[Science]   [BAS home]   [Met home]   [Beowulf home]   [Techdb index]	Antarctic Meteorology

Please note: the files here are from the Hadley Centre and presumably remain their intellectual property, assuming the concept applies The files are "read-only" in this place: don't try the "enter problem" etc features... it won't work.

UM System Problems

Select keyword, or page down to browse database

Known bug: do not resize page, after having selected a problem title

MODEL

OCEAN

UMUI

SCRIPTS

RECONFIGURATION

UTILITIES

GENERAL

MPP

Back to ....

Enter problem

Delete entry

Append entry

Edit entry

View problem database

MODEL

OCEAN

UMUI

SCRIPTS

RECONFIGURATION

UTILITIES

GENERAL

MPP

MODEL

Title: Barrier hang caused by TIMER in atmos code

Author: S.D.Mullerworth

First submitted: Tue Dec 22 11:05:27 GMT 1998

TIMER was changed at 4.5 to include a barrier, however, if only some
PEs run a particular piece of code with a timer in it, then the
PEs start to become out of sync. The solution is to change timer call
to a value above 100 so that the barrier call is not made.

This problem occurs in Hydrology 5A decks (Section 8 Moses I, 
HadCM3 option) where certain subroutines are not called by PEs 
without land points. This is resolved by modset gsm1f406.

[formerly problem 404, re-entered due to system problems MJH]

______________________________________________

Title: STASH accumulations/means with times list

Author: Paul Selwood

First submitted: Tue Jan 5 10:47:20 GMT 1999

As problem 408 for vn4.4. Time processing (but not time-series)
with times lists rather than frequencies do not work at vn4.5
either deadlocking or just not doing the processing.

Fixed at vn4.5 by gps0f406

______________________________________________

Title: Must use ask1f406 with vn 2B of Section 9

Author: Stephen Cusack

First submitted: Wed Jan 6 11:00:20 GMT 1999

Version 2B of Section 9 was built into the UM at vn4.5, for the new
RHcrit parametrization. Only HadCM4 developers run with this version
of section 9. The new deck called CALRHC2B calls SQRT at four
locations.
For optimisation purposes, at the review stage the argument passed
to the SQRT function was defined to be the difference between two
quantities. When these two quantities are identical, the argument to
the SQRT can become negative at the bit level (I'm not sure if the
number is -0.0, or if it's actually less than zero). Therefore, the
model crashes with a fatal arithmetic error. This is quite a rare
occurrence, but anyone doing a HadCM4-type model integration MUST
use ask1f406 in the 4.6 mods directory on the T3E.

______________________________________________

Title: MOSES II : STASH workspace not initialised in section 19

Author: Richard Betts

First submitted: Fri Feb 5 11:48:07 GMT 1999

*** THIS PROBLEM ONLY AFFECTS MOSES II ***

In section 19, STASH workspace is uninitialised.
Mod abx1f406 initialises the workspace to zero.

This mod is not required when section 19 is not in use
(ie: A19_0A is set in UPDEFS)

______________________________________________

Title: MOSES II : redundant code in SETMODL

Author: Richard Betts

First submitted: Fri Feb 5 12:07:51 GMT 1999

*** THIS PROBLEM ONLY AFFECTS MOSES II ***

SETMODL contains code to reset the version mask for section 19.
This code was written before section 19 came into use for 
vegetation, and although the original requirement for it no 
longer exists, it causes a problem with the extraction of
diagnostics from the new section 19.  

abx2f406 removes the redundant code.

This mod is not required when section 19 is not in use
(ie: A19_0A is set in UPDEFS)

______________________________________________

Title: MOSES II : STASH frequency for section 19 diagnostics

Author: Richard Betts

First submitted: Fri Feb 5 12:20:35 GMT 1999

*** ONLY AFFECTS MOSES II ***

PRELIM1 sets incorrect start times and frequencies for diagnostics
available on phenology or TRIFFID timesteps.  PHENOL_PERIOD and
TRIFFID_PERIOD are available in units of days but are used as if
they are in units of atmosphere timesteps, and the start time is set 
to the first atmosphere timestep.

abx3f406 sets new variables A_PHENOL_STEP and A_TRIFFID_STEP to
hold the phenology and TRIFFID periods in timesteps, and sets
the start time to the first phenology/TRIFFID timestep accordingly.

