HR: 1340h
AN: H13H-1399 [Abstracts]
TI: Parallelization of a Fully-Distributed Hydrologic Model using Sub-basin Partitioning
AU: * Vivoni, E R
EM: vivoni@nmt.edu
AF: Department of Earth and Environmental Science,
New Mexico Institute of Mining and Technology, 801 Leroy Place, Socorro, NM 87801
United States
AU: Mniszewski, S
EM: smm@lanl.gov
AF: Computer and Computational Science,
Los Alamos National Laboratory, Technical Area 3, Los Alamos, NM 87545
United States
AU: Fasel, P
EM: pkf@lanl.gov
AF: Computer and Computational Science,
Los Alamos National Laboratory, Technical Area 3, Los Alamos, NM 87545
United States
AU: Springer, E
EM: everetts@lanl.gov
AF: Earth and Environmental Science,
Los Alamos National Laboratory, Technical Area 51, Los Alamos, NM 87545
United States
AU: Ivanov, V Y
EM: viva@mit.edu
AF: Department of Civil and Environmental Engineering,
Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139
United States
AU: Bras, R L
EM: rlbras@mit.edu
AF: Department of Civil and Environmental Engineering,
Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139
United States
AB:
A primary obstacle towards advances in watershed simulations has been the limited computational capacity available to most
models. The growing trend of model complexity, data availability and physical representation has not been matched by adequate
developments in computational efficiency. This situation has created a serious bottleneck which limits existing distributed
hydrologic models to small domains and short simulations. In this study, we present novel developments in the parallelization
of a fully-distributed hydrologic model. Our work is based on the TIN-based Real-time Integrated Basin Simulator (tRIBS),
which provides continuous hydrologic simulation using a multiple resolution representation of complex terrain based on a
triangulated irregular network (TIN). While the use of TINs reduces computational demand, the sequential version of the model
is currently limited over large basins (>10,000 km2) and long simulation periods (>1 year). To address this, a parallel
MPI-based version of the tRIBS model has been implemented and tested using high performance computing resources at Los Alamos
National Laboratory. Our approach utilizes domain decomposition based on sub-basin partitioning of the watershed. A stream
reach graph based on the channel network structure is used to guide the sub-basin partitioning. Individual sub-basins or
sub-graphs of sub-basins are assigned to separate processors to carry out internal hydrologic computations (e.g.
rainfall-runoff transformation). Routed streamflow from each sub-basin forms the major hydrologic data exchange along the
stream reach graph. Individual sub-basins also share subsurface hydrologic fluxes across adjacent boundaries. We demonstrate
how the sub-basin partitioning provides computational feasibility and efficiency for a set of test watersheds in northeastern
Oklahoma. We compare the performance of the sequential and parallelized versions to highlight the efficiency gained as the
number of processors increases. We also discuss how the coupled use of TINs and parallel processing can lead to feasible
long-term simulations in regional watersheds while preserving basin properties at high-resolution.
DE: 1805 Computational hydrology
DE: 1816 Estimation and forecasting
DE: 1847 Modeling
DE: 1860 Streamflow
DE: 1879 Watershed
SC: Hydrology [H]
MN: Fall Meeting 2005