Theory and application of discrete hidden Markov models in plant community ecology and restoration

The simple stationary Markov model (SMM) is a good model of some ecological dynamics, such as classic Clementsian succession in time toward a climax. It is likely, however, that ecological sequences are the result of underlying ecological processes that determine observational probability and that the dynamics of the observed community state may mask the dynamics of the underlying (hidden) ecological state. The application of hidden Markov models (HMM) is proposed for two reasons: (1) they can have multiple states with observations that need not converge on a stable configuration and (2) the hidden states allow for the consideration of underlying ecological processes. A comparative analysis is made between the well-known SMM and the lesser-known HMM using a range of hypothetical species response types with concentration on the prediction of ecological observation sequences and the detection of underlying ecological processes. The HMM provides similar predictive ability to that of the SMM in the case of monotonic dynamics but shows considerably improved performance for non-monotonic responses. The HMM also provides increased interpretive capabilities. (Abstract shortened by UMI.)...