Important capabilities of DART

In this section we discuss the capabilities of DART that may be of interest to the user. This is a partial list of all of the functionality that is available in DART, and additional capabilities and improvements are continually being added.

As mentioned above, DART allows for both OSSE and OSE systems of models large and small. This allows users to test both theoretical limits of DA, models, and observations with idealized experiments as well as to improve actual real-world forecasts of chaotic systems with real observations.

Models supported by DART

A full list of models can be found here, but in brief the models supported by DART include:

Model

Latest version

Model

Latest version

lorenz_63

Manhattan

lorenz_84

Manhattan

lorenz_96

Manhattan

lorenz_96_2scale

Manhattan

lorenz_04

Manhattan

simple_advection

Manhattan

bgrid_solo

Manhattan

WRF

Manhattan

MPAS

Manhattan

ATM

Manhattan

ROMS

Manhattan

CESM

Manhattan

CAM-FV

Manhattan

CAM-CHEM

Manhattan

WACCM

Manhattan

WACCM-X

Manhattan

CICE

Manhattan

CM1

Manhattan

FESOM

Manhattan

NOAH-MP

Manhattan

WRF-Hydro

Manhattan

GCCOM

Lanai

LMDZ

Lanai

MITgcm_ocean

Lanai

NAAPS

Lanai

AM2

Lanai

CAM-SE

Manhattan

CLM

Manhattan

COAMPS

Lanai

COSMO

Lanai

Dynamo

Lanai

GITM

Lanai

Ikeda

Lanai

JULES

Lanai

MPAS_ocean

Lanai

null_model

Lanai

openggcm

Lanai

PARFLOW

Lanai

sqg

Lanai

TIE-GCM

Lanai

WRF-CHEM

Lanai

ECHAM

Prior to Lanai

PBL_1d

Prior to Lanai

MITgcm_annulus

Prior to Lanai

forced_barot

Prior to Lanai

pe2lyr

Prior to Lanai

ROSE

Prior to Lanai

CABLE

Prior to Lanai

The models listed as “Prior to Lanai” will take some additional work to integrate with a supported version of DART; please contact the dart @ ucar.edu team for more information. The versions listed as “Lanai” will be ported to the Manhattan version of DART depending on the needs of the user community as well as the availablity of resources on the DART team.

Observation converters provided by DART

Given a way to compute the expected observation value from the model state, in theory any and all observations can be assimilated by DART through the obs_seq.out file. In practice this means a user-defined observation converter is required. DART provides many observation converters to make this process easier for the user. Under the directory DART/observations/obs_converters there are multiple subdirectories, each of which has at least one observation converter. The list of these directories is as follows:

Observation

Directory

Format

Atmospheric Infrared Sounder satellite retrievals

AIRS

HDF-EOS

Advanced Microwave Sounding Unit brightness temperatures

AIRS

netCDF

Aviso: satellite derived sea surface height

Aviso

netCDF

Level 4 Flux Tower data from AmeriFlux

Ameriflux

Comma-separated text

Level 2 soil moisture from COSMOS

COSMOS

Fixed-width text

Doppler wind lidar

DWL

ASCII text

GPS retrievals of precipitable water

GPSPW

netCDF

GSI observation file

GSI2DART

Fortran binary

Global Temperature-Salinity Profile Program (GTSPP)

GTSPP

netCDF

Meteorological Assimilation Data Ingest System (MADIS)

MADIS

netCDF

MIDAS ionospheric obs

MIDAS

netCDF

MODIS satellite retrievals

MODIS

Comma-separated text

NCEP PREPBUFR

NCEP/prep_bufr

PREPBUFR

NCEP ASCII observations

NCEP/ascii_to_obs

NCEP text files

ROMS verification observations

ROMS

netCDF

Satellite winds from SSEC

SSEC

ASCII text

Sea surface temperature

SST

netCDF

Special Sensor Ultraviolet Spectrographic Imager (SSUSI) retrievals

SSUSI

netCDF

World Ocean Database (WOD)

WOD

World Ocean Database packed ASCII

National Snow and Ice Data Center sea ice obs

cice

Binary sea ice

VTEC Madrigal upper atmospheric obs

gnd_gps_vtec

ASCII text

GPS obs from COSMIC

gps

netCDF

Oklahoma Mesonet MDF obs

ok_mesonet

Oklahoma Mesonet MDF files

QuikSCAT scatterometer winds

quikscat

HDF 4

Radar reflectivity/radial velocity obs

Radar

WSR-88D (NEXRAD)

MODIS Snowcover Fraction obs

snow

General text

Text file (e.g. spreadsheet) obs

Text

General text

Total precipitable water from AQUA

tpw

HDF-EOS

Automated Tropical Cyclone Forecast (ATCF) obs

Tropical Cyclones

Fixed width text

LITTLE_R obs

var

little-r

MM5 3D-VAR radar obs

var

MM5 3D-VAR 2.0 Radar data files

Data assimilation algorithms available in DART

DART allows users to test the impact of using multiple different types of algorithms for filtering, inflation/deflation, and covariance localization.

DART offers numerous filter algorithms. These determine how the posterior distribution is updated based on the observations and the prior ensemble. The following table lists the filters supported in DART along with their type (set by filter_kind in input.nml under the “assim_tools_nml” section):

Filter #

Filter Name

References

1

EAKF (Ensemble Adjustment Kalman Filter)

Anderson, J. L., 2001. [1] Anderson, J. L., 2003. [2] Anderson, J., Collins, N., 2007. [3]

2

ENKF (Ensemble Kalman Filter)

Evensen, G., 2003. [4]

3

Kernel filter

4

Observation Space Particle filter

5

Random draw from posterior

None. IMPORTANT: (contact dart @ ucar.edu before using)

6

Deterministic draw from posterior with fixed kurtosis

None. IMPORTANT: (contact dart @ ucar.edu before using)

7

Boxcar kernel filter

8

Rank Histogram filter

Anderson, J. L., 2010. [5]

9

Particle filter

Poterjoy, J., 2016. [6]

DART also has several inflation algorithms available for both prior (the first value in the namelist) and posterior (the second value in the namelist). The following table lists the inflation “flavors” supported in DART along with their type number (set by inf_flavor in input.nml under the “filter_nml” section):

Flavor #

Inflation flavor name

References

0

No inflation

n/a

1

(Not Supported)

n/a

2

Spatially-varying state-space (Gaussian)

Anderson, J. L., 2009. [7]

3

Spatially-fixed state-space (Gaussian)

Anderson, J. L., 2007. [8]

4

Relaxation to prior spread (posterior inflation only)

Whitaker, J.S. and T.M. Hamill, 2012. [9]

5

Enhanced spatially-varying state-space (inverse gamma)

El Gharamti M., 2018. [10]

DART has the ability to correct for sampling errors in the regression caused by finite ensemble sizes. DART’s sampling error correction algorithm (and localization algorithm) is described in Anderson, J.L., 2012 [11] Sampling error correction can be turned on or off via the sampling_error_correction variable in the input.nml under the “assim_tools_nml” section.

The following covariance localization options are available (set by select_localization in input.nml under the “cov_cutoff_nml” section):

Loc #

Localization type

References

1

Gaspari-Cohn eq. 4.10

Gaspari, G. and Cohn, S. E., 1999. [12]

2

Boxcar

None

3

Ramped boxcar

None

The following image depicts all three of these options:

cutoff_fig

References