André et al. (1990) summarizies results of using a simplified hydrological model of Girard and Boukerma (1985) with HAPEX-MOBILHY data. The longer term measurement program that combined energy balance observations and soil moisture profile measurements (Goutorbe et al., 1989) permitted calibration and testing of the large scale hydrologic model because the longer times scales associated with changes in soil moisture in the root zone were observed. Monthly averages of runoff, rainfall, soil moisture storage and ET were compared to observations over a 10 x 10 km grid. Even though the model computed a 27% larger total in moisture extraction than was measured, the model did predict the north-south gradient in ET. A more physically-based scheme of the surface layer water budget was introduced by Ottlé et al. (1989) that included a vegetation component. Comparison of model derived soil moisture with observations indicated a significant improvement in simulated water content. Moreover, satellite images providing maps of water deficit were used to verify model output of the distributed water content over the study area.
Another validation study of the longer term water budget with the HAPEX-MOBILHY data was performed by Mahfouf (1990). The land surface scheme developed by Noilhan and Planton (1989) was used because it contains relatively few parameters while preserving the basic physics controlling the energy and water transport in the soil-plant-atmosphere continuum. The scheme has been validated on a diurnal basis with HAPEX-MOBILHY data (Jacquemin and Noilhan, 1990). To test its capability in simulating the water balance of the soil profile, especially in the root zone, requires simulations for much longer time periods, namely weeks to months. The one dimensional version of the French Weather Service Mesoscale Model (Bougeault et al., 1989; 1991) provided the atmospheric parameters for the surface scheme. For six sites having a range in soil types and cover, the land-surface scheme satisfactorily reproduced ET and soil water content. By having the land-surface scheme linked to the planetary boundary layer, local air temperature and relative humidity could be compared to observations. The evolution of these meterorological variables were realistic for most sites, except for those whose local environmental conditions differed markedly from the surronding region. This study is one of the first attempts at linking the hydrologic cycle to an atmospheric model at the mesoscale.
A more recent attempt of incorporating a hydrologic
model within an atmospheric prediction model is presented
by Smith et al (1994). Validation of the model output of
soil-water content was not feasible over the whole model
domain (4.8 x 10
km
). However, the soil hydrology
model output of several grid elements were compared to soil
moisture measurements collected during the the NASA
Multisensor Airborne Campaign for HYDROlogy (MAC-HYDRO)
in 1990 (Jackson et al., 1994). This experiment
took place over the Mahantango Creek Watershed maintained
by the U.S. Department of Agriculture, Agricultural
Research Service (USDA-ARS). The temporal trend of
daily soil moisture simulated by the model for the upper
15-cm followed the overall pattern of the measurements
averaged over the 7 x 20 km sampling area. In general they
concluded that improvement in the model mainly lies in the
acquisition of more detailed precipitation data. This may
come from satellite and ground-based radar systems such as
the NEXt Generation Weather Radar (NEXRAD) systems
(Crum and Alberty, 1993). In addition, satellite-derived
cloud cover would enhance model reliability in computing
the fluxes via better simulations of regional solar radiation;
but, at present they concluded that it is too costly and time
consuming to integrate into atmospheric models.
Besides ARME, other longer term measurements over a single site have been used in assessing the water and energy balance simulations at the GCM-scale. One such measurement program lasting 3 years was conducted in a fallow savannah in the Sahel, the Sahelian Energy Balance Experiment (SEBEX) (Wallace et al., 1991). These data were used to evaluate the performance of the United Kingdom Meteorological Office General Circulation Model in simulating the climate of the Sahel (Dolman et al., 1993). The testing of the three-dimensional GCM was performed with the monthly average data while a one-dimensional version of the GCM (Dolman and Gregory, 1992) was used to study the hourly and daily performance. The model overestimated monthly solar radiation by 20% during the dry season. There was also an overestimation in monthly evaporation at the start of the rainy season and a more rapid decline during the dry months compared to observations. The analysis of the 1-D version of the model allowed a detailed examination of the land surface-atmosphere parameterizations. Comparison of daily evaporation with observations (Gash et al., 1991) indicated adjustments to the conductances in the model were necessary to obtain reasonable agreement. This, however, is a diagnostic correction and would not necessarily lead to improved simulations in the future. Therefore, in many natural environments supporting multiple vegetation species, micrometerological measurements need to be combined with plant physiological measurements (Dolman et al., 1993). This will permit better calibration of the physical parameters of multiple-source land surface energy balance models (e.g., Dolman, 1993) and lead to long term simulations of ET. These more reliable ET estimates can then be used in water balance calculations with hydrologic models. However, a major difficulty both in the modeling and in validation for this region is the precipitation component. The Sahel region has significant spatial and temporal gradients in rainfall. This is compounded by the fact that the density of raingage networks are relatively low.
Although there has been a marked improvement in
modeling the surface-atmosphere exchanges over different
land types and climates, several of the above studies suggest
that the modeling of precipitation is still largely inadequate.
The EPSAT-NIGER (Estimation des Precipitations par
SATellite au Niger) experiment was implemented in 1988 as
a means of monitoring the long term rainfall for the Sahel
(Lebel et al., 1992), and to serve as an important first step
in the implementation of the 1992 HAPEX-Sahel experiment.
The EPSAT-NIGER raingage network has 100
stations spread over 16,000 km
, which is 20 times more
dense than the operational network. Preliminary analyses
and results from three years of data (1989-1991) have been
reported (Taupin et al., 1993). Differences in precipitation
estimates as large as 50% were obtained using a 200 km
raingage network versus the operational 1,600 km
raingage
network. At the 200 km
grid size, smoothing of the
rainfallfields given by the sparse raingage networks typically
available resulted in errors reaching 100%. These results
indicate that it will be difficult to assess the spatial
distribution of seasonal/annual precipitation over the Sahelian
region with the present operational raingage network. Hence, the
water balance of basins 100 km
will be difficult to model
and validate. Taupin et al. (1993) point out that their
analysis indicates large variability in annual rainfall appears
to be related to the significant spatial variation in the number
of rainfall events. This information may be retrieved from
the European Meteorological satellite (Meteosat). Therefore
it may be possible to use satellite data to monitor the
occurrence of these events, and with high density experimental
raingage networks develop relationships between
number of occurances and cumulative rainfall on an annual
or seasonal basis.