Skip to main content
Log in

Regularized gaussian mixture models for effective short-term forecasting of rainfall patterns

  • Reviewed Original Papers
  • Published:
e&i Elektrotechnik und Informationstechnik Aims and scope Submit manuscript

Abstract

In this paper, we report about the application of regularized Gaussian mixture models in the task of predicting rainfall patterns based on ground radar images. The mixture models are used to obtain a parametric estimate of the rainfall density distribution that permits prediction through linear extrapolation of the parameters. Regularization is introduced to ensure the identifiability of the parameters and the linear relationship needed for prediction. On a set of 6 independent image sequences we demonstrate that this type of regularization can indeed significantly improve the prediction performance. Short-term prediction of rainfall patterns is extremely important for meteorological and environmental reasons. The type of prediction described in this paper is being used in a commercial package based on a low-cost ground radar system designed to provide optimal in-flow forecasts for sewage treatment plants.

Zusammenfassung

Dieser Artikel berichtet über die Anwendung Gauß’scher Mischmodelle auf die Vorhersage von Regenfallmustern aus Radaraufnahmen. Die Mischmodelle werden zur parametrischen Schätzung von Regenfalldichteverteilungen verwendet, die Vorhersage durch lineare Extrapolation erlauben. Die Regularisierung dient zum Erzwingen der Identifizierbarkeit der Parameter und ihres linearen Zusammenhangs, der zur Vorhersage benötigt wird. Anhand von sechs unabhängigen Bildsequenzen wird demonstriert, dass diese Art von Regularisierung tatsächlich die Vorhersageperformanz signifikant verbessern kann. Kurzzeitvorhersage von Regenfällen ist aus meteorologischen und umweltrelevanten Gründen von enormer Wichtigkeit. Die Art der Vorhersage, wie sie in diesem Artikel beschrieben wird, wird in einem kommerziellen Paket auf der Basis eines kostengünstigen Radarsystems eingesetzt, das zu Zwecken der optimalen Vorhersage des Abwasserflusses in Wasserwiederaufbereitungsanlagen dient.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

References

  1. Bishop, C. M.: Neural networks for pattern recognition. Oxford: Clarendon Press. 1995.

    Google Scholar 

  2. Broomhead, D. A., Lowe, D.: Multivariable functional interpolation and adaptive networks Complex Systems 2 (1988), pp. 321–355.

    MATH  MathSciNet  Google Scholar 

  3. Denœux, T., Rizand, P.: Analysis of radar images for rainfall forecasting using neural networks, Neural Computing and Applications 3 (1995), pp. 50–61.

    Article  Google Scholar 

  4. Einfalt, T., Denœux, T., Jacquet, G.: A radar rainfall forecasting method designed for hydrological purposes. Journal of Hydrology 114 (1990), pp. 229–244.

    Article  Google Scholar 

  5. Møller, M.: Efficient training of feed-forward neural networks. Ph. D. Thesis. Aarhus University, Denmark. 1993.

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Rights and permissions

Reprints and permissions

About this article

Cite this article

Dorffner, G., Schellner, K. & Prem, E. Regularized gaussian mixture models for effective short-term forecasting of rainfall patterns. Elektrotech. Inftech. 118, 371–378 (2001). https://doi.org/10.1007/BF03157842

Download citation

  • Issue Date:

  • DOI: https://doi.org/10.1007/BF03157842

Keywords

Schlüsselwörter

Navigation