Abstract: In forest dynamics models, the intensive computation and load involved in the simulation of seed dispersal can become unbearably huge for large-scale forest analysis. To solve this problem, we propose a multi-resolution algorithm to compute seed dispersal on GPU. By exploiting the computation parallelism of seed dispersal, the computation of the whole forest plot is divided into multiple small plot cells, which are computed independently by parallel threads on GPU. To further improve the calculation efficiency with limited threads scale for GPU computation, we propose a hierarchical method to cluster the plot cells into a multi-resolution form according to the biological curves of tree seed dispersal. Experimental results show that our algorithm not only greatly reduces computational time but also obtains comparably correct results as compared to the naive GPU algorithm, which makes it especially suitable for large-scale forest modeling.

Key words: GPU, Seed dispersal, Large-scale, Multi-resolution, Data clustering

Please provide your name, email address and a comment

Your Name: Affili: E-Mail Address:

Comment (Please comment in English):


Verification Code(click to change a pic):