![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Parameter Estimation In Ensemble Data Assimilation To Characterize Model Errors In Surface-Layer Schemes Over Complex Terrain |
VerfasserIn |
Joshua Hacker, Jared Lee, Lili Lei |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250098531
|
Publikation (Nr.) |
EGU/EGU2014-14216.pdf |
|
|
|
Zusammenfassung |
Numerical weather prediction (NWP) models have deficiencies in surface and boundary layer
parameterizations, which may be particularly acute over complex terrain. Structural and
physical model deficiencies are often poorly understood, and can be difficult to identify.
Uncertain model parameters can lead to one class of model deficiencies when they are
mis-specified. Augmenting the model state variables with parameters, data assimilation can
be used to estimate the parameter distributions as long as the forecasts for observed variables
is linearly dependent on the parameters. Reduced forecast (background) error shows that the
parameter is accounting for some component of model error. Ensemble data assimilation has
the favorable characteristic of providing ensemble-mean parameter estimates, eliminating
some noise in the estimates when additional constraints on the error dynamics are
unknown.
This study focuses on coupling the Weather Research and Forecasting (WRF)
NWP model with the Data Assimilation Research Testbed (DART) to estimate the
Zilitinkevich parameter (CZIL). CZIL controls the thermal “roughness length”
for a given momentum roughness, thereby controlling heat and moisture fluxes
through the surface layer by specifying the (unobservable) aerodynamic surface
temperature.
Month-long data assimilation experiments with 96 ensemble members, and grid spacing
down to 3.3 km, provide a data set for interpreting parametric model errors in complex
terrain. Experiments are during fall 2012 over the western U.S., and radiosonde, aircraft,
satellite wind, surface, and mesonet observations are assimilated every 3 hours. One
ensemble has a globally constant value of CZIL=0.1 (the WRF default value), while a second
ensemble allows CZIL to vary over the range [0.01, 0.99], with distributions updated via the
assimilation.
Results show that the CZIL estimates do vary in time and space. Most often, forecasts are
more skillful with the updated parameter values, compared to the fixed default values,
suggesting that the parameters account for some systematic errors. Because the parameters
can account for multiple sources of errors, the importance of terrain in determining
surface-layer errors can be deduced from parameter estimates in complex terrain; parameter
estimates with spatial scales similar to the terrain indicate that terrain is responsible for
surface-layer model errors. We will also comment on whether residual errors in the state
estimates and predictions appear to suggest further parametric model error, or some other
source of error that may arise from incorrect similarity functions in the surface-layer
schemes. |
|
|
|
|
|