Investigating the Influence of Spatial and Temporal Granularities on Agent-Based Modeling

Abstract

Epidemic agent‐based models (ABMs) simulate individuals in artificial societies that are capable of movement, interaction, and transmitting disease among themselves. ABMs have been used to study the spread of disease at various spatial and temporal scales ranging from small communities to the world, over days, months, and years. The representations of space and time often vary between different epidemic ABMs and can be influenced by factors such as the size of a modeled population, computational requirements, population environments, and disease‐related data. The influence that the representations of space and time have on epidemic ABMs is difficult to assess. Here we show that the finest representations of space and time—termed spatial and temporal granularities (STGs)—in a parsimonious ABM affect speed, intensity, and spatial spread of a synthetic disease. Specifically, we found disease spread faster and more intensely as spatial granularity is coarsened, whereas disease spread slower and less intensely as temporal granularity is coarsened in a parsimonious ABM. Our study is the first to use the same epidemic ABM to examine the influence of STGs. Our results demonstrate that STGs influence ABM dynamics including early disease burnout and that an interrelationship exists between the coarsening of STGs and the speed and intensity at which disease spreads. Our parsimonious ABM is extended based on a structured community model and we found STGs also influence ABM dynamics in a more realistic context that includes hierarchical movement. Broadly, our study serves as a basis for further inquiry toward the influence of space–time representations on more realistic models that include multiscale mobility, routine movements (e.g., commuting), and heterogeneous population distributions.

Geoexpression: A Petri network framework for representing geographic process concurrency

“…Therefore, providing suitable polymorphism over both geo‐objects and geo‐fields for all functions, in addition to ensuring high computational efficiency for any implementation, and concurrent accessibility by different threads, is a very difficult challenge. Internally performing conversions between geo‐atomic field and geo‐atomic object representations within a model function introduces potential errors related to spatial and temporal granularities and the modifiable area unit problem (Shook & Wang, ). Further research will be necessary to determine, from a computing standpoint, if polymorphic geographic modeling functions should be developed in lieu of separate models for different data ontologies.…”

“…Computational models of geographic processes are used to understand complex geographic dynamics. These models are often coarsened due to limitations in fine‐scale understanding, data, and computational capabilities (Shook & Wang, ). Models may be created with different spatial and temporal granularities to confine process models to some representative scale.…”

An integrated framework of global sensitivity analysis and calibration for spatially explicit agent‐based models

“…To address these challenges, we have developed an integrated framework of global sensitivity analysis and calibration (GSA‐CAL), and applied the framework to a spatially explicit ABM of influenza transmission as a case study. The framework takes advantage of cyberGIS and associated high‐performance computing to enable computationally intensive simulations for GSA and calibration (Shook & Wang, 2015; Wang, 2010; Wright & Wang, 2011). This cyberGIS approach to GSA could help to resolve the computational intensity arising from Monte Carlo simulations required prior to performing UA‐GSA (Kang et al., 2020).…”

Future Developments in Geographical Agent‐Based Models: Challenges and Opportunities

“…Agent‐based models were first formally proposed in the early 1990s (e.g., Epstein and Axtell 1996) but their lineage goes back much further to the development of models of individual locational decision‐making in the 1950s and 1960s in the influential work of Hagerstrand (1953), Donnelly et al (1964), and Schelling (1969) among others. They are now reaching a point of acceptance as a research tool across the geographical and social sciences, exploring such phenomena as epidemiology (Shook and Wang 2015), invasive species (Anderson and Dragićević 2020), settlement patterns (Bura et al 1996), and segregation (Benenson and Hatna 2011) (see Polhill et al 2019 for further discussion on the applications of ABMs and their use in policy).…”