Accession Number:

ADA277765

Title:

A Neural Network Forecast Demonstration

Descriptive Note:

Corporate Author:

NATIONAL WEATHER SERVICE FORECAST OFFICE BISMARCK ND

Personal Author(s):

Report Date:

1990-11-01

Pagination or Media Count:

10.0

Abstract:

This experiment demonstrates the efficacy of using neural networks to solve certain forecast problems. There is a presumption, of course, that similar weather patterns will lead to similar weather events in time. This, as we all know, is not always the case. But, as we become more detailed in describing a pattern i.e., using more and more meteorological parameters we can only improve this approach. To analyze and recognize these complex patterns we will need to rely more and more on neural networks and their advanced pattern recognition capabilities. Imagine a neural network whose input is the actual 500 mb heights and whose output is the 12-48 hour forecast of heights at each grid point. Also, imagine that its training pairs include all the upper air data from the last 30 years

Subject Categories:

  • Meteorology
  • Computer Hardware
  • Computer Systems
  • Cybernetics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE