[Dart-dev] [6402] DART/trunk/doc/html: filled in the non-backwards-compatible changes, the new models,
nancy at ucar.edu
nancy at ucar.edu
Tue Aug 13 13:49:46 MDT 2013
Revision: 6402
Author: nancy
Date: 2013-08-13 13:49:45 -0600 (Tue, 13 Aug 2013)
Log Message:
-----------
filled in the non-backwards-compatible changes, the new models,
new obs, new diagnostics, etc sections of the release notes.
added the GITM model to the diffs document. getting closer.
Modified Paths:
--------------
DART/trunk/doc/html/Lanai_diffs_from_Kodiak.html
DART/trunk/doc/html/Lanai_release.html
-------------- next part --------------
Modified: DART/trunk/doc/html/Lanai_diffs_from_Kodiak.html
===================================================================
--- DART/trunk/doc/html/Lanai_diffs_from_Kodiak.html 2013-08-13 17:41:33 UTC (rev 6401)
+++ DART/trunk/doc/html/Lanai_diffs_from_Kodiak.html 2013-08-13 19:49:45 UTC (rev 6402)
@@ -27,8 +27,8 @@
</tr>
</table>
+<a href="#Changes">Changes to Core DART routines</a> /
<a href="#NewModels">New Models or Changes to Existing Models</a> /
-<a href="#Changes">Changes to Core DART routines</a> /
<a href="#ForwardOps">New or Changed Forward Operators</a> /
<a href="#ObsConvert">Observation Converters</a> /
<a href="#Diagnostics">New or Updated DART Diagnostics</a> /
@@ -53,6 +53,80 @@
</P>
<!--==================================================================-->
+
+<A NAME="Changes"></A>
+<div class="top">[<a href="#">top</a>]</div><hr />
+<H2>Changes to Core DART routines</H2>
+<P>
+This section describes changes in the basic DART library routines
+since the Kodiak release.
+</P>
+
+<UL>
+<LI>Added a completely new random number generator based on the Mersenne
+Twister algorithm from the GNU scientific library. It seems to have better
+behavior if reseeded frequently, which is a possible usage pattern if
+perfect_model_obs is run for only single steps and the model is advanced in
+an external script. As part of this code update all random number code was
+moved into the random_seq_mod and random_nr_mod is deprecated.</LI>
+<LI>Perfect_model_obs calls a seed routine in the time manager now that
+generates a consistent seed based on the current time of the state. This
+makes subsequent runs give consistent results and yet separate runs don't get
+identical error values.</LI>
+<LI>Added random number generator seeds in several routines to try to get
+consistent results no matter how many MPI tasks the code was run with. This includes:
+<UL>
+<LI>cam model_mod.f90, pert_model_state()</LI>
+<LI>assim_tools_mod.f90, filter_assim(), filter kinds 2, 3, and 5</LI>
+<LI>wrf model_mod.f90, pert_model_state()</LI>
+<LI>adaptive_inflate_mod.f90, adaptive_inflate_init(), non-deterministic inf</LI>
+</UL>
+</LI>
+<LI>There is a new &filter_nml namelist item: enable_special_outlier_code.
