Network design for verification of ceiling and visibility forecasts

Cloud ceiling height and visibility forecasts are of importance for use in flight planning. This is particularly true for the general aviation community, where pilots who are not instrument rated must avoid flying in areas with poor visibility conditions. These forecasts rely on data obtained from A...

Full description

Saved in:

Bibliographic Details
Published in:	EnvironMetrics (London, Ont.) Vol. 17; no. 6; pp. 575 - 589
Main Authors:	Gilleland, Eric, Fowler, Tressa L.
Format:	Journal Article Conference Proceeding
Language:	English
Published:	Chichester, UK John Wiley & Sons, Ltd 01-09-2006 Wiley
Subjects:	Animal, plant and microbial ecology Biological and medical sciences cloud ceiling height coverage design Earth sciences Earth, ocean, space Engineering and environment geology. Geothermics Exact sciences and technology flight rules forecast verification Fundamental and applied biological sciences. Psychology General aspects. Techniques METAR Methods and techniques (sampling, tagging, trapping, modelling...) Pollution, environment geology spatial correlation visibility United States USA algorithms cloud ceiling height visibility Flight testing spatial correlation clouds networks North America coverage design lead Weather METAR planning instruments correlation flight rules forecast verification strategy Community Distance
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Cloud ceiling height and visibility forecasts are of importance for use in flight planning. This is particularly true for the general aviation community, where pilots who are not instrument rated must avoid flying in areas with poor visibility conditions. These forecasts rely on data obtained from Aviation Routine Weather Report (METAR) stations. Two issues arise in the verification of ceiling and visibility forecasts: (i) forecasts with short‐to‐no lead time (i.e., nowcasts) require use of the same METAR observations for forecast creation and verification, and (ii) placement of reporting stations may introduce bias in the verification analysis. Independent data is necessary for an accurate verification, but because the same data are used to verify the nowcasts as are used to create the nowcasts, the data are not independent. A common strategy is to divide the data into testing and training sets to obtain quasi‐independent data. The second issue arises from the fact that the METAR stations are unevenly spaced so that a forecast verification is potentially biased for areas with numerous stations; thereby giving little or no information on forecast performance elsewhere. In this paper, we present a network thinning strategy to obtain a reasonable subset network of METAR stations that is more ‘regularly’ spaced. The strategy involves three steps: (i) an initial inspection of graphs of pair‐wise agreement between stations by distance, (ii) a coverage design algorithm, and finally (iii) inspection of the distributions of verification scores among various designs and the original network. We give test examples for two different regions in the United States. In each case, it seems reasonable to reduce the network size by nearly half of the original number of stations. The resulting designs are regularly spaced, and verification results for the original network are very similar to those for the thinned networks. Copyright © 2006 John Wiley & Sons, Ltd.
Bibliography:	ArticleID:ENV765 istex:A3EA544613EB61810B93248A91F301387DCF987C ark:/67375/WNG-3R3X6CV2-6 Federal Aviation Administration National Science Foundation ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 ObjectType-Article-2 ObjectType-Feature-1
ISSN:	1180-4009 1099-095X
DOI:	10.1002/env.765