no (1 5 Min) 
Figure 2: Time series of watershed average soil moistures in the surface zone updated using the 
PBMR measurements (represented by A). 
However, in order to properly determine the weights, there is a need to explore the error structures 
of the hydrologic model and the remote sensors. 
The remote sensing soil moisture information used in the above simulation is based on the PBMR. 
Because of the small size of the watershed and the mixed land cover conditions resulted from agricul 
tural practice, the only reliable soil moisture information that PBMR can provide is the watershed 
average values. The SAR, on the other hand, yields spatial soil moisture patterns with a higher res 
olution (on the order of 2500 m 2 for MAC-HYDRO’90). The effect of the resolution of soil moisture 
pattern upon simulated evaporation is examined by comparing the temporal variations of watershed 
average evaporation based on the two different sensors. We found that the average evaporation is 
not sensitive to the resolution of soil moisture information, especially during the first 7 days of the 
simulated period. This result is expected because at the initial stage of the dry-down period, the soils 
remain to be atmospheric-controlled. As the soils dry down, the margin of the difference between 
these two simulations becomes larger. However, because the catchment under study is rather small 
and the lack of an energy balance component in the model, it is difficult to pinpoint the underlying 
mechanism responsible for the discrepancy. Research efforts are currently directed toward the incor 
poration of an energy balance component and a vegetation parameterization in the model. 
4 CONCLUSIONS 
Spatial distributions of soil moisture over an agricultural watershed with a drainage area of 60 ha 
were derived from two NASA microwave remote sensors. These information were used to determine 
the initial condition for a distributed hydrologic model based on a 1-D local water balance model. 
Simulated hydrologic variables over a period of 12 days were compared with field observations and 
with model predictions based on a streamflow-derived initial condition. It was concluded that the 
remotely sensed soil moistures can be used as a feedback to correct the state of the hydrologic model. 
A Kalman filter framework is developed to take advantage of the multi-temporal remotely sensed 
data. There is a need to investigate the underlying error structures of the various remote sensors. 
For a small region such as the WD38 subwatershed, the higher data resolution provided by the SAR 
does not have a great impact upon the simulations. 
463