Wind observations and forecasts over oceans have traditionally
been difficult to perform due to sparse data. In the past decade,
scatterometer observations of oceanic winds have improved
this data gap, but have not been used routinely in operational
models. This paper describes the implementation of a numerical
model which has been performing this task twice a day since June
1997. This model, called the Quasi-Lagrangian Model (QLM), is a
hydrostatic primitive equation model originally developed by
Mathur (1983; 1991) and recently enhanced by Barker (1998).
The scatterometer data is downloaded from the following
three satellites between 12-20Z and 00-08Z: the TOPEX altimeter,
the ERS-2 scatterometer, and the SSM/I microwave radiometer.
The scatterometer winds are then adjusted to 10-m height and
1-min averages. At 08Z and 20Z, the QLM is
initialized using 00Z or 12Z data, and then nudged with additional
land and scatterometer data during the next 8 hour period.
The nudging improves model initialization, with the added benefit
of generating dynamically consistent wind fields in nearby
data-void regions, and producing temporally correct winds
between satellite passes. Once the nudging is complete, the model
continues forecasting an additional 16 hours.
Forecasts are typically completed within one hour after model
start, and are downloaded to the web site
http://www.usersystems.com. Currently operational forecasts are
performed for the CONUS region with a grid spacing of 80 km
and 3-min time step. These forecasts are done on a Gateway 2000
300 MhZ personal computer in a Windows NT environment (not unix).
Data is downloaded from a satellite link onto the NT, although
automated tasks and the satellite data
assimilation are performed on an SGI networked to the NT.
It is found that the satellite data improves (sometimes
dramatically) the wind forecasts for these last 16 hours,
especially over oceanic regions. Other forecast parameters
(temperature, moisture, rainfall, etc.) also are improved
with wind nudging. QLM case studies with and without
scatterometer nudging will be presented at the conference.
Continuous modifications are being performed regarding
nudging techniques and data assimilation periods. Extended
forecasts out to 48 h are being evaluated. Other future
improvements include a portable version which can be run
regionally around the world with higher resolution.
In addition, development for operational MM5 forecasts with
satellite nudging will begin in Summer 1998 for the southern
United States. A discussion of these QLM modifications, and the
MM5 modeling efforts will be presented at the conference.