Surface runoff and sediment dynamics in arid and semi-arid regions


Lange, J.; Leibundgut, C.

International Contributions to Hydrogeology 23: 115-150

2003


Surface flow plays an important role in the ecological balance of dry areas, being responsible for the distribution of renewable water resources and for enhanced sediment dynamics. The present chapter first provides an overview of dominant processes, followed by a practical concept which applies process knowledge to water harvesting. Different techniques for quantifying surface runoff and sediment dynamics are then introduced. Direct measurements are compared with indirect estimation tools, the latter providing valuable alternatives in areas with missing data. Finally, to illustrate the general characteristics as they are now known, case studies from different parts of the globe are presented. Since the processes occurring and methods applied are pre-defined by scale and climate, the studies are grouped accordingly. To a certain degree all the presented findings are site specific. However, if the scale and climatic regime are similar, the principal results may be translated to other locations. They then are of particular value in the search for adequate research tools or for first approximations.

CHAPTER
4
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
Jens
Lange
and
Chris
Leibundgut
Institute
of
Hydrology,
University
of
Freiburg,
Fahnenbergplatz,
Freiburg,
Germany
ABSTRACT:
Surface
fl
ow
plays
an
important
role
in
the
ecological
balance
of
dry
areas,
being
responsible
for
the
distribution
of
renewable
water
resources
and
for
enhanced
sediment
dynamics.
The
present
chapter
first
provides
an
overview
of
dominant
processes,
followed
by
a
practical
concept
which
applies
process
knowledge
to
water
harvesting.
Different
techniques
for
quantifying
surface
runoff
and
sediment
dynamics
are
then
introduced.
Direct
measurements
are
compared
with
indirect
estimation
tools,
the
latter
providing
valuable
alternatives
in
areas
with
missing
data.
Finally,
to
illustrate
the
general
characteristics
as
they
are
now
known,
case
studies
from
different
parts
of
the
globe
are
presented.
Since
the
processes
occurring
and
methods
applied
are
pre
-defined
by
scale
and
climate,
the
studies
are
grouped
accordingly.
To
a
certain
degree
all
the
presented
findings
are
site
specific.
However,
if
the
scale
and
climatic
regime
are
similar,
the
principal
results
may
be
translated
to
other
locations.
They
then
are
of
particular
value
in
the
search
for
adequate
research
tools
or
for
first
approximations.
4.1
GENERAL
ASPECTS
4.1.1
Runoff
generation
processes
In
humid
regions
there
is
an
obvious
excess
of
precipitation
over
the
seasonally
integrated
water
need
for
an
abundant
plant
cover.
Different
runoff
generation
processes
(e.g.
runoff
from
saturated
areas,
piston
-flow
effects,
macropore
fl
ow
and
the
slow
outflow
of
large
groundwater
bodies)
sustain
the
fl
ow
of
perennial
rivers.
On
the
other
hand,
semi
-arid
zones
can
be
viewed
as
those
where
a
favourable
water
balance
is
achieved
only
seasonally.
During
the
wet
season
most
precipitation
infiltrates
to
refill
underground
storages
emptied
during
the
long
dry
period.
Humid
runoff
generation
processes,
dependent
on
the
abundance
of
water,
lose
significance;
runoff
is
increasingly
generated
as
infiltration
excess
overland
fl
ow
following
the
ideas
of
Horton
(1933).
Conditions
for
this
process
are
even
more
favourable
in
arid
areas,
mainly
as
a
result
of
the
absence
of
a
developed
soil
and
vegetation
cover
and
exposure
of
impervious
surfaces.
Surface
runoff
hence
achieves
renewed
significance
in
desert
environments
(Gat
1980).
Overland
fl
ow
may
be
defined
as
fl
ow
of
water
over
the
land
surface
towards
a
stream
channel
and
as
the
initial
phase
of
surface
runoff
in
dry regions.
If
rainfall
intensity
at
any
time
during
a
storm
exceeds
the
infiltration
rate
of
a
soil,
water
accumulates
on
and
near
the
surface.
The
soil
infiltration
rate
usually
declines
exponentially
with
time
reaching
a
constant
final
value.
When
surface
depressions
are
filled,
water
spills
over
to
run
116
Understanding
water
in
a
dry
environment
40
E
E
30-
(i)
a
20-
c
O
0
10-
c
0
J3
.030
ho
A
.
...
.......
is
B
iktt
db.o
El
....
a °
C
;i1P,
e.
it
.
b.
0..,7
4R.fix
A
RAtit
............
0
20
40
60
80
Accumulated
rain
(mm)
Figure
4.1.
Typical
infiltration
rates
for
Mediterranean
(A),
semi
-arid
(B)
and
arid
(C)
sites
along
a
climatic
transect
(Lavee
et
al.
1998).
downslope.
On
plane
surfaces
(e.g.
paved
urban
areas
or
a
laboratory
fl
ume)
a
thin
film
or
even
sheet
of
fl
owing
water
may
develop,
often
termed
sheet
fl
ow.
On
natural
slopes,
however,
topographic
irregularities
direct
most
runoff
water
into
lateral
concentrations
of
fl
ow.
Following
anastomosing
paths,
these
concentrations
often
give
the
appearance
of
fl
ow
in
a
wide
braided
channel
and
hence
no
simple
description
or
modelling
of
overland
fl
ow
hydraulics
is
possible
(Emmet
1978).
Experimental
fieldwork
by
Lavee
et
al.
(1998)
in
the
Near
East
has
shown
the
climatic
dependence
of
surface
fl
ow
generation.
Sites
were
located
along
a
climatic
transect,
from
the
Mediterranean
(600
mm
annual
precipitation)
through
the
semi
-arid
(300
mm
annual
precipitation)
to
an
arid
climate
(100
mm
annual
precipitation).
Organic
matter
content
and
the
stability
of
soil
aggregates
generally
decreased
with
aridity.
As
a
consequence
infiltration
rates
also
decreased,
as
observed
from
a
set
of
rainfall
simulation
experiments
(Figure
4.1).
Widespread
infiltration
was
the
dominant
process
in
the
Mediterranean
climate
area
and
Hortonian
overland
fl
ow
dominated in
the
arid
area.
The
transitional
semi-
arid
area
was
characterized
by
a
mosaic
-like
pattern
of
patches
contributing
and
accepting
surface
water.
Similar
patterns
of
different
hydro
-ecological
and
vegetational
characteristics
were
also
found
for
semi
-arid
areas
in
the
Sahelian
Zone
of
Northern
Africa
(Bromley
et
al.
1997).
In
general
these
spatial
patterns
are
highly
vulnerable
to
anthropogenic
or
climatic
changes,
and
highlight
the
delicate
ecological
balance
of
semi
-arid
environments.
Other
experimental
studies
also
emphasise
the
transitional
character
of
semi
-arid
zones.
Martinez
-Mena
et
al.
(1998)
studied
the
natural
hydrological
response
of
four
(0.3-
0.75
ha)
micro
-catchments
in
semi
-arid
Spain
over
a
three
year
period.
In
more
degraded
areas
with
fine
textured
and
poorly
permeable
soils
Hortonian
overland
fl
ow
was
found
to
be
the
dominant
runoff
generation
process.
In
soils
with
coarser
texture
runoff
occurred
only
after
saturation.
In
terms
of
runoff
generation
the
environment
could
be
separated
into
areas
where
humid
processes
prevail
and
those
where
arid
processes
dominate.
The
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
117
runoff
response
for
the
latter
area
was
more
accentuated
(9%
runoff
coefficient,
3.6
mm
as
threshold
for
runoff
initiation)
than
for
the
former
(<3%,
8
mm
respectively).
In
a
similar
environment
Bergkamp
(1998)
studied
a
cultivated,
terraced
slope.
He
found
that
in
extreme
natural
events
overland
fl
ow
was
generated
on
several
parts
of
the
slope,
but
did
not
reach
the
channel.
Observed
runoff
at
the
catchment
scale
had
to
be
attributed
to
areas
adjacent
to
the
stream
bed
or
uncultivated
parts
of
the
catchment.
In
drier,
truly
arid
regions
(annual
rainfall
below
100
mm)
plant
cover
is
only
concen-
trated
in
small
patches
and
in
most
areas
organic
matter
is
totally
absent
on
the
ground
surface.
The
surface
soil
is
largely
the
first
point
of
contact
by
rainfall.
Physical
and
chemical
properties
of
surficial
material
thus
play
a
primary
role
in
runoff
generation.
Two
different
landscapes
may
be
distinguished:
terrain
with
thick
unconsolidated
sediments
of
aeolian
or
fluvial
origin,
and
terrain
where
rocky
or
debris
mantled
slopes
dominate.
With
increasing
aridity
bare
rock
or
scree
slopes
are
increasingly
important.
Only
a
minority
of
global
deserts
are
covered
by
aeolian
sands;
for
example
15%
of
the
Sahara
(Mabutt
1977).
In
rocky
deserts
underlying
rocks
are
usually
exposed
or
covered
by
either
a
thin
veneer
of
debris
or
shallow
lithosols.
At
the
base
of
most
slopes
colluvium
accumulates.
Infilt-
ration
rates
of
bare
rock
surfaces
are
low
(about
1
mm
of
threshold
for
runoff
initiation
and
1
to
5
mm
hr
-1
final
rate)
and
vary
little
due
to
differences
in
rock
type
or
jointing
(Schick
1988).
However,
infiltration
characteristics
may
differ
significantly
with
slope
position.
On
rocky
upslope
areas,
for
example,
large
amounts
of
runoff
may
be
generated
immedi-
ately
after
the
onset
of
rain.
Infiltration
rates
of
the
colluvial
base,
on
the
other
hand,
are
an
order
of
magnitude
higher
and
allow
losses
of
large
amounts
of
runoff
originating
from
upslope
areas
(Yair
1992).
Different
aspects
of
runoff
generation
and
fl
ow
discontinuity
were
studied
on
a
limestone
slope
in
the
arid
northern
Negev
desert,
Israel,
and
are
discussed
in
a
later
case
study
(see
Section
4.3.2).
Infiltration
characteristics
for
unconsolidated
sediments
depend
mainly
on
grain
size
distribution
and
the
tendency
to
surface
sealing.
In
many
deserts
final
infiltration
rates
for
bare
coarse
sands
exceed
100
mm
hr
-1
(Agnew
&
Anderson
1992).
Nevertheless,
even
within
an
arid
sand
dune
area
a
biological
topsoil
crust
may
reduce
infiltration
signifi-
cantly
and
lead
to
runoff
(Yair
1990);
the
formation
of
surface
crusts
is
more
accentuated
in
sediments
with
a
high
silt
and
clay
content.
During
high
intensity
storms
rain
drops
disperse
the
soil
matrix
and
form
a
stable
surface
layer
of
reduced
permeability.
Most
runoff
in
silty
arid
and
semi
-arid
terrain
is
caused
by
these
crusts.
The
effect
of
rock
fragments
on
surficial
properties
is
more
complex;
these
have
been
reported
to
both
increase
and
decrease
infiltration
rates
and
amounts
(Brakensiek
&
Rawls
1994).
Accounting
for
all
losses
(slope-
and
channel
losses),
estimates
for
the
threshold
of
run-
off
initiation
in
small
(ca.
2
km
2
)
desert
catchments
depend
on
lithology
and
differ
between
4.5
and
11
mm
(Table
4.1).
Despite
this
variability
comparisons
of
maximum
probable
fl
oods
in
small
desert
catchments
yield
similar
results
in
markedly
different
lithologies.
One
may
therefore
assume
that,
at
least
in
small
arid
catchments,
differences
in
terrain
char-
acteristics
only
leave
their
mark
in
normal
or
large
fl
ood
events.
During
truly
high
magni-
tude
events
they
tend
to
be
blurred
and
rainfall
parameters
are
more
decisive
(Schick
1988).
4.1.2
Wadi
fl
ow
and
transmission
losses
Some
of
the
largest
rivers
in
the
world
(e.g.
Nile
and
Indus)
fl
ow
partly
in
arid
and
semi-
arid
regions.
These
perennial
streams
are
important
water
resources
but
originate
from
118
Understanding
water
in
a
dry
environment
Table
4.1.
Infiltration
characteristics
of
different
terrain
types
in
an
arid
catchment
(after
Lange
et
al.
1999).
Terrain
type
Initial
loss
(mm)
Final
infiltration
rate
(mm
11
-1
)
Limestone
plateau**
4.5
5
Dissected
limestone
plateau**
7
15
Steep
active slope
10
30
Dissected
loessial
colluvium
10
20
Loessial
plateau
7.5
15
Sandy
crusted
plain
9
15
Sandy
vegetated
plain
11
50
Pleistocene
terrace**
6
8
Early
Holocene
terrace**
7
12
Late
Holocene
terrace**
11
40
Flint
plateau
6
10
Dissected
fl
int
plateau
8
20
Marly
sediment
9.5
15
Lisan
Marl,
uncovered
7.5
12
Lisan
Marl,
covered
6
8
Marine
Jurassic
7.5
10
Active
alluvium*
Agricultural
area*
Disturbed
area*
Badlands
on
marl**
9.5
20
Iron
crust
7
20
*
Terrain
where
all
rain
infiltrates.
**
Terrain
type
with
performed
field experiment.
more
humid
areas.
Streams
which
originate
within
arid
and
semi
-arid
lands
remain
dry
for
most
of
the
year.
They
fl
ow
only
occasionally
as
a
result
of
runoff
generating
rainstorms.
Graf
(1988)
distinguished
three
kinds
of
ephemeral
wadi
fl
oods.
Flash
fl
oods
limited
to
small
catchments
(<100
km
2
)
are
produced
by
convective
rainstorm
cells.
Hydrographs
are
characterized
by
a
rapidly
rising
limb,
a
sharp
peak
and
an
equally
sharp
falling
limb.
Single
-peak
fl
oods
in
larger
catchments
are
generated
by
regional
rain
systems
(e.g.
frontal
or
tropical
rain).
Multiple
peak
fl
oods
result
if
multiple
precipitation
events
occur,
or
different
tributaries
of
a
dendritic
channel
system
are
active.
Schick
(1988)
reviewed
the
characteristics
of
fl
ashy
high
magnitude
fl
oods
in
rocky
arid
catchments
located
in
the
southern
Negev
and
in
Sinai.
These
are
geomorphologically
highly
efficient,
modifying
the
landscape
by
erosion
and
depositing
huge
amounts
of
debris.
Flow
rises
from
insignificant
levels
to
high,
short
fl
ood
peaks
within
several
minutes
forming
'walls
of
water'.
Direct
observation
of
one
of
these
fl
oods
indicated
supercritical
fl
ow
with
fl
oat
determined
surface
velocities
of
about
5
m
5
-1
.
