Interpreting outputs of agent-based models using abundance–occupancy relationships

Abstract

a b s t r a c tReliable assessments of how human activities affect wild populations are essential for effective natural resource management. Agent-based models provide a powerful tool for integration of multiple drivers of ecological systems, but selecting and interpreting their output is often challenging. Here, we develop an indicator (the AOR-index) based on the abundance-occupancy relationship to facilitate the interpretation of agent-based model outputs. The AOR-index is based on the distribution of individuals in the landscape translated into the number of individuals in each cell of a regular grid. The proportion of grid cells with at least one individual is used to quantify occupancy and the mean number of individuals in occupied cells is used to quantify abundance. The AOR-index is a two-dimensional index giving the relative change in abundance and occupancy in response to a scenario (e.g. a change in land use or climate). We systematically modify a digital version of a real landscape to produce a set of artificial landscapes differing only in the degree of landscape fragmentation. We test how these different landscapes affect the AOR-index of six model animal species in four different land use scenarios using an agent-based model framework (ALMaSS). Our results suggest that the AOR-index is a sensitive tool to demonstrate how different species respond to particular land-use scenarios. The bird and mammal species generally showed larger responses than the invertebrates and changes in abundance and occupancy were often of different magnitude. The different responses are caused by species-specific habitat requirements and dispersal abilities, but the importance of such life history traits depend on landscape structure. Hence, predictions of species-specific responses to land-use changes in terms of abundance and occupancy are greatly improved by incorporation in a model framework taking spatial and temporal dynamics into account. The AOR-index facilitates the evaluation of multiple scenarios and allows for multi-species assessments. Its use, however, still requires identified management goals in order to evaluate scenario responses.

Towards a landscape scale management of pesticides: ERA using changes in modelled occupancy and abundance to assess long-term population impacts of pesticides

“…To alleviate this problem, ALMaSS output was used to create an index developed from the Abundance to Occupancy Relationship, AOR (Gaston et al, 2000), often studied in macro-ecology. The AOR-index as measurement endpoint has the advantage that it provides a clear picture of the changes in range and density of animals relative to a baseline condition (Hoye et al, 2012). Previously, ALMaSS results have been expressed as changes in local abundance and spatial distribution as described by the univariate Ripleys K(r) (Jepsen et al, 2005).…”

“…Occupancy was quantified by overlaying the landscape by a regular grid and quantifying the proportion of grid cells containing (super-) individual female beetles using the procedure described by Hoye et al (2012. The aim of this procedure is to obtain a grid-cell size large enough to allow more than one (super-) individual to be present in each grid cell but small enough also to avoid occupancy and abundance being identical. Two rules were used to identify the grid cell size: 1) in the baseline scenario approximately 50% of the cells should be occupied; 2) if possible within the above constraints the grid size chosen should result in a mean occupancy of >5.…”

Applying a biocomplexity approach to modelling farmer decision‐making and land use impacts on wildlife

“…The impact of the experimental scenarios on each species was evaluated using two measures: a change in a mean population size in the last 5 years of a simulation relative to the baseline and the abundance‐occupancy relationship, i.e., AOR‐index (Høye, Skov, & Topping, ). The numbers of animals used to determine AOR—index are recorded on June 1st for all species except the carabid beetle (April 16th) and on January 1st for the population size in each year of a simulation; in both cases the numbers refer to adult females.…”

Complex Adaptive Systems Modeling: A multidisciplinary Roadmap

“…Qu et al present a digital orange tree simulation using an ABM (Qu et al 2012). Another work by Høye et al demonstrates systematic modification of the digital version of a real landscape to produce artificial landscapes (Høye et al 2012). Iantovics presents a large-scale hybrid medical diagnosis system (Iantovics 2012).…”