CREM ABM Initiatives
Coral reef decision-makers need to know how much of what kinds of stressors or management interventions applied in what areas and in what combinations are likely to result in what kinds of impacts to the reef and its dependents. Deterministic predictions of specific variable values (e.g. biomass or harvests of a given species) are not likely to be reliable in such a complex and spatially heterogeneous system. CREM's approach is to use multi-scale agent-based models that capture the major elements and interactions of the coral reef and relevant socioeconomic systems. By varying model inputs across the ranges of their uncertainties in model iterations, we will map the ‘potential outcome spaces’ associated with disturbances and interventions among sets of variables, in terms of what is likely versus what is improbable.
Specifically, CREM's primary objective is to develop, verify, validate and evaluate a multi-scale, extensible, adaptable simulation system which captures the major features of a given coral reef biophysical-socioeconomic system and provides useful guidance for participatory management decision-making within a GIS framework.
Conceptual Approach
The scale-dependent nature of reef ecosystems calls for a
multi-scale modeling approach, as recommended by H.T. Odum and
others, but rarely implemented. Agent-based models have the
advantages that they account for individual or aggregate group
variability, minimize reliance on assumptions of equilibrium and
homogeneity (the mean field assumption), easily adapt to spatial and
GIS-based applications, and facilitate interdisciplinary modeling by
circumventing the need for a common currency. CREM will build models as
set of coupled sub-models at the whole reef (at 0.5 km2
resolution) and patch (at 1
cm2 resolution) scales. The whole-reef sub-model will be
grid-based, with each grid containing “blocks” of biomass of
plankton, algae, seagrass, coral, herbivores, planktivores,
piscivores and detritovores. Human interactions will involve
fishers, tourists, polluters and others as identified in field
studies. The patch sub-model framework will be run with several
broad setups to represent the major reef habitats, in the manner of
sampling with large quadrats. The patches will be based on
continuous (non-grid) space and include the most abundant
characteristic macro-species, as well as biomass blocks for smaller
organisms. Human interactions will be included here as in the larger
scale. Runs of the large-scale sub-model will be alternated with
runs of the patch sub-models, each providing inputs to the other.
The ABM models in the CREM Lab are always subjected to extensive verification, in terms of sensitivity analyses, the use of test-based module development, etc. The validation process involves parameterization and field testing in conjunction with a multi-year interdisciplinary participatory coastal management project focused on some Caribbean states. CREM's previous ecological and socioeconomic field surveys on the regions of interest will be used for further validation in a retrospective context. Validation in general will involve the matching of potential outcome spaces to realized outcomes using statistical approaches. Evaluation will involve determining the usefulness of the model in the process of resource planning and implementation of the plan on some study sites in the Caribbean.
The CREM Lab has already developed online innovative GIS frameworks, a large-scale ABM reef model, and fine-resolution models of coral dynamics, coral disease propagation and macro-algal interactions. CREM intends to further develop these products and new components into an ecosystem-based decision-support tool. This decision-support tool will be designed specifically for reef managers and researchers and will include other components such as economic, quality of life, hydrodynamic, connectivity and watershed modules.
We will soon make public some of the ABM models that CREM had been working one. Please return to this page again.