In
these
envir-
onments
many
characteristics
of
small
fl
ashy
fl
oods
are
preserved
even
in
large
catch-
ments,
where
peak
fl
ows
of
up
to
1650
m
3
5
-1
are
reached.
In
arid
and
semi
-arid
areas
evaporation
is
pronounced
and
the
time
span
between
single
fl
ood
events
is
usually
long.
Ephemeral
fl
oods
hence
mostly
travel
on
a
dry
bed
allowing
significant
infiltration
losses
into
the
channel
alluvium
on
their
way
downstream.
These
transmission
losses,
which
result
in
a
downstream
decrease
in
runoff,
have
been
observed
for
many
years
(e.g.
Renard
&
Keppel
1966;
Wheater
2002).
During
a
fl
ood
event
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
119
different
processes
(e.g.
air
entrapment,
scour
and
fill)
are
active,
constantly
modifying
the
intensity
and
amount
of
transmission
loss.
Dunkerley
and
Brown
(1999),
for
example,
directly
observed
a
discrete
fl
ow
event
in
a
small
desert
stream
in
western
Australia.
Transmission
loss
totally
consumed
the
fl
ow
over
7.6
km
resulting
in
a
rapid
loss
rate
of
13.2%
per
km
with
marked
spatial
variability.
Different
factors
governing
fl
ow
losses
were
observed,
including
abstractions
of
fl
ow
to
pools,
scour
holes
and
other
low
points
along
the
channel,
overflow
abstractions
into
dead-end
channel
filaments,
and
extensive
mud
drapes
settling
on
sand
bars
and
other
porous
channel
materials.
Silt
carried
by
fl
ood
waters
can
effectively
seal
the
alluvial
surface
even during
fl
ood
events
at
unexpectedly
high
fl
ow
velocities
(Crerar
et
al.
1988).
The
interplay
of
scour
and
fill
is
complex
and
not
fully
understood,
which
prohibits
accurate
quantification
of
loss
volumes
and
makes
reach
-scale
studies
essential
for
further
insight
(Wheater
et
al.
1997).
At
the
reach
scale
Knighton
and
Nanson
(1994)
compared
the
water
balances
of
about
30
individual
events
in
two
reaches
of
Cooper
Creek,
Australia,
of
32
and
more
than
400
km
long.
Plotting
outflow/inflow
ratios
against
event
magnitudes
they
obtained
striking
differences.
In
the
short
32
km
reach,
ratios
increased
with
event
magnitude
approaching
a
near
constant
value
in
medium
to
high
fl
ows,
whereas
the
same
events
caused
a
highly
non-linear
pattern
in
the
long
reach.
There
outflow/inflow
ratios
increased
with
event
magnitude
only
to
an
intermediate
maximum.
With
further
increase
they
fell
rapidly
to
a
fairly
constant
level
before
again
rising
at
only
very
high
discharges.
The
intermediate
maximum
was
explained
by
the
authors
as
bank
full
discharge
in
primary
channels
with
maximum
fl
ood
transmission
efficiency.
Only
during larger
events
was
the
fl
oodplain
considered
to
act
as
an
additional
area
of
losses.
Later
on
the
same
authors
found
other
indirect
evidence
for
a
stage
-dependent
fl
ooding
of
over
-bank
areas.
Using
the
same
data
set,
they
related
catchment
travel
time
to
discharge
and
again
obtained
an
intermediate
maximum,
equated
as
the
onset
of
fl
ood
plain
fl
ow
with
maximum
channel-
fl
oodplain
interactions
(Knighton
&
Nanson
2001).
Applying
a
non
-calibrated
fl
ow
routing
scheme
to
a
150
km
arid
channel
reach
of
the
Kuiseb
river,
in
the
Namib
Desert,
Namibia,
Lange
et
al.
(2002)
showed
similar
characteristics
of
reach
-scale
channel
trans-
mission
losses.
Channel
transmission
losses
were
deliberately
excluded
from
the
scheme.
By
doing
this,
transmission
losses
could
be
identified
at
the
downstream
end
of
the
reach,
when
model
simulation
results
(including
their
uncertainty
range)
plotted
significantly
higher
than
measured
discharges.
This
methodology
showed
directly
during
fl
ow
events
that
channel
transmission
losses
concentrate
during
high
discharge
peaks
and
are
minor
during
small
to
medium
fl
ows.
In
the
Kuiseb
river
this
behaviour
was
also
attributed
to
a
fl
ow
-dependent
scour
of
clogging
silt
layers
and,
more
important,
to
enhanced
water
losses
in
fl
ooded
over
-bank
areas.
4.1.3
Sediment
dynamics
Arid
and
semi
-arid regions
are
characterized
by
intense
physical
weathering
and
sparse
vegetation
cover.
High
rainfall
intensities
are
prominent
during
most
runoff
producing
storm
events.
All
these
factors
promote
soil
dispersion
by
rain
splash,
with
subsequently
generated
Hortonian
overland
fl
ow
removing
the
dispersed
soil
particles.
Further
down
-
slope
rills
and
gullies
are
carved
into
the
land
surface
by
concentrated
fl
ow.
These
erosional
features
are
most
recognizable
in
badlands,
characterized
by
intensely
dissected
landscapes
on
unconsolidated
or
poorly
cemented
material.
Throughout
the
arid
and
120
Understanding
water
in
a
dry
environment
semi
-arid
zone
such
badlands
develop
on
marls,
shales
and
silty
-clay
formations,
enhanced
by
sparse
vegetation
and
marked
seasonal
climatic
variations.
Long-term
annual
denuda-
tion
rates
may
cover
a
wide
range
(0.4-4
mm
yr
-1
)
even
in
morphologically
similar
areas,
producing
a
discrepancy
that
cannot
be
completely
explained
(Bryan
&
Yair
1982).
Eroded
material
makes
up
the
huge
sediment
load
of
ephemeral
wadi
fl
oods
and
is
contributed
from
either
adjacent
slopes
or
derived
from
unstable
beds
and
banks
in
the
wadis.
Small
particles
may
travel
for
long
distances
downstream
as
suspended
load
before
settling
to
the
bed,
with
bed
load
of
larger
particles
moving
by
traction
on
the
river
bed.
Walling
and
Webb
(1996)
have
provided
a
global
overview
of
annual
suspended
sediment
yield
within
the
world's
rivers
(Figure
4.2).
Catchment size,
relief,
loess
cover
and
climate
were
the
predominant
factors
modifying
the
global
pattern.
Since
arid
and
semi
-arid
precipitation
and
runoff
events
are
variable
both
in
time
and
space,
caution
should
be
exercised
in
comparing
annual
sediment
yields.
Nevertheless,
some
interesting
features
resulted.
Despite
decreasing
annual
runoff,
semi
-arid
regions
are
characterised
by
higher
values
of
sediment
yield
than
most
temperate
regions,
pointing
to
enhanced
sediment
dynamics
during
single
runoff
events.
Within
arid
deserts
relatively
low
values
are
deter-
mined,
revealing
the
effect
of
the
rare
occurrence
of
wadi
fl
oods.
Data
on
bed
load
are
globally
scarce.
Sediment
budgeting
in
a
small
hyperarid
catchment
indicated
the
impor-
tance
of
bed
load
in
rocky
desert
environments
(Schick
&
Lekach
1993).
During
a
decade
two-thirds
of
the
sediment
was
made
up
by
bed
load
and
more
than
four
-fifths
by
the
sand
-and
-larger
fractions.
4.1.4
Water
harvesting
adapted
to
aridity:
the
concept
of
micro
-catchments
The
quantity
of
water
available
is
a
principal
constraint
in
arid
zone
agriculture.
Civili-
sations
settling
in
desert
environments
had
two
options:
they
either
conveyed
water
from
C
Secknent
yield
It
Van
yr
'j
1000
750
5CIP
'
250
:
'CO
!
--
1
Desert
arid
permanent
ice
Figure
4.2.
World
wide
distribution
of
sediment
yield
(Walling
&
Webb
1996).
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
121
distant
regions
or
developed
techniques
to
exploit
the
meagre
local
resource.
Ancient
desert
civilisations
usually
did
not
have
the
skills
or
resources
to
transport
water
over
long
dis-
tances.
Hence,
in
many
arid
regions
local
water
harvesting
techniques
date
back
for
more
than
2000
years.
This
ancient
knowledge
is
also
applicable
today,
providing
a
cheap
alternative
for
agriculture
adapted
to
limited
rainfall;
the
concept
of
micro
-catchments
(Shanan
et
al.
1970).
The
general
principle
behind
this
concept
is
to
modify
the
surface
in
order
to
use
runoff
water
more
efficiently,
achieving
increased
agricultural
yields.
Runoff
water
from
small
(less
than
10
ha),
mostly
rocky
'contributing'
catchments
is
collected
and
spread
over
a
soil
covered
'receiving
-cultivated'
area.
This
leads
to
the
introduction
of
a
'catchment
-
cultivated
ratio'
(ratio
between
contributing
and
receiving
catchments)
depending
on
the
local
environmental
setting
(Table
4.2).
The
concept
of
micro
-catchment
systems
has
the
following
principal
advantages:
Simple
construction
at
low
cost;
Use
of
low
-salinity
runoff
water;
Easy
operation
and
cheap
maintenance.
In
the
contributing
catchments
slope
runoff
should
be
collected
before
it
reaches
rills
and
gullies,
and
the
depth
of
fl
ow
should
be
less
than
2
mm
with
velocities
below
7
cm
s
-1
.
Thereby,
the
maximum
length
of
overland
fl
ow
may
be
related
to
slope
gradient
(Table
4.3).
Planning
the
layout
of
micro
-catchments
must
take
into
account
the
quantity
and
frequency
of
runoff,
the
permeability
of
the
soil
and
tolerance
of
the
crop
to
drought
and
standing
water.
Detailed
surveys
of
soils
and
topography
should
therefore
be
accompanied
Table
4.2.
Catchment
-cultivated
ratios
in
different
environments.
Setting
Catchment
-cultivated
Source
ratio
Sorghum
in
semi
-arid
North
Dakota,
USA
2
Haas
et
al.
1966
Orchards
in
the
semi
-arid
Beersheva-Plain,
Israel
1.5-6
Hillel
1967
Different
field
crops
and
pasture
plants
in
the
Around
20
Evenari
et
al.
1982
arid
Negev
Desert,
Israel
Table
4.3.
Maximum
recommended
length
of
overland
fl
ow
on
a
bare
desert
surface
(Shanan
&
Tadmor
1979).
Slope
gradient
(%)
Max.
recommended
length
of
overland
fl
ow
(m)
1
25
3
14
5
11
10
8
20
5
40
4
122
Understanding
water
in
a
dry
environment
by
measurements
of
the
real
runoff
producing
potential
for
several
(3-5)
years
on
runoff
plots.
From
the
results
of
such
a
measuring
campaign,
Shanan
and
Schick
(1980)
quantified
initial
losses
(the
amount
of
storm
rainfall
lost
until
runoff
initiation)
in
the
rocky
Northern
Negev
Desert,
Israel.
In
this
region,
with
100
mm
of
annual
rainfall,
about
5
mm
of
rain
was
lost
during
each
storm
in
a
1-7
ha
catchment
due
to
crust
wetting
(2.5
mm)
and
overland
fl
ow
losses
(2.5-3
mm).
4.2
ASSESSMENT
TECHNIQUES
4.2.1
Direct
measurement
of
fl
ow
and
sediment
Runoff
gauging
stations
are
rare
in
arid
and
semi
-arid
environments,
since
in
most
countries
the
necessary
financial
means
for
installation
and
maintenance
are
restricted.
Recent
studies
in
Africa,
for
example,
have
demonstrated
that
networks
of
hydrological
observing
stations
are
in
decline
(Sehmi
&
Kundzewicz
1997).
Continuously
recording
standard
gauges
only
measure
the
water
stage
by
fl
oats
or
pressure
transducers.
In
ephemeral
streams
fl
ows
with
depths
of
>
0.5
m
cause
particles
of
up
to
gravel
size
to
become
unstable
on
channel
beds
(Graf
1988).
Direct
measurement
of
velocity
by
current
meters
becomes
difficult
or
even
impossible
during
most
high
fl
ows
because
of
this
bed
and
bank
instability.
A
common
alternative
is
to
compute
discharge
indirectly
using
the
slope
—area
method.
This
adapts
a
uniform
-flow
equation
(the
Manning
equation)
using
channel
characteristics,
water
—surface
profiles
and
a
roughness
coefficient.
The
fact
that
cross-sectional
geometry
may
permanently
change
during
a
fl
ood
event
is
therefore
neglected.
As
a
result,
gauged
stream
fl
ow
data
in
most
arid
and
semi
-arid
channels
have
a
high
inherent
uncertainty.
Moreover,
gauging
stations
are
frequently
destroyed
by
the
huge
erosive
power
of
high
magnitude
events
and
data
records
are
incomplete
(e.g.
Greenbaum
et
al.
1998).
Field
measurements
of
suspended
sediment
load
comprise
the
collection
of
water
and
sediment
samples
during
fl
ood
events.
Mean
concentration
of
sediment
is
multiplied
by
simultaneously
measured
fl
ow
volumes
to
obtain
the
total
suspended
load.
Direct
measurement
of
bed
load
is
more
difficult
and
realistic
results
are
highly
dependent
on
particle
size
and
the
sampling
instrument.
The
relative
efficiency
of
different
samplers
ranges
from
less
than
50%
to
more
than
100%
(Graf
1988).
Reservoirs
facilitate
the
most
accurate
determination
of
total
yield
from
a
catchment,
since
both
suspended
sediment
and
bed
load
are
completely
trapped
(e.g.
Schick
&
Lekach
1993).
4.2.2
Indirect
estimation
tools
Paleoflood
technique
Peak
stages
of
historic
and
recent
fl
oods
may
be
reconstructed
by
paleoflood
techniques
using
morphological
and
sedimentary
evidence
(Baker
1987).
Where
fl
ow
boundaries
produce
markedly
reduced
fl
ow
velocities
during
times
of
high
fl
ow,
suspended
sediment
is
deposited
forming
slack
water
sediments.
These
deposits
are
minimum
paleostage
indicators,
since
some
fl
ood
water
must
have
been
above
them
at
the
time
of
emplace-
ment.
Absolute
high
water
indicators,
including
scour
marks
and
lines
of
silt,
driftwood
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
123
or
organic
debris,
are
more
accurate.
Due
to
reduced
biological
activity
both
slack
water
deposits
and
high-water
marks
may
be
preserved
for
long
periods
in
arid
and
semi
-arid
settings.
