A spatially independent seed dispersal design for forest landscape models

Seed dispersal is a spatially continuous process occurring over areas ranging from tens to thousands of square meters. Current seed dispersal designs (CDD) in forest landscape models (FLMs) have notable limitations. CDD may fail to disperse seeds outward when cell size exceeds dispersal distance and commonly assumes seed sources are centrally located within cell. The problems can accumulate spatially and temporally leading to large prediction errors. We propose a spatially independent dispersal design (SID) that overcomes the pitfalls of CDD. SID assumes seed sources occur anywhere in a cell and tracks the time required for a tree species to traverse a cell. This technically allows SID to operate at any cell sizes in FLMs. We validated the SID approach using artificial landscapes to isolate dispersal patterns from confounding factors such as environmental heterogeneity and competition, ensuring that dispersal patterns solely reflect SID and species attributes. We applied SID to a temperate forest landscape to evaluate the uncertainties in predicted tree species distributions. We demonstrated that the SID algorithm achieves spatial independence when predicting tree species distributions and their changes. In simulations of the temperate forest landscape, the average difference in percent cover across spatial resolutions ranging from 100 to 400 m was less than 5% under SID, compared to over 12% under CDD. Previous prediction studies using the CDD algorithm at coarse resolutions may result in deviations exceeding 20% in predicting the percent cover of mid- and late-successional species. The SID algorithm enhances the simulation realism of seed dispersal in FLMs, especially at coarse spatial resolutions. The design flexibility makes it suitable for integration into broader terrestrial biosphere models, improving predictions under climate change and disturbances.