+If .true. the DART quality control code will call a separate subroutine at
+the end of filter.f90 to evaluate the outlier threshold. The user can add
+code to that routine to change the threshold based on observation type or
+values as they wish. If .false. the default filter outlier threshold code
+will be called and the user routine ignored.</LI>
+<LI>Filter will call the end_model() subroutine in the model_mod for
+the first time. It should have been called all along, but was not.</LI>
+<LI>Added a time sort routine in the time_manager_mod.</LI>
+<LI>Avoid a pair of all-to-all transposes when setting the inflation mean and
+sd from the namelist. The new code finds the task which has the two copies
+and sets them directly without a transpose. The log messages were also moved
+to the end of the routine - if you read in the mean/sd values from a restart
+file the log messages that printed out the min/max values needed to be after
+the read from the file.</LI>
+<LI>Reordered the send/receive loops in the all-to-all transposes to scale
+better on yellowstone.</LI>
+<LI>Remove a state-vector size array from the stack in
+read_ensemble_restart(). The array is now allocated only if needed and then
+deallocated. The ensemble write routine was changed before the Kodiak release
+but the same code in read was apparently not changed simply as an
+oversight.</LI>
+<LI>If the ensemble mean is selected to be written out in dart restart file
+format, the date might not have been updated correctly. The code was fixed to
+ensure the ensemble mean date in the file was correct.</LI>
+<LI>filter writes the ensemble size into the log file.</LI>
+<LI>Reorganized the code in the section of obs_model_mod that prints out the
+time windows, with and without verbose details. Should be clearer if the next
+observation is in or out of the current assimilation window, and if the model
+needs to advance or not.</LI>
+<LI>Added a fill_inflation_restart utility which can write a file with a
+fixed mean and sd, so the first step of a long assimilation run can use the
+same 'start_from_restart_file' as subsequent steps.</LI>
+<LI>Added new location module options:
+<UL>
+<LI>Channel coordinate system</LI>
+<LI>[0-1] periodic 3D coordinate system</LI>
+<LI>X,Y,Z 3D Cartesian coordinate system</LI>
+<LI>2D annulus coordinate system</LI>
+</UL>
+</LI>
+</UL>
+
+
+<!--==================================================================-->
<br />
<br />
<A NAME="NewModels"></A>
@@ -125,6 +199,7 @@
</LI>
<LI>Added support for the 'SQG' uniform PV two-surface QC+1 spectral model.</LI>
<LI>Added support for a flux-transport solar dynamo model.</LI>
+<LI>Added support for the GITM upper atmosphere model.</LI>
<LI>Added support for the NOAH land model.</LI>
<LI>Added support for the NAAPS model.</LI>
<LI>Added model_mod interface code for the NOGAPS model to the SVN repository.</LI>
@@ -138,80 +213,6 @@
<!--==================================================================-->
-<A NAME="Changes"></A>
-<div class="top">[<a href="#">top</a>]</div><hr />
-<H2>Changes to Core DART routines</H2>
-<P>
-This section describes changes in the basic DART library routines
-since the Kodiak release.
-</P>
-
-<UL>
-<LI>Added a completely new random number generator based on the Mersenne
-Twister algorithm from the GNU scientific library. It seems to have better
-behavior if reseeded frequently, which is a possible usage pattern if
-perfect_model_obs is run for only single steps and the model is advanced in
-an external script. As part of this code update all random number code was
-moved into the random_seq_mod and random_nr_mod is deprecated.</LI>
-<LI>Perfect_model_obs calls a seed routine in the time manager now that
-generates a consistent seed based on the current time of the state. This
-makes subsequent runs give consistent results and yet separate runs don't get
-identical error values.</LI>
-<LI>Added random number generator seeds in several routines to try to get
-consistent results no matter how many MPI tasks the code was run with. This includes:
-<UL>
-<LI>cam model_mod.f90, pert_model_state()</LI>
-<LI>assim_tools_mod.f90, filter_assim(), filter kinds 2, 3, and 5</LI>
-<LI>wrf model_mod.f90, pert_model_state()</LI>
-<LI>adaptive_inflate_mod.f90, adaptive_inflate_init(), non-deterministic inf</LI>
-</UL>
-</LI>
-<LI>There is a new &filter_nml namelist item: enable_special_outlier_code.
