Evolutionary optimisation techniques to estimate input parameters in environmental emergency modelling

Kerstin Wendt, Mónica Denham, Ana Cortés, Tomàs Margalef

Research output: Contribution to journalArticleResearchpeer-review

4 Citations (Scopus)

Abstract

Parameter estimation in environmentalmodelling is essential for input parameters, which are difficult or impossible to measure. Especially in simulations for disaster propagation prediction, where hard real-time constraints have to be met to avoid tragedy, the additionally introduced computational burden of advanced global optimisation algorithms still hampers their use in many cases and poses an ongoing challenge. In this chapter we demonstrate how modifications of a Genetic Algorithm (GA) are able to decrease time-consuming fitness evaluations and hence to speed up parameter calibration. Knowledge from past observed catastrophe behaviour is used to guide the GA during various phases towards promising solution areas resulting in a fast convergence. Together with parallel computing techniques it becomes a viable estimation approach in environmental emergency modelling. Encouraging results were obtained in predicting forest fire spread. © 2011 Springer-Verlag Berlin Heidelberg.
Original languageEnglish
Pages (from-to)125-143
JournalStudies in Computational Intelligence
Volume359
DOIs
Publication statusPublished - 8 Aug 2011

Keywords

  • environmental modelling
  • forest fire spread prediction
  • input parameter estimation
  • knowledge guided Genetic Algorithm

Fingerprint Dive into the research topics of 'Evolutionary optimisation techniques to estimate input parameters in environmental emergency modelling'. Together they form a unique fingerprint.

  • Cite this