Error propagation in an integrated spatially-explicit individual-based model

Abstract

Error propagation is an important consideration in individual-based modeling, but it has been considered insufficiently studied. We investigated the propagation of error due to uncertainty in initial conditions using a previously published spatially-explicit individual-based model that simulates infestation of sorghum by an invasive pest aphid, Melanaphis sorghi (Theobald) [sorghum aphid; previously published as sugarcane aphid, Melanaphis sacchari (Zehntner)]. We initialized the model with aphids in three alternative initial infestation locations using one of three neighboring cells and analyzed the resulting model outputs in three pairwise scenario comparisons. The spatio-temporal patterns of aphid infestation, as estimated by timing and probability of first infestation, were statistically significantly different between the scenarios, but the differences were locally restricted and scenario-dependent. In particular, the two pairwise differences between scenarios originating from neighboring cells indicated that error propagation through the studied system depends not only on the physical distance between the alternative initialization cells (i.e., the extent of the initial spatial uncertainty), but also on the actual location of the model initialization cells, likely reflecting differences in the environmental characteristics of those locations. Despite the statistical significance, the differences were small from the perspective of practical application with few highly localized exceptions. The spatio-temporal trajectories of the propagating error indicate that, within the examined range, the spatial uncertainty in model initialization has a low effect on the timing and probability of first infestation, and the propagation of the error in the observed variables is limited by the system itself.