-If .true. the DART quality control code will call a separate subroutine at
-the end of filter.f90 to evaluate the outlier threshold. The user can add
-code to that routine to change the threshold based on observation type or
-values as they wish. If .false. the default filter outlier threshold code
-will be called and the user routine ignored.</LI>
-<LI>Filter will call the end_model() subroutine in the model_mod for
-the first time. It should have been called all along, but was not.</LI>
-<LI>Added a time sort routine in the time_manager_mod.</LI>
-<LI>Avoid a pair of all-to-all transposes when setting the inflation mean and
-sd from the namelist. The new code finds the task which has the two copies
-and sets them directly without a transpose. The log messages were also moved
-to the end of the routine - if you read in the mean/sd values from a restart
-file the log messages that printed out the min/max values needed to be after
-the read from the file.</LI>
-<LI>Reordered the send/receive loops in the all-to-all transposes to scale
-better on yellowstone.</LI>
-<LI>Remove a state-vector size array from the stack in
-read_ensemble_restart(). The array is now allocated only if needed and then
-deallocated. The ensemble write routine was changed before the Kodiak release
-but the same code in read was apparently not changed simply as an
-oversight.</LI>
-<LI>If the ensemble mean is selected to be written out in dart restart file
-format, the date might not have been updated correctly. The code was fixed to
-ensure the ensemble mean date in the file was correct.</LI>
-<LI>filter writes the ensemble size into the log file.</LI>
-<LI>Reorganized the code in the section of obs_model_mod that prints out the
-time windows, with and without verbose details. Should be clearer if the next
-observation is in or out of the current assimilation window, and if the model
-needs to advance or not.</LI>
-<LI>Added a fill_inflation_restart utility which can write a file with a
-fixed mean and sd, so the first step of a long assimilation run can use the
-same 'start_from_restart_file' as subsequent steps.</LI>
-<LI>Added new location module options:
-<UL>
-<LI>Channel coordinate system</LI>
-<LI>[0-1] periodic 3D coordinate system</LI>
-<LI>X,Y,Z 3D Cartesian coordinate system</LI>
-<LI>2D annulus coordinate system</LI>
-</UL>
-</LI>
-</UL>
-
-
-<!--==================================================================-->
-
<A NAME="ForwardOps"></A>
<div class="top">[<a href="#">top</a>]</div><hr />
<H2>New or changed Forward Operators</H2>
Modified: DART/trunk/doc/html/Lanai_release.html
===================================================================
--- DART/trunk/doc/html/Lanai_release.html 2013-08-13 17:41:33 UTC (rev 6401)
+++ DART/trunk/doc/html/Lanai_release.html 2013-08-13 19:49:45 UTC (rev 6402)
@@ -35,6 +35,7 @@
<a href="#NewFeatures">New Features</a> /
<a href="#NewModels">New Models</a> /
<a href="#ChangedModels">Changed Models</a> /
+<a href="#NewFOs">New Forward Operators</a> /
<a href="#NewObs">New Observations</a> /
<a href="#NewDiagnostics">New Diagnostics and Documentation</a> /
<a href="#NewUtilities">New Utilities</a> /
@@ -226,8 +227,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="Installation"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Installation</h2>
<p>
@@ -278,8 +279,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="compilers"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h3>Requirements: an F90 Compiler</h3>
<p>
@@ -301,8 +302,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="netCDFlib"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h3>Requirements: the <em class=file>netCDF</em> library</h3>
<p>
@@ -346,8 +347,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="download"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Downloading the distribution.</h2>
<p>
@@ -403,8 +404,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="customizations"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Customizing the build scripts -- Overview.</h2>
<p>
@@ -561,8 +562,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="building"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Building the Lorenz_63 DART project.</h2>
<p>DART executables are constructed in a <em class=file>work</em>
@@ -804,8 +805,8 @@
<!--==================================================================-->
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="Running"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Running Lorenz_63.</h2>
<p>
@@ -1685,8 +1686,8 @@
<!--==================================================================-->
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="Workshop"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>The Tutorial.</h2>
<p>
The <em class=file>DART/tutorial</em> documents are an excellent way
@@ -1695,8 +1696,8 @@
</p>
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="matlab"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<!--==================================================================-->
<h2>Matlab® Diagnostics</h2>
<p>The output files are netCDF files, and may be examined with many
@@ -1825,8 +1826,8 @@
</table>
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="discussion"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<!--==================================================================-->
<h2>Bias, filter divergence and covariance inflation (with the L63 model)</h2>
@@ -1894,8 +1895,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="syntheticobservations"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<!--==================================================================-->
<h2>Synthetic Observations</h2>
<p>Synthetic observations are generated from a `perfect' model integration,
@@ -1922,8 +1923,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="CurrentUsers"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Notes for Current Users</h2>
<p>
@@ -1952,8 +1953,8 @@
<!--==================================================================-->
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
<a name="Nonbackward"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
<h2>Non-backwards Compatible Changes</h2>
<p>
@@ -1965,7 +1966,7 @@
<ol>
<li>The DART system uses a new random number generator based on
-the Mersenns Twister algorithm from the GNU scientific library.
+the Mersenne Twister algorithm from the GNU scientific library.
It is believed to have better behavior in general, and in particular
when it is frequently reseeded, as may be the case in some
perfect_model_obs experiments. The seed in perfect_model_obs
@@ -2041,468 +2042,240 @@
</ol>
-<p>
-For completeness, the following were changes in the Kodiak
-release (SVN revision number 5018, 15 June 2011), which were
-not backwards compatible with the previous Jamaica branch
-(SVN revision number 2884, 30 April 2007).