Bedrock
-confined
reaches
are
preferable,
since
they
accommodate
high
discharges
by
largely
increasing
fl
ow
depth
and
show
a
high
cross-sectional
stability.
Peak
stages
are
transformed
to
paleodischarge
estimates
by
slope
-area
calculations
or
hydraulic
modelling;
e.g.
by
the
HEC-2
package
(Hydraulic
Engineering
Center
1982).
Since
absolute
ages
of
slack
water
deposits
may
be
assessed
using
radiocarbon
dating,
paleoflood
hydrology
may
augment
existing
fl
ood
records
and
result
in
a
more
accurate
fl
ood
frequency
analysis.
Deterministic
rainfall
-runoff
models
Deterministic
rainfall
-runoff
models
may
also
be
used
to
simulate
and
analyse
the
fl
ood
response
of
arid
and
semi
-arid
catchments.
However,
most
existing
approaches
depend
heavily
on
calibration
with
measured
stream
fl
ow
data
(e.g.
Hughes
1997;
Wheater
et
al.
1997;
Al-Weshah
2002a).
As
such,
they
are
limited
to
gauged
catchments
and
include
the
uncertainties
of
hydrometric
data
collection.
El-Hames
and
Richards
(1998)
applied
a
robust,
physically
-based
rainfall
-runoff
model
to
a
170
km
2
arid
catchment
in
Saudi
Arabia.
They
achieved
reasonable
simulations,
but
a
minimum
of
calibration
was
still
required.
Only
more
recently
has
a
fully
non
-calibrated
approach
been
developed
for
high
magnitude
fl
oods
in
arid
rocky
desert
catchments
(Lange
et
al.
1999).
Since
only
field
-
based
parameters
were
included,
applications
become
possible
in
ungauged
catchments
using
other
recent
techniques
(e.g.
rainfall
radar,
remote
sensing),
thus
overcoming
the
need
for
model
calibration.
These
latter
two
models
are
described
in
later
case
studies.
Runoff
regression
models
Regression
and
regionalization
may
facilitate
the
estimation
of
fl
ood
magnitudes.
A
world
wide
comparison
of
fl
ood
frequency
curves
resulted
in
apparent
similarities
in
arid
and
semi
-arid
catchments
and
it
was
proposed
by
Farquharson
et
al.
(1992)
to
treat
them
as
a
homogeneous
region.
First
the
mean
annual
fl
ood
(MAF)
was
calculated:
MAF
=
1
In
Qi
(1)
where
Q,
denotes
the
annual
maximum
fl
ood
series
with
n
values.
Then
the
fl
ood
series
at
each
gauging
station
was
reduced
to
dimensionless
form
by
dividing
by
the
MAF
before
fitting
to
a
General
Extreme
Value
(GEV)
distribution.
For
all
arid
regions
the
50
-year
fl
ood
was
4.51
times
higher
than
the
MAF,
while
for
the
100-
and
500
-year
fl
ood
this
factor
was
6.15
and
12.28
respectively.
As
a
second
step
all
MAFs
(m
3
s
-1
)
in
arid
catchments
were
related
to
catchment
area
A
(km
2
)
by
regression
analysis:
MAF
=
1.87A
°
578
(2)
As
r
2
(the
coefficient
of
determination)
did
not
exceed
0.55,
this
equation
should
only
be
used
as
an
initial
estimate.
To
obtain
entire
design
hydrographs
a
study
by
Walters
(1989b)
can
be
consulted.
In
this,
Walters
related
the
factors
of
design
hydrographs
1/2PKW
(`width'
of
the
hydrograph
at
1/2
peak
stage,
min),
1/4PWK
(`width'
of
the
hydro
-
graph
at
1/4
peak
stage,
min)
and
the
recession
constant
k
to
catchment
area
A
(km
2
)
by
the
following
equations:
1/2PKW
=
2.578A°5°1
(3)
124
Understanding
water
in
a
dry
environment
1/4PKW
=
2.508A
0617
k
=
0.025A
°.441
(4)
(5)
Equations
(3)-(5)
should
again
only
serve
as
initial
estimates,
as
r
2
-values
were
in
the
same
order
of
magnitude
as
for
eqn.
(2).
Graphical
relationships
between
catchment
area
and
maximum
discharge
have
also
been
determined
for
different
regions
in
Israel
(Shentsis
et
al.
1997);
arid
regions
showed
by
far
the
highest
values
and
at
a
certain
catchment
size
the
discharge
-area
relationships
exhibited
a
negative
trend
due
to
the
pronounced
effect
of
channel
transmission
losses.
Assessment
of
channel
transmission
losses
Channel
transmission
losses
can
be
regarded
as
important
contributors
to
groundwater
recharge
in
arid
and
semi
-arid
areas.
Studies
by
Hughes
and
Sami
(1992)
of
infiltration
processes
within
the
channel
alluvium
indicated
large
differences
within
small
distances.
Point
measurements
of
infiltration
rates
have
also
been
made
by
Parissopoulos
and
Wheater
(1992).
Kuells
et
al.
(1995)
combined
these
physical
measurements
with
artificial
tracers
(rhodamine)
making
preferential
fl
ow
paths
apparent.
Mean
infiltration
rates
for
channel
alluvium
may
cover
wide
ranges
starting
from
44
mm
hr
-1
in
loess
-covered
channels
(Shanan
1975)
to
more
than
400
mm
hr
-1
in
silty
-gravel
channels
(Kuells
et
al.
1995).
In
indirect
studies
by
Wallace
and
Renard
(1967),
a
measured
rise
in
the
groundwater
table
near
ephemeral
streams
was
related
to
fl
ow
events.
Using
this
relationship
a
groundwater
model
could
be
constructed
and
calibrated
for
an
alluvial
aquifer
recharged
by
channel
transmission
losses.
Limited
information
on
fl
ow
and
channel
characteristics
was
then
combined
with
groundwater
level
measurements
to
estimate
recharge.
As
a
further
example,
Alderwish
and
Dottridge
(1995)
reliably
estimated
groundwater
recharge
for
two
wadis
in
Yemen
using
the
finite
-difference
groundwater
model
MODFLOW
(McDonald
&
Harbaugh
1988)
and
piezometric
measurements
over
one
dry
and
one
wet
season.
Details
of
this
work
are
given
in
Section
4.3.4.
If
hydrometric
data
are
available
upstream
and
downstream
in
a
channel
reach,
inflow
and
outflow
volumes
may
be
compared
and
transmission
losses
quantified.
Then,
by
means
of
regression
analysis,
volumes
of
transmission
loss
may
be
related
to
fl
ow
and
channel
characteristics.
Including
estimates
of
lateral
inflows
from
tributary
catchments,
Sorman
and
Abdulrazzak
(1997)
used
this
approach
for
a
channel
reach
in
the
Tabalah
catchment,
Saudi
Arabia
(see
Section
4.3.3).
Lane
(1985)
analysed
different
magnitude
fl
ows
in
14
channel
reaches
throughout
Arizona,
Kansas,
Nebraska
and
Texas,
USA,
and
developed
a
simple
estimation
tool
for
transmission
losses
in
semi
-arid
channels.
The
outflow
volume
Q
0
(acre-feet)
of
a
channel
reach
is
given
by
the
expression:
{0
P0
Pi
Q0
a
±
bP,
P
o
<
P,
(6)
where
P,
is
the
volume
of
inflow
(acre-feet),
P
o
the
threshold
volume
(acre-feet)
and
a,
b
are
parameters.
P,
is
obtained
by
hydrological
analysis
with
the
remaining
unknowns
calculated
according
to:
b
=
(7)
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
125
a
=
[A
I
(1
-
B)](1
-
P
o
=
-a/b
(8)
(9)
where
x
(mi)
and
w
(ft)
are
length
and
width
of
the
channel
reach.
A,
B
and
k
are
obtained
from:
A
=
-0.00465KDi
(10)
k
=
-1.09
ln
[1
-
0.0545KD
i
PI
1
]
(11)
B
=
e
-k
(12)
where
D,
is
the
duration
of
inflow
(hr)
and
K
is
the
effective
hydraulic
conductivity
estimated
according
to
Table
4.4.
Regression
models
for
sediment
dynamics
The
Universal
Soil
Loss
Equation,
USLE
(Wischmeier
&
Smith
1978)
is
the
most
widely
and
sometimes
inappropriately
used
estimator
of
slope
erosion
and
slope
sediment
production
because
it
is
well
defined,
has
published
guidelines
and
offers
apparent
com-
patibility
with
numerous
studies
in
many
parts
of
the
world.
For
principally
agricultural
areas
the
soil
loss
A
(kg
M
-2
s
-1
)
is
given
by:
A
=
0.224RKLSCP
(13)
However,
Graf
(1988)
has
stated
that
use
of
the
USLE
in
non-agricultural
dry
land
settings
should
only
serve
as
an
initial
estimate.
This
becomes
apparent
from
the
large
range
for
the
single
parameter
values
he
estimated.
The
rainfall
erositivity
factor
R
ranges
from
20
to
50
and
the
soil
erodibility
factor
K
from
about
0.05
for
sandy
soils
to
about
0.6
for
silty
ones.
The
slope
length
factor
L
and
the
slope
gradient
factor
S
may
be
combined
as
a
topographic
factor
LS.
For
a
slope
100m
long,
LS
varies
from
about
0.18
for
a
1%
gradient
to
about
15
for
a
30%
slope.
The
cropping
management
factor
C
lies
between
45
for
areas
with
no
vegetation
and
about
4
for
areas
which
are
80%
covered
with
shrubs.
Finally,
the
erosion
control
practice
factor
P
lies
in
the
0.1-0.3
range
for
natural
slopes.
Table
4.4.
Relationship
between
bed
material
characteristics
and
hydraulic
conductivity
K
(Lane
1985).
Bed
material
group
Bed
material
characteristics
Effective
hydraulic
conductivity
K
(in
11
-1
)
1
Very
high
loss
rate
2
High
loss
rate
3
Moderately
high
loss
rate
4
Moderate
loss
rate
5
Insignificant
to
low
loss
rate
Very
clean
gravel
and
large
sand
Clean
sand
and
gravel,
field
conditions
Sand
and
gravel
mixture
with
low
silt
-clay
content
Sand
and
gravel
mixture
with
high
silt
-clay
content
Consolidated
bed
material;
high
silt
-clay
content
>5
2-5
1-3
0.25-1
0.001-0.1
126
Understanding
water
in
a
dry
environment
Table
4.5.
Vegetation
community
values
for
F
c
(Miraki
1983).
Vegetation
community
F
c
Protected
forest
0.2
Unclassed
forest
0.4
Arable
areas
0.6
Shrub
and
grass
0.8
Waste
area
1.0
For
large
semi
-arid
and
arid
catchments
in
India,
Miraki
(1983)
used
measurements
of
sedimentation
in
32
reservoirs
to
define
the
following
regression
equation:
V
=
1.182
x
0
6
A
1.026
p
1.289
Q
0.287
s
0.075
Dd
0.3986.422
(14)
where
V
is
the
volume
of
sediment
yield
(m
3
yr
-1
),
A
the
drainage
area
(km
2
),
P
the
annual
rainfall
(cm),
Q
the
mean
annual
runoff
(10
6
m
3
),
S
the
catchment
slope,
Dd
the
drainage
density
(km
-1
)
and
Fc
an
spatially
weighted
erodibility
factor
as
defined
in
Table
4.5.
Tracer
techniques
Tracer
techniques
facilitate
various
studies
of
hydrological
surface
processes.
Hydrograph
separation
by
natural
tracers
may
be
used
to
analyse
runoff
generation
and
in
most
investigations
two
runoff
components
are
separated:
Event
water
and
pre
-event
water
(Sklash
&
Farvolden
1979):
Qev
I
Qges
=
(Cges
Cprev)
I
(Cev
Cprev)
(15)
where
Q
g
,
is
total
runoff
(m
3
s
-1
),
Q,
the
amount
of
the
component
event
water
(m
3
s
-1
),
Q
p
,
the
amount
of
the
component
pre
-event
water
(m
3
s
-1
)
and
C
ges
,
C,
and
Cprev
are
the
respective
concentrations
(mg
L
-1
).
As
prerequisite
for
a
successful
application
of
this
method
the
following
conditions
must
be
met:
The
concentrations
of
the
event
(e.g.
rainfall)
and
pre
-event
(e.g.
groundwater)
com-
ponents
are
measurable
and
differ
significantly.
There
are
no
additional
runoff
components.
The
spatial
and
temporal
variations
in
the
event/pre-event
tracer
concentrations
must
be
determinable.
In
general,
conservative
environmental
tracers
(ions
such
as
Cl
-
,
Si0
3
4
-
and
the
isotopes
18
0
and
2
H)
can
be
used
if
the
prerequisites
are
met.
In
contrast
to
most
humid
catchments,
storm
fl
ow
in
semi
-arid
east
Africa
was
clearly
dominated
by
event
water
(Sandstrom
1996).
The
movement
of
sediment
may
be
quantified
more
accurately
by
the
environmental
tracer
137
Cs
(e.g.
Wise
1980)
than
by
regression
models.
The
30
-year
half-life
of
this
isotope
facilitates
accurate
dating
in
recent
time
spans
(since
1954),
which
is
intermediate
between
short-term
(10
0
-10
1
years)
process
studies
and
long-term
(10
3
-10
8
)
dating
techniques.
In
areas
unaffected
by
fall
-out
deposition
from
the
Chernobyl
accident,
137
Cs
was
derived
from
historic
atmospheric
tests
of
nuclear
weapons
with
marked
peaks
in
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
127
1957-59,
1962-64
and
1971
(Ritchie
et
al.
1974).
If
deposited
on
mineral
soils
it
is
rapidly
and
strongly
adsorbed
by
clay
and
organic
colloids.
This
results
in
an
exponential
decrease
in
concentration
with
increasing
depth
below
soil
surface.
Redistribution
only
takes
place
in
association
with
sediment
particles.
In
dry
land
environments
137
Cs
is
mainly
deposited
in
two
systems:
On
hillslopes
and
within
channel
sediments.
Ritchie
et
al.
(1974)
derived
the
following
regression:
Y
=
1.6X
°
'
68
(16)
where
Y
is
radionuclide
loss
as
a
percentage
of
input
and
X
is
the
soil
erosion
(tonnes
ha
-1
yr
-1
).
Comparisons
of
the
137
Cs
content
for
adjacent
undisturbed
and
eroded
sites
may
provide
quantitative
estimates
of
soil
erosion
rates
(Walling
&
Quine
1991).
If
mea-
surements
in
fl
uvial
sediment
are
combined
with
those
within
catchments,
studies
of
sediment
source
areas
and
sediment
routing
in
channel
systems
are possible.
