The proximity of the WMO station (locations P4 and
T4) to the SPCFLUX93 site allowed us to implement and test an original
method (named CREADB and described in Ruti et al., 1997 and in Cassardo
et al., 1998) to derive from synoptic observations (which are wide-spread)
the necessary data to drive SVAT (Soil Vegetation Atmosphere Transfer)
schemes.
As know, a SVAT model needs boundary conditions:
air pressure, temperature, humidity information (relative, specific, dew–point,
etc.), wind vector, precipitation, solar radiation or cloudiness. CREADB
method can use synoptic data to produce the necessary information for SVAT
models.
We tested this procedure during SPCFLUX93. The
results showed that air temperature and relative humidity data are reproduced
in a good way, wind velocity in a satisfactory way, while precipitation
is somehow smoothed, due to the fact that synoptic data only provided the
information of cumulated precipitation in the last 6 hours. From these
conclusions, we can state that CREADB can be safely used for climatological
purposes.
Sensible, latent and ground-atmosphere heat fluxes,
evaluated with the above mentioned procedures and the net radiation data
measured in location T1B, were compared with the output coming from a simulation
run with the biospheric model Land Surface Process Model (LSPM, version
“2” of 1996) of Cassardo et al. (1995b). The assumed vegetation type was
short grass. As initial soil temperatures, “climatological values”
for SPC site were used. Regarding initial values of soil moistures, we
used a particular method in order to avoid the so-called “spin-up problem”:
in fact, as known, the climatic system can require a few months before
the initial values of the soil moisture fields are forgotten and this can
be an estimate of the time needed by a climate model to adjust realistically
to initial errors in the soil moisture distribution. Then, we used the
values coming from the long time simulation (6 months from January to June
1993) carried out on SPC data and described in Ruti et al. (1997), where
it was demonstrated that, after this long time simulation, LSPM reached
its equilibrium state in June, and then the June values can be considered
as unaffected by spin-up problems. Finally, the boundary conditions necessary
for the above mentioned 6-months run (2 m temperature, 2 m relative humidity,
2 m wind, sea level pressure, precipitation and cloud coverage) were extracted
and arranged from synop data by using CREADB algorithm (see § 6.3).
The necessary radiation input (long and short wave) for the two models
were simulated using the empirical package of LSPM, taken from Page (1986).
In Figures 23a are reported the time series of
observations (points) and LSPM output (solid line) during the whole period
of SPCFLUX93 campaign for net radiation. Generally speaking, the agreement
between data and model predictions was good. Some overestimates and underestimates
of midday values in clear and cloudy sky conditions were due to errors
in the interpolation of cloudiness from synoptic observations (from which
the radiation inputs are calculated) or to inaccuracy of the global radiation
parameterisation in cloudy days.
Figures 23b and 23c show sensible and latent heat
fluxes observations (points) and LSPM output (solid line). From these plots,
we can infer that LSPM predictions of both heat fluxes were close to the
observed data in their whole range. Only a few maximum values of the latent
heat flux were underestimated, particularly in concurrence with large errors
in the evaluation of net radiation or in cloudy sky conditions. For this
reason, the Bowen ratio predicted by LSPM is larger than the observed one.
Figure 23d shows the heat flux G0 at the ground-atmosphere
interface evaluated from observations using eq. 3 (points) compared with
the LSPM calculations (solid line). The observations resulted in good agreement
with the predictions, with the exception of a little overestimate during
daytime around noon and the presence of strong negative minima (calculated)
at sunrise.
Thermal wave propagation into soil
The data referring to temperature measurements
at 7 levels in the first meter of soil allowed us to test different soil
parameterisation schemes. In particular, we compared two widely used soil
schemes: the (5-level) multi-layer scheme, used for instance in the
LSPM (Cassardo et al., 1995b), and the (3-level) force-restore scheme,
used for instance in the BATS (Dickinson et al., 1986). This comparison
was performed running BATS and LSPM (driven with the same initial conditions)
for a 6 months simulation (starting from 1st January 1993). Synoptic data
arranged with CREADB method (see § 6.3) were used as initial and boundary
conditions. In Ruti et al. (1997) we demonstrated that the multilayer scheme
produced a more accurate estimate of the thermal wave propagation with
respect to the results coming from the use of the classic force-restore
method.