Appendix B. Water Balance Model (WBM) Description
(Vörösmarty 1998)
The WBM simulates soil moisture variations,
evapotranspiration, and runoff on single grid cells using biophysical data sets
that include climatic drivers, vegetation, and soil parameters. The state
variables are determined by interactions among time-varying precipitation,
potential evaporation, and soil water content. The original model has been
described in detail by Vörösmarty and coworkers (Vörösmarty et al. 1989,
Vörösmarty et al. 1996, Vörösmarty and Moore 1991). The governing equations can
be summarized as follows:
Equation variables:
dWs/dt change
in soil moisture
p probability
that a day has a wetting event
Pa available
precipitation (rainfall and snowmelt) [mm day-1]
g(Ws) unitless
soil drying function
Ep potential
evaporation [mm day-1]
Dws soil moisture deficit equal
to amount of water required within a time step to fill soil to its water
holding capacity and simultaneously satisfy potential evaporation [mm day-1]
Es estimated
actual evapotranspiration [mm day-1]
Xr excess
rainfall (available for runoff and runoff detention pool) [mm day-1]
Pr rainfall
Xs excess
snowfall (available for runoff and runoff detention pool) [mm day-1]
Ms snowmelt
Ws soil
moisture [mm]
dWs/dt =
pPa{g(Ws) + [1-g(Ws)]e-(Ep/Pa)
– e-(Dws/Pa)} – Epg(Ws) Pa < Ep
(original equation = -g(Ws)(Ep
– Pa))
= Pa
– Ep Ep
< Pa < Dws
= Dws
– Ep Dws
< Pa
Es =
Pa – dWs/dt Pa < Ep
= Ep Ep <
Pa
Xr =
0 Pa
< Dws
= Pr
–Dws Dws
< Pa
Xs =
0 Ms
< Dws
= Ms
– Dws Dws
< Ms
Equation variables:
a empirical constant
Ws soil
moisture [mm]
Wc soil
and vegetation-dependent available water capacity [mm]
g(Ws) = 1
– e (-aWs/Wc) / 1 – e-a
a = 5.0 so that the drying
curve would resemble that of Pierce (1958) when g(Ws) = Es/Ep
is plotted as a function of Ws/Wc during periods of no
precipitation.
Equation variables:
p probability
that a day has a wetting event
u = 0.005 best
fit for broad geographic range
Prm monthly
rain
p = 1 – e-u(Prm)
Assumes that there is an exponential distribution of rainfall amounts during those days that are rainy within a month.
Equation variables:
Pr rainfall
Prm monthly
rain
nd number of days in a month
Pr = Prm/ndp
During periods of snowmelt recharge we add Ms directoly to Prm, increasing the probability of soil wetting. We have assumed that the within-month variability in potential evaporation is much less than for rainfall and snowmelt recharge and can be approximated by Ep = Epm/nd, where Epm is computed potential evaporation.
Equation variables:
Dr rainfall
runoff detention pool
Rr rainfall-derived
runoff emerging from the grid cell
Xr excess
rainfall (available for runoff and runoff detention pool) [mm day-1]
b and g empirical
constants set to 0.0167 [day-1] and 0.5 respectively
dDr/dt = ( 1 – g) Xr – bDr
Rr = gXr
+bDr
Snowpack
Snowpack accumulates when monthly temperatures are below
–1.0 oC. Snowmelt is a prescribed
function of temperature and elevation. Snowmelt occurs at or
above –1.0 oC.
Equation variables:
Ps snowfall
[mm day –1]
Ks accumulated
snowpack [mm water equivalent]
Ms snowmelt
excess [mm day –1]
Ep potential
evaporation [mm day-1]
T monthly
temperature
At or below 500 m elevation, Ms = Ks in the first month where T >= -1.0 oC.
dKs/dt = Ps – Ep T < -1.0 oC
dKs/dt = -Ms –Ep T >= -1.0 oC
Above 500 m elevation, snowmelt proceeds over 2 months with 0.5 of Ks lost within each month.
A detention pool is tracked which is used to generate runoff associated with snowmelt.
Rs snowmelt-derived
runoff [mm month-1]
Ds snowmelt-derived
detention pool
Xs excess
snowfall (available for runoff and runoff detention pool) [mm day-1]
dDs/dt = Xs – Rs
Rt = Rs +Rr
At or below 500 m:
Rs = 0.1 Ds month = 1 of T >+ -1/0 oC
Rs = 0.5 Ds month > 1 of T >= -1.0 oC
Above 500 m:
Rs = 0.1 Ds month = 1 of T >= -1.0 oC
Rs = 0.25 Ds month = 2 of T >= -1.0 oC
Rs = 0.5 Ds month > 2 of T >= -1.0 oC
The expected monthly changes in the pools are calculated as
the average daily change multiplied by nd, the number of days in
each month. Likewise, the associated water fluxes computed by the WBM are
initially expressed as a daily average for the duration of each month. These
also are multiplied by nd to obtain corresponding monthly values.
The time varying changes in Ws, Ks, Dr and Ds
are solved using a fifth-order Runge-Kutta integration technique (International
Mathematical and Statistical Libraries, Houston TX).