Artificial
tracers
(such
as
the
fluorescence
dyes
uranine
or
rhodamine)
may
provide
direct
insight
to
processes
during
real
fl
ood
events.
Lange
et
al.
(1998)
chose
this
technique
to
study
infiltration
losses
in
a
dry
wadi
bed.
In
order
to
not
be
fully
dependent
on
rarely
occurring
natural
fl
ood
events,
an
artificial
fl
ash
fl
ood
was
studied
in
a
small
arid
stream
channel,
Nahal
Shahmon,
Israel.
A
reservoir
was
built,
filled
with
water
(550
m
3
)
and
breached
artificially.
Three
different
artificial
tracers
were
injected
at
different
times
of
the
fl
ood
and
surface
water
samples
were
taken
at
two
locations
downstream.
Water
moving
within
the
alluvium
was
also
sampled
with
the
help
of
fl
uocapteurs
(small
steel
-mesh
bags
containing
adsorptive
charcoal)
distributed
at
different
locations
within
the
alluvial
body.
The
fl
uocapteurs
enabled
identification
of
the
different
pathways
of
the
two
dye
tracers
downstream,
and
the
maximum
depth
of
infiltration
could
be
approximated.
This
sampling
technique
is
promising
for
future
studies
in
arid
rivers,
since
fl
uocapteurs
can
remain
in
place
for
a
long
time
without
maintenance.
Towards
the
end
of
fl
ow,
exfiltrating
water
from
the
channel
alluvium
was
observed
with
a
high
tracer
concentration.
4.3
CASE
STUDIES
4.3.1
Introduction
Dry
land
research
has
been
performed
in
many
parts
of
the
globe.
However,
since
real
progress
depends
on
the
existence
of
long,
detailed
and
reliable
hydrological
data
sets,
this
concentrates
major
research
findings
to
a
rather
small
number
of
experimental
sites
and
catchments.
Prominent
examples
are
outlined
in
this
chapter;
others
are
detailed
by
(e.g.)
Al-Weshah
(2002b)
and
Alhamid
and
Reid
(2002).
Hydrological
processes
investigated
and
methods
applied
are
predefined
by
the
study
scale.
Most
hillslope-scale
studies
address
processes
of
runoff
generation
and
sediment
mobilization,
while
in
catchment
-scale
studies
transmission
losses,
water
balances
and,
finally,
hydrological
modelling
are
more
relevant.
Apart
from
scale,
research
findings
should
also
be
classified
according
to
the
prevailing
climatic
regime.
Compared
to
the
almost
barren
extreme
deserts,
semi
-arid
landscapes
are
at
least
periodically
vegetated.
Hillslope-scale
processes,
in
particular,
are
then
entirely
different
from
those
in
deserts,
and
require
different
approaches
to
understand
and
quantify
surface
runoff
and
sediment
dynamics.
128
Understanding
water
in
a
dry
environment
To
a
certain
degree
all
research
findings
are
site
specific.
However,
if
the
scale
and
climatic
regime
are
similar,
principal
results
may
be
translated
to
other
locations.
They
are
then
especially
valuable
in
the
search
for
adequate
research
tools
or
for
first
approxi-
mations.
The
following
case
studies
are
thus
grouped
according
to
scale
and
in
each
the
climatic
regime
is
explicitly
stated.
The
final
two
examples
provide
regional
overviews
and
include
a
variety
of
climatic
conditions.
4.3.2
The
hillslope
scale
—runoff
generation
and
sediment
dynamics
Classification
of
runoff
generation
in
semi
-arid
west
Africa
Casenave
and
Valentin
(1992)
summarized
the
results
of
sprinkling
experiments
carried
out
on
87
plots
in
arid
and
semi
-arid
areas
in
five
states
of
West
Africa.
Analysing
the
factors
governing
infiltration
and
runoff
generation,
they
arrived
at
a
hierarchical
classi-
fication system
for
runoff
generation
based
on
eleven
unit
surfaces;
homogeneous
bodies
of
land
surface
defined
by
sprinkling
experiments
(Table
4.6).
For
each
unit
surface
the
depth
of
generated
surface
runoff
L
f
(mm)
was
determined
by
the
regression
equation:
L
f
=
AP
u
BIK
CP
u
IK
D
(17)
where
P
u
is
depth
of
rainfall
(mm),
A
—D
are
regression
coefficients
typical
for
each
unit
surface
(Table
4.6)
and
IK
is
a
dimensionless
index
quantifying
initial
moisture:
IK„
=
(IK„_i
P„
0
5t
(18)
Table
4.6.
Unit
surfaces,
identification
criteria
and
regression
coefficients
(Casenave
&
Valentin
1992).
Unit
Identification
criteria
surface
A
B
C
D
Cl
Cultivated,
vesicular
porosity
in
the
crust
<5%
C2
Cultivated,
vesicular
porosity
in
the
crust
5-30%
C3
Cultivated,
vesicular
porosity
in
the
crust
>30%
TW
Non
-cultivated,
>20%
earthworm
casts
and
>30%
termite
harvesting
constructions
W
Non
-cultivated,
>20%
earthworm
casts
and
<30%
termite
harvesting
constructions
DRY
Non
-cultivated,
<20%
earthworm
casts
and
<40%
of
coarse
fragments
without
crust
or
with
a
drying
crust
ST2
Non
-cultivated,
<20%
earthworm
casts
and
<40%
of
coarse
fragments
covered
with
a
structural
crust
of
two
micro
-layers
ST3
Non
-cultivated,
<20%
earthworm
casts
and
<40%
of
coarse
fragments
covered
with
a
structural
crust
of
three
micro
-layers
SED
Non
-cultivated,
<20%
earthworm
casts
and
<40%
of
coarse
fragments
covered
with
a
sedimentation
crust
ERO
Non
-cultivated,
<20%
earthworm
casts
and
<40%
of
coarse
fragments
sealed
with
a
erosion
crust
G
Non
-cultivated,
<20%
earthworm
casts
and
>40%
of
coarse
fragments
with
the
major
part
embedded
into
a
structural
crust
0.2
0.03
0.004
3
0.35
0.04
0.004
3
0.9
0.05
0.002
10
0.05
0.01
0.001
1
0.1
0.05
0.002
3
0.3
0.01
0.003
8
0.5
0.02
0.004
10
0.85
0.01
0.003
8
0.8
0.08
0.001
12
0.95
0.09
0.001
9
0.99
0.05
0.001
6
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
129
where
IK„
is
the
value
before
rain
n,
IK,„
1
the
value
before
rain
n
-
1,
P
n
_i
the
depth
of
rain
n
-
1,
and
t
the
time
(d)
separating
rain
n
and
n
-
1.
IK
approaches
0
for
dry
soils,
is
20
in
wet
soils
and
exceeds
80
at
soil
saturation.
The
spatial
distribution
of
these
unit
surfaces
was
mapped
in
15
small
watersheds
and
runoff
L
f
was
calculated
using
spatially
weighted
coefficients
in
eqn.
(17).
Due
to
the
different
scales
the
observed
runoff
L
o
(mm)
deviated
from
that
calculated
and
a
second
step
was
required:
L
o
=
1.0694
-
1.51
(19)
The
findings
of
this
study
clearly
illustrate
the
dominant
role
of
surface
conditions
(type
of
crust
and
biological
activity)
for
runoff
generation
in
semi
-arid
West
africa.
However,
the
need
for
a
second
regression
before
application
to
small
catchments
precludes
accurate
runoff
prediction
from
ungauged
areas.
The
semi
-arid
loess
region,
north
China
The
loess
region
in
north
China
is
characterized
by
a
semi
-arid
continental
climate.
Mean
annual
precipitation
is
about
500
mm
with
most
of
the
rain
falling
during
heavy
storms
in
the
monsoon
season
(June
to
September).
Hilly
areas
are
covered
by
Pleistocene
loess
with
a
thickness
reaching
200
m.
Two
main
characteristics
have
attracted
researchers
during
the
last
decades:
enhanced
runoff
production
due
to
pronounced
soil
surface
crusting,
and
the
high
sediment
yields
from
the
area,
these
being
amongst
the
highest
in
the
world.
Applying
two
alternative
techniques,
Zhang
et
al.
(1998)
attempted
to
verify
existing
estimates
of
erosion
rates
on
cultivated
loess
areas.
These
very
high
rates
(in
the
range
10,000-16,000
t
km
-2
yr
-1
)
were
based
on
empirical
relationships
(relating
erosion
rates
to
slope
angle
and
length)
derived
from
erosion
plot
studies.
Erosion
was
independently
quantified
using
rill
volume
data
and
137
Cs
as
a
tracer
(Figure
4.3).
The
results
showed
good
agreement
for
the
downslope
variation
with
visible
signs
of
rilling
occurring
30
m
Erosion
rate
(kg
m
2
yr-1)
10
8.
6-
4
2-
Rills
1
"Cs
Empirical
0
•,•
••.*'
25
20
15
cy)
10
0
8
a)
5
E
0
0
10
20
30
40
50
60
70
80
90
Distance
from
hilltop
[m]
Figure
4.3.
Erosion
rates
along
a
slope
determined
by
different
methods.
Upper
parts
of
the
slope
are
gently
sloping,
while
lower
parts
are
steep
(Zhang
et
al.
1998).
130
Understanding
water
in
a
dry
environment
downslope
from
the
hilltop
reaching
a
maximum
about
50
m
downslope.
Further
down
rilling
intensity
was
limited,
as
the
larger
rills
incised
harder
loess
underlying
the
plough
layer.
A
similar
downslope
pattern
was
determined
using
137
Cs.
The
absolute
quantities
were
different,
however,
because
the
137
Cs-values
represented
long-term
means
while
the
rills
developed
during
only
one
major
event.
Nevertheless,
both
techniques
provided
a
more
accurate
spatial
pattern
of
erosion
rates.
They
indicated
that
existing
empirical
relationships
overestimate
erosion
rates
on
steep
slopes
(lower
parts in
Figure
4.3),
while
on
gently
sloping
fields
the
opposite
is
true
(upper
parts
of
Figure
4.3).
Li
et
al.
(1995)
conducted
simulated
rainfall
experiments
using
sprinkling
devices
on
runoff
plots
in
different
topographic
zones.
Moderate
slopes
(5°
gradient)
on
a
hilltop
were
planted
with
typical
field
crops,
and
were
the
areas
where
the
highest
infiltration
rates
and
longest
ponding
times
were
recorded.
On
steeper
cultivated
hill
slopes
(25°
gradient)
moderate
values
occurred.
The
lowest
infiltration
and
the
shortest
ponding
time
were
registered
for
a
steep
gully
wall
(35°
gradient)
with
a
natural
shrub
vegetation.
Due
to
the
limited
sample
size
the
results
could
only
be
treated
as
general
tendencies
and
were
related
to
a
range
of
enviromental
conditions
including
gradient,
soil
type
and
land
use.
Using
the
same
methodology,
Luk
et
al.
(1993)
studied
more
thoroughly
the
effects
of
antecedent
crusting
and
sealing
on
soil
and
water
losses.
They found
that
on
crusted
surfaces
the
initial
runoff
rate
was
enhanced,
while
the
initial
sediment
concentration
was
reduced.
The
net
effect
on
soil
loss,
however,
was
quite
limited.
The
surprisingly
small
differences
in
soil
loss
were
attributed
to
two
main
reasons.
First,
higher
runoff
rates
and
transport
capabilities
were
balanced
by
an
enhanced
shear
strength
of
the
topsoil.
Secondly,
it
was
assumed
that
new
surface
seals
developed
during
the
experiments.
The
studied
crust
effect
was
therefore
limited
to
storms
of
short
duration,
diminishing
in
longer
storms
due
to
the
formation
and
possibly
more
uniform
distribution
of
surface
seals.
Runoff
plots
in
semi
-arid
northern
Iraq
Hussein
(1996)
studied
runoff
generation
and
soil
loss
in
six
adjacent
runoff
plots
in
the
semi
-arid
zone
of
northern
Iraq.
Mean
annual
rainfall
based
on
23
years
of
measurement
was
333
mm,
and
soils
were
classified
as
Xerollic
Calciorthids
with
a
balanced
grain
size
distribution.
The
soil
was
tilled
at
the
beginning
of
the
rainy
season,
and
then
left
bare
to
account
for
the
effect
of
surface
sealing
and
crusting.
Runoff
volumes
and
sediment
concentrations
were
measured
in
collecting
tanks
at
the
plot
outlets.
For
single
events
a
complex
model
applying
principles
of
stream
power
and
conservation
of
mass
and
energy
could
be
reasonably
fitted
to
measured
values
of
runoff
and
soil
loss.
The
validity
of
two
widely
used
simple
approaches
to
estimate
runoff
and
soil
loss
was
also
tested.
The
Curve
Number
Method
(SCS
1985)
adequately
accounted
for
the
storm
volume
but not
for
the
intensity
distribution
of
rainfall.
It
therefore
over
-
predicted
runoff
during
storms
of
low
intensity,
but
under
-predicted
runoff
during
high
intensity
storms
(Figure
4.4).
Alternatively,
the
Universal
Soil
Loss
Equation,
USLE
(Wischmeier
&
Smith
1978),
was
designed
to
predict
long-term
annual
soil
loss
caused
by
sheet
and
rill
erosion
in
the
USA.
Unlike
predicted
soil
losses,
measured
values
obtained
for
single
storms
in
the
Iraq
study
also
included
non
-erosive
events
of
very
low
intensities.
As
a
result,
predicted
and
measured
values
of
seasonal
soil
loss
differed
greatly.
It
was
thus
proposed
that
simple
estimation
methods
of
runoff
and
soil
loss
should
be
modified
to
account
for
typical
rainfall
and
soil
loss
characteristics
in
semi
-arid
regions.
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
131
Predicted
Runoff
(mm)
30
25
20
15
10
0
0
0
o
o
0
0
0
0
0
5
10
15
20
25
30
Observed
Runoff
(mm)
Figure
4.4.
Comparison
of
predicted
(SCS
method)
and
observed
runoff
on
small
plots
in
northern
Iraq
(Hussein
1996).
The
semi
-arid
Dinosaur
badlands,
Canada
The
Dinosaur
badlands
in
Alberta,
Canada
are
located
in
a
semi
-arid
area
with
a
mean
annual
precipitation
of
330
mm.
During
summer
they
form
the
major
source
of
sediment
to
the
Red
Deer
river,
when
sediment
delivery
to
the
river
invariably
coincides
with
fl
ow
in
the
ephemeral
tributaries
following
sporadic
rain
storms.
Erosion
rates
are
high
with
an
average
of
3
mm
yr
-1
and
peak
values
of
13
mm
yr
-1
(Campbell
1981).
