Hillslope Hydrology

Streams are the “conduits” of the surface and subsurface runoff

generated in watersheds. GW-SW interaction needs to be
understood from the watershed perspective.
precipitation
During
D i a storm
t event,
t different
diff t hill
hillslopes
l
generate runoff at different timing and
Q2
amounts. The main channel integrates Q1
hillslope inputs.
Hydrology of 1st order streams are
controlled by hillslope processes.
Q3
Channel processes are more
i
important in
i higher-order
hi h d streams.
Q for main channel
Objectives
1. Understand hillslope runoff processes and the roles of GW.
2. Develop tools for baseflow analysis.
1

Runoff mechanisms
In this lecture, we take a broad view of hillslope – any parts of a
watershed that can potentially generate runoff.
“Rain” in the following refers to the fraction of rain that reaches
the ground after interception loss.

1. Horton overland flow, or infiltration excess overland flow.

The rainfall intensity exceeds the infiltrability of soil surface, even
though the underlying soil is still unsaturated (see next page).
The lower bound of soil infiltrability is saturated hydraulic
conductivity, which is usually > 10-5 m s-1 in undisturbed soil.
Rainfall intensit
intensity of most storm e
events
ents are smaller than 20
20-30
30
mm/hr.
→ Horton overland flow is rare in most environments.
Exceptions?

1
 Horton overland flow during an intense Soil was unsaturated
(≅ 50 mm/hr) rain event in Tanzania. under the sheet flow.

Horton overland flow over the

frozen soil near Calgary, Canada
3

2. Subsurface storm flow Pre-storm soil

moisture profile
During non-storm periods, the flow
of 1st order stream is sustained by
baseflow (BF). BF
0 θs 0 θs 0 θs
rain

Horton overland flow (HOF) may

occur when the soil surface is
severely disturbed or frozen, but
BF such cases are rare.
0 θs 0 θs 0 θs

Infiltration causes the WT to rise,

particularly in the areas where
the capillary fringe is close to the
surface. This increases the
BF+SSSF
hydraulic gradient, resulting in
0 θs 0 θs 0 θs
subsurface storm flow (SSSF).
Modified from Dunne and Leopold (1978.
Water in Environmental Planning. p.263)
4

2
3. Saturation overland flow
As the storm continues, the WT
keeps rising and eventually reaches
DPS
the surface in riparian areas. This
RF
causes a further increase in SSSF,
BF+SSSF
and return flow (RF) of groundwater 0 θs 0 θs 0 θs
to the surface.
In addition. any rain falling on saturated areas becomes surface
runoff, called direct precipitation onto saturated area (DPS).
The sum of RF and DPS is called saturation
overland flow.

Saturated areas grow during a

storm, meaning that runoff is
generated in variable source
areas. The extent of saturated Saturated area
areas also has seasonal pre-storm
50 m
variability. end of storm
Modified from Dunne and Leopold (1978, p.268)
5

Contribution of surface and subsurface runoff

rain event
Different runoff mechanisms contribute
discharge (mm hr-1)

different amounts (and quality) of water DPS + RF

at different timing.
Return flow ((RF),
), subsurface storm shallow SSSF
flow (SSSF), and baseflow (BF) all
occur under the WT. → Groundwater! deep SSSF + BF

Physical distinction between storm

time (hr)
flow and baseflow is very fuzzy. Modified from Dunne & Leopold, p.266.
rain
Using stable isotopes (2H, 18O), “event”
L s-1)

event water water resulting from the storm,

water, storm can be
discharge (L

distinguished from pre-event water

pre-event water from the storage.
Pre-event water contributes typically
time (day) 50-75 % of storm flow in low-order
Modified from Buttle (1994. Progress watersheds (e.g. Buttle, 1994).
in Physical Geography, 18:26) 6

3
There must be some processes that push the pre-event water very
quickly from the soil (and groundwater) to streams. One of the
biggest challenges in today’s hillslope hydrology is to understand
these processes (Weiler and McDonnell, 2004. J. Hydrol. 285: 3-18).

T
Transmissivity
i i it feedback
f db k
In classical hydrogeology, transmissivity T (m2 s-1) is defined by:
T = Ks × (aquifer thickness)
When Ks(z) varies with depth, the definition is modified to:
zB
T ( zW ) = ∫zW
K s ( z )dz Ks(z)
0
Ks is
i integrated
i t t d from
f the
th WT to
t a relatively
l ti l
impermeable surface, e.g. unweatherd bedrock, zW
clay, permafrost, etc. Note that T increases as
the WT rises toward the surface.
Why do we not include unsaturated K? zB

In general Ks in shallow (< 5 m) sediments decreases with depth

(e.g. Buttle, 1994). Below are examples of organic soil from the
subarctic region (left) and glacial till from the prairie region (right)
in Canada, measured by various methods described below.
Ks decreases dramatically with depth.
→ T(zW) is very sensitive to zW, especially near the surface.
surface
Ks (m d-1) Ks (m d-1)
1 10 100 1000 10-5 10-4 10-3 10-2 0.1 1 10
0 0

0.1 1
depth (m)

h (m)

