Title:
|
PARALLEL CALIBRATION OF SPATIAL DYNAMIC MODELS IN TERRAME |
Author(s):
|
Saulo Henrique Cabral Silva, Joubert de Castro Lima, Tiago Garcia de Senna Carneiro |
ISBN:
|
978-989-8533-06-7 |
Editors:
|
Hans Weghorn, Leonardo Azevedo and Pedro IsaĆas |
Year:
|
2011 |
Edition:
|
Single |
Keywords:
|
Parallel calibration, spatial dynamic models, TerraME |
Type:
|
Short Paper |
First Page:
|
451 |
Last Page:
|
456 |
Language:
|
English |
Cover:
|
|
Full Contents:
|
click to dowload
|
Paper Abstract:
|
Two of the main problems in calibrating spatial dynamic computational models are: (i) the huge runtime of a simulation, (ii) and few variables that can be simulated in a sequential calibration process. There are several methods to calibrate spatio-temporal models. The three most popular are: Monte Carlo Method, Genetic Method and Least Squares Method. Spatio-temporal computational simulators like Swarm, Stella, TerraME, Dinamica-Ego, Repast and Vensim have one or more calibration methods, but none of them can execute models or calibrations in parallel, i.e., none of them can be executed efficiently in shared memory computer architectures. In this paper, we extend TerraME, by introducing parallel calibration of spatial dynamic models using Monte Carlo and Genetic Methods. Our results demonstrate that parallelism can increase the number of variables being calibrated and reduce the runtime of calibration process. In general, TerraME with parallel calibration has 75-80% of linear speedup when the number of cores is equal the number of disks in a unique machine and 50-60% of linear speedup, otherwise. |
|
|
|
|