Release of VIC 4.0.4
June 4, 2003: VIC release 4.0.4beta r2
This covers bugs found during tests with the snow algorithm.
solve_snow.c: Bug found by KAC
Counter for number of days since last snowfall was
incremented twice in the MELTING update. This has been
fixed.
solve_snow.c: modification by KAC
Added check so that MELTING flag is only TRUE if melt
occurs in the melt season - currently this is defined
between March 1 and October 1. Otherwise the MELTING
flag can trigger rapid very early season melt
write_model_state.c, read_initial_model_state.c, open_state_file.c,
check_state_file.c: Modification by KAC
Modified to handle both ASCII and BINARY state files.
NOTE: ASCII mode is designed to make it easier to create
a state file for initializing a point model. It includes
all features of the Binary state file, but values are
truncated so the resulting restart will not be identical.
If you need an exact restart, use the Binary files. Also
removed ice content from the state file as it is computed
at the beginning of each time step, so storing its value
is unnecessary.
April 23, 2003: VIC release 4.0.4beta r1
This covers bug fixes found by beta testers and fixed in the version of
the code bundled with this file.
surface_fluxes.c: (found by Ingjerd Haddeland)
Indexing fix sent SNOW_STEP to calc_surf_energy_bal rather
than the model time step, meaning that without snow the
evaporation was computed for SNOW_STEP hours rather than a
full day. This was fixed by introducing step_inc to
index the arrays, while step_dt keeps track of the correct
time step.
March 27, 2003: VIC release 4.0.4beta
This release covers patches for several bugs found in 4.0.3, which were
never formally released (i.e. the downloadable source code was modified,
but no major announcement was made). It also includes other fixes and
modifications that have been identified as being needed prior to releasing
version 4.1.0, which will involve several significant changes (including
lakes & wetlands, spatial snow & frost, and a closed canopy energy balance).
Modifications:
Snow albedo update: (found by Keith)
In previous releases, the snow albedo function has been hyper
sensitive to trace amounts of snowfall during the melt period.
Whenever new snow falls or the cold content falls below 0, the
albedo is switched from the ablation curve to the accumulation
curve. This curve is then followed until the cold content
exceeds 0, indicating it is in the spring melt season. This is
fine when accounting for thin early season snowpacks or mid-
season melt events, however, a cold snap or light dusting of
snow should not reset the snowpack albedo to much higher winter
values for days or weeks at a time. This release of the model
monitors the state of pack with the variable MELTING. This
flag keeps the snowpack on the ablation curve unless new snow-
fall exceeds a threshold (TraceSnow in user_def.h) indicating
that the top of the snowpack should be represented by the albedo
of the new snow.
Model initialization: (found by Ulysses and Keith)
In previous releases, the initialization of soil layer moist[]
and ice[] was within a second set of loops for band and
vegetation, using the same counters. Because of the dual use
of counters, initialization was not completed correctly. The
primary effect of this was that thermal node values beyond the
first vegetation type were not correctly initialized, which
caused the model to crash during some simulations with full
energy or frozen soils active. Without frozen soil, most
simulations would compensate for the problem within their
spin-up time, so it is unlikely that this bug impacts any
simulations not employing frozen soil.
Snow time step: (found by Andy, et al.)
The snow algorithm needs to run sub-daily for the energy balance
components to function properly. This means that for daily
simulations, the snow model must be solved at a finer (sub-daily)
time step. In the previous release, initialize_atmos.c stored
sub-daily met data in each days variable using positions (e.g.
0,1,..8 for 3 hour data). In surface_fluxes.c the model indexed
the sub-daily time steps used by the snow algorithm with hours
(e.g. 0,3,6,...21 for three hour data). This means the arrays
were incorrectly indexed and the resulting model simulations
would be wrong. The fix implemented here has been tested under
several model configurations and is deemed the official version.
WARNING: There are several versions of this fix circulating,
please update your code to this version - the previous fixes
may not work in all circumstances!
