A Neural Network Forecast Demonstration

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

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 1990
Accession Number
ADA277765

Entities

People

  • Richard S. Leblang

Organizations

  • National Weather Service

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Accuracy
  • Artificial Intelligence
  • Computations
  • Computers
  • Detectors
  • Expert Systems
  • Fuzzy Logic
  • Neural Networks
  • North America
  • North Dakota
  • Pattern Recognition
  • Recognition
  • Training
  • Weather Forecasting

Fields of Study

  • Environmental science

Readers

  • Aviation Safety and Air Traffic Management
  • Educational Psychology
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks