DocumentCode
326170
Title
Genetic algorithm method for determining temperature profiles
Author
Buczak, Anna L. ; Barrett, J.J.
Author_Institution
Allied-Signal Inc., Morristown, NJ, USA
fYear
1998
fDate
4-9 May 1998
Firstpage
142
Lastpage
147
Abstract
A genetic algorithm is used to determine atmospheric temperature profiles from infrared observations of ambient carbon dioxide gas in the atmosphere. The observation range is divided into four range bins and the goal is to predict with reasonable accuracy the temperature in each range bin. An overdetermined system of equations corrupted by measurement noise describes the horizontal temperature profile. The use of a classical linear least squares fit to solve the equations results in predicted temperature errors over 30 K for small amounts of measurement noise. To overcome this problem, a genetic algorithm was developed using fitness function correction terms based on the authors´ experience with the atmospheric temperature profiles and it was tested on a simulated data system. Excellent results were obtained demonstrating the usefulness of the genetic algorithm approach for solving this type of problems
Keywords
atmospheric temperature; genetic algorithms; geophysics computing; least squares approximations; noise; radiometry; 30 K; CO2; ambient CO2 gas; atmospheric temperature profile determination; classical linear least squares fit; fitness function correction terms; genetic algorithm; horizontal temperature profile; infrared observations; measurement noise; simulated data system; Accuracy; Atmosphere; Atmospheric measurements; Carbon dioxide; Equations; Genetic algorithms; Least squares methods; Noise measurement; System testing; Temperature distribution;
fLanguage
English
Publisher
ieee
Conference_Titel
Evolutionary Computation Proceedings, 1998. IEEE World Congress on Computational Intelligence., The 1998 IEEE International Conference on
Conference_Location
Anchorage, AK
Print_ISBN
0-7803-4869-9
Type
conf
DOI
10.1109/ICEC.1998.699491
Filename
699491
Link To Document