459 
( 4 ) 
T 1/2 
where 9 U represents the rewetted moisture content which is equal to the saturation soil moisture 
minus entrapped air fraction, K u is the hydraulic conductivity evaluated at 9 U \ and D is the soil 
moisture diffusivity. 
2.1.2 Percolation and Capillary Rise. Percolation during the storm event and capillary rise dur 
ing the interstorm period are assumed to be in a steady state. Employing the Brooks-Corey soil 
characteristic relationships (Brook and Corey, 1974), the net vertical flux in the transmission zone is 
estimated using the expression developed by Salvucci (1993): 
where g is the steady-state flow rate in the transmission zone, positive for capillary rise and negative 
for percolation. z x is the depth of the local water table. Tp c is the air-bubbling tension head, ip sz 
is the tension head of the surface zone. A = 2 + 3 B, where B is the Brooks-Corey soil pore size 
distribution parameter. 
2.1.3 Exfiltration. Analogous to the treatment of infiltration, we can estimate the exfiltration rate, 
f e , by taking the minimum of exfiltration capacity, /*, and potential évapotranspiration, e p : 
Neglect the effect of gravity and again applying the time condensation approximation, the exfil 
tration capacity can be expressed as a function of cumulative exfiltration, F e , and the initial condition 
at the start of the interstorm event: 
where S e is the desorptivity and is given by Milly (1986) as a function of initial soil moisture 9^: 
spatially-distributed data in model parameters, inputs and outputs. Catchment topography is rep 
resented by digital elevation model (DEM) information. The local model is applied to each grid 
fe = min[f*,e p } 
( 6 ) 
2.1.4 Runoff. Both infiltration excess and saturation excess runoffs are accounted for in the model. 
Saturation excess runoff occurs when rain falls on the saturated grid elements which are either ad 
jacent to the stream, or have a small storage deficit that can be easily satisfied during the storm. 
Infiltration excess runoff is generated on those parts of the watershed where p > /*. No flow routing 
is carried out in this model. Thus, the total streamflow volume for the watershed is obtained by 
summing the contribution to surface runoff from each grid element in the computational domain and 
the subsurface base flow Q j, described below. 
2.2 Watershed-Scale Water Balance Model 
Given the local water balance model presented in the previous section, we employ an aggregation 
scheme similar to that used in Famiglietti (1992) to construct a model suitable for use in watershed- 
scale hydrologic simulations. A raster-based geographic information system (GIS) is used to manage