-</p>
+<!--==================================================================-->
-<ol>
+<a name="NewFeatures"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
+<h2>New Features</h2>
-<li> &filter_nml used to have a single entry to control whether
-to read in both the inflation values and standard deviations
-from a file or use the settings in the namelist. The old namelist
-item, <em class=code>inf_start_from_file</em>, has been replaced by
-two items that allow the inflation values and the standard deviation
-to be read in separately. The new namelist items are
-<em class=code>inf_initial_from_file</em> and
-<em class=code>inf_sd_initial_from_file</em>. See the
-filter namelist documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/filter/filter.html#Namelist">Website</a>
-or <a href="../../filter/filter.html#Namelist">local file</a>
-for more details.
-</li>
-
-<li>The WRF/DART converter program used to be called
-<em class=program>dart_tf_wrf</em>, had no namelist,
-and you entered <tt>T</tt> or <tt>F</tt> to indicate
-which direction you were converting. Now we have
-<em class=program>dart_to_wrf</em> and
-<em class=program>wrf_to_dart</em> (documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/wrf/WRF_DART_utilities/dart_to_wrf.html">Website</a>
-or <a href="../../models/wrf/WRF_DART_utilities/dart_to_wrf.html">local file</a>)
-each with a namelist to control various options.
-</li>
-
-<li>The CAM/DART converter programs used to be called
-<em class=program>trans_sv_pv</em> and <em class=program>trans_pv_sv</em>,
-with no namelists, and with several specialized variants
-(e.g. <em class=program>trans_pv_sv_time0</em>).
-Now we have
-<em class=program>cam_to_dart</em> (documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/cam/cam_to_dart.html">Website</a>
-or <a href="../../models/cam/cam_to_dart.html">local file</a>)
-and <em class=program>dart_to_cam</em> (documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/cam/dart_to_cam.html">Website</a>
-or <a href="../../models/cam/dart_to_cam.html">local file</a>)
-each with a namelist to control various options.
-</li>
-
-<li>The <em class=code>obs_def_radar_mod.f90</em> radar observation
-module was completely rewritten and the namelist has changed substantially.
-See the module documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/obs_def/obs_def_radar_mod.html">Website</a>
-or <a href="../../obs_def/obs_def_radar_mod.html">local file</a>
-for details.
-
-For example, the defaults for the old code were:
-<pre><tt>
-&obs_def_radar_mod_nml
- convert_to_dbz = .true. ,
- dbz_threshold = 0.001 ,
- apply_ref_limit_to_obs = .false. ,
- reflectivity_limit_obs = 0.0 ,
- lowest_reflectivity_obs = -888888.0,
- apply_ref_limit_to_state = .false. ,
- reflectivity_limit_state = 0.0 ,
- lowest_reflectivity_state = -888888.0 /
-</tt></pre>
-
-and the new ones are:
-
-<pre><tt>
-&obs_def_radar_mod_nml
- apply_ref_limit_to_obs = .true. ,
- reflectivity_limit_obs = 0.0 ,
- lowest_reflectivity_obs = 0.0 ,
- apply_ref_limit_to_fwd_op = .true. ,
- reflectivity_limit_fwd_op = 0.0 ,
- lowest_reflectivity_fwd_op = 0.0 ,
- dielectric_factor = 0.224 ,
- n0_rain = 8.0e6 ,
- n0_graupel = 4.0e6 ,
- n0_snow = 3.0e6 ,
- rho_rain = 1000.0 ,
- rho_graupel = 400.0 ,
- rho_snow = 100.0 ,
- allow_wet_graupel = .false.,
- microphysics_type = 3 ,
- allow_dbztowt_conv = .false. /
-</tt></pre>
-</li>
-
-
-<li>The WRF &model_mod namelist has changed. It now
-requires a <em class=code>wrf_state_variables</em> list to choose which
-WRF fields are put into the state vector. The order of the
-field names in the list sets the order of the fields in the
-state vector. See the WRF model_mod documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/wrf/model_mod.html#Namelist">Website</a>
-or <a href="../../models/wrf/model_mod.html#Namelist">local file</a>
-for details. Although they haven't been removed
-from the namelist, the following items have no effect
-on the code anymore:
-
<ul>
-<li>num_moist_vars</li>
-<li>surf_obs</li>
-<li>soil_data</li>
-<li>h_diab</li>
-</ul>
-
+<li>Customizable Outlier-Threshold Handling
+ <ul><li>Filter contains code to compute whether an observation should
+ not be assimilated because the forward operator mean is too different
+ from the observation value. This is done uniformly for all observation
+ values and types. To customize this computation (e.g. to allow all obs
+ of a particular type to be assimilated without having to pass the outlier
+ threshold test), there is a new namelist item
+ <em class=code>enable_special_outlier_code</em> in the &filter_nml namelist
+ that enables a call to a subroutine at the end of the filter.f90 source file.