This mod is not required when section 19 is not in use
(ie: A19_0A is set in UPDEFS)

______________________________________________

Title: MOSES II : sections 3 and 19 diagnostics corruption

Author: Richard Betts

First submitted: Fri Feb 5 16:06:25 GMT 1999

*** THIS PROBLEM ONLY AFFECTS MOSES II ***

In section 19, the call to STASH includes 3 as the
section number as a result of cut and pasting code from section 3.
This leads to diagnostics from both sections 3 and 19 being 
corrupted when either phenology or TRIFFID are enabled  section
19 diagnostics are unset, while section 3 diagnostics are too small
by factors of 24, 48, 240 or 480.

abx0f406 corrects the section number in the STASH call.

This mod is not required when section 19 is not in use
(ie: A19_0A is set in UPDEFS)

______________________________________________

Title: TRIFFID: initialising leaf turnover rate

Author: Richard Betts

First submitted: Tue Feb 16 12:24:45 GMT 1999

*** THIS PROBLEM ONLY AFFECTS TRIFFID ***

Due to an incorrect IF..THEN..ELSE arrangement in INITVEG,
the accumulation prognostic for leaf turnover rate is not
initialised.  This is corrected in abx4f406.  This mod is
only required when TRIFFID (A19_2A) is in use.

______________________________________________

Title: Time interpolation not allowed for effective vegetation parameters

Author: Richard Betts

First submitted: Tue Feb 23 13:49:29 GMT 1999

This is only an issue when prescribing seasonal effective 
vegetation paraameters.

Currently the reconfiguration does not allow the effective vegetation 
parameters to be interpolated in time, probably because they are 
assumed to be constant.  This prevents the prescription of 
seasonally-varying vegetation parameters with update frequencies
other than 30 days.  The mod xbx0f406 enables time interpolation 
for these fields.

If FIXED effective vegetation parameters are being used (as is 
normal practice) then this mod is not necessary.

______________________________________________

Title: Values of test diagnostics (15,231-234) incorrect for MPP

Author: Rick Rawlins

First submitted: Mon Mar 1 12:37:25 GMT 1999

Routine TESTDIAG generating test diagnostics had not been updated for
MPP and so values of latitude and longitude calculated as part of the
analytic test function are not appropriate. This is corrected by modset
$UMDIR/vn4.6/mods/source(arr0f406).

All previous versions valid with MPP (vn4.3,vn4.4) would have the same problem.

______________________________________________

Title: Correct Restartability problem in autopp_tidyup after model archiving failure.

Author: W Roseblade

First submitted: Thu Mar 18 12:52:38 GMT 1999

When a run fails during the transfer of archive data to the IBM the 
restart file is transferred to the directory:-
/u/m20/cprod/oparcherr.

However, if a subsequent archive transfer failure occurs during 
autopp_tidyup after the first failure then the restartfile is no 
longer present and thus cannot be restarted by the operators.

The solution to this was to add a Step 4 to autopp_tidyup which is
executed if an error was found in this script with the archive
transfer process (AUTOPPCODE set to non zero). This then
calls restartinfo where a new restartfile is generated.

Also, for the first run "failure" the restartfile move to OPARCHERR
code was taken out of qsmain and put back but in submitchk.

Problem fixed in mod gwr4u406.

______________________________________________

Title: Overwriting of U with Item 208 section 4

Author: S.D.Mullerworth

First submitted: Thu Mar 18 14:53:17 GMT 1999

An overwriting error has been discovered in the atmosphere component
of the UM at vn4.5. Selection of the diagnostic item 208 section 4
"VISIBILITY AFTER LS PREC IN LEVEL M" results in a serious overwriting
of the prognostic U field if "Total Aerosol fields" in Atmos->Model
Config -> Prognostic variable. Unfortunately, the overwriting doesn't
cause a crash - it just sets all negative U points to 0.1.

The fix should be used in all atmos or coupled runs.

