Astron. Astrophys. 364, 83-101 (2000)

Gaussian decomposition of the Leiden/Dwingeloo survey

I. Decomposition algorithm

U. Haud

Tartu Observatory, 61602 Tõravere, Tartumaa, Estonia (urmas@aai.ee)

Received 3 July 2000 / Accepted 7 September 2000

Abstract

A new, fully automatic computer program for decomposition into Gaussian components of large 21-cm HI line surveys is described. To solve, at least partly, some problems inherent to such decompositions, special attention is paid on the following features:

1. Several quite different solutions may often fit the observations almost equally well. To choose from these solutions, it is supposed that general properties of the hydrogen distribution are somewhat correlated at neighboring sky positions and therefore the program tries to find for corresponding profiles also similar decompositions.

2. With increasing complexity of the observed profiles the number of Gaussians in decompositions usually grows rapidly and the values of their parameters become mutually dependent. To reduce this problem, special means have been used to keep the number of Gaussians as small as possible.

The recent Leiden/Dwingeloo Survey of the galactic neutral hydrogen has been used as test data for this decomposition program. These tests have demonstrated that the program is able to decompose fully automatically all the profiles in this survey in a reasonable time, spending on an average about per profile on PII PC. The comparison of the obtained results with those by other authors shows that, in general, this program needs less Gaussians per profile than used in comparison sets. The analysis of the stability of the decomposition results seems to indicate that the major source of the uncertainties is the observational noise, but also the precise behavior of the program has certain influence on final results.

Key words: methods: data analysis surveys ISM: atoms, ions radio lines: ISM

This article contains no SIMBAD objects.

Contents

• 1. Introduction
• 2. Data preparation
  • 2.1. Comparison of repeated observations
  • 2.2. Noise level from the Savitzky-Golay filtering
  • 2.3. Noise level from signal-free regions of the profiles
  • 2.4. Comparison of the noise estimates and the weights
• 3. Gaussian decomposition
  • 3.1. Adding new Gaussians
  • 3.2. Removing Gaussians from decomposition
  • 3.3. Selecting the next profile
• 4. Stability of the obtained decompositions
  • 4.1. Comparison with other authors
  • 4.2. Internal stability of results
• 5. Conclusions
• Acknowledgements
• References

© European Southern Observatory (ESO) 2000

Online publication: December 15, 2000
helpdesk.link@springer.de