The
badlands
are
formed
in
Upper
Cretaceous
strata
comprising
intercalated
lagoonal
shales
and
sandstones
of
variable
character.
Tunnel
erosion
is
a
widespread
phenomenon,
with
tunnel
diameter
varying
from
a
few
millimetres
to
several
metres.
Hodges
and
Bryan
(1982)
carried
out
simulated
rain
storm
experiments
on
ten
experi-
mental
microcatchments
covering
33
lithological
units
and
examples
of
most
small-scale
topographic
forms.
Average
rainfall
intensity
was
29
mm
hr
-1
with
a
usual
duration
of
30
min,
conditions
which
recurred
in
the
area
every
two
to
five
years.
Results
indicated
that
runoff
generation
in
this
environment
was
complex
and
depended
ultimately
on
lithologic
variations.
Sandstones
and
silts
generated
Hortonian
overland
fl
ow
almost
immediately
after
rain started,
regardless
of
antecedent
moisture.
Runoff
from
shale
units
was
more
variable
and
could
be
delayed
by
9-23
min
on
a
dry
and
3-15
min
on
a
wet
surface,
indicating
the
importance
of
antecedent
moisture.
Thus,
during
most
natural
rain
storms
runoff
almost
entirely
came
from
sandstones
and
silts.
Only
during
extreme
events,
which
occurred
once
or
twice
a
year
and
appeared
to
be
of
greater
geomorphic
signi-
ficance,
did
the
shale
units
also
respond.
De
Boer
(1992)
studied
the
influence
of
scale
on
runoff
generation. Measurements
were
carried
out
in
nested
catchments
representing
three
scale
levels:
the
microscale
(<1
m
2
of
experimental
plots),
the
sub
-basin
scale
(sub
-basins
of
1882
and
2104
m
2
,
respectively)
132
Understanding
water
in
a
dry
environment
and
the
mesoscale
(catchments
of
7.9
and
20.2
ha,
respectively).
In
accordance
with
the
sprinkling
experiments
listed
above,
lithology
dominated
runoff
response
at
the
micro
-
scale
causing
completely
different
runoff
coefficients.
In
the
sub
-basins,
runoff
generated
on
sandstone
units
started
quickly
after
the
onset
of
rain;
the
threshold
of
fl
ow
was
not
sensitive
to
antecedent
moisture
as
represented
by
the
antecedent
precipitation
index
API
(Linsley
et
al.
1982).
At
the
mesoscale,
however,
the
threshold
of
fl
ow
clearly
depended
on
API
(Figure
4.5).
This
latter
relationship
was
caused
by
alluvial
deposits
that
were
only
present
in
mesoscale
channels,
causing
water
losses
by
infiltration.
Under
dry
conditions
trans-
mission
losses
therefore
increased
the
threshold
of
fl
ow
at
the
mesoscale;
this
value
could
not
be
translated
from
studies
at
a
smaller
scale.
Studies
on
sediment
dynamics
yielded
similar
results
(De
Boer
&
Campbell
1989).
Deep
tunnel
systems
existed
only
in
meso-
scale
catchments
and
dominated
their
behaviour
by
an
increase
in
sediment
concentration.
The
time
of
fl
ow
initiation
in
the
tunnels
could
be
determined
even
more
accurately
by
a
change
in
the
relationship
between
sulphate
and
electrical
conductivity
(De
Boer
&
Campbell
1990).
A
limestone
slope
in
the
arid
Negev
desert,
Israel
Long-term
monitoring
of
runoff
generation
processes
has
been
carried
out
at
the
Sede
Boger
experimental
site
(Yair
et
al.
1978;
Yair
1992).
The
experimental
watershed
drains
2.15
ha
of
a
typical
limestone
desert
slope.
The
upper
headwater
area
is
dominated
by
extensive
rocky
outcrops,
almost
devoid
of
any
soil
cover
and
the
base
of
the
slopes
is
covered
by
a
colluvial
mantle
which
thickens
rapidly
down
slope.
Alluvial
terraces
and
reaches
are
located
near
the
channels.
The
experimental
network
was
designed
for
auto-
matic
and
simultaneous
measurement
of
rainfall,
and
both
hillslope
and
channel
runoff.
Channel
runoff
is
recorded
at
two
stations.
While
the
upper
channel
drains
only
the
rocky
headwater
area,
the
lower
one
is
located
near
the
outlet
of
the
experimental
catchment
and
15
o
No
runoff
Runoff
10
E
o
o
ok
1
ID
0
'400
0
ft
06
0
'
.•
.
0
5
10
Total
Rainfall
(mm)
15
Figure
4.5.
Flow
generation
in
mesoscale
catchments
of
the
Dinosaur
badlands:
dashed
line
indicates
threshold
of
fl
ow
(De
Boer
1992).
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
133
collects
runoff
contributions
from
all
different
units.
To
facilitate
more
detailed
analysis
of
runoff
generation,
runoff
plots
equipped
with
sprinkling
devices
were
nested
within
the
catchment.
During
infiltration
tests
on
1.5
m
2
plots,
the
infiltration
rates
for
rocky
slopes
approached
0
mm
hr
-1
after
only
8
min.
The
same
tests
on
lower
colluvial
mantle
resulted
in
infiltr-
ation
rates
of
15
mmhr
1
reached
after
30
min,
showing
that
colluvial
slopes
were
char-
acterized
by
a
considerably
higher
infiltration
rate
and
porosity.
This
was
confirmed
by
different
runs
of
sprinkling
experiments
when
fl
ow
lines
were
mapped
inside
a
runoff
plot
(Figure
4.6).
The
responses
of
rocky
and
colluvial
units
were
markedly
different.
Runoff
was
generated
in
the
rocky
area
even
at
low
rain
intensities.
The
threshold
rain
intensity
was
as
low
as
5
mmhr
1
for
a
short
duration
of
3
min
only;
on
the
colluvial
area
all
low
intensity
rainfall
was
lost
by
infiltration.
Moreover,
most
of
the
fl
ow
lines
were
discon-
tinuous,
disappearing
at
the
interface
between
the
rocky
upslope
area
and
colluvial
base.
Hydrological,
pedological
and
botanical
studies
(Yair
&
Lavee
1985)
indicated
that
this
fl
ow
discontinuity
increased
the
amount
of
available
water
by
a
factor
of
two
or
three
in
the
transition
zone
between
rocky
and
colluvial
areas.
During
an
extended
low
intensity
rain
storm
in
December
1988,
the
combined
effect
of
fl
ow
discontinuity
and
transmission
losses
into
the
channel
alluvium
became
apparent
(Figure
4.7).
Total
rain
amounting
to
35.1
mm,
with
intermittent
rainfall
during
three
consecutive
days,
resulted
in
several
separate
fl
ows.
The
upper
channel
draining
the
rocky
headwater
area
recorded
runoff
following
each
spell
of
rain,
whereas
the
lower
channel
at
the
outlet
of
the
catchment
was
active
only
after
the
third
rain
spell.
First
run
15
mm
hr
-1
30
min
510
m
505
m
500
m
Second
run
30
mm
hr
-1
30
min
Third
run
60
mm
hr
-1
30
min
0
10m
Figure
4.6.
Flowlines
on
a
limestone
slope;
upper
part
rocky,
lower
part
colluvial
(Yair
1992).
134
Understanding
water
in
a
dry
environment
500
-
400
-
_J
a
_c
0
'L
)
Li
300-
200
-
100
-
Rainfall
:
12.2
mm
Runoff
volume
27.7
L
Specific
yield:
3.4
L
M
-2
Runoff
coefficient:
28%
0
1111
12:00
Upper
Channel
I
I
18:00
00:00
06:00
12:00
26.12.88
18:00
00:00
500
-
400-
E
300-
a)
a
0
200
-
100
-
Rainfall
:
3.7
mm
Runoff
volume
3.1
L
Specific
yield:
0.14
L
m
2
Runoff
coefficient:
3.8%
0
12:00 18:00
Lower
Channel
I
I I
I
00:00
06:00
12:00 18:00
00:00
26.12.88
Figure
4.7.
Rainstorm
of
December
1988
(Yair
1992).
4.3.3
The
catchment
scale
—transmission
losses,
water
balance
and
modelling
The
semi
-arid
Walnut
Gulch,
Arizona,
USA
The
150
km
2
Walnut
Gulch
catchment
was
established
as
an
experimental
basin
in
1953.
Located
in
southeastern
Arizona,
the
climate
is
semi
-arid
and
most
of
the
rainfall
derives
from
multi
-cellular
convective
thunderstorms
of
high
intensity,
short
duration
and
limited
areal
extent.
Typical
fl
ash
fl
oods
result,
which
decrease
significantly
downstream
due
to
transmission
losses
(Figure
4.8).
Most
of
the
catchment
consists
of
gently
rolling
rangeland
underlain
by
Quaternary
sediments.
Vegetation
is
dominated
by
native
shrubs
on
the
slopes
and
grasses
in
the
upland
areas.
Runoff
is
measured
at
the
catchment
outlet,
and
from
25
sub
-catchments,
by
a
combination
of
permanent
structures
and
water
level
recorders.
Additionally,
95
rainfall
recorders
are
scattered
throughout
the
catchment
(Osborn
&
Renard
1970).
This
unique
gauging
network has
facilitated
numerous
basic
studies,
thus
improving
the
understanding
of
semi
-arid
hydrology.
Michaud
and
Sorooshian
(1994a)
used
data
on
the
spatial
rainfall
distribution
to
study
the
error
in
runoff
simulations
that
might
be
caused
by
interpolation
of
coarse
rain
gauge
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
135
Flume
1
Flume
2
Flume
6
0
4
km
120
Flume
6:
248
107
m'
100
4
-
80
C)
O
60
O
a
40
20
2ff
400
E
E
200
"
a
2
5
°
Flume
2:
197
m
3
,
73
m
e
2
15
Flume
1
:
156
m
3
,
56
m
3
S'
10
20
30
40
min
4
5h
Figure
4.8.
A
typical
storm
event
in
Walnut
Gulch
(Goodrich
et
al.
1997).
networks.
They
recommended
a
minimum
density
of
one
gauge
per
4
km
2
for
semi
-arid
catchments
with
localized
convective
rain
cells.
Different
kinds
of
rainfall
-runoff
models
were
also
tested
(Michaud
&
Sorooshian
1994b).
A
comparison
for
the
complete
catch-
ment
showed
that
a
complex
distributed
model,
especially
developed
for
this
environ-
ment,
could
be
reasonably
applied
without
calibration.
After
calibration
the
model
precision
improved
only
marginally,
but
the
complex
model
was
then
no
more
superior
than
a
simpler,
distributed
alternative.
Calibrated
rainfall
-runoff
models
of
different
complexity
have
also
been
applied
to
several
tributaries
within
Walnut
Gulch
(e.g.
Grayson
et
al.
1992;
Karnieli
et
al.
1994;
Wheater
et
al.
1997).
Gauged
hydrographs
at
the
limits
of
different
channel
reaches
facilitated
systematic
reach
-scale
studies
on
channel
transmission
losses.
Lane
et
al.
(1971)
used
multi
-linear
regression
of
outflow
on
inflow
volumes
with
additional
channel
parameters.
Differential
equations
relating
inflow
volumes
to
transmission
losses
were
subsequently
developed
and
became
a
standard
method
for
semi
-arid
channels
in
the
USA
(see
Section
4.2.2).
In
a
more
recent
study,
the
linearity
of
basin
response
as
a
function
of
scale
was
also
investigated
(Goodrich
et
al.
1997).
Results
indicated
that
runoff
volumes
and
peak
discharges
increased
proportionally with
catchment
area
only
up
to
the
critical
transition
136
Understanding
water
in
a
dry
environment
threshold
area
of
about
37-60
ha.
For
larger
catchments
the
effects
of
channel
trans-
mission
losses
and
fractional
storm
area
coverage
changed
this
relationship.
The
semi
-arid
Luni
basin,
India
The
34,866
km
2
Luni
catchment
is
located
in
the
northwestern
dry
part
of
India
(Figure
4.9).
Average
annual
values
of
rainfall
and
potential
evapotranspiration
vary
from
300
to
600
mm
and
170
to
2060
mm,
respectively.
Rainfall,
which
mostly
occurs
during
the
rainy
season
from
June
to
September,
is
characterized
by
a
high
spatial
and
temporal
variability
typical
for
desert
regions.
The
eastern
part
of
the
catchment
is
characterized
by
hilly
and
rocky
piedmont
comprising
igneous
and
metamorphic
rocks
of
Precambrian
and
Palaeozoic
age,
whereas
the
western
part
is
covered
by
Pleistocene
alluvium
and
Holocene
sands.
Sharma
(1997)
reported
environmental
problems
due
to
rising
water
demand
and
uncoordinated
water
resources
development
efforts.
Flash
fl
oods
and
severe
droughts
have
afflicted
the
catchment
for
a
long
time.
Furthermore,
many
medium
to
minor
reservoirs
were
sited
for
short-term
benefits
without
due
consideration
being
given
to
climatic,
geomorphologic,
lithological
and
hydrological
conditions.
As
a
result,
suspended
sediment
loads
from
fl
oods
reduce
the
storage
capacity
of
reservoirs
by
1.9-7.8%
annually,
stored
water
often
becomes
saline
and
downstream
reaches
suffer
from
water
shortage.
Since
1979
the
Luni
and
its
tributaries
have
been
gauged
at
more
than
30
stations,
facilitating
a
number
of
different
hydrological
studies.
Sharma
and
Murthy
(1996)
used
79
observed
hydrographs
to
relate
time
-to
-peak
and
peak
discharge
to
catchment
area
and
effective
rainfall.
The
recession
limb
was simulated
using
a
single
leaky
reservoir.
This
simple
model
could
be
reasonably
fitted
to
gauged
events
in
the
headwaters.
A
simple,
lumped
routing
procedure
was
used
to
facilitate
stream
fl
ow
routing,
account-
ing
for
transmission
losses
within
the
ephemeral
channels (Sharma
et
al.
1994).
With
the
y
Stream
gauging
-station
---
Rainfall
isohyet
(mm)
0
50
km
300
-
300
/-*
500
V
"""s.
400
I
600
500
600
Air
I
India
Figure
4.9.
The
Luni
basin,
India
(Sharma
&
Murthy
1996).
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
137
help
of
a
differential
equation
transmission
losses
were
described
as
a
function
of
inflow,
distance,
channel
width,
time
parameters
of
fl
ow
and
effective
hydraulic
conductivity.
Model
parameters
for
single
channel
reaches
were
determined
using
linear
regression
on
observed
inflow
-outflow
data.
Outflow
hydrographs
could
then
be
reasonably
approxi-
mated
by
a
triangular
shape.