+ That subroutine can be customized by the user to do any computation required.
+ See the filter namelist documentation
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/filter/filter.html#Namelist">
+ Website</a> or <a href="../../filter/filter.html#Namelist">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>The WRF model_mod now computes geometric heights
-instead of geopotential heights. It also uses the staggered
-grids as read in from the <em class=file>wrfinput_dNN</em> file(s)
-instead of interpolating in the non-staggered grid to get
-individual cell corners.
+<li>Fill inflation restart files
+ <ul><li>There is a new utility that will write inflation restart files based on
+ values read from the console. This enables multi-step runs to start with the
+ 'read inflation values from a file' option set to .true. for all steps instead of
+ having to change the namelist after the first cycle.
+ See the documentation
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/adaptive_inflate/fill_inflation_restart.html">Website</a>
+ or <a href="../../adaptive_inflate/fill_inflation_restart.html">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>The code in <em class=file>filter.f90</em> was corrected
-to match the documentation for how the namelist item
-<em class=code>input_qc_threshold</em> is handled.
-(See filter namelist documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/filter/filter.html#Namelist">Website</a>
-or <a href="../../filter/filter.html#Namelist">local file</a>.)
-In the Jamaica release, observations with
-incoming data QC values greater than
-or equal to the namelist setting were discarded.
-Now only incoming data QC values greater than the
-<em class=code>input_qc_threshold</em> are discarded
-(values equal to the threshold are now kept).
+<li>New location module options
+ <ul><li>There are additional options for the model and observation coordinate systems.
+ Note that only a single location option can be chosen and all observations and all
+ model locations must use that coordinate system. New options include:
+ <ul><li>Channel coordinate system</li>
+ <li>[0-1] periodic 3D coordinate system</li>
+ <li>X,Y,Z 3D Cartesian coordinate system</li>
+ <li>2D annulus coordinate system</li>
+ </ul>
+ See the documentation
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/location/location_mod.html">Website</a>
+ or <a href="../../location/location_mod.html">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>The <em class=program>merge_obs_seq</em> utility
-has been replaced by the more comprehensive
-<em class=program>obs_sequence_tool</em> utility.
-See the documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/obs_sequence/obs_sequence_tool.html">Website</a>
-or <a href="../../obs_sequence/obs_sequence_tool.html">local file</a>.
+<li>Missing values in state
+ <ul><li>In some models there are values which are not valid in all ensemble members.
+ With this release it is possible to indicate that it is ok to have missing values
+ in the state vector. There is a namelist item <em class=code>allow_missing_in_state</em>
+ in the &assim_tools_nml namelist. Setting this to .true. will allow DART to
+ avoid updating any state vector items in which one or more of the ensemble members
+ have a missing value. Inflation cannot be enabled with this option, and all
+ forward operators must test for and be prepared to return with a failed forward
+ operator code if any of the values it requires are missing.
+ See the documentation
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/assim_tools/assim_tools_mod.html#Namelist">Website</a>
+ or <a href="../../assim_tools/assim_tools_mod.html#Namelist">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>The prepbufr observation converter was located in the
-<em class=file>DART/ncep_obs</em> directory in the
-last release. It has been moved to be with the other
-programs that convert various types of observation files
-into DART format. It is now located in
-<em class=file>DART/observations/NCEP</em>.
+<li>Different task layout options
+ <ul><li>The ensemble manager has a new option to distribute MPI tasks round robin
+ across the available nodes instead of assigning them sequentially. The first N
+ tasks, where N is the ensemble size, require more memory than other tasks.
+ Distributing them round-robin may allow assigning more tasks per node with a more
+ uniform memory usage. This may result in a small decrease in performance at runtime,
+ but it might allow using fewer nodes for the job and thus reduce the job cost.
