GSWP-2 Initialization and Spin-up
The
ten-year GSWP-2 baseline period spans the calendar years 1986-1995. However,
the initial time for integrations is 1 July 1982. The reason is to
provide a spinup period so that the models will be adequately equilabrated
and have a reasonably realistic climate state at the beginning of 1986 that
both reflects climate anomalies in the preceeding years, and is consistent
with the model itself.
Initial conditions will be posted on the GSWP-2 data servers, but they will
be described here. The start date is in boreal summer, so that all
ice-free land points can be safely initialized with no snow cover. Likewise,
canopy interception and other surface water stores should begin at zero.
As in GSWP-1, soil wetness at all points and all layers shoulde be
initialized at 75% of saturation. Soil temperature should be initialized
at the mean June air temperature at all layers. For models that predict
vegetation, LAI, NDVI, and FPAR are available from ISLSCP-II from the beginning
of 1982. Data from July 1982 can be used to initialize the vegetation
state. To summarize:
- Snow and other surface water stores = 0
- Soil wetness = 75% of saturation
- Soil temperature = mean June air temperature (provided)
- Vegetation = July 1982 (provided)
Spin-up will be performed using data beginning 0300UTC 1 July 1982. There
are two phases to the spin-up period. In the first phase, LSS integrations
will loop through the first 12 months of forcing data until the modeler is
satisfied that soil moisture and temperature has spun up and sufficiently
equilibrated.
A lesson from GSWP-1 pilot project was that this spin-up process overly amplifies
the impact of climate anomalies from that year on the land surface state
variables. Therefore, the models will then proceed with their integrations
forward from July 1983 – December 1985 so as to converge to a realistic “land
climate” at the start of the evaluation period. The 10-year baseline
integration, which will be evaluated within the group of GSWP participants
and later released to the community at large, covers the 10-year period from
0000UTC 1 January 1986 up to 0000UTC 1 January 1996.
Not all of the data sets for the spin-up period are available from ISLSCP-II.
COLA has all of the required NCEP/DOE reanalysis fields for the period,
but not all of the same observed data sets used for hybridization of the
data are available. Also, the 3-hourly SRB radiation data are not avaiable
before 1986. As a result, there will be some differences in the forcing
data between the spin-up period and the 10-year baseline period:
- Radiation: Hybrid of NCEP/DOE 3-hourly surface downward shortwave
and longwave with SRB, based on the monthly-3-hourly systematic errors found
in the NCEP/DOE data for 1986-1995.
- Precipitation: GPCC gauge data are not available for hybridization, so the CRU precipitation and its less-dense network will be used instead.