Search

Abstract: This paper will discuss TRIPLEX-Flux, a process-based model that estimates net ecosystem production (NEP) as well as analyzing the level of sensitivity of the model''s response by simulating CO2 flux in an existing old black spruce BOREAS site in central Canada. The research objectives were: (1) to test the TRIPLEX-Flux model simulations against flux tower measurements; and (2) to examine parameter and input variable effects on model response via sensitivity analysis. Validation of NEP data at 30min intervals derived from tower and chamber measurements showed that the NEP data from the model corresponded well with the measured NEP from the BOREAS site (R 2 >0.65). Sensitivity analysis demonstrated different levels of sensitivity between morning and noon periods and from the current to doubled atmospheric CO2 concentrations. Additionally, the comparison of different algorithms to calculate stomatal conductance showed that the NEP predicted by the model using the iteration algorithm was consistent with the results using a constant C i/C a of 0.74 for current and 0.81 for doubled CO2 concentrations. Varying parameter and input variable values by ±10% resulted in a similar model response between morning and noon periods (less than or equal to 27.6% and 27.4%, respectively). Most parameters were more sensitive at noon than they were in the morning except for those that were correlated with air temperature, suggesting that air temperature has considerable influence over model sensitivity for these parameters/variables. Air temperature effect was greater under doubled than it was under current atmospheric CO2 concentrations. In contrast, model sensitivity to CO2 decreased under doubled CO2 concentrations. [Copyright &y& Elsevier]

Abstract: In this paper we use Markov chain Monte Carlo (MCMC) techniques to carry out Bayesian inference for piecewise stochastic logistic growth models using discretely observed data sets, which allows us to fit models for time series data, including data on fish productions and yields, with structural changes. The estimation framework involves the introduction of latent data points between each pair of observations, and the use of MCMC techniques, based on the Gibbs sampling algorithm, in conjunction with the Euler–Maruyama discretization scheme. These methods are used to sample from the posterior distribution using exact bridges, allowing estimation of the model parameters including switching point(s). We apply our methods to examples involving both simulated data and real data for fisheries resources management. [Copyright &y& Elsevier]

Abstract: Modeling and simulation is essential for understanding complex ecological systems. However, knowledge of the structure and behavior of these systems is limited, and models must be revised frequently as our understanding of a system improves. Moreover, the dynamic, spatial distribution of activity in very large systems necessitates mapping natural mechanisms as logically as possible onto computer structures. This paper describes theoretical and algorithmic tools for building component-based models and simulations of dynamic spatial phenomena. These methods focus attention on and exploit the irregular distribution of activity in ecological processes. We use the DEVS formalism as the basis for a component-based approach to modeling spatially distributed systems. DEVS is a mathematical theory of discrete-event systems that is well suited for describing large systems that are described by small parts with irregular, short-range interactions. This event-based approach to modeling leads naturally to efficient simulations algorithms which focus on the active parts of a large model. Ecological modeling benefits from these efficient the simulation algorithms and the reusability of the model’s basic components. Our event-based method is demonstrated with a physics-based model of fire spread. [Copyright &y& Elsevier]

Ecological network analysis (ENA) is a systems-oriented methodology to analyze within system interactions used to identify holistic properties that are otherwise not evident from the direct observations. Like any analysis technique, the accuracy of the results is as good as the data available, but the additional challenge is that the data need to characterize an entire ecosystem''s flows and storages. Thus, data requirements are substantial. As a result, there have, in fact, not been a significant number of network models constructed and development of the network analysis methodology has progressed largely within the purview of a few established models. In this paper, we outline the steps for one approach to construct network models. Lastly, we also provide a brief overview of the algorithmic methods used to construct food web typologies when empirical data are not available. It is our aim that such an effort aids other researchers to consider the construction of such models as well as encourages further refinement of this procedure. [Copyright &y& Elsevier]