+ See the documentation for the <em class=code>layout</em> and
+ <em class=code>tasks_per_node</em> in the &ensemble_manager_nml namelist
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/ensemble_manager/ensemble_manager_mod.html#Namelist">Website</a>
+ or <a href="../../ensemble_manager/ensemble_manager_mod.html#Namelist">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>The sampling error correction generator program in
-<em class=file>DART/system_simulation</em> now has a
-namelist &full_error_nml. See the documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/system_simulation/system_simulation.html">Website</a>
-or <a href="../../system_simulation/system_simulation.html">local file</a>
-for more details. Tables for 40 common ensemble sizes are
-pregenerated in the
-<em class=file>DART/system_simulation/final_full_precomputed_tables</em>
-directory, and instructions for generating tables for other ensemble
-sizes are given.
+<li>Different MPI communication options
+ <ul><li>The ensemble manager has 3 new options for the order in which the
+ communication is done when transposing the ensemble of state vectors.
+ There is a new namelist option in the &ensemble_manager_nml called
+ <em class=code>communication_configuration</em> which can have the values 1-4.
+ If DART is running slower than expected, try the various options and see
+ which is fastest on your hardware. The fastest value depends on the
+ MPI library implementation, the type and speed of interconnect, the
+ processor speed, and node memory size and so it is almost impossible
+ to recommend a value without doing timing tests on the target system.
+ See the documentation in the &ensemble_manager_nml namelist
+ <a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/ensemble_manager/ensemble_manager_mod.html#Namelist">Website</a>
+ or <a href="../../ensemble_manager/ensemble_manager_mod.html#Namelist">local file</a>
+ for more details.</li></ul>
</li>
+<br />
-<li>Most <em class=file>work</em> directories now have
-a <em class=program>quickbuild.csh</em> script
-which recompiles all the executables
-instead of a
-<em class=program>workshop_setup.csh</em> script.
-(Those directories used in the tutorial have both.)
-To control whether <em class=program>filter</em> is
-compiled with or without MPI (as a parallel program or
-not) the <em class=program>quickbuild.csh</em> script
-takes the optional arguments <em class=code>-mpi</em> or
-<em class=code>-nompi</em>.
+<li>Several more places where large arrays were put on the stack
+have been removed, decreasing the total amount of stack required by DART.
</li>
-<li>The <em class=program>preprocess</em> program was
-changed so that any obs_def files with module definitions
-are directly included in the single
-<em class=file>obs_def_mod.f90</em> file.
-This means that as you add and delete obs_def modules
-from your &preprocess_nml namelist and rerun
-<em class=program>preprocess</em> you no longer have to
-add and delete different obs_def modules from your
-<em class=file>path_names_*</em> files.
-</li>
+</ul>
-<li>The utilities module now calls a function in the mpi_utilities code to
-exit MPI jobs cleanly. This requires that non-mpi programs now
-include the <em class=file>null_mpi_utilities_mod.f90</em> file
-in their <em class=file>path_names_*</em> files.
-</li>
-<li>The <em class=file>DART/mpi_utilities</em> directory as distributed
-now works with all compilers except for gfortran. In <em class=file>
-DART/mpi_utilities</em> is a <em class=program>./fixsystem</em> script that
-when executed will change the source files so they will compile with gfortran.
-Previous releases compiled with gfortran as distributed but
-no other compilers.
-</li>
+<!--==================================================================-->
-<li>The GPS Radio Occultation observation forward operator code now
-requires a namelist, <em class=code>&obs_def_gps_nml</em>. See the
-GPS documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/obs_def/obs_def_gps_mod.html#Namelist">Website</a>
-or <a href="../../obs_def/obs_def_gps_mod.html#Namelist">local file</a>
-for details on what to add. All <em class=file>input.nml</em> files in
-the repository have had this added if they have the GPS module in their
-<em class=code>&preprocess_nml</em> namelist.
-</li>
+<a name="NewModels"></a>
+<div class="top">[<a href="#">top</a>]</div><hr />
+<h2>New Models</h2>
-</ol>
+<ul>
-<!--==================================================================-->
-
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
-<a name="NewFeatures"></a>
-<h2>New Features</h2>
-
+<li>CESM framework components
+<ul><li>DART now supports running CESM components CAM, POP, and CLM under the CESM framework.
+Setup scripts are provided to configure a single or multiple component assimilation.