/u/um1/vn4.6/mods/source/asm1f406

______________________________________________

Title: Updating land fields from ancillary files

Author: Richard Betts

First submitted: Wed Mar 24 15:59:23 GMT 1999

There is confusion in the chain of subroutines REPLANCA - READFLDS - 
READ_MULTI - GENERAL_SCATTER_FIELD over whether some fields are 
land-only or full fields, which means that when (for example) the
snow-free albedo field is updated, the ancillary data is incorrectly 
scattered to the PEs and values such as zero and the missing data 
indicator end up in the wrong place (thus causing a model failure).  
The same would occur when updating other land fields, ie: all soil & 
veg ancillaries.

What happens is this:

i) REPLANCA calls READFLDS which calls READ_MULTI, which reads in the
ancillary data on PE 0.  For snow-free albedo (and all veg and soil
ancillaries), the field is stored in the ancillary file on all points
even though the prognostic field in the model is on land points only
(this is true in both the standard ancillary file and my own file with
monthly albedos).

ii) READ_MULTI calls GENERAL_SCATTER_FIELD to scatter the data to all
PEs.

iii) GENERAL_SCATTER_FIELD checks the grid code to see whether it is a
land or full field.  The grid code for the snow-free albedo prognostics
specifies it as a land-only field, so FROM_LAND_POINTS is called to
expand it to a full field.  Since it is already a full field, the first
2381 points get treated as if they were the land points, and the
remaining 4627 points are discarded.

iv) GENERAL_SCATTER_FIELD scatters the data to all PEs then calls
TO_LAND_POINTS to try to get the data back to a land field.

v) Back in REPLANCA, TO_LAND_POINTS is called again because REPLANCA
knows that the original ancillary field was on the full grid.  However,
the field has already been corrupted in GENERAL_SCATTER_FIELD so the
final result contains lots of zeros and missing data.

A fix is in preparation.

______________________________________________

Title: Timer Reports Incorrect CPU Times with SHMEM_NAM

Author: Bob Carruthers

First submitted: Wed Apr 7 11:16:21 GMT 1999

The CPU times reported by TIMER at the end of a job are
in general wrong when using shmem_nam.  This arises because the
default mode at the end of a job appears to be to get data.  Before
PE 0 can guarantee to have collected the data from the other PE's
for printing, the other PE's overwite it with the data from the next
routine.  This occurs whether true CPU time or elapsed time are being
used as the sample.

Fixes are available in:

$UMDIR/vn4.4/mods/fixes/fbc0f404

$UMDIR/vn4.6/mods/source/gbc0f406

______________________________________________

Title: DIF_CTL Fails with fpe when L_TRACER_THETAL_QT is true

Author: Bob Carruthers

First submitted: Wed Apr 7 11:48:47 GMT 1999

RECIP_RS_SQUARED_DELTAP is uninitialised in DIF_CTL when called
with L_TRACER_THETAL_QT set to true.  Fixes are in:

$UMDIR/vn4.4/fixes/fbcdf404

$UMDIR/vn4.6/source/gbcdf406

______________________________________________

Title: Diagnostic - Omega on pressure levels (202,12)

Author: Dave Robinson

First submitted: Tue May 18 10:31:18 GMT 1999

During testing of 4.5 for operational implementation, this
diagnostic caused model runs to crash due to NaN values in
the haloes. The modset gdr1f406 add extra swapbounds and sets
unused bottom rows to zero.

-------------------------------------------------------------------

Traceback caused by this failure. It crashes in STACCUM called
from TEMPORAL.

SIGNAL: Floating point exception (invalid floating point operation)

 Beginning of Traceback (PE 31):
  Interrupt at address 0x8001eabf4 in routine '???'.
  Called from line 492 (address 0x8002028c4) in routine 'TEMPORAL'.
  Called from line 4163 (address 0x8001fab20) in routine 'STWORK'.
  Called from line 3881 (address 0x8001ef7b4) in routine 'STASH'.
  Called from line 6906 (address 0x800021288) in routine 'ATM_DYN'.
  Called from line 3366 (address 0x800053810) in routine 'ATM_STEP'.
  Called from line 3986 (address 0x8002256a8) in routine 'U_MODEL'.
  Called from line 3055 (address 0x800233270) in routine 'UM_SHELL'.
  Called from line 475 (address 0x800000c98) in routine '$START$'.
 End of Traceback.

