Monday, 8 January 2018
Exhibit Hall 3 (ACC) (Austin, Texas)
Climate change adaptation would benefit from identification of suitable areas for minor crops, which would facilitate introduction of a new crop into an existing cropping system. Simple climate suitability models such as the EcoCrop model have been used to assess climate suitability of crops using gridded climate data, which would take a considerably long time especially at a high spatial resolution, e.g., 1 km. The objectives of this study were to apply a high-performance computing approach to the assessment of climate suitability for a large gridded climate data. A script was written to perform the climate suitability assessment for a chinese cabbage using the EcoCrop model under a parallel computing environment in R. The Snow package was used to provide the high performance computing functionality. As the number of CPU (Central Processing Unit) core increased, the wall clock time reduced considerably. Still, the speedup diminished with increasing numbers of CPU core. For example, the wall clock time reduced by 90% with 16 CPU cores. However, it took twice as much as time to compute the climate suitability index compared with theoretical time for the given number of CPU. The total running time reduced mostly in calculation of suitability. For example, a computation time for calculating suitability reduced by about 90% whereas that for reading climate data did by 70% using 16 CPU cores. In a further study, implementation of a parallel computing system based on the MPI (Message Passing Interface) would help assess climate suitability using a large datasets such as global climate data or data at a very high spatial resolution, e.g., 30 m.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner