Close this window

Short-Term Ceiling and Visibility Forecast Systems

(last updated 8/3/2005)

Overview

Two experimental short-term (1 to 12 hours) ceiling and visibility forecast methodologies have been developed in an effort to improve upon persistence forecasts:

• An observations-based (hereafter referred to as OBS-based) guidance system developed by Vislocky and Fritsch that uses a network of surface observations as the primary source of predictive information to improve very short-term forecasts of aviation weather parameters
• A forecasting method for ceiling heights and visibilities using a high resolution NWP model (the Penn State/NCAR MM5) developed by Porter and Seaman

The OBS-based system is discussed below. For information about the MM5 system, please see Porter, Christopher, W. and N. L. Seaman, 1996. (Short-term high-resolution forecasting of cloud ceiling heights and visibilities. Preprints, 6th Conference on Aviation Weather Systems, 331-336)

OBS-based System Overview

The OBS-based system was developed and tested for predicting very short-term ceiling and visibility conditions at various locations in the eastern U.S. The ultimate goal of the research is to show that the use of a network of observations not only improves skill levels greater than what can be obtained by using persistence (even for 1-hr projections), but also produces skill levels above and beyond the traditional MOS method.

Three forecast methods are compared in this study:

1. An OBS-based forecast system in which the future weather conditions are predicted through a statistical relationship between a network of surface weather observations at the initial time in conjunction and climatological variables
2. A traditional MOS-based technique in which future weather conditions are statistically correlated to NWP model output, climatic variables, and the latest surface observation from the forecast site
3. Use of persistence climatology (otherwise known as conditional persistence or conditional climatology) in which a future weather variable for a given location is predicted through a statistical relationship with the initial value of that variable from the forecast site and several climatic terms

Each forecast method was tested on two years of independent data for 25 stations in the eastern U.S. Two parameters (ceiling and visibility) were forecast for eight thresholds, three lead times (1, 3, 6 hr), and two initialization times (0300 and 1500 UTC).

Verifications show that the OBS-based method:

• Is far superior to persistence climatology at all lead times (5%-20% improvement)
• Outperforms the MOS-based technique at the 1- and 3-hr projections with skill increases averaging 4%
• Performance of the OBS-based and MOS-based systems is similar at the 6-hr projection, which appears to be near the crossover point when the MOS guidance becomes more important than the observations in terms of predictive input.

Advantages

One of the more notable aspects of these results is the gain over persistence climatology made by the OBS-based technique, even at the 1-hr lead-time. Since persistence climatology is widely known as a formidable competitor in the very short range, output from an OBS-based system would provide useful guidance to operational forecasters and users on a routine basis.

In addition, with an OBS-based guidance system, the automated procedure can be designed to provide updated predictions every hour as new observations arrive. Also, the system is portable, relatively easy to implement, is not tied to any NWP model, and can be run in only a few seconds on a mere personal computer.

Improving the Technique

The accuracy and operational utility of the OBS-based techniques can be improved significantly by:

• Including several other predictive sources such as radar data, satellite imagery, and upper-air observations
• Using an optimal combination of the OBS-based and MOS-based predictions, either by using the OBS-based forecast as a potential predictor in a MOS-based equation or by computing the optimal blend on a short sample of independent data
• By finding a better way to incorporate motion into the observations. In this study, only the climatological movement of the observations is implicit in the regression equations.
• Using observational predictors from fixed points on a regularly analyzed grid instead of predictors from specific stations. This would result in a more accurate method of substituting for missing observations than the simplified nearest-neighbor approach used in this study. Most importantly though, such a strategy would effectively eliminate the adverse effects of station openings and closings on the operational system and would permit introduction of the vast new array of automated sensors. Thus, once such an operational OBS-based system is initially established, relatively little maintenance would be required.

References

Leyton, Stephen M., and J. M. Fritsch, 2000. Short-term probabilistic forecasts of ceiling and visibility utilizing high-frequency surface observations. Preprints, 9th Conference on Weather Analysis and Forecasting/15th Conference on Numerical Weather Prediction, 4 pp.

Vislocky, Robert L., J. Michael Fritsch, 1997: An Automated, Observations-Based System for Short-Term Prediction of Ceiling and Visibility. Weather and Forecasting: 12, No. 1, pp. 31-43
http://ams.allenpress.com/amsonline/?request=get-document&issn=1520-0434&volume=012&issue=01&page=0031

Vislocky, Robert L., J. Michael Fritsch, 1995: Generalized Additive Models versus Linear Regression in Generating Probabilistic MOS Forecasts of Aviation Weather Parameters. Weather and Forecasting: 10, No. 4, pp. 669-680
http://ams.allenpress.com/amsonline/?request=get-pdf&file=i1520-0434-010-04-0669.pdf