______________________________________________

Title: Global-Mes Parallel running

Author: Dave Robinson

First submitted: Tue May 18 10:41:41 GMT 1999

The modset gdr0f406 is required if running the Global (with 3DVAR)
and Mesoscale in parallel - ie, only in operational runs. It is
included in the global operational modset for 4.5.

It corrects the initial forecast time relative to T+0 that is sent
to the text file associated with parallel running.

______________________________________________

Title: incrementing number of timesteps since TRIFFID

Author: Richard Betts

First submitted: Tue Jul 6 13:09:23 GMT 1999

There is a bug in the way the carbon fluxes are corrected after 
TRIFFID has been called.  The intention is to apply the correction 
in the first call to boundary layer after TRIFFID, using the counter 
A_INTHD(23).  The problem is that this counter is incremented
immediately before the call to the veg section  therefore, since
boundary layer is called before the veg section in the atmos timestep,
the BL section doesn't see A_INTHD(23).EQ.1 until TWO 
timesteps after the call the TRIFFID.  This will contribute to the 
carbon imbalance seen in the model.  To correct this, ABX5F406 moves
the update of A_INTHD(23) to the beginning of the physics timestep.

______________________________________________

Title: Corruption of disturbed fraction diagnostic

Author: Richard Betts

First submitted: Thu Jul 8 13:37:51 GMT 1999

The inclusion of the unset variable PSLEVEL in the code to copy the 
disturbed fraction field to STASH causes the diagnostic to appear 
stripey.  This is corrected in ABX6F406.

______________________________________________

Title: need fix aff1f406 to run IAU with mixed phase microphysics

Author: Bruce Macpherson

First submitted: Thu Jul 15 09:44:32 GMT 1999

In an IAU run for 3DVar assimilation, using the mixed phase microphysics,
the model will generate an erroneous moisture field and almost certainly
 fall over.  The fix aff1f406 corrects the use of the ice array input to THETL_QT

______________________________________________

Title: Increased timings in U_MODEL

Author: Dave Robinson

First submitted: Fri Jul 16 13:11:46 GMT 1999

At 4.5, a flush of Unit 6 was added to the end of every timestep
in U_MODEL. This has increased the timings of U_MODEL in operational
and climate runs.

The modset $UMDIR/vn4.6/mods/source/gdr2f406 removes this flush.

Also, the call to flush has caused occasional crashes due to a t3e I/O
problem.

Therefore you should use this modset in all 4.5 runs.

______________________________________________

Title: Infinite resubmission fix

Author: Steve Mullerworth

First submitted: Fri Jul 23 13:45:10 GMT 1999

If you run with automatic resubmission and a target end date that is
not a timestep boundary (eg running for 1 month and 30 minutes with
a 20 minute timestep), the final run will resubmit infinitely. 

Modset gsm7f406 causes the final run to abort instead. 23/07/99

______________________________________________

Title: need fix aff2f406 when assimilating visibility

Author: Bruce Macpherson

First submitted: Tue Nov 16 10:56:05 GMT 1999

Without the fix, the IAU increments from a VAR analysis can cause aerosol to go negative.
If running with Edwards-Slingo radiation, then the model can fall over.

______________________________________________

Title: SW Diagnostic 213 (Convective Cloud at lit points) with anvils

Author: J. M. Edwards

First submitted: Thu Nov 25 10:01:49 GMT 1999

When anvils are used this diagnostic should sample the convective
cloud amount in the top convective layer. The current addressing
of the array of convective cloud fractions gives a field of zeros
or may cause a model failure, depending on the number of wet and
cloudy levels used. The modset $UMDIR/vn4.6/mods/source/adb3f406
should be used to correct this. The evolution of the model is
bit-comparable with this modset. This modset would also work at 4.4
if used in conjunction with $UMDIR/vn4.5/mods/source/adr1f405 which
initializes convective cloud amounts outside convecting levels.