+See:
<ul>
-<li>Inflation Damping
- <ul><li>Handles the case where observation density is irregular in time,
- e.g. areas which were densely observed at one point are no longer
- observed. Adaptive inflation values can grow large where the observations
- are dense, and if that region is no longer observed the inflation is
- not recomputed. Inflation damping shrinks the inflation values and
- compensates for this.
- See the inflation documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/filter/filter.html#Inflation">
- Website</a> or <a href="../../filter/filter.html#Inflation">local file</a>
- for more details and paper references.</li></ul>
+<li>
+<a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/models/CESM/model_mod.html">Website</a> or
+<a href="../../models/CESM/model_mod.html">local file</a> for multi-component assimilation,
</li>
-<li>Sampling Error Correction
- <ul><li>Compensates for the numbers of ensembles being small compared to the
- number of degrees of freedom in the system. See the last item in this section
- of the documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/filter/filter.html#GettingStarted">Website</a>
- or <a href="../../filter/filter.html#GettingStarted">local file</a>
- for more details.</li></ul>
+<li>
+<a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/models/cam/model_mod.html">Website</a> or
+<a href="../../models/cam/model_mod.html">local file</a> for CAM single component assimilation
</li>
-<li>Adaptive Localization and Localization Diagnostics
- <ul><li>See a discussion of localization-related issues
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/assim_tools/assim_tools_mod.html#Localization">Website</a>
- or <a href="../../assim_tools/assim_tools_mod.html#Localization">local file</a>.</li></ul>
+<li>
+<a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/models/POP/model_mod.html">Website</a> or
+<a href="../../models/POP/model_mod.html">local file</a> for POP single component assimilation
</li>
-<li>Scale height vertical localization option in 3d models
- <ul><li>See a discussion of specifying vertical localization in terms of scale height
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/location/threed_sphere/location_mod.html#Namelist">Website</a>
- or <a href="../../location/threed_sphere/location_mod.html#Namelist">local file</a>.
- See the <a href="http://en.wikipedia.org/wiki/Scale_height">Wikipedia page</a>
- for a discussion of how scale height is defined. Note that
- there is no support in the diagnostic Matlab routines for observations
- using scale height as the vertical coordinate.</li></ul>
+<li>
+<a href="https://proxy.subversion.ucar.edu/DAReS/DART/trunk/models/clm/model_mod.html">Website</a> or
+<a href="../../models/clm/model_mod.html">local file</a> for CLM single component assimilation
</li>
-<li>CAM supports FV code, PBS scripting
- <ul><li>See details on the features of the CAM/DART system
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/cam/model_mod.html">Website</a>
- or <a href="../../models/cam/model_mod.html">local file</a>. </li></ul>
-</li>
-<li>Boxcar Kernel Filter Option
- <ul><li>See how to select this filter option in the namelist
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/assim_tools/assim_tools_mod.html#FilterTypes">Website</a>
- or <a href="../../assim_tools/assim_tools_mod.html#FilterTypes">local file</a>. </li></ul>
-
-</li>
-<li>Option for "undefined vertical location" for obs using the 3d sphere locations
- <ul><li>See how to specify this option when creating observations
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/location/threed_sphere/location_mod.html">Website</a>
- or <a href="../../location/threed_sphere/location_mod.html">local file</a>. </li></ul>
-</li>
-<li>Schedule module for repeated time intervals
- <ul><li>See documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/time_manager/schedule_mod.html">Website</a>
- or <a href="../../time_manager/schedule_mod.html">local file</a>. </li></ul>
-</li>
-<li>Support for 2 different Mars calendars in time manager
- <ul><li>Gregorian Mars</li>
- <li>Solar Mars</li>
- </ul>
-</li>
-<li>Code corrections to make the smoother run correctly
-</li>
-<li>Forward operators now have access to the ensemble number and
-the state time if they want to make use of this information
- <ul><li>See the "Get Expected Obs From Def" section of the obs_def documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/obs_def/obs_def_mod.html">Website</a>
- or <a href="../../obs_def/obs_def_mod.html">local file</a>
- for details on how to use these values. This change is fully backwards-compatible
- with existing forward operator code.</li></ul>
-</li>
-<li>Option to output all echo of namelist values to a separate log file
