Orchard level simulation of fruit tree light interception

摘要

Several functional-structural plant models have been built for studying fruit trees, most of which focus on single trees. Simulating an entire orchard has been a difficulty due to two challenges: one is in computing, limited by processor capability and memory capacity, as well as the sequential nature of conventional approaches; the other is the modelling of plant-plant and plant-environmental interactions. The purpose of this work is to address the first challenge using high-performance computing tosimulate an orchard with fine-scale growth details and efficient implementation. A cluster of "thin" computing nodes with multiple processing cores are responsible for simulating individual trees in parallel. Then a "fat" computing node, with many morecores and larger memory capacity, is used to integrate the individual trees into an orchard. The time consumed by this orchard-level simulation is similar to that of a single tree, which significantly improves the efficiency of virtual-experiment implementation. Our first application of the orchard-level simulation is to investigate the optimal interval between neighboring trees for light interception efficiency. A 4x4 orchard with evenly distributed trees is simulated for this investigation, suggestingthat 2 m might be an optimal tree interval. In future work, the orchard-level simulation will also allow evaluation of the impact on light interception efficiency of other factors, such as pruning, row orientation, cross planting, and unusual terrains,setting ideal targets for genetic, physiological and orchard-management studies.