______________________________________________

Title: Prevent occasional failures in COEX packing

Author: Steve Mullerworth

First submitted: Mon Nov 29 15:25:49 GMT 1999

The function CRI2IBM in COEX is not declared and there is no
IMPLICIT NONE in the deck. This causes occasional crashes when
it returns a real value to integer IER.

gsm9f406 adds an INTEGER declaration

______________________________________________

Title: Reduce memory in very high res climate runs

Author: Steve Mullerworth

First submitted: Mon Nov 29 17:09:59 GMT 1999

Failures due to lack of memory below subroutine MEANCTL could
be fixed by inclusion of 

/u/um1/vn4.6/mods/source/gsm8f406

which reduces the size of an oversized array. 

Essential in 1/3 degree ocean jobs that have climate meaning.

______________________________________________

Title: Gregorian calendar - model integration ends at wrong time.

Author: Matt Huddleston

First submitted: Tue Jun 13 14:48:29 GMT 2000

There is a bug in INITTIME that effects models using the Gregorian
calendar.  It occurs when (a) The model MODEL_BASIS_TIME is not
the 1st January together with (b) some component  of RUN_TARGET_END 
in months.

The calculation of TARGET_END_STEP is corrupted as the number of
steps is calculated from the 1st January regardless (in TIME2SEC).
The outcome is an integration that ends a day or two too early or 
too late.

Richard Hill is having a look to develop a suitable mod.

______________________________________________

Title: Fourier damping too strong

Author: William Ingram

First submitted: Thu Aug 31 08:48:21 GMT 2000

The Fourier damping used to maintain stability at higher latitudes is
too strong by a factor of 4 at 4.5 & all preceding releases, which can
cause instabilities at the limit of filtering.
awi2f406 corrects this.  (~t20eb/mods/awi2f406_4.4 at 4.4,
~t20eb/mods/awi2f406_before4.4 at earlier releases still.)
The excess filtering means that what with hindsight seem very low 
filtering safety factors have been used: with this bug corrected 
one can expect to need larger filtering safety factors.
For further details see newsgroup posting.

______________________________________________

Title: PP file reinitialisation for Gregorian calendar

Author: M Huddleston

First submitted: Tue Jan 23 11:21:37 GMT 2001

PP files are not reinitialised if the time step is longer
than 59 minutes and the Gregorian calendar is used.  A simple
mod below fixes this problem by changing the logic in
SETTSCTL (NB Wont work for timesteps > 23 hrs 59 mins!)

Mod : /u/m20/home/mea/t20hh/um_mods/vn4.6/gmh0f406

______________________________________________

Title: Problem in MPP filtering in unusual configurations

Author: William Ingram

First submitted: Thu Apr 12 09:47:51 BST 2001

The chunking of the MPP filtering at 4.4 & 4.5 is not safe in all 
cases - it is possible (though not when running anything close to 
any normal Met Office configuration) to need filtering but have less 
ffts to do than PEs to do them, so that MPP_FILTER is called on some 
PEs with the number of ffts to do, which is the size of a couple of 
dynamically allocated arrays, zero, but the first element of one of 
those arrays referenced (in the T3E case where the trouble was found 
the second of these caused failure: with other compilers the first 
might, or the code might execute but over-write).
AWI3F406 corrects this.

______________________________________________

Title: Advection creating extrema

Author: William Ingram

First submitted: Tue Dec 11 09:11:08 GMT 2001

The advection scheme is designed so that half the increment physically
due to motion from one grid-box to the next is added in to the 
*source* grid-box, clearly unphysical but with numerical benefits.  
But this can lead to run-away instability if the source grid-box
has much less heat capacity than the downstream one, as the source 
grid-box gets increments proportional to how much it differs from the 
downstream one.  I presume it is not normally a problem because:
1) There is about as much air flowing into the grid-box, which at 
least is having half the counteracting effect it physically should.
2) The physics will tend to damp it out.
3) The problem needs a substantial local anomaly to get started.
4) The heat capacity constraint, taken with the normal grid, means 
it can probably only occur with strong vertical motion away from the 
model boundary, which will not normally be common.
but it did occur at the S Pole (no physics) in a run with a strong
inversion at adjacent points (so (1) was not so effective) &, I 
assume, orographically forced strong ascent low down.