Water
quality
and
sediment
concentration
were
also
determined
in
addition
to
measurements
of
water
quantity
(Choudhari
&
Sharma
1984).
An
initial
water
sample
was
taken
at
the
onset
of
fl
ow,
with
subsequent
samples
collected
at
irregular
intervals
and
with
significant
changes
in
the
discharge
until
the
fl
ow
ceased.
Highest
values
of
total
dissolved
salts
(TDS)
were
recorded
at
the
onset
of
the
first
runoff
event
of
the
year
followed
by
a
rapid
decrease
during
the
rising
and
recession
fl
ow
stages.
Due
to
different
solubilities,
single
ions
showed
different
behaviours.
While
concentrations
of
Na+
and
Cl
-
were
diluted
at
higher
discharge
rates,
Ca
2
+
showed
the
opposite
trend
and
and
Mg
2
+
remained
relatively
uniform.
A
comparison
of
the
chemistry
for
the
different
tribu-
taries
showed
that
the
mineralogical
signature
and
weatherability
adequately
explained
the
stream
water
ionic
composition.
In
another
study
sediment
yields
of
the
Luni
basin
were
compared
with
other
catch-
ments
in
the
Indian
arid
zone
(Sharma
1996).
Sandstone
outcrops
generated
highest
sediment
yields,
followed
by
limestone,
quartzite
and
shale.
Moreover,
sediment
loads
increased
with
rainfall
and
drainage
basin
slope.
Promotion
of
vegetation
on
the
slopes
and
the
construction
of
check
dams
reduced
sediment
yields
by
65-94%.
The
arid
Nahal
Zin,
Israel
Nahal
(Wadi)
Zin
is
one
of
the
major
arteries
draining
the
northern
Negev
Desert
in
Israel
to
the
Dead
Sea.
The
upper
part
of
this
1400
km
2
arid
catchment
drains
the
plateaus
of
the
northern
Negev
Highlands.
Shallow
rocky
soils
dominate
the
terrain
and
the
valleys
are
filled
with
loessial
silty
sediments.
In
the
lower
catchment
loose
sediments
and
reaches
covered
by
thick
channel
alluvium
can
also
be
found.
Within
the
catchment
average
annual
rainfall
is
60-90
mm;
localized
storms
occurring
as
convective
cells
are
mainly
responsible
for
the
high
magnitude
fl
oods
(Schick
1988).
Greenbaum
et
al.
(1998)
analysed
one
high
magnitude
rain
storm
fl
ood
which
occurred
in
October
1991.
The
rainfall
pattern
was
determined
using
temporal
intensity
patterns
at
a
point,
intensity
factor
maps
(Kelway
&
Herbert
1969),
and
a
spatial
evaluation
of
the
time
at
which
maximum
rainfall
intensity
occurred
at
a
given
point.
The
results
were
subsequently
verified
by
rainfall
radar
data.
This
thorough
analysis
yielded
a
completely
different
picture
of
rainfall
intensities
than
might
have
been
portrayed
if
based
on
available
rain
gauge
data
only
(Figure
4.10).
The
fl
ood
hydrograph
was
reconstructed
using
techniques
of
paleoflood
hydrology.
These
were
based
on
sedimentological
evidence
of
fine-grained
fl
ood
deposits
in
back
fl
ooded
tributaries
as
well
as
on
other
stage
indicators.
The
HEC-2
procedure
(Hydro-
logical
Engineering
Center
1982)
was
used
to
determine
water
surface
profiles
and
paleodischarge
estimates.
A
triple
peak
hydrograph
was
reconstructed,
which
reached
a
maximum
discharge
of
550
m
3
s
-i
with
an
estimated
recurrence
interval
of
40-50
years.
Lange
et
al.
(1999)
used
this
detailed
information
on
the
high
magnitude
fl
ood
to
develop
a
non
-calibrated
rainfall
-runoff
model.
Model
parameters
were
determined
using
only
previous
experimental
results,
measured
field
data
and
information
from
maps
and
aerial
photographs.
Appropriate
tools
to
derive
and
to
process
the
spatially
distributed
138
Understanding
water
in
a
dry
environment
/
0
Q
20
10
40
30
30
.
44
Dead
ea
1
/.
10
!
10
Isohyet,
mm
=I
20-30
mm
I=
35
mm
35-50
mm
/
Figure
4.10.
Total
storm
rainfall
in
Nahal
Zin.
Calculated
pattern
based
only
on
available
station
data
(left)
and
reconstructed
pattern
(right)
(Greenbaum
et
al.
1998).
800
600
'1
ME
400
Ct
Ct
200
0
15
00
00:00
:
Reconstructed
peaks
:
Simulations
Time
[hours]
09:00
Figure
4.11.
Simulated
hydrographs
and
reconstructed
peaks
at
three
gauging
stations
(1-3)
(Lange
et
al.
1999).
model
input
(weather
radar,
air
photos
and
GIS)
were
used.
The
two
major
processes
governing
the
development
of
large
desert
fl
oods,
generation
of
Hortonian
runoff
on
the
surfaces
and
fl
ow
losses
along
the
channel
network,
were
integrated
in
a
spatially
distributed
way.
Model
uncertainty
was
assessed
independently
for
each
of
the
13
input
parameters.
All
model
simulations
fell
within
the
uncertainty
ranges
of
the
reconstructed
peak
discharges
(Figure
4.11).
From
the
fully
distributed
structure
of
the
model
it
could
be
shown
that
during
this
fl
ood
only
one
large
upper
tributary
reacted
simultaneously
with
the
main
channel.
Lower
tributaries
produced
preceding
peaks
which
wet
the
channel
alluvium
before
the
main
fl
ood
arrived;
transmission
losses
therefore
lost
their
signifi-
cance
downstream.
The
arid
Tabalah
catchment,
Saudi
Arabia
The
1,270
km
2
Tabalah
catchment
drains
eastwards
towards
the
inner
part
of
Saudi
Arabia
and
provides
an
extensive
data
base
comprising
at
least
14
rainfall
recorders,
three
runoff
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
139
stations,
one
climate
station
and
five
observation
wells.
In
this
arid
catchment,
with
an
exceptionally
well
established
data
base,
Abdulrazzak
et
al.
(1989)
attempted
to
quantify
individual
water
balance
components.
Crystalline
rocks
and
basalts
did
not
offer
significant
groundwater
storage
and
the
main
aquifer
was
located
in
alluvial
deposits
beneath
the
channel.
As
a
consequence,
single
runoff
events
(three
to
eight
per
year)
had
to
be
regarded
as
an
important
water
resource.
Over
a
period
of
two
years
water
balance
components
were
determined
independently
for
13
runoff
producing
and
13
non
-runoff
producing
events.
Mean
values
for
these
events
indicated
that
only
3%
of
annual
precipitation
became
surface
runoff,
of
which
70-75%
contributed
to
recharge
of
the
shallow
alluvial
aquifers.
As
a
whole,
63%
of
the
annual
rain
was
lost
by
evaporation
and
30-32%
was stored
as
soil
moisture
in
the
unsaturated
zone.
The
accuracy
of
the
calculated
water
balance,
however,
strongly
depended
on
correct
estimation
of
the
mean
spatial
precipitation
and
potential
evaporation.
In
a
second
study
Sorman
and
Abdulrazzak
(1993,
1997)
more
closely
examined
recharge
to
an
alluvial
aquifer
within
the
Tabalah
catchment;
the
volume
of
transmission
losses
was
determined
for
27
events
for
a
study
reach
of
about
60.5
km.
Gauged
volumes
of
upstream
inflow
were
combined
with
tributary
runoff
contributions
and
subsequently
related
to
the
downstream
outflow
volumes.
Tributary
input
was
estimated using
measured
precipitation
and
the
simple
SCS-formula
(SCS
1985)
for
runoff
generation.
Relationships
between
the
magnitude
of
transmission
losses
and
fl
ow
parameters
were
obtained
by
regression
analysis.
It
was
noted
that
transmission
losses
were
highly
correlated
to
inflow
volume
(r
2
=
0.902):
T
L
=
0.19
+
0.47VUP
(20)
where
T
L
and
Vup
are
the
volumes
(MCM)
of
transmission
loss
and
upstream
inflow
into
the
reach,
respectively.
The
measured
rise
in
groundwater
table
for
the
underlying
alluvial
aquifer
(GWR,
m)
was
related
to
the
volume
of
transmission
losses
by:
GWR
=
0.005
+
0.59471
(21)
with
an
r
2
of
0.78.
Reliable
estimates
of
transmission
loss
and
associated
groundwater
table
rise
could
hence
be
achieved
knowing
only
the
inflow
volume.
However,
these
equations
remain
site
-specific,
and
may
require
further
verification
and
modification
using
additional
data
from
other
arid
channels.
To
shed
more
light
on
evaporation
processes
from
bare
soils
in
wadi
channels,
Sorman
and
Abdulrazzak
(1995)
used
different
approaches
on
an
experimental
reach
within
the
Tabalah
catchment.
They
first
determined
potential
evaporation
using
various
methods.
Concurrent
soil
moisture
data
were
obtained
by
neutron
probe
measurements
directly
in
the
channel
alluvium.
From
these
measurements
groundwater
recharge
due
to
direct
infiltration
of
rain
water
could
thus
be
excluded.
When
ponded
water
as
a
result
of
wadi
fl
ow
was
recorded,
the
wetting
front
reached
the
shallow
groundwater
table
and
recharged
the
shallow
subsurface
reservoir.
Daily rates
of
actual
evaporation
were
estimated
by
comparing
soil
moisture
contents
at
different
times.
The
values
reached
3.0
mm
day
-1
immediately
after
fl
oods
or
rain,
but
quickly
decreased
to
1.5
mm
day
-1
in
dry
periods.
After
a
long
dry
season
actual
evaporation
could
decrease
to
as
low
as
0.1-0.2
mm
day
-1
.
These
values
were
confirmed
by
measured
data
from
an
artificial
rainfall/infiltration
experiment.
140
Understanding
water
in
a
dry
environment
Discharge
(m'
70
50
50
40
60
50
40
Observed
Simulated
30
40
30
20
30
20
-
20
10
0
10
10'
0
100
200
300
400
0
200
400
600
0
0
200
400
600
800
Time
(min)
Figure
4.12.
Observed
and
simulated
hydrographs
for
three
events
in
the
upper
Tabalah
catchment
(El-Hames
&
Richards
1998).
El-Hames
and
Richards
(1998)
used
the
knowledge
and
data
described
above
to
develop
and
apply
a
physically
-based
rainfall
-runoff
model
to
the
upper
part
(170
km
2
)
of
the
Tabalah
catchment.
Kinematic
wave
theory
and
a
full
solution
of
the
St.
Venent
equations
were
used
to
account
for
fl
ood
routing
along
slopes
and
in
wadi
channels,
respectively.
A
numerical
scheme
solving
the
Richards
equation
was
coupled
with
these
fl
ood
routing
techniques
to
account
for
infiltration
losses.
In
the
absence
of
reliable
spatially
distributed
rainfall
data,
full
calibration
of
the
model
was
rather
difficult.
However,
the
model
could
still
reproduce
three
storm
runoff
events
with
a
fairly
high
degree
of
precision
(Figure
4.12).
The
arid
Wadi
Ghat,
Saudi
Arabia
The
597
km
2
Wadi
Ghat
catchment
drains
the
Asir
Mountain
range,
in
southwest
Saudi
Arabia,
to
the
Red
Sea.
Annual
precipitation
is
322
mm
and
approximately
90%
of
the
catchment
consists
of
rocky
outcrops
and
shallow
soils.
Intense
convectional
rain
storms
produce
typical
fl
ash
fl
oods
with
steep
rising
limbs
and
rapid
recession
to
zero
base
fl
ow.
Walters
(1989a)
reported
on
a
unique
fl
ood
event which
occurred
on
23
April
1985.
Three
rain
gauges
located
within
the
catchment
and
seven
in
its
surrounding
area
facilitated
reconstruction
of
the
spatial
rainfall
distribution.
During
four
hours
a
rain
cell
crossed
the
catchment
from
west
to
east.
Maximum
recorded
rainfall
(177.4
mm)
occurred
within
three
hours,
reaching
an
intensity
of
115.4
mm
hr
-1
in
the
southeastern
part
of
the
catchment.
This
intense
rainfall
caused
a
large
fl
ash
fl
ood
at
the
Wadi
Ghat
gauging
station
(Figure
4.13).
After
a
modest
increase
there
was
a
tremendous
rise
to
the
peak
of
3200
m
-3
5
-1
.
Comparing
fl
ood
volume
with
spatially
weighted
rainfall
as
determined
by
the
Theissen
polygon
method,
resulted
in
a
runoff
coefficient
of
27.2%.
Frequency
analysis
indicated
that
the
hourly
rainfall
had
a
return
period
which
exceeded
200
years,
while
the
recorded
peak
discharge
could
be
expected
on
average
once
every
143
years.
The
arid
Barrier
range,
Australia
Most
ephemeral
channels
draining
western
fl
anks
of
the
Barrier
Range
in
western
New
South
Wales,
Australia,
are
ungauged.
The
greatest
volume
of
storm
runoff
results
from
synoptic
scale
depressions
rather
than
from
local
convective
storms.
Mean
annual
rainfall
and
potential
evaporation
total
190
and
more
than
2000
mm,
respectively.
The
Barrier
Range
is
composed
of
crystalline
basement
and
sedimentary
sequences,
while
the
fl
anking
plains
are
covered
by
accumulated
piedmont
sediments.
In
the
absence
of
gauged
stream
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
141
Discharge
(m
3
51)
Time
(hours)
3000
-
12:00
14:00
18:00
20:00
E
20
L
40
2000
-
60
1000
-
0
/
12:00
14:00
16:00 18:00
Time
(hours)
Figure
4.13.
Flash
fl
ood
at
Wadi
Ghat
(Walters
1989a).
fl
ow
data,
Dunkerley
(1992)
surveyed
channel
form
and
bed
material
at
sequential
sites
along
two
lowland
stream
reaches.
Cross
-sections
were
mapped
at
intervals
of
about
1
km
and
channel
gradients,
highest
water
surface
elevations
and
the
maximum
grain
size
transported
during
fl
ow
events
were
determined.
From
these
parameters
critical
fl
ow
velocities
(required
to
transport
the
largest
boulders),
mean
fl
ow
velocities
and
discharges
were
estimated.
All
parameters
significantly
declined
downstream.
The
estimates
of
fl
ood
discharge
could
not
substitute
actual
fl
ow
data,
since
they
were
based
on
calculated
velocities
without
fully
understanding
the
fl
ow
mechanics.
However,
their
striking
downstream
decline
provided
valuable
indications
of
channel
processes
and
emphasized
the
importance
of
transmission
losses
in
these
environments
(Figure
4.14).