- <ul><li>See the utilities module documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/utilities/utilities_mod.html#Namelist">Website</a>
- or <a href="../../utilities/utilities_mod.html#Namelist">local file</a>
- for how to select where the contents of all namelists are output.</li></ul>
-</li>
-<li>Large file support for netCDF
- <ul><li>See the
- <a href="http://www.unidata.ucar.edu/software/netcdf/faq-lfs.html">Unidata
- netCDF documentation</a> pages for more information about what large file
- support gives you and what it is compatible with.</li></ul>
-</li>
-<li>Better support for adaptive localization
- <ul><li>See the Localization section of the assim_tools documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/assim_tools/assim_tools_mod.html#Localization">Website</a>
- or <a href="../../assim_tools/assim_tools_mod.html#Localization">local file</a>
- for details.</li></ul>
-</li>
-<li>Option to localize with different distances based on observation type
- <ul><li>See the Localization section of the assim_tools documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/assim_tools/assim_tools_mod.html#Localization">Website</a>
- or <a href="../../assim_tools/assim_tools_mod.html#Localization">local file</a>
- for details.</li></ul>
-</li>
-<li>The error handler can take up to 3 lines of text so you can give more
-informative error messages on exit
- <ul><li>See the utilities module documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/utilities/utilities_mod.html#Interface">Website</a>
- or <a href="../../utilities/utilities_mod.html#Interface">local file</a>
- for details.</li></ul>
-</li>
-<li>Option to output ensemble mean in restart file format when filter exits
- <ul><li>See the filter program namelist documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/filter/filter.html#Namelist">Website</a>
- or <a href="../../filter/filter.html#Namelist">local file</a>
- for details.</li></ul>
-</li>
-<li>The start of a suite of forecast verification and evaluation tools
- <ul><li>See the verification tool documentation
- <a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/obs_sequence/obs_seq_verify.html">Website</a>
- or <a href="../../obs_sequence/obs_seq_verify.html">local file</a>
- for details.</li></ul>
-</li>
-<li>Performance improvement in the internal transposes for very large state
-vectors. all_vars_to_all_copies() now has a single receiver and multiple
-senders, which is much faster than the converse.
-</li>
-<li>Better support for users who redefine R8 to be R4, so that filter runs
-in single precision. Fixed code which was technically correct but numerically
-unstable in single precision when computing variance and covariances.
-</li>
-<li>Fixed a case in the 3D sphere locations code which made it possible that
-some observations and state variables at higher latitudes might not be
-impacted by observations which were barely within the localization cutoff.
-</li>
-<li>The observation type table at the top of all obs_seq files
-now only contains the types actually found in the file.
-</li>
-<li>When one or more ensemble members fail to compute a valid forward operator,
-the prior and/or posterior mean and standard deviation will be set to MISSING_R8
-in the output obs_seq.final file in addition to setting the DART QC flag.
-</li>
-<li>Use less stack space by allocating large arrays instead of
-declaring them as local (stack) variables in routines
-</li>
-<li>The copyright has changed from GPL (GNU) to an NCAR-specific one
-which is found <a href="http://www.image.ucar.edu/DAReS/DART/DART_download">here</a>.
-</li>
</ul>
-
-
-<!--==================================================================-->
-
-<div><hr /><p align=right><a href="#"><small>[top]</small></a></p></div>
-<a name="NewModels"></a>
-<h2>New Models</h2>
-
-
+Documentation for the model:
<ul>
-
-<li>POP Ocean Model
-<ul><li>DART interface documentation
-<a href="https://proxy.subversion.ucar.edu/DAReS/DART/releases/Kodiak/models/POP/model_mod.html">Website</a> or
-<a href="../../models/POP/model_mod.html">local file</a>.
-Documentation for the model itself
-<a href="http://www.cesm.ucar.edu/models/ccsm2.0.1/pop/">in CESM</a>
-and <a href="http://climate.lanl.gov/Models/POP/">stand-alone version from
-Los Alamos</a>.
-</li></ul>
+<li>the user's guide for CESM version 1.1.1:<br />
+ <a href="http://www.cesm.ucar.edu/models/cesm1.1/cesm/doc/usersguide/book1.html">
+ http://www.cesm.ucar.edu/models/cesm1.1/cesm/doc/usersguide/book1.html</a>
+</li>
+<li>the page that explains how to download the release code:<br />
+ <a href="http://www.cesm.ucar.edu/models/cesm1.1/cesm/doc/usersguide/x388.html">
@@ Diff output truncated at 40000 characters. @@
More information about the Dart-dev
mailing list