 |
| Titel |
Multi-scale Data Assimilation for Large Scale Hydrologic Modeling Applications |
| VerfasserIn |
E. F. Wood, M. Pan |
| Konferenz |
EGU General Assembly 2009
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 11 (2009) |
| Datensatznummer |
250024296
|
|
|
|
|
|
| Zusammenfassung |
| Recent developments in hydrologic modeling have resulted in continental-to-global modeling
domains at high temporal and spatial resolutions. For example, global hydrologic
simulations have been performed at 0.5 degree level or finer, and regionally (over United
States) simulations have been performed at 1 km resolutions. As the number of
modeling pixels increase rapidly (and expected to reach at least 105 to 106 grids), so
increases the dimension and complexity of the data assimilation problem associated
with such applications. Traditional data assimilation computational approaches
normally cubically with problem size, making large problems practically impossible
to solve. So it is critical to develop efficient assimilation techniques to meet this
challenge.
In this presentation a computationally efficient assimilation system for large scale
hydrologic modeling applications is presented based on a multi-scale autoregressive (MAR)
framework. The MAR framework was developed for high dimensional signals with
multi-scale features and provides an efficient filtering procedure for the optimal estimation
(data assimilation) of high dimensional dynamic systems. An ensemble version of the
multi-scale filtering algorithm, the ensemble multi-scale filter (EnMSF), is utilized. The
EnMSF relies on Monte Carlo samples, making this technique suitable for a range
of geosciences data assimilation problems. The EnMSF is implemented within a
hydrologic data assimilation system that runs a land surface model and assimilates
remotely sensed soil moisture. Assimilation experiments are carried out over the
Arkansas-Red river basin in central U.S. (645,000 sq km), using the Variable Infiltration
Capacity (VIC) model with a computing grid of 1062 pixels. Two assimilation
experiments are presented: one is driven by meteorological forcing fields downscaled from
NOAA/NCEP’s Climate Forecast System (CFS) ensemble seasonal climate forecasts; and
the second experiment is driven by ensembles generated from remotely sensed
rainfall data (the TRMM 3B42-RT product.) The results not only confirms both the
efficiency and accuracy of the multi-scale assimilation method, but also shows the great
potential of the multi-scale assimilation system for large-scale applications driven by
either coarse-scale atmospheric model forecasts or remote sensing observations. |
|
|
|
|
|
|