Parallel algorithm for an age-structured population model
Complex ecosystem simulation models such as those used for the prediction of global warming and spreading of deserts require high computer memory and fast computational speed for efficient execution. Constraints imposed by limited memory and speed often put restriction on model resolutions that may...
Uloženo v:
| Hlavní autoři: | , , |
|---|---|
| Médium: | Proceedings Paper |
| Jazyk: | English |
| Vydáno: |
Penerbit Universiti Sains Malaysia
2015
|
| Témata: | |
| On-line přístup: | http://agris.upm.edu.my:8080/dspace/handle/0/10014 |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo otaguje tento záznam!
|
| Shrnutí: | Complex ecosystem simulation models such as those used for the prediction of global warming and spreading of deserts require high computer memory and fast computational speed for efficient execution. Constraints imposed by limited memory and speed often put restriction on model resolutions that may result in unsatisfactory model output. In this paper, we study the Unionid mussel communities in the United States of America, using age-structured spatially explicit modelling approach. The use of such models has given researchers numerous insights into the mechanisms at work in the mussel ecosystems, as well as proving to be useful in ecological risk assessment studies. Special features of the mussel life-cycle make possible the construction of an efficient parallel algorithm for the mussel community model, implemented on a Maspar Mp-2 machine. The algorithm parallelizes the river cells, of which there are some 4,000, with each cell being assigned a microprocessor to simulate its local mussel community. The dynamics of the mussels in all cells are computed simultaneously, with inter-processor communication been kept to an appropriate level, occuring mainly during the dispersal and settling period of the glochidia juveniles-which are facilitated by the movements of fish host. The glochidia are optimally grouped into cohorts. Comparison with a sequential algorithm and suggestion for further research will be discussed. |
|---|