Software Open Access
Joe Hamman; Bart Nijssen; Ted Bohn; Wietse Franssen; yixinmao; homefc; Diana Gergel; Bin Peng; The Gitter Badger; apcraig
Release date: (September 2, 2016)
This is a major update from VIC 4. The VIC 5.0.0 release aims to have nearly identical physics as VIC 4.2 while providing a clean, refactored code base supporting multiple drivers. There are a number of new features, bug fixes, and backward incompatible changes. See the VIC Github page for more details on the changes included in this release.New Features:
Although the physics and model behavior of VIC 5.0.0 should be nearly identical to VIC 4.2, the source code has undergone a major cleanup and reorganization. We have separated the physical core ("vic_run") from the driver source code. This work has improved the extensibility and readability of the model.
Classic Driver (GH#7)
The Classic Driver provides similar functionality as VIC 4, including ASCII and binary I/O, and a time-before-space evaluation loop order. The Classic Driver is maintained for two main reasons:
Image Driver (GH#7)
The Image Driver adds a number of features to the user interface of the VIC model. Most notably, it uses a space-before-time evaluation loop order, netCDF I/O, and parallelization using MPI. Image Driver specific documentation can be found here.
Constants File (GH#192)
Earlier versions of VIC included many hard-coded parameters and constants. We have consolidated these constants into a single structure and developed an input file that allows users to modify parameters at run-time. See here for more information.
A set of logging Macros have been added to all drivers and vic_run. The logging level can be set in the driver Makefile via the LOG_LVL variable. The logging Macros provide the filename and line number in the source code to aid in debugging. Additionally, when compiler support is available, a traceback is printed when VIC exits during runtime. When the LOG_DIR variable is provided in the global parameter file, VIC will write its log(s) to log files instead of printing to stdout.
Sub-hourly Timestep (GH#188)
Previous versions of VIC were limited to a minimum timestep of one hour. The units of the VIC timestep have been changed from hours to seconds and the minimum timestep is now one second. If you intend on running VIC at a timestep of less than one hour, we suggest extensive testing.
Calendar Support (GH#188)
Earlier versions of VIC used the standard Gregorian calendar. Because many modern climate models use non-standard calendars, we have implemented all CF compliant calendars. The standard Gregorian calendar remains the VIC default. See the documentation for individual drivers for how to set the calendar option (e.g. classic.
Sample Datasets (GH#387)
The VIC_sample_data repository contains the necessary input datasets (forcings and parameters) to run short simulations of the VIC model for both the classic and image driver.
Tests Datasets (GH#79)
See https://github.com/UW-Hydro/VIC/issues/79 for more information. A temporary location of the test data is here: ftp://ftp.hydro.washington.edu/pub/gergel/VIC5_test_data/
Testing and Continuous Integration (GH#190)
A comprehensive testing platform has been implemented and is available for public use along with the VIC model. A small subset of the test platform is run on Travis-CI, which facilitates continuous integration of the VIC test platform. More information on the test platform is here.
Run-time profiling and timing (GH#442)
A timing module has been added to VIC in order to assess the computational cost and throughput of the VIC model. New output variables (OUT_TIME_VICRUN_WALL and OUT_TIME_VICRUN_CPU) document the time spent in vic_run for each variable. Additionally, a timing table is printed to LOG_DEST at the end of each simulation.
The format of ASCII forcing and output files has changed in VIC 5. These changes were motivated by the desire to improve simulation metadata tracking and reproducibility of VIC simulations.
Classic Driver Global Parameter Options
A number of global parameter options have changed for the Classic Driver, relative to VIC 4.
State files now include seconds (GH#464)
Classic Driver Output Variables (GH#352)
Computation of potential evapotranspiration (PET) has been simplified, reducing the number of output variables from 6 to 1. The following output variables have been removed:
These have been replaced by:
Removed MTCLIM (GH#288)
Previous versions of VIC used MTCLIM to generate missing forcing variables required to run VIC. This led to confusion by many users and considerably more complex code in the Classic Driver. VIC forcings are now required to be provided at the same time frequency as the model will be run at (SNOW_STEPS_PER_DAY).
As part of this change, the following options have been removed from the Classic Driver:
As part of this change, the following output variables have been removed from the Classic Driver:
In the future, we would like to provide a stand-alone version of MTCLIM that produces subdaily meteorological forcings. We are looking for community support for this feature (GH#17)
Removed LONGWAVE and SHORTWAVE forcing types (GH#379).
Previous versions of VIC allowed users to specify either LONGWAVE or LWDOWN to denote the incoming longwave radiation flux and SHORTWAVE or SWDOWN to denote the incoming shortwave radiation flux. We have removed these duplicate options, standardizing on the more descriptive LWDOWN and SWDOWN.
Similarly, output variables OUT_NET_LONG and OUT_NET_SHORT have been replaced with OUT_LWNET and OUT_SWNET, respectively.
Changed the name of the variable VEGCOVER to FCANOPY, since this more accurately captures the meaning of the term (i.e., the fractional area of the plant canopy within the veg tile). Similarly changed OUT_VEGCOVER to OUT_FCANOPY.
Similarly, changed the names of the following global parameter file options:
Miscellaneous fixes to lake module (GH#425)
Several lake processes (aerodynamic resistance, albedo, latent/sensible heat fluxes, net radiation, etc) were reported incorrectly or not at all in output files. This has been fixed. In addition, in the absence of an initial state file, lake temperatures were initialized to unrealistic temperatures (the air temperature of the first simulation time step). To fix this, we now initialize the lake temperature to annual average soil temperature.
Fix for computation of soil layer temperatures when soil thermal nodes do not reach the bottom of the soil column. (GH#467)
Previously, if the soil thermal damping depth was shallower than the bottom of the deepest soil layer, and FROZEN_SOIL==TRUE, VIC would abort when estimating layer ice contents because it could not estimate a layer temperature if the thermal nodes did not completely span the layer. Now, a layer temperature is estimated even when thermal nodes do not completely span the layer, and the error no longer occurs.
Previously, VIC did not produce the same results (fluxes and states) if a simulation was separated into multiple shorter-period runs by saving the state variables and restarting. This was due to:
The MTCLIM algorithm resulted in slightly different sub-daily meteorological variable values for different lengths of forcings (MTCLIM is deprecated in the current version)
A few bugs resulting in inexact restart. The following bugs have been fixed:
Two flux variables energy.LongUnderOut and energy.snow_flux are now saved to the state file.
!!!Note This is a temporary solution to ensure exact restart. A better way of handling these two flux variables needs to be done in the future (see GH#479)
Fix for binary state file I/O (GH#487)
Fixed a bug so that the binary format state file I/O works correctly.
Fix for a physical constant (water heat capacity) (GH#574)
Fixed a bug where volumetric heat capacity of water should be used in func_canopy_energy_bal (previously specific heat capacity was used).