Parallelization of forest landscape model to improve computational efficiency and simulation realism

Forest landscape models (FLMs) are computationally intensive because of complex spatial interactions simulated. The current FLMs use sequential processing that simulates from the upper left pixel of the landscape to the lower right pixel. Sequential processing has series shortcomings among which simulation time and realism are the bottlenecks. In this study, we present a parallel processing design embedded in the LANDIS forest landscape model. Specifically, we apply a spatial domain decomposition approach that assigns pixel subsets to individual cores, enabling parallel execution of species- and stand-level processes on each core, while dynamically reallocating subsets across cores to execute landscape-level processes, i.e., seed dispersal. We compare the simulation results between parallel and sequential processing to evaluate the effectiveness and performance of the new design. Our result showed that when the number of pixels reaches millions parallel processing will save about 32.0–64.6 % of the time than sequential processing, for a 200-year simulation at 10-year time step. When the simulation time step is 1 year for a 200-year simulation, parallel processing can save 64.6–76.2 % of the time compared with sequential processing. Parallel processing improves simulation realism because it simulates multiple blocks simultaneously and performs multiple tasks, which is closer to the reality of species-level, stand-level, and seed dispersal processes. This study highlights the potential of parallel processing in improving the computational efficiency and simulation realism of FLMs.

Abstract licence: CC BY-NC

40 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.