______________________________________________

Title: Too little Fourier filtering can bring trouble

Author: William Ingram

First submitted: Tue Dec 11 09:25:01 GMT 2001

The model's Fourier damping is designed solely considering CFL-type 
problems from E-W flow, but
1) It does in fact damp out other polar instabilities which caused 
rapid failure when I ran the model with such a short timestep the 
filtering switched off entirely & without anything else to remove 
them.
2) It created failures because of inconsistency between the N-S flow 
on adjacent rows when I ran the model with a timestep so short that 
the filtering switched on only very close to the pole, giving very 
different N-S flows on adjacent rows, when the model was simulating
quite strong cross-polar flow.
Setting a large FILTERING_SAFETY_FACTOR fixed the latter at least
(as a hand-edit, since the umui will not allow values large enough 
to help).

______________________________________________

OCEAN

Title: Unprotected calls to SWAPBOUNDS relating to ice model code

Author: R. Hill

First submitted: Thu Apr 1 08:25:14 GMT 1999

OSWAPDIAGS contains two new SWAPBOUNDS calls at UM4.5
which relate to ice model fields (CARYSALT and OCEANHEATFLUX).
These calls are unconditional with the result that
they are called regardless of whether the ice model is activated
or not. The dimensioning of the arrays is such that these
spurious calls do not have any adverse effect on results
(ie no memory overwriting occurs) - the problem is one of
extra unnecessary work and code consistency.
CARYSALT is also fully dimensioned in OSWAPDIAGS. Its
dimensions should be conditional on the presence of the ice model.

Modset ORH0F406 corrects this. 

______________________________________________

Title: Ocean Assimilation inconsistencies

Author: R Hill

First submitted: Thu Jun 24 09:49:47 GMT 1999

Ocean assimilation will fail if TIMER is switched on at UM4.5.
This is due to introduction at 4.5 of a version of TIMER which 
requires synchronisation at each TIMER call across all PEs.
The structure of the assimilation code do not always lend 
itself to this. A modset is available which 
replaces these TIMER calls with the non synchronizing
versions.

Also, OARFVS1A uses GC_IBCASTS to broadcast INTEGER data.
These should use GC_RBCAST for consistency, although the 
code has always worked correctly in the past because the 
UM has, to date, always assumed the same word length for 
INTS and REALS.

Modset orh1f406 fixes both these problems.

______________________________________________

Title: correction for LAM ocean using Gill boundary condition

Author: Sarah Ineson

First submitted: Tue Jul 6 16:04:58 GMT 1999

Row 2 meridional flux divergence for LAM models using the Gill 
lateral boundary condition is incorrect at vn4.5 only

modset osi0f406 corrects this 

______________________________________________

Title: Correct WSX and WSY at PE halos

Author: R. Hill

First submitted: Wed Jul 21 13:00:38 GMT 1999

WSXM and WSYM are incorrectly set at UM 4.5 (and 4.4 for which
a fix modset already exists).
A correction was introduced at UM 4.5 to fix this,
however a conflicting modset reinstated the original 
problem, resulting in no net change.
Modset orh2f406 removes the spurious occurrence of the
incorrect code, leaving just the correct version.
All UM 4.5 ocean models using wind stresses would be 
advised to use this mod. 

______________________________________________

Title: Correct reference latitude on rotated ocean grids

Author: R. Hill

First submitted: Wed Mar 15 10:15:33 GMT 2000

U.M. 4.5 contains code to allow the MPP ocean
model code to use rotated grids and obtain bit comparable results
on any number of PEs. In particular a mod was added at 4.5 to compute
an average reference latitude for use in DENSCOEF (for calculating
an average value of acceleration due to gravity) - calculating
the coefficients used in the equation of state. While tests with
existing models suggest a consistent reference latitude is used
on all PEs, it is not necessarily the "average" latitude intended 
by the original non-mpp version of the code. Further, it may not
always be bit reproducible for all possible model domains.

Modset orh3f406 corrects the calculation to use an "average" latitude
in the form intended by the original code. This will not change
model answers to any appreciable extent, but should avoid any
potential bit reproducibility problems.

orh3f406 is recommended for all rotated grid ocean models.

