Search Results

You are looking at 1 - 4 of 4 items for

  • Author or Editor: H. Glahn x
  • Refine by Access: All Content x
Clear All Modify Search
H. R. Glahn

Abstract

Two adaptive logic models and a training algorithm for each are described. These models are tested on a meteorological prediction problem with the use of large developmental and test data samples. Discriminant analysis is used for comparison. It is found that ceiling heights at Washington National Airport could be forecast better with the discriminant analysis technique than with the adaptive logic models.

Full access
H. R. Glahn and J. O. Ellis

Abstract

Full access
William H. Klein and Harry R. Glahn

Experience over the past decade has shown that objective forecasts of local weather elements can best be obtained by using statistical methods to complement the raw output of numerical prediction models. One of the most successful techniques for accomplishing this is called Model Output Statistics (MOS). The MOS method involves matching observations of local weather with output from numerical models. Forecast equations are then derived by statistical techniques such as screening regression, regression estimation of event probabilities, and the logit model. In this way the bias and inaccuracy of the numerical model, as well as the local climatology, can be built into the forecast system. MOS has been applied by the Techniques Development Laboratory to produce automated forecasts of numerous weather elements including precipitation, temperature, wind, clouds, ceiling, visibility, and thunderstorms. In this paper, the derivation and operational application of MOS forecasts for each of these elements are discussed. Many of the products are transmitted nationwide over facsimile and/or teletypewriter; others are provided for internal use within the National Weather Service. Ultimately, a completely automated, computer-worded, local weather forecast will be produced routinely as part of a program for Automation of Field Operations and Services (AFOS).

Full access
N. E. Westcott, S. D. Hilberg, R. L. Lampman, B. W. Alto, A. Bedel, E. J. Muturi, H. Glahn, M. Baker, K. E. Kunkel, and R. J. Novak

In the midwestern United States, the summertime rise in infection rate by the West Nile virus is associated with a seasonal shift in the abundance of two mosquito populations, Culex restuans and Culex pipiens. This seasonal shift usually precedes the time of the peak infection rate in mosquitoes by 2–3 weeks and generally occurs earlier in the summer with above normal temperatures and later in the summer with below-normal temperatures. Two empirical models were developed to predict this seasonal shift in mosquito species, or the “crossover,” and have been run operationally since 2004 by the Midwestern Regional Climate Center located at the Illinois State Water Survey. These models are based on daily temperature data and have been verified by use of a unique dataset of daily records of mosquito species abundance collected by the Illinois Natural History Survey. An unfortunate characteristic of the original temperature models was that the crossover date often was reached with little or no lead time. In 2009, the models were modified to incorporate National Weather Service (NWS) model output statistics (MOS) 10-day temperature forecasts. This paper evaluates the effectiveness of these models to predict the crossover date and thus the period of increased risk of West Nile virus in the Midwest.

For the 8-yr period from 2002 to 2009, 6 yr had at least one model predicting the crossover within one week of the actual crossover date, and for 7 yr at least one of the model predictions was within 2 weeks of the actual crossover date. Incorporation of MOS temperature forecasts for a 10-day period, although not substantially changing the predicted crossover date, greatly improved the forecast lead time by about 9 days. From a disease management point of view, this improvement in advanced notice is significant. In 2009, there was an unprecedented early crossover date and a failed forecast. The poor forecast was likely caused by an unusually early summer prolonged and intense heat wave, followed immediately by a record cold July.

Full access