Data driven modelling and simulation of wetland dynamics

A wetland is an ecosystem, which is inundated by water. Several biogeochemical processes related to wetlands place it into an important environmental system class. This study employs for the first time, a data-driven methodology, specifically one based on system identification, to develop a nonlinear-autoregressive-exogenous-input model structure corresponding to the tropical wetland dynamics, as an alternative to the classical topographic modelling approaches. Altogether 94 wetland sites from Amazon, Africa, and Asia constitute the model structure development. The topographical variations among the wetlands entail a unique set of model parameters for each wetland site, which is achieved by using the approximate Bayesian computation sequential Monte Carlo method. The performance of the single model structure is subsequently evaluated by simulating all the sample wetland sites. The obtained wetland model structure in this study can be used to parameterise any tropical wetland site as well as simulate their dynamical characteristics.