______________________________________________

Title: Non-seaice ocean models

Author: R. Hill

First submitted: Fri Nov 3 12:23:25 GMT 2000

At UM 4.5 a number of SWAPBOUNDS calls were introduced
to BLOKINIT relating to ice model variables: CARYSALT/HEAT
and FLUXTOICE. However the calls are unconditionally
called. Since the arrays in question ARE conditionally
dimensioned, inappropriate memory access may occur
in ocean models which do not use seaice.
Modset frh1f406 corrects this. 
It should be used in all non-seaice ocean models at 4.5

______________________________________________

Title: Out of bounds opertions in QUICK scheme

Author: R. Hill

First submitted: Mon Jan 15 10:17:15 GMT 2001

Modset orh4f406 (written by D. Storkey) corrects
out of bounds operations in ADV_SOURCE when using the
Quick scheme.

The problem is not usually fatal because the
oob memory usually contains zeros, though 
the effect on the validity of results is unknown.

This correction will be applied at UM 5.2. 

______________________________________________

Title: Bug in ICEDRIFT cyclic conditions

Author: R Hill

First submitted: Mon Jan 21 11:01:01 GMT 2002

See entry under UM version 5.2.

______________________________________________

UMUI

Title: Spurious setting of L_FLUXCORR

Author: R Hill

First submitted: Tue Aug 15 09:08:34 GMT 2000

L_FLUXCORR is the logical switch controlling flux correction in the
ocean model. This switch is only available to coupled ocean-atmos
models. In an ocean-only model the switch is greyed out in the UMUI.
However, it's value is still picked up when processing the umui job.
Thus, if the switch is set to "Flux correction applied (coupled
models only)", but access to the switch is greyed out because
the atmosphere is turned off, the logical L_FLUXCORR is still set to
true by the umui. This will cause failures concerning unrecognised
or missing fields during dump reading either during reconfiguration
or during the main model run.

In order to get round this it is currently necessary to 
temporarily activate the atmosphere to access to the flux 
correction switch.
Turn it off and then deactivate the atmosphere. 

______________________________________________

SCRIPTS

Title: Re-compiling after BUILDSECT=true has previously failed.

Author: Rosalyn Hatcher

First submitted: Fri Feb 12 16:10:12 GMT 1999

After a run with BUILDSECT=true has failed with a compile error, the
next run that is submitted runs on even though there is not a complete
set of .o files in the section-version (Rebuild) directory, and then
has missing externals in the run.

Mod gro1u406 fixes this.

______________________________________________

Title: Mod to prevent Camelot failure causing disks to be filled

Author: Steve Mullerworth

First submitted: Thu Dec 9 17:43:46 GMT 1999

If camelot fails it kills server. But model does not detect server
failure so will run on till disks are full.

Use

/u/um1/vn4.6/mods/scripts/gsm1u406

______________________________________________

RECONFIGURATION

Title: Problems with deep soil temperature

Author: William Ingram

First submitted: Wed Oct 17 10:23:11 BST 2001

1) When the dump to be output from reconfiguration has more soil 
levels than the one input, and configuring from an ancillary file is 
not specified, so that there is no source of data for the extra 
levels, the problem is not trapped & the added levels are output 
filled with random numbers - in most cases apparently valid REALs,
so the model runs but develops an odd surface climate or delayed
numerical problems.
This has apparently been the case for a long time, probably always.
At 5. there is code to deal with it.

2) In one case at 4.5 a run specifying that deep soil temperatures
are to be reconfigured from an ancillary file when the input dump had 
fewer levels than the the file & the output dump failed with a 
floating point exception in INANCILA. Other such cases have worked -
the standard number of soil levels was increased with MOSES.

______________________________________________

UTILITIES

Title: Comparing vn4.5 and pre-vn4.5 dumps with cumf

Author: D.Goddard

First submitted: Tue Dec 22 11:32:06 GMT 1998

When comparing a pre-vn4.5 dump with a vn4.5 dump, be sure to state the
pre-vn4.5 dump first.

There is a bug in the indexing which means that the wrong field is used
from  the second file for STASH code 22  or later for MOSES dumps.

