 |
| Titel |
Large Scale Terrestrial Modeling: A Discussion of Technical and Conceptual Challenges and Solution Approaches |
| VerfasserIn |
M. Rahman, T. Aljazzar, S. Kollet, R. Maxwell |
| Konferenz |
EGU General Assembly 2012
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 14 (2012) |
| Datensatznummer |
250065206
|
|
|
|
|
|
| Zusammenfassung |
| A number of simulation platforms have been developed to study the spatiotemporal
variability of hydrologic responses to global change. Sophisticated terrestrial models demand
large data sets and considerable computing resources as they attempt to include
detailed physics for all relevant processes involving the feedbacks between subsurface,
land surface and atmospheric processes. Access to required data scarcity, error and
uncertainty; allocation of computing resources; and post processing/analysis are some of
the well-known challenges. And have been discussed in previous studies dealing
with catchments ranging from plot scale research (102m2), to small experimental
catchments (0.1-10km2), and occasionally medium-sized catchments (102-103km2).
However, there is still a lack of knowledge about large-scale simulations of the coupled
terrestrial mass and energy balance over long time scales (years to decades). In
this study, the interaction between subsurface, land surface, and the atmosphere
are simulated in two large scale (>104km2) river catchments that are the Luanhe
catchment in the North Plain, China and the Rur catchment, Germany. As a simulation
platform, a fully coupled model (ParFlow.CLM) that links a three-dimensional
variably-saturated groundwater flow model (ParFlow) with a land surface model (CLM) is
used. The Luanhe and the Rur catchments have areas of 54,000 and 28,224km2
respectively and are being simulated using spatial resolutions on the order of 102 to 103m
in the horizontal and 10-2 to 10-1m in the vertical direction. ParFlow.CLM was
configured over computational domains well beyond the actual watershed boundaries
to account for cross-watershed flow. The resulting catchment models consist of
up to 108 cells which were implemented over more than 1000 processors each
with 512MB memory on JUGENE hosted by the Juelich Supercomputing Centre,
Germany. Consequently, large numbers of input and output files were produced for each
parameter such as; soil hydraulic characteristics, land cover, geology, hydraulic
conductivity, and atmospheric parameters. For example, to simulate the Luanhe
catchment for one year at one hour time step, the number of input and output files
approach 108 files based on the most simple implementation of parallel I/O. Storage
requirements quickly approach terabytes in production simulations, while single data
files often exceed available memory. Handling these large data sets in terms of
pre-/post-processing and visualization demands considerable computing resources and
new data analysis tools. The difficulties will increase considerably at even finer
lateral spatial resolution and longer simulation times, which will be required in the
discussed studies and are planned at 100m and 101 years. In this article we present in
detail the main challenges and technical issues inherent in simulating the interaction
between land surface, subsurface and climate parameters of catchments at the basin
scale. We also suggest approaches to overcome some of the discussed difficulties. |
|
|
|
|
|
|