Mean
rates
of
discharge
loss
(19.8
m
3
5
-1
km
-1
and
17.0
m
3
s
-1
km
-1
)
were
in
surprisingly
close
agreement
for
the
two
wadis.
4.3.4
Regional
comparisons
Floods
and
transmission
losses
in
Yemen
Farquharson
et
al.
(1996)
studied
surface
water
resources
in
Yemen.
The
mountainous
nature
of
the
country,
rising
from
sea
level
to
over
3700
m
within
about
100
km,
strongly
affects
the
hydrology.
The
orographic
influence
on
rainfall
is
strong
and
mean
annual
values
vary
from
50
to
800
mm.
In
most
dry
upland
areas
a
unique
practice
of
bunding
and
terracing
impedes
immediate
runoff
and
erosion,
permitting
reliance
on
rain
-fed
agri-
culture.
Short
data
records,
typical
for
arid
and
semi
-arid
regions,
made
assessment
of
the
available
water
resources
even
more
difficult.
Long-term
runoff
records
were
available
for
only
a
few
main
wadis
with
only
two
sites
providing
data
for
more
than
20
years.
142
Understanding
water
in
a
dry
environment
7
Ln
of
discharge
Ln
of
discharge
6
-a
..
... ..
..
.
11,-
-----
a
-
5
s---------
-
vi-
-----
--
--------
-------
i
--
a.
--
--
-----
4
......
ir ....
-
"
...
3
.................
..
2
Ln
Qmax
1
Ln
mean
Q
Fowlers
Creek
0
0
6
5-
4
3
2
1
-
2
4
6
8
10
12
14
16
Distance
from
uplands
(km)
......
.
••••
..
....
....
.....
R.
.. .....
.....
X
.. ..
.........
.....
Ln
Qmax
Ln
mean
Q
......
'ff•
.....
..
.........
....
-
M*
.......
Sandy
Creek
1
2
3
4
5
6
Distance
from
uplands
(km)
7
Figure
4.14.
Discharge
estimates
for
two
ephemeral
channels
in
Australia (Dunkerley
1992).
Since
rainfall
records
covered
a
longer
time
period
and
were
thought
to
be
more
reliable,
they
were
used
to
prolong
runoff
records
for
the
northern
part
of
the
country
with
the
help
of
a
rainfall
-runoff
model.
Characteristic
zones
for
surface
runoff
were
classified
to
account
for
the
effect
of
terracing
and
water
absorption
into
deep
alluvium
layers;
the
model
used
the
SCS
Curve
Number
method
(SCS
1985)
and
was
calibrated
for
six
different
wadis.
By
incorporating
existing
sparse
and
sporadic
fl
ow
records
in
the
southern
and
eastern
parts
of
the
country,
a
countrywide
relationship
between
mean
annual
rainfall
and
runoff
was
established
(Figure
4.15).
Large
variations
in
the
runoff
coefficients
were
related
to
differences
in
topography
and
land
use.
However,
even
in
the
most
arid
areas
runoff
coefficients
did
not
fall
below
5%.
Alderwish
and
Dottridge
(1995)
selected
catchments
in
the
intermontane
Sana'
a
basin
to
quantify
the
yearly
recharge
from
two
ephemeral
streams.
They
first
collected
all
the
available
geological,
hydrogeological
and
hydrochemical
information.
With
this
extensive
set
of
subsurface
data
they
calibrated
the
finite
-difference
groundwater
model
MODFLOW
(McDonald
&
Harbaugh
1988)
in
a
steady-state,
reproducing
the
general
pattern
of
groundwater
fl
ow.
Using
the
steady-state
model
as
a
starting
point,
a
transient
calibration
was
performed
including
four
fl
oods
in
the
first
rainy
season
of
1993.
Detailed,
regular
water
level
measurements
in
the
alluvial
aquifer
had
already
been
initiated.
The
boundary
conditions
from
the
steady-state
calibration
(aquifer
thickness
and
hydraulic
conductivity)
were
left
Mean
annual
runoff
(mm)
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
143
50
-
40 -
30
-
20
-
10-
0
A
NW
NE
A
South
Others
r
i
0
A
A
A
A
A
0
200
400
600
Mean
annual
rainfall
(mm)
Figure
4.15.
Mean
annual
rainfall
and
runoff
in
Yemen
(Farquharson
et
al.
1996).
Table
4.7.
Calculated
recharge
volumes
for
two
wadis
in
Yemen
(Alderwish
&
Dottridge
1995).
Location
Catchment
area
Recharge
(m
3
)
from
Recharge
(m
3
)
from
(km
2
)
groundwater
model
channel
water
balance
Wadi
Dhar
13.4
35,718
37,635
Wadi
Alsir
182.7
594,720
548,575
unchanged.
By
a
trial
and
error
procedure,
stream
bed
hydraulic
conductance
(represent-
ing
the
degree
of
hydraulic
connectivity
between
wadi
and
alluvial
aquifer)
and
specific
yield
of
the
alluvium
were
adjusted
(calibrated)
until
the
observed
groundwater
hydro
-
graphs
could
be
reproduced
by
MODFLOW.
As
a
result,
recharge
from
the
two
wadis
during
a
rainy
season
could
be
quantified;
these
compared
well
with
independent
esti-
mates
of
recharge
derived
from
computations
of
the
surface
water
balance
(Table
4.7).
High
magnitude
floods
in
the
entire
United
States
Costa
(1987)
analysed
the
largest
historic
rainfall
-runoff
fl
oods
from
small
catchments
(0.39-370
km
2
)
in
the
entire
USA;
twelve
were
identified
to
be
the
largest
ever
measured.
For
all
of
them,
post
-flood
surveys
and
subsequent
calculations
by
the
slope
-area
method
provided
indirect
measurements
of
peak
discharge.
All
events
plotted
on
a
smooth
curve
of
drainage
area
versus
discharge,
and
occurred
in
the
semi
-arid
and
arid
parts
of
the
western
and
southwestern
United
States
(Figure
4.16).
Mean
annual
precipitation
in
these
catchments
ranged
from
114
to
676
mm.
However,
the
greatest
measured
rainfall
inten-
sities
for
short
durations
were
measured
in
humid
parts
of
the
United
States
and
did
not
coincide
with
the
maximum
fl
ood
peaks.
Estimated
composite
Manning's
n
values
ranged
from
0.028
to
0.048
with
a
mean
of
0.038.
Mean
peak
fl
ow
velocities
and
maximum
Froude
numbers
for
the
entire
channel
cross
sections
were
3.47-9.97
m
5
-1
and
0.81-2.49
respectively.
This
indicated
critical
or
even
supercritical
conditions
during
peak
fl
ow.
Shear
stresses
and
unit
stream
powers
144
Understanding
water
in
a
dry
environment
10000
1000
a)
0
_c
U
100
10
Texas
,
0
11
4
/;.-
Texas
0.
1 '.
Colorado
0
Nevada
...**,•*
Texas
Arizona
Oregon
o
Oregon
.••••••
•*.
r
§
t
Nevada
'
o
Utah..*
Nevada
e
Washington
1
1
1
1
11111
1
1 1
111111
1 1
1 1
11111
1 1 1
I
1111
0.1
10
Drainage
area
(km')
100
100C
Figure
4.16.
Envelope
curves
(based
on
data
that
existed
in
1890,
1939
and
1985)
of
the
largest
rainfall
-runoff
fl
oods
measured
in
the
United
States
(Costa
1987).
were
several
hundred
times
greater
than
for
fl
oods
in
large
rivers
(e.g.
Mississippi,
Amazon).
Morphometric
catchment
characteristics
did
not
differ
markedly
from
other
fl
ash
-flood
prone
areas.
Exceptions
could
only
be
found
in
relatively
high
elongation
ratios,
indicating
short
compact
basins
and
relatively
low
drainage
densities.
It
was
concluded
that
the
location
of
the
largest
historic
fl
ash
-flood
peak
discharges
from
small
basins
appeared
to
be
controlled
by
an
optimum
combination
of
high
rainfall
intensity,
and
basin
physiographic
and
morphologic
characteristics.
Key
environmental
conditions
were:
(a)
the
rapid
surface
runoff
response
typical
for
arid
and
semi
-arid
climate,
due
to
an
abundance
of
exposed
bedrock
and
restricted
soil
and
vegetation
cover;
and
(b)
prevailing
short,
steep
and
rugged
catchments.
4.4
CONCLUSIONS
AND
RECOMMENDATIONS
Surface
fl
ow
plays
a
decisive
role
in
the
hydrological
balance
of
dry
areas,
as
it
is
responsible
for
the
redistribution
of
its
rare
renewable
water
resources.
Unlike
in
most
humid
areas,
fl
oods
mobilise
huge
amounts
of
sediment
leaving
fluvial
forms
indelibly
imprinted
on
the
landscape.
Two
main
hydrological
processes
dominate
their
occurrence:
The
generation
of
Hortonian
overland
fl
ow
on
dry
land
terrain,
and
transmission
losses
into
the
dry
alluvial
beds
of
ephemeral
channels.
Similar
to
dry
land
rainfall,
these
processes
are
characterized
by
an
accentuated
spatial
heterogeneity
making
them
difficult
to
assess
for
catchment
-scale
problems.
The
present
chapter
has
presented
an
overview
of
processes
and
techniques
for
their
assessment.
Case
studies
show
that
dry
land
research
has
been
performed
in
many
parts
of
the
globe.
Using
empirical
regression
equations,
general
results
may
be
translated
to
similar
catchments
in
order
to
obtain
initial
estimates.
However,
caution
must
be
exercised
when
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
145
equations
lack
relations
to
characteristics
of
the
area,
such
as
eqns.
(2)-(5)
and
(19)-(21),
especially
when
based
on
only
a
few
independent
data
sets.
Others,
where
surface
char-
acteristics
are
included
(eqns.
6-14
and
18),
have
a
more
general
character
and
should
be
given
priority
where
possible.
For
sound
hydrological
analyses
all
equations
should
be
corroborated
by
local
field
data,
which
does
not
necessarily
require
establishing
an
extens-
ive
hydrometric
network.
In
particular,
the
paleoflood
techniques
described
in
Section
4.2.2
and
applied
in
case
studies
(Sections
4.3.3
and
4.3.4),
offer
lower
cost
alternatives.
With
ongoing
progress
in
computer
technology,
hydrological
models
will
be
increas-
ingly
used
also
in
arid
and
semi
-arid
areas.
For
these,
similar
aspects
apply
as
for
regres-
sion
relations.
Simple,
lumped approaches
depend
on
calibration
with
local
runoff
data.
As
such,
their
parameters
comprise
uncertainties
of
hydrometric
measurements
and
depend
on
local
catchment
characteristics.
Their
use
should
thus
be
restricted
to
catchments
with
long
data
records
and
parameter
values
should
not
be
directly
transferred.
If
calibration
is
minimized,
the
general
applicability
of
a
model
increases
(see
Section
4.3.3).
However,
the
approach
should
then
be
kept
as
simple
as
possible,
since
the
number
of
input
param-
eters
required
may
limit
general
applicability.
Recent
techniques
(e.g.
remote
sensing,
GIS)
have
already
facilitated
parameter
determination
for
a
rather
simple
but
completely
non
-calibrated
rainfall
-runoff
model
(see
Section
4.3.3),
making
it
applicable
for
ungauged
catchments.
Finally,
episodic
dry
land
fl
oods
offer
an
often
unused
potential
for
the
development
of
meagre
arid
and
semi
-arid
water
resources.
Their
wise
use,
e.g.
for
a
controlled
recharge
of
local
and
regional
aquifers,
may
help
bridge
times
of
drought
and
safeguard
the
often
non-renewable
water
resources
in
deep
aquifers
for
future
generations.
4.5
LIST
OF
SYMBOLS
1/4PWK
'Width'
of
the
hydrograph
at
1/4
peak
stage
min
1/2PKW
'Width'
of
the
hydrograph
at
1/2
peak
stage
min
A
Soil
loss
kg
111
-2
S
-1
A
Drainage
area
km
2
C
Cropping
management
factor
cev
Tracer
concentration
in
the
event
water
mg
L
-1
C
g
es
Tracer
concentration
in
the
runoff
mg
L
-1
Cprev
Tracer
concentration
in
the
pre
-event
water
mg
L
-1
D
d
Drainage
density
km
-1
D,
Flow
duration
hr
F
c
Spatially
weighted
erodibility
factor
GWR
Measured
rise
in
groundwater
table
m
IK
Index
quantifying
initial
moisture
k
Recession
constant
-
K
Effective
hydraulic
conductivity
in
hr
-1
K
Soil
erodibility
factor
-
L
Slope
length
factor
-
If
Generated
plot
surface
runoff
mm
L
o
Observed
catchment
runoff
mm
LS
Topographic
factor
146
Understanding
water
in
a
dry
environment
MAF
Mean
annual
fl
ood
m
3
S
-1
Q
Mean
annual
runoff
10
6
m
3
P
Erosion
control
practice
factor
-
P
Annual
rainfall
cm
Po
Threshold
runoff
volume
acre-feet
P
1
Inflow
volume
to
a
channel
reach
acre-feet
Pu
Rainfall
depth
mm
R
Rainfall
erositivity
factor
-
S
Slope
gradient
factor
-
S
Catchment
slope
T
L
Volume
of
transmission
loss
MCM
Qev
Amount
of
event
water
m
3
S
-1
Qges
Total
runoff
m
3
S
-1
Q,
Annual
maximum
fl
ood
m
3
S
-1
Qo
Outflow
volume
of
a
channel
reach
acre-feet
Qprev
Amount
of
pre
-event
water
m
3
S
-1
Volume
of
sediment
yield
m
3
yr
-1
VUP
Volume
of
inflow
into
a
channel
reach
MCM
w
Width
of
a
channel
reach
feet
x
Length
of
a
channel
reach
feet
X
Soil
erosion
t
ha
-1
yr
-1
Y
Radionuclide
loss
(
137
Cs)
%
REFERENCES
Abdulrazzak,
M.J.,
Sorman,
A.U.
&
Alhamas, A.S.
1989.
Water
balance
approach
under
extreme
arid
conditions
a
case
study
of
Tabalah
Basin,
Saudi
Arabia.
Hydrological
Processes
3:
107-122.
Agnew,
C.
&
Anderson,
E.
1992.
Water
Resources
in
the
Arid
Realm.
Routledge,
London.
Al-Weshah,
R.A.
2002a.
Rainfall
-runoff
analysis
and
modelling
in
wadi
systems.
In:
H.S.
Wheater
&
R.A.
Al-Weshah
(eds),
Hydrology
of
Wadi
Systems.
