2), climatology, pilot reports (PIREPS) and persistence are being integrated in the system to provide an analysis and forecasts of C&V over the CONUS.

RAP scientists have also worked with the Naval Research Laboratory to build and run NRL’s cloud classification algorithm locally. The algorithm will be run in real-time so that products can be verified and RAP’s science and engineering group can develop an understanding of how the classification products will influence the ceiling and visibility system. Remote sensing data serves as an additional input into the ceiling and visibility analyses.

An important component of the expert algorithm development is the construction and execution of a winter weather field program in the northeastern United States. This initiative supports scientific efforts to understand the formation of fog and low stratus in the Northeast corridor. RAP scientists have investigated and defined the sensors necessary for a successful field program. R. Tardif and J. Cole, both members of RAP’s ceiling and visibility team, met with representatives of Rutgers University to survey the Rutgers field site and found that the site met the requirements for instrumentation. R. Tardiff worked on defining an updated fog climatology for the New York area. The focus was on the Newark (EWR) airport since the instrumented site at Rutgers University is believed to be most representative of this airport. This study will help assess the characteristics of fog events and will serve as a basis for evaluating the relevance the possible approaches to improve fog forecasts. Data compiled by (NCAR-RAP) indicated an average of about 1400 hours of fog per year for the 1977-1990 period. The frequency of fog events in the area was fairly uniform. Nevertheless, some spatial variability was observed, possibly related to the various fog types (radiative, advection etc.) preferentially occurring in different areas. Also, data indicated a slight seasonal variability of the number of hours of observed fog. Maximum observed hours of fog occurred in December and May at the Newark airport.

[top of page]

C. Tebaldi explored statistical applications derived using historical datasets.  This research is in support of the development of obs-based statistical forecast techniques. She has investigated the use of neural networks, flexible discriminant analysis and logistic regression in forecasting ceiling and visibility at various locations throughout the United States. In an effort to evaluate these approaches, she has utilized Briar Score, MSE, Probability of Detection (POD) and False Alarm Rate (FAR) metrics.  Preliminary results have revealed that persistence remains competitive and is a favorable predictor (Figure J1).

Figure J1.  FAR plotted against POD. Blue, green, and black lines represent logistic regression, neural network. and discriminant anlaysis, respectively. Persistence performance is represented by the intersection point of the horizontal and vertical blue lines.

Finally, B. Brown, T. Fowler, M. Chapman, C. Tebaldi, J. Mahoney (FSL) and M. Kay (FSL) have explored verification approaches and methods that can be used for verification of national-scale ceiling and visibility forecasts and diagnoses. Some of these approaches are already being used in ongoing verification of operational forecasts, referred to as the Airmen’s Meteorological Advisories (AIRMETs) of instrument-flight-rule (IFR) conditions (i.e., IFR AIRMETs) issued by the Aviation Weather Center (AWC), by the Forecast System Laboratory’s Real-Time Verification System (RTVS; Mahoney et al. 1997). In addition, some of the methods are being used for ongoing feedback to the current automated product. Their investigation considered general concepts of verification, as well as specific approaches that can be applied for various types of predictors and predictands (i.e., categorical, probabilistic, continuous).

REFERENCES

Petty K.,  B. Carmichael, G. Wiener, M. Petty, and M. Limber, 2000: A fuzzy logic system for the analysis and prediction of cloud ceiling and  visibility. Preprints Ninth Conference on Aviation, Range, and Aerospace Meteorology. Orlando, Fl. Amer.  Meteor.  Soc., 331-333.

Mahoney, J.L., J.K. Henderson, and P.A. Miller, 1997: A Description of the Forecast Systems Laboratory's Real-Time Verification System (RTVS).  Preprints, 7th Conference on Aviation, Range, and Aerospace Meteorology, Long Beach, American Meteorological Society (Boston), J26-J31.

[top of page]