0.2 2
depth

0.3 3
tracer
0.4 lab test 4 core
GP slug test
0.5 5
Quinton et al. (2008. Hydrol. Process., Parsons et al. (2004. Hydrol. Process., 18:2011)
22:2829) 8

4
Tracer Tests Lab tests: “Cube” method
(Quinton et al., 2008) (Beckwith et al. 2003, Hydrol. Process., 17: 89)
Δx

v = Δx / (travel time)
npv = Ks Δh/Δx
Ks = npv / (Δh/Δx)
Piezometer slug test
Guelph permeameter (GP)
Hayashi
y & Quinton ((2004. Can. J. Soil Sci.,, 84: 255))
5 cm H0

unsaturated

saturated

Ks(z) (mm hr-1) T(zW) (mm hr-1)

Simple models are commonly
used to represent Ks(z) functions: 0 50 100 0 10 20
0
Ks = Ks0 exp(-fz)
Ks = Ks0 (1 – z/zB)n-1
h (m)

where Ks0 is the value at z=0

( -11)
depth

f : decay
d constant
t t (m 05
0.5
zB: depth to bedrock (m)
n : dimensionless constant exp.
See Wigmosta and Lettenmaier (1999. linear
Water Resour. Res., 35: 255) 1
In this example, f = 5 m-1, zB = 1 m, and n = 2 (linear).

Integrating Ks(z) from the bedrock to the WT,

T(zW) = (Ks0/f)[exp(-fzW) – exp(-fzB)] ≅ (Ks0/f)exp(-fzW) = T0exp(-fzW)
or
T(zW) = (Ks0zB/n)(1 – zW/zB)n = T0(1 – zW/zB)n
where T0 is the value of T at zW = 0 (i.e. entire soil is saturated).

10

5
In many hillslope studies, Ks was shown to be very high near the
surface due to macropores, piping, etc., and decrease exponentially.
→ Rapid rise in the WT during a storm causes a fast flow of pre-
event water.

Fill and spill mechanism

Hillslopes commonly have
relatively impermeable boundaries
restricting groundwater flow, such
as unweathered bedrock.
The storage in bedrock “hollows” must be filled to the top before
th continuous
the ti flow
fl pathways
th for
f subsurface
b f storm
t flow
fl (and
( d
return flow) develop.
This process is called “fill and spill” runoff in the recent literature.
(e.g. Freer et al., 2002: Water Resour. Res., 38, 1269; Spence and Woo, 2003. J. Hydrol.,
279: 151).

11

Scotty Creek watershed in Northwest

Territories, Canada
This hillslope is underlain by permafrost,
which thaws in summer to 0.5-1 m below
the ground surface.
The frost table (FT) has irregular shape
shape.
(Wright et al., 2009. Water Resour. Res., 45: W05414)

Fill-and-spill model simulations for a gentle

slope (not in the photo) is shown below.
water saturation water saturation
FT topography with 5 mm of rain with 30 mm of rain
0.5 m

0.01

saturated
5m thickness (m)
12

6
 Detailed study in Panola Mountain
watershed in Georgia, USA (Tromp-van Meerveld
and McDonnell, 2006. Water Resour. Res., 42, W02411).
12 m

9 mm
Larger storm events caused saturation
of larger areas of bedrock hollows and
channels
channels.
There seems to be a threshold (50%?).
29 mm
Above this value, subsurface channels
may become connected.

subsurface sttorm
61 mm flow (mm)

0 100
saturated hillslope area (%)
13

Distributed hydrological model

Complex subsurface geometry and water-table dynamics may be
captured by physically-based, distributed hydrological models.
Such a model could couple the three-dimensional Richards
equation for subsurface flow and channel-hydraulics equations
for surface flow. The “blueprint” for the model was proposed by
Freeze and Harlan (1969. J. Hydrol., 9: 237) and implemented by
VanderKwaak & Loague (2001. Water Resources Research, 37: 999).
∂h ∂h
Simpler models use the Dupuit- Q x = −T Q y = −T
∂x ∂y
Forchhemier approximation with
transmissivity function T(zW) to move
water laterally
y on g
grid cells ((e.g.
g
Wigmosta & Lettenmaire, 1999).

These models may still be missing key processes, and it will never
be possible to collect sufficient field data to populate the models.
→ More realistic approach is offered by well calibrated, lumped
hydrological models (Beven, 2002. Hydrol. Process., 16: 189).
14

7
 Distributed Hydrological Models

Numerical solution of the 3-D Richards MIKE SHE model

equation (VanderKwaak & Loague, 2001) (www.dhisoftware.com)

100-1000 m

10-100 km

15

Lumped Hydrological Models

Precipitation
Lumped (or conceptual) models are
commonly represented by series of
reservoirs with outlets.
The example in the left is PREVAH model
of Gurtz et al. (2003. Hydrol. Process., 17: 297).
Storage and outflow characteristics of
reservoirs are controlled by a set of
empirical parameters, which needs to be
calibrated by comparing modeled
discharge hydrograph with observed one.
Ad
Advantages
t off llumped
d model?
d l?
discharge (mm d-1)

observed
model
Disadvantages?

16