FROZEN_SOIL active flag: (found by Ed and Justin)
The cause of the problem is a bug in the code that occurs when
the global frozen soils flag (FROZEN_SOILS) is set to true but
the individual cell frozen soil flag (FS_ACTIVE) is set to
false. This causes the change in soil heat storage to be
calculated incorrectly. This was fixed by adding additional
conditions within frozen_soil.c and initialize_model_state.c,
which verify the FS_ACTIVE is TRUE before running
estimate_layer_ice_content. This avoids the problem of the
soil layer ice content being set to a positive value, ignored
by the rest of the model.
Vapor pressure: (Keith)
All internal vapor pressure calculations are now done in
Pascals. Previous release versions, switched between Pa
and kPa, so this simply removes the extra step. The input
file format is unchanged, so there should be no change to
model output (model might run slightly faster, but it is
also unlikely that this will be witnessed by a normal user).
Constant dew despite changing LAI: (Dave Peterson)
The modification of read_vegparam.c to update LAI based on
a grid cell specific value did not change the values of Wdmax.
Wdmax values were computed in read_veglib.c based on the
default LAI values, so they did not necessarily reflect the
actually LAI values used for the grid cell. Values for Wdmax
are now computed in read_vegparam.c whenever GLOBAL_LAI are
provided. The effects of this change will change in magnitude
based on how different the cell LAI values are from those in
the default library file.
DRY_TIME error: (Reinur Schnur)
DRY_TIME in dist_prcp.c was incremented by the time step in
hours. Then to see if the current rain was part of the same
storm or the start of a new one, DRY_TIME was checked to see
if it was greater than or equal to 24/dt. This compares
DRY_TIME in hours to the model time step. The "/dt" has been
removed, so now DRY_TIME is checked versus the hours since the
last storm.
State file: (KAC)
*** WARNING: This may require modifications to your global file ***
The state file has been modified to account for model updates.
It has also been converted to write binary files - this makes
them less convenient to edit, but means that model starts using
the same forcing, soil and vegetation files will produce the
exact same results. There has also been a slight change in how
the global file is set up to restart the model. The global
file should now have the same year, month, day and hour as the
original global file - the VIC model will compute the number of
records to skip at the beginning to reach the point where the
model state was saved. This means that calculations to yield
sub-daily metrological forcings from daily forcings will produce
the exact same forcing values -> this also means that restarted
simulations will be exactly the same as the original run. Slight
variations in the model results were also introduced because the
method for storing soil node depths led to the possibility of
very small differences in dz_node for the restarted model.
Previous versions also did not store the snowpack cold content,
this meant that for restarts during winter, the snowpack albedo
might start on the ablation curve (as cold content was initialized
to 0 and not to a value less than 0) rather than the accumulation
curve. This could lead to slight differences in the snowpack
if no new snow fell and the snowpack was not melting - but
after 10 years of simulations the differences were minor. As
noted above the new version of the state file should allow the
model to be restarted and to produce exactly the same results as
the original complete result. If there are cases where this is
not true, please report. If you edit the read/write model state
functions - BE VERY CAREFUL to edit Nbytes to reflect any changes.
COMPUTE_TREELINE: (KAC and LCB)
This is an added feature which computes the treeline elevation
in the current grid cell and does not include vegetation
fractions with overstory in the grid cell averages for snow bands
that exceed the treeline elevation. This feature was added
to the model to reduce the appearance of "glaciers" in high
elevation snow bands. It computes average annual July air
temperatures using the temperature data from the atmospheric
forcing files (WARNING - elevation of treeline may change if
the period of forcing data used changes). It then lapses the
average annual July air temperature to locate the elevation
at which it equals 10C. This is assumed to be the treeline,
so vegetation types with overstory in snow bands with average
elevations higher than this, are excluded from the grid cell
average variables in put_data.c. For the time being full
energy and water balances are still computed for these
vegetation fractions, and no attempt is made to verify that
a snow band has a non-overstory vegetation type that can be
expanded to represent the coverage area lost due to the
exclusion of the overstory fraction.