UNESCO
IHP-V
Technical
Documents
in
Hydrology
55:
87-111.
Al-Weshah,
R.A.
(ed.)
2002b.
Water
Resources
of
Wadi
Systems
in
the
Arab
World:
Case
Studies
(Jordan,
Egypt,
Sudan,
Morocco,
Tunisia).
IHP
Arab
Wadi
Hydrology
Network,
UNESCO
Cairo
Office,
12.
Alderwish,
A.
&
Dottridge,
J.
1995.
Modelling
infiltration
from
ephemeral
wadi
fl
ows
in
the
Sana'a
basin,
Yemen.
In:
P.L.
Younger
(ed.),
Modelling
River
Aquifer
Interactions.
British
Hydrological
Society
Occasional
Paper
6,
pp.
4-16.
Alhamid,
A.A.
&
Reid,
I.
2002.
Sediment and
the
vulnerability
of
water
resources.
In:
H.S.
Wheater
&
R.A.
Al-Weshah
(eds),
Hydrology
of
Wadi
Systems.
UNESCO
IHP-V
Technical
Documents
in
Hydrology
55:
37-55.
Baker,
V.R.
1987.
Paleoflood
hydrology
and
extraordinary
fl
ood
events,
Journal
of
Hydrology
96:
79-99.
Bergkamp,
G.
1998.
A
hierarchical
view
of
the
interactions
of
runoff
and
infiltration
with
vegetation
and
microtopography
in
semiarid
shrublands.
Catena
33:
201-220.
Brakensiek,
D.L.
&
Rails,
W.J.
1994.
Soil
containing
rock
fragments:
effects
on
infiltration.
Catena
23:
99-110.
Bromley,
J.,
Brouwer,
J.,
Barker,
A.P.,
Gaze,
S.R.
&
Valentin,
C.
1997.
The
role
of
surface
water
redistribution
in
an
area
of
patterned
vegetation
in
a
semi
-arid
environment,
south-west
Niger.
Journal
of
Hydrology
198:
1-29.
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
147
Bryan,
R.B.
&
Yair,
A.
1982.
Perspectives
on
studies
of
badland
geomorphology.
In:
R.
Bryan
&
A.
Yair
(eds),
Badland
Geomorphology
and
Piping.
Geo
Abstracts,
Norwich,
UK,
pp.
1-12.
Campbell,
I.A.
1981.
Spatial
and
temporal
variations
in
erosion
measurements.
IAHS
Publications
133:
447-456.
Casenave,
A.
&
Valentin,
C.
1992.
A
runoff
capability
classification
system
based
on
surface
features
criteria
in
semi
-arid
areas
of
West
Africa.
Journal
of
Hydrology
130:
231-249.
Choudhari,
J.S.
&
Sharma,
K.D.
1984.
Stream
salinity
in
the
Indian
arid
zone.
Journal
of
Hydrology
71:
149-163.
Costa,
J.E.
1987.
Hydraulics
and
basin
morphometry
of
the
largest
fl
ash
fl
oods
in
the
conterminous
United
States.
Journal
of
Hydrology
93:
313-338.
Crerar,
S.,
Fry,
R.G.
&
Slater,
G.
1988.
An
unexpected
factor
affecting
recharge
from
ephemeral
river
fl
ows
in
SWA/Namibia.
In:
I.
Simmers
(ed.),
Estimation
of
Natural
Groundwater
Recharge.
Reidel,
Dordrecht,
The
Netherlands,
pp.
11-28.
De
Boer, D.H.
1992.
Constraints
on
spatial
transference
of
rainfall
-runoff
relationships
in
semi
-arid
basins
drained
by
ephemeral
streams.
Hydrological
Science
Journal
37:
491-504.
De
Boer, D.H.
&
Campbell,
I.A.
1989.
Spatial
scale
dependence
of
sediment
dynamics
in
a
semi-
arid
badland
drainage
basin.
Catena
16:
277-290.
De
Boer,
D.H.
&
Campbell,
I.A.
1990.
Runoff
chemistry
as
an
indicator
of
runoff
sources
and
routing
in
semi
-arid,
badland
drainage
basins.
Journal
of
Hydrology
121:
379-394.
Dunkerly,
D.L.
1992.
Channel
geometry,
bed
material,
and
inferred
fl
ow
conditions
in
ephemeral
stream
systems,
Barrier
Range,
western
N.S.W.
Australia.
Hydrological
Processes
6:
417-433.
Dunkerley,
D.
&
Brown,
K.
1999.
Flow
behaviour,
suspended
sediment
transport
and
transmission
losses
in
a
small
(sub
-bank
-full)
fl
ow
event
in
an
Australian
desert
stream.
Hydrological
Pro-
cesses
13:
1577-1588.
El-Hames,
A.S.
&
Richards,
K.S.
1998.
An
integrated,
physically
based
model
for
arid
region
fl
ash
-flood
prediction
capable
of
simulating
dynamic
transmission
loss.
Hydrological
Processes
12:
1219-1232.
Emmet,
W.W.
1978.
Overland
fl
ow.
In:
M.J.
Kirkby
(ed.),
Hillslope
Hydrology.
Wiley,
New
York,
pp.
145-176.
Evenari,
M.,
Shanan,
L.
&
Tadmor,
N.
1982.
The
Negev,
the
Challange
of
a
Desert.
Harvard
University
Press,
Cambridge,
Massachusetts.
Farquharson,
F.A.K.,
Meigh,
J.R.
&
Sutcliffe,
J.V.
1992.
Regional
fl
ood
frequency
analysis
in
arid
and
semi
-arid
areas.
Journal
of
Hydrology
138:
487-501.
Farquharson,
F.A.K.,
Plinston,
D.T.
&
Sutcliffe,
J.V.
1996.
Rainfall
and
runoff
in
Yemen.
Hydrological
Science
Journal
41:
797-811.
Gat,
J.R.
1980.
The
relationship
between
surface
and
sub
-surface
waters:
water
quality
aspects
in
areas
of
low
precipitation.
Hydrological
Science
Bulletin
25:
257-267.
Goodrich,
D.C.,
Lane,
L.J.,
Shillito,
R.M.,
Miller,
S.N.,
Syed,
K.H.
&
Woolhiser,
D.A.
1997.
Linearity
of
basin
response
as
a
function
of
scale
in
a
semi
-arid
watershed.
Water
Resources
Research
33:
2951-2965.
Graf,
W.D.
1988.
Fluvial
Processes
in
Dryland
Rivers.
Springer
Verlag,
Berlin.
Grayson,
R.B.,
Moore,
I.D.
&
McMahon,
T.A.
1992.
Physically
based
hydrologic
modeling,
1.
A
terrain
-based
model
for
investigative
purposes.
Water
Resources
Research
28:
2639-2658.
Greenbaum,
N.,
Margalit,
A.,
Schick,
A.P.
&
Baker,
V.R.
1998.
Reconstruction
of
a
high
magni-
tude
rainstorm
-flood
in
Nahal
Zin
a
large
hyperarid
catchment
in
the
Negev
Desert,
Israel.
Hydrological
Processes
12:
1-23.
Haas,
H.J.,
Willis,
W.O.
&
Boatwright,
G.O.
1966.
Moisture
storage
and
spring
wheat
yields
on
level
bench
terraces
as
influenced
by
contributing
area
cover
and
evaporation
control.
Agricultural
Journal
58:
297-299.
Hillel,
D.
1987.
Runoff
Inducement
in
Arid
Lands.
Faculty
of
Agriculture,
the
Hebrew
University,
Rehovot,
Israel.
Hodges,
W.K.
&
Bryan,
R.B.
1982.
The
influence
of
material
behaviour
on
runoff
initiation
in
the
Dinosaur
Badlands,
Canada.
In:
R.B.
Bryan
&
A.
Yair
(eds),
Badland
Geomorphology
and
Piping,
pp.
13-46.
Horton,
R.E.
1933.
The
role
of
infiltration
in
the
hydrological
cycle.
EOS,
American
Geophysical
Union
Transactions
14:
446-460.
148
Understanding
water
in
a
dry
environment
Hughes,
D.A.
1997.
Southern
African
TRIEND'-The
application
of
rainfall
-runoff
models
in
the
SADC
region.
WRC
Report
235/1/97,
Grahamstown,
South
Africa.
Hughes,
D.A.
&
Sami,
K.
1992.
Transmission
losses
and
associated
moisture
dynamics
in
a
semi-
arid
ephemeral
channel
system
in
southern
Africa.
Hydrological
Processes
6:
45-53.
Hussein,
M.H.
1996.
An
analysis
of
rainfall,
runoff
and
erosion
in
the
low
rainfall
zone
of
northern
Iraq.
Journal
of
Hydrology
181:
105-126.
Hydrological
Engineering
Center.
1982.
HEC-2
Water
Surface
Profiles:
Program
User's
Manual.
US
Army
corps
of
Engineers,
Davis.
Karnieli,
A.M.,
Diskin,
M.H.
&
Lane,
L.J.
1994.
CELMODS
a
semi
-distributed
model
for
conversion
of
rainfall
into
runoff
in
semi
-arid
watersheds.
Journal
of
Hydrology
157:
61-85.
Kelway,
P.S.
&
Herbert,
S.I.
1969.
Short
term
rainfall
analysis.
Weather
24:
342-354.
Knighton,
A.D.
&
Nanson,
G.C.
1994.
Flow
transmission
along
an
arid
zone
anastomosing
river,
Copper
Creek,
Australia.
Hydrological
Processes
8:
137-154.
Knighton,
A.D.
&
Nanson,
G.C.
2001.
An
event
-based
approach
to
the
hydrology
of
arid
zone
rivers
in
the
channel
country
of
Australia.
Journal
of
Hydrology
254:
102-123.
Kuells,
C.,
Leibundgut,
Ch.,
Schwartz,
U.
&
Schick,
A.P.
1995.
Channel
infiltration
study
using
dye
tracers.
IAHS
Publications
232:
429-436.
Lane,
L.J.
1985.
Transmission
losses.
In:
SCS
(ed.),
National
Engineering
Handbook,
Ch.
19,
pp.
1-21.
Lane,
L.J.,
Diskin,
M.H.
&
Renard,
K.G.
1971.
Input-output
relationships
for
an
ephemeral
stream
channel
system.
Journal
of
Hydrology
13:
22-40.
Lange,
J.,
Leibundgut,
Ch.,
Greenbaum,
N.
&
Schick,
A.P.
1999.
A
noncalibrated
rainfall
-runoff
model
for
large,
arid
catchments.
Water
Resources
Research
35:
2161-2172.
Lange,
J.,
Leibundgut,
Ch.,
Grodek,
T.,
Lekach,
J.
&
Schick,
A.P.
1998.
Using
artificial
tracers
to
study
water
losses
of
ephemeral
fl
oods
in
small
arid
streams.
IAHS
Publications
247:
31-40.
Lange,
J.,
Mostert,
A.
&
Wessels,
C.
submitted.
Dynamics
of
transmission
losses
in
a
large
arid
stream
channel.
Journal
of
Hydrology.
Lavee,
H.,
Imeson,
A.C.
&
Sarah,
P.
1998.
The
impact
of
climate
change
on
geomorphology
and
desertification
along
a
mediterranean-arid
transect.
Land
Degradation
and
Development
9:
407-
422.
Li,
G.,
Luk,
S.H.
&
Cai,
Q.G.
1995.
Topographic
zonation
of
infiltration
in
the
hilly
loess
region,
north
China.
Hydrological
Processes
9:
227-235.
Linsley,
R.K.,
Kohler,
M.A.
&
Paulhus,
J.L.H.
1982.
Hydrology
for
Engineers.
McGraw
Hill,
New
York.
Luk,
S.H.,
Cai,
Q.G.
&
Wang,
G.P.
1993.
Effects
of
surface
crusting
and
slope
gradient
on
soil
and
water
losses
in
the
hilly
loess
region,
north
China.
Catena
Supplement
14:
29-45.
Mabutt,
J.A.
1977.
Desert
Landforms.
Australian
National
Univiversity
Press,
Canberra.
Martinez
-Mena,
M.,
Albaladejo,
J.
&
Castillo,
V.M.
1998.
Factors
influencing
surface runoff
generation
in
a
Mediterranean
semi
-arid
environment:
Chicamo
watershed,
SE
Spain.
Hydrological
Processes
12:
741-754.
McDonald,
M.G.
&
Harbaugh,
A.W.
1988.
A
modular
three
dimensional
finite
-difference
groundwater
fl
ow
model.
Techniques
of
Water
Resources
Investigations,
Report:
TWI
06
-Al,
U.S.
Geological
Survey,
Reston,
VA,
USA.
Michaud,
J.
&
Sorooshian,
S.
1994a.
Comparison
of
simple
versus
complex
distributed
runoff
models
on
a
midsized
semi
-arid
watershed.
Water
Resources
Research
30:
593-605.
Michaud,
J.
&
Sorooshian,
S.
1994b.
Effect
of
rainfall
-sampling
errors
on
simulations
of
desert
fl
ash
fl
oods.
Water
Resources
Research
30:
2765-2775.
Miraki,
G.D.
1983.
Sediment
Yield
and
Deposition
Profiles
in
Reservoirs.
PhD
Thesis,
University
of
Roorkee,
India.
Osborn,
H.G.
&
Renard,
K.G.
1970.
Thunderstorm
runoff
on
the
Walnut
Gulch
Experimental
Watershed,
Arizona.
USA.
IAHS
Publications
96:
456-464.
Parissopoulos.
G.A.
&
Weather,
H.S.
1992.
Experimental
and
numerical
infiltration
studies
in
a
wadi
stream
bed.
Hydrological
Science
Journal
37:
27-37.
Renard,
K.G.
&
Keppel,
R.V.
1966.
Hydrographs
of
ephemeral
streams
in
the
Southwest.
Journal
of
the
Hydraulics
Division
92:
33-52.
Ritchie,
J.C.,
Spraberry,
J.A.
&
McHenry,
J.R.
1974.
Estimating
soil
erosion
from
the
redistribution
of
fallout
137Cs.
Proceedings
Soil
Science
Society
of
America
38:
137-139.
Surface
runoff
and
sediment
dynamics
in
arid
and
semi
-arid
regions
149
Sandstrom,
K.
1996.
Hydrochemical
deciphering
of
streamflow
generation
in
semi
-arid
east
Africa.
Hydrological
Processes
10:
703-720.
Schick,
A.P.
1988.
Hydrological
aspects
of
fl
oods
in
extreme
arid
environments.
In:
V.R.
Baker
(ed.),
Flood
Geomorphology.
Wiley,
New
York,
pp.
189-203.
Schick,
A.P.
&
Lekach,
J.
1993.
An
evaluation
of
two
ten-year
sediment
budg