Hidden Markov models for spatial pattern analysis

Abstract

Hidden Markov models for spatial pattern analysis. Hidden Markov models (HMM) constitute a flexible modelling tool, originally used in the  field  of  automated  speech  recognition,  that  have  found  wide  application  in  the  last  years  in  many  scientific  and  technical  problems, although their use in ecology is still scarce. In this review, the essential elements of HMM are described, the basic algorithms that facilitate their estimation are presented and some recent applications are pointed out, with emphasis on the possibilities that HMM offer in analysing complex  spatial  patterns,  as  they  allow  incorporating a  priori  information  about  the  system  into  the  modelling  process.  An  example  of application is presented where HMM are used to model vegetation transects with presence-absence data, aimed at analysing disturbances in the spatial distribution of the vegetation after a wildfire in a semiarid zone.