Ensemble Adaptive Data Assimilation Techniques Applied to Land-Falling North American Cyclones - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 551

23.3 Application of EnKF Observation Targeting

to Land-Falling Mid-Latitude Cyclones

We now apply the EnKF observation targeting methodology of Ancell and
Hakim ( 2007a ) to understand the nature of the impacts of hypothetical observations
beyond those of routine data for land-falling mid-latitude cyclones on the west coast
of North America. Land-falling cyclones routinely produce heavy precipitation
and high winds in coastal regions, and their predictability characteristics are thus
important to understand. Mass and Dotson ( 2010 ) describe some of the most intense
cyclones to strike the west coast of North America, and discuss the major societal
impacts they produced. McMurdie and Mass ( 2004 )and We d a m e t a l . ( 2009 )
describe how short-term forecast errors by deterministic operational numerical
models can be large for storms impacting the west coast. The purpose of this
study is to demonstrate how adaptive assimilation techniques can be applied
retrospectively to cases of land-falling cyclones as a research tool to investigate the
role of additional observations within a specific assimilation system and observing
network.

23.3.1 Details of the EnKF and the Forecast Model

The EnKF used in this study is an ensemble square-root filter that assimilates
observations serially ( Whitaker and Hamill 2002 ) and was created at the University
of Washington ( Torn and Hakim 2008 ). The 80-member EnKF runs on a 6-h update
cycle on the modeling domain shown in Fig. 23.1 at 36-km grid spacing with 37
vertical levels. The routine observations that are assimilated are cloud-track wind
(typically from 1,000 to 4,000 total), acars aircraft wind and temperature (typically
from 1,000 to 4,000 total), radiosonde wind, temperature, and relative humidity
(typically around 1,500 total), and surface wind, temperature, and altimeter data
(typically from 7,000 to 10,000 total). The EnKF uses both Gaspari-Cohn horizontal
localization ( Gaspari and Cohn 1999 ) and posterior inflation to address sampling
error and to avoid filter divergence ( Anderson and Anderson 1999 ). The inflation
and localization parameters used in this study are the same as those in To r n an d
Hakim ( 2008 ) which were tuned over a similar domain to produce appropriate
spread and minimum ensemble mean errors. Boundary conditions were perturbed
around the Global Forecasting System (GFS) analyses and forecasts using the fixed
covariance perturbation method of Torn et al. ( 2006 ).
The EnKF was cycled for 6 months from 0000 UTC October 1, 2009 to 1800
UTC March 31, 2010, and extended ensemble forecasts were run to 24-h forecast
time to capture a number of wintertime land-falling cyclones. The forecast model
used here is the Advanced Research Weather Research and Forecasting (WRF-
ARW) model Version 3.0.1.1 ( Skamarock et al. 2008 ). The WRF physics used are
the Mellor-Yamada-Janjic (MYJ) planetary boundary layer scheme ( Janjic 1990 ,
1996 , 2002 ), the Kain-Fritsch cumulus parameterization ( Kain and Fritsch 1990 ,