Development of an EPIC parallel computing framework to facilitate regional/global gridded crop modeling with multiple scenarios: A case study of the United States

Abstract

Crop models are increasingly used to evaluate crop yields at regional/global scales. These applications require the integration and processing of very large data sets in order to explore the implications of land management options across spatially heterogeneous scales. These modeling involve the combination of large spatially explicit data sets for climate, biophysical and crop management variables as well as significant computational capacity for regional/global scale simulations. As a result, the application of crop models at regional/global scales is challenging due to the requirements for input data, calibration, validation and simulation setups appropriate for thousands to millions of spatial points. Not surprisingly, the implementation of these models across large areas using fine-scale grids can be limited by computational time requirements. To reduce the large computational load of an agroecosystem simulation process for regional and global scales, we developed an EPIC Parallel Computing Framework (EPCF) to facilitate regional/global gridded crop modeling. The EPCF can make full use of the CPU resources of the workstation through parallel processing. For future users, only a few lines of additional code modification are needed to convert the single process code to parallel computing code. Parallel processing in one machine makes it easy to handle the whole system without the overhead and expertise required for a distributed system. EPCF is a system that provides not only the ease of development but also cost efficiency.