310 R.J. Harding et al. Agricultural and Forest Meteorology 100 2000 309–322
water potential, vapour pressure deficit, temperature and photon flux density for several coniferous forests.
Based on the analysis of Jarvis, Stewart 1988 derived a similar description in terms of variables generated
by a meteorological model e.g. soil moisture, air tem- perature, vapour pressure deficit and solar radiation.
Similar formulations have been produced for conifer- ous forest Prince and Black, 1989; Dang et al., 1997,
for tropical rain forest Dolman et al., 1991; Wright et al., 1996, for savannah Huntingford et al., 1995 and
for grassland Stewart and Verma, 1992. Huntingford and Cox 1997 confirmed, through the use of uncon-
strained neural network fitting techniques applied to data from Thetford Forest, UK, that these functional
dependences are of the correct form to describe the variation of stomatal conductance.
The JarvisStewart functions are, however, empiri- cal descriptions of complex physiological responses.
Recently, more mechanistic, coupled models of pho- tosynthesis and transpiration e.g. Collatz et al., 1992
have been included within land surface schemes of GCMs. Typical of these is the MOSES Meteoro-
logical Office Surface Energy Scheme model of the UK Meteorological Office. The canopy conductance
sub-model within MOSES was developed using data from the Kansas grassland, FIFE sites, a C
4
grass, Cox et al., 1998. MOSES has also been tested suc-
cessfully for a savannah site within HAPEX-Sahel Blyth et al., 1999. This paper describes a similar
but more extensive testing and calibration of MOSES on a 3 year run of evaporation and soil moisture
data taken over a pasture grassland in the UK a C
3
grass. These observations have provided the op- portunity to compare the performance of the land
surface model in different, contrasting years and to check the applicability, and thus predictive ability, of
a calibration made with one year’s data in subsequent years.
2. Data
2.1. Introduction The dataset is very extensive, containing over
three years of simultaneous measurements of surface fluxes, surface meteorology and soil moisture. The
years 1995–1997 are the most complete and they are described and modelled here. The three years had in-
teresting contrasts: 1995 and 1996 had below average rainfall and 1997 average rainfall. In the summers
of 1995 and 1996 the grass showed signs of water stress, whereas 1997 saw a dry spring but wet summer
months with no obvious visual signs of soil moisture stress.
2.2. Site description The pasture field at Wallingford, Oxfordshire, UK
51
◦
36
′
N 1
◦
7
′
E is shown in Fig. 1, along with the positions of the flux measurements, soil moisture
measurements and additional meteorological mea- surements. The field is permanent pasture classified
as an Lolium perenne–Cynosauriis cristatus grassland community Rodwell, 1993. It was grazed intermit-
tently with cattle during the period of measurement, such that the vegetation height varied from 10 to
30 cm. The field is flat, with a slight slope 1.5 m in 300 m from the flux mast towards the River Thames.
The soils are a 4.5 m thick layer of calcareous gravels and sands overlain between 0.5 and 1 m of a clay
loam Jarvis et al., 1984. Well level and soil mois- ture measurements showed that the groundwater was
below 3 m for the study period, and therefore outside the range of the grass roots. The field is bounded by
the river to the west, a row of mature trees to the south, a two storey building to the east and a mix of
trees and buildings to the north Fig. 1. The uninter- rupted fetch over grass exceeded 300 m to the west,
but reduced to 100 m to the north, east and south-east. The fetch over grass was therefore not perfect in all
directions and this showed itself in the measured mo- mentum exchange and calculated roughness lengths,
as described further. However the field is fairly typical of the UK and the measured energy closure suggests
the fetch is adequate see Section 4.2.
2.3. Soil moisture measurements Regular soil moisture measurements are available
from 1993 to 1997. Measurements were made using a neutron probe Bell, 1977 at a single tube. Readings
were taken every 10 cm in the top metre and at every 20 cm below this depth, down to 3 m, the readings
were taken approximately weekly.
R.J. Harding et al. Agricultural and Forest Meteorology 100 2000 309–322 311
Fig. 1. A diagram showing the position of the pasture field and associated instrumentation, all areas other than trees and buildings are grassland.
2.4. Flux and meteorological measurements Eddy correlation measurements were made of evap-
oration, λE W m
− 2
, sensible heat flux, H W m
− 2
and momentum flux using Hydra systems Shuttle- worth et al., 1988. The Hydra system consists of a
one-dimensional sonic anemometer, an infrared hy- grometer, a fine-wire thermocouple thermometer and a
sensitive cup anemometer to measure horizontal wind fluctuations, all monitored at 10 Hz. In this case the
sensor head was mounted at 3.2 m above the ground. The Hydra systems were recalibrated every year. Be-
cause the site was used for testing systems, as well as routine monitoring, there were often two, or some-
times three, systems in the field. The agreement be- tween individual systems was generally excellent.
The Hydra systems output the mean fluxes hourly. These data were unfortunately not always continuous.
There were inevitable gaps in the data due to rain on the sensors, rapid changes of temperature and humid-
ity, and sensor and logger failures. The Hydra logger software automatically detects when the data are un-
reliable and only data which was coded without fault were included in the analysis. In 1995 and 1996 over
50 of the data were so coded over 4000 h, in 1997 60 of data were coded without error. The automatic
quality control system was fairly stringent and proba- bly rejected some data which were of acceptable qual-
ity, however even with this there remains a large body of data available for modelling. Apart from equipment
malfunctions the major reason for missing data is the presence of raindrops on the sensors particularly the
hygrometer. This effect may bias the measurements. For this reason the measurements must be regarded
as a description of the transpiration, and not overall evaporation.
312 R.J. Harding et al. Agricultural and Forest Meteorology 100 2000 309–322
The Hydra system also recorded hourly average hu- midity, air temperature, wind speed and net radiation.
The humidity was recorded with a Vaisala RH sen- sor, the air temperature with the thermocouple sensor
also used for the measurement of temperature fluc- tuations and the net radiation, with a REBS Q6 ra-
diometer. To provide the continuous record of forcing data required for the model runs, these data were sup-
plemented by data from an automatic weather station AWS operated at a site 200 m north-east of the flux
tower Fig. 1. Comparisons showed that the agree- ment between the data taken at the flux mast and the
AWS was excellent. There was a small underestimate of the net radiation by the AWS at high radiation lev-
els which was almost certainly due to inadequacies in the design of the AWS Shenk radiometer. However the
hours of high radiation were sufficiently few that the totals from the two radiometers agreed well.
3. Model description