______________________________________________

Title: Convieee and old C90 ocean dumps

Author: D.M. Goddard

First submitted: Tue Feb 2 13:04:29 GMT 1999

Convieee at vn 4.5 fails to correct the compressed field indices when
converting from a c90 dump with compressed ocean fields to a T3E dump
with no compressed ocean fields. Modset udg0f406 corrects this.
A script to run the corrected executable is available at 
~t11dg/um/vn4.5/utils/ieee

______________________________________________

Title: qxhistreport output wrong

Author: Steve Mullerworth

First submitted: Mon Mar 1 11:42:00 GMT 1999

The current output from qxhistreport does not print out the 
"Current Restart dump(s) name" correctly. A modset has been lodged
to correct this (usm1f406) and the executable has been rebuilt with
this included.

To pick up the new version, switch on "Using test small executables"
in Submodel Independent -> General Config... in the UMUI. This adds
/u/um1/umtest/vn4.5/exec to your pathname in UM runs.

______________________________________________

Title: Failure of ieee -32e option with fieldsfiles

Author: D.M. Goddard

First submitted: Mon Jul 12 15:39:37 GMT 1999

If ieee is executed on fieldsfiles anc converting to 32 bit with 
unpacking of WGDOS fields, it will fail when writing last field.

Solution: 

Take a copy of the ieee script /u/um1/vn4.5/utils/ieee and edit
the execution line as follows:

Replace 

$EXEC/convieee $PRECISION <$USTSNUM

with

$EXEC/convieee.wgdos $PRECISION <$USTSNUM 

Then run with new script.

______________________________________________

GENERAL

Title: TSTMSK bug - unavailable diagnostics not rejected

Author: Steve Mullerworth

First submitted: Tue Jan 19 12:22:08 GMT 1999

One or two bugs in TSTMSK are corrected by gsm2f406. If a diagnostic
is requested that is marked as unavailable in the UMUI STASH panel,
duplicate logic in the UM should also reject it, but in some instances
there are bugs in this logic.

______________________________________________

Title: Restartability when mean reference time not a P1 mean period

Author: Steve Mullerworth

First submitted: Mon Mar 1 12:50:05 GMT 1999

Runs using climate meaning are not restartable if the mean reference 
time does not coincide with a P1 mean period. eg if the mean ref time
is 1st Dec, and the run is started with an NRUN from the 21st Dec 
with a period 1 mean of 1 month.

The problem was caused by a wrong assumption of the sign of the 
OFFSET_DUMPSim variable which is corrected by 

/u/um1/vn4.6/mods/source/gsm5f406

______________________________________________

Title: Wrong filenames for reinitialised boundary datasets

Author: Dave Robinson

First submitted: Mon Jul 19 15:22:52 GMT 1999

The filenaming convention for reinitialised boundary datasets
was changed at 4.5 and an error introduced in the process.

To get the correct filenames, the following modset is required.

$UMDIR/vn4.6/mods/source/gdr3f406

Required in any model run (atmos or ocean) setting up 
reinitialised boundary datasets. Not required in operational
suite.

______________________________________________

Title: Non-portability of some PORTIO2A C routines

Author: Paul Selwood

First submitted: Tue May 23 13:37:55 GMT 2000

Some routines in PORTIO2A use Cray extensions in declaring 
character arrays in a Fortran-like manner. ie
char fname[*len+1] 
where len originates from the function argument list.
This means that PORTIO2A will not compile on some platforms
(alpha and HP for example).

The fix is to revert to ISO C compliant behaviour thus:-
char *fname 
fname = calloc(*len+1,1) 
.
.
.
free(fname) 

This is implemented in gps1c406 which is recommended for
non-Cray builds.

______________________________________________

MPP

Title: Climate meaning on 1x1 PE causes overwriting

Author: Steve Mullerworth

First submitted: Mon Nov 29 11:17:59 GMT 1999

Maximum field size in ACUMPS calculates maximum global field size,
but this is smaller than local field size+halos when on 1x1 PE leading
to overwriting.

Use gsm3f406 which is a general rewrite of the climate meaning code
that does not incorporate this bug.

______________________________________________