ساختار و تجزیه و تحلیل حساسیت از مدل های شکارچی - طعمه مبتنی بر فرد

The expensive computational cost of sensitivity analyses has hampered the use of these techniques for analysing individual-based models in ecology. A relatively cheap computational cost, referred to as the Morris method, was chosen to assess the relative effects of all parameters on the model’s outputs and to gain insights into predator–prey systems. Structure and results of the sensitivity analysis of the Sumatran tiger model – the Panthera Population Persistence (PPP) and the Notonecta foraging model (NFM) – were compared. Both models are based on a general predation cycle and designed to understand the mechanisms behind the predator–prey interaction being considered. However, the models differ significantly in their complexity and the details of the processes involved. In the sensitivity analysis, parameters that directly contribute to the number of prey items killed were found to be most influential. These were the growth rate of prey and the hunting radius of tigers in the PPP model as well as attack rate parameters and encounter distance of backswimmers in the NFM model. Analysis of distances in both of the models revealed further similarities in the sensitivity of the two individual-based models. The findings highlight the applicability and importance of sensitivity analyses in general, and screening design methods in particular, during early development of ecological individual-based models. Comparison of model structures and sensitivity analyses provides a first step for the derivation of general rules in the design of predator–prey models for both practical conservation and conceptual understanding.

Predator–prey interaction is one of the classic ecological issues that has been extensively described by mathematical models and increasingly simulated by means of spatially explicit computer models. This interaction is frequently described as numerical responses at the population level and as functional responses at the individual level. For the latter, the Holling Type II function [1] was found to adequately describe empirical observations in many species [2] and [3], and is most commonly applied in the mathematical description, but also has been adapted to simulation models. Current developments in individual-based models (IBMs) in ecology have opened new opportunities for testing the suitability of theoretical predator–prey interaction concept for the analysis of natural predator–prey systems and for practical conservation [4], [5] and [6]. IBMs have frequently been used to understand and predict population dynamics that emerge from individual traits. Examples include the prediction of population dynamics arising from food availability in water flea [7] or the role of individual home-range maintenance behaviour [8] in the assessment of population persistence of the Iberian lynx [9], tiger [10], and the Florida panther [11] for conservation purpose. One of the fundamental processes in the development of IBMs is the model analysis. This step involves various approaches and techniques such as the robustness test, statistical analysis, sensitivity analysis, etc. [6]. In spite of the large number of studies employing IBMs for ecological and evolutionary processes that have been published in the last two decades [5], very few have been concerned with evaluating individual-based models by means of sensitivity analyses. In fact, sensitivity analyses might improve ecological models by investigating uncertainties in the parameters, helping us to take inference from the results, to understand the model itself, and to gain insight into the systems represented by the model [6] and [12]. IBMs sometimes involve many uncertain parameters during model development. To identify those parameters, which will have a major influence on the output of a model, the sensitivity of selected parameters is usually tested using the traditional “one factor at a time” (OAT) method [13]. For example, Karanth and Stith [14] as well as Nilsson [15] tested the effect of prey density and size on the dynamics of predator population or predation behaviour, while MacCarthy et al. [16] studied the effect of fecundity and the initial number of birds on the population viability of the helmeted honeyeater. A comprehensive sensitivity analysis of all parameters is considered to be a computational process that is not feasible for complex IBMs. Therefore, this kind of analysis is only recommended for relatively simple IBMs [6]. In addition, the use of the sensitivity analysis for IBMs has been neglected due to missing links between the purpose of IBMs and the inferences taken from the results of sensitivity analysis, as well as the usefulness of robustness tests for IBMs [6]. Sensitivity analysis methods vary with different techniques, ranging from local to global and from quantitative to qualitative sensitivity analysis. Among these techniques, screening methods have been recommended to deal with highly complex models [17] and [12]. In the study presented, the importance of sensitivity analysis as a crucial part in the early development of any ecological individual-based model is addressed. Two predator–prey models of different complexities were chosen in order to derive ecological implications for the particular predator–prey systems, to deduce possible generalisations for the parameterisation of such models and to test the feasibility of two sensitivity analysis methods during the process of IBM development.

The role of sensitivity analysis in the development of ecological models has often been neglected. Moreover, expensive computational costs may have hindered a comprehensive analysis of complex individual-based models. In this study, the Morris method served as a feasible instrument for gaining a sound overview of the relative importance of model parameters, using reasonable computational efforts. Factor screening should be recognised as an important step during the development of individual-based models. Screening methods, accompanied by the simple technique of gradually changing parameter values, provide a useful tool for gaining insight into both simple as well as complex ecological models.