Ensemble Data Assimilation for Very Large Atmosphere, Ocean and Coupled Models with the Data Assimilation Research Testbed

The Data Assimilation Research Testbed (DART) is a community facility for ensemble
data assimilation developed and supported by the National Center for Atmospheric
Research. DART provides a comprehensive suite of software, documentation, examples
and tutorials that can be used for ensemble data assimilation research, operations, and
education. Scientists and software engineers from the Data Assimilation Research
Section at NCAR are available to actively support DART users who want to use existing
DART products or develop their own new applications. Current DART users range from
university professors teaching data assimilation, to individual graduate students working
with simple models, through national laboratories doing operational prediction with large
state-of- the-art models. DART runs efficiently on many computational platforms ranging
from laptops through thousands of cores on the newest supercomputers.

This poster describes some sample applications of DART for atmospheric, oceanic, and
coupled data assimilation. The first example is a cycling limited-area mesoscale ensemble
data assimilation system using DART with the WRF-ARW model. This system is run in
real-time at 15 km resolution and provides initial conditions for a 10-member ensemble
of 3 km resolution forecasts. The second example is a short global ocean ensemble
reanalysis using DART with the POP ocean model at 1/10 of a degree resolution. A final
example discusses the development of a fully-coupled ensemble data assimilation for the
CAM atmosphere and POP ocean components of NCAR's CESM climate system model.
This will include a discussion of some of the challenges in implementing fully-coupled
data assimilation in a community system like DART.

All three of these models have very large state vectors that need to be efficiently stored
and accessed throughout the data assimilation process. New supercomputer architectures
are tending towards larger numbers of processors but with less memory for each. Recent
advances in DART have addressed the need to run with a smaller memory footprint on a
higher node count by utilizing MPI-2 one-sided communication to do non-blocking
asynchronous access of distributed data. Benefits of the new DART implementation will
be discussed.