response
Lewin,
of
studies. Initial support came especially from exper-
imental physical modelling of network development
small catchments similarly indicated that physically
continuous terraces could be asynchronous, whilst
lower surfaces upstream could be temporally equiv-
Geomorphology 69 (2005)
1. Introduction
Catchment complex response, broadly interpreted
to mean differential internal catchment processing of
sediment transfer in response to uniform external
forcing stimuli, has been noted in a variety of
and sediment transfer (see Schumm, 1979, pp. 493?
497). In these experiments, incision in the lower part
of an experimental catchment led to rejuvenation of
upstream tributaries and their enhanced sediment
delivery then led to aggradation in the lower incised
reaches. Later experiments on terrace formation in
Received 23 February 2004; received in revised form 19 January 2005; accepted 22 January 2005
Available online 16 March 2005
Abstract
The CAESAR (Cellular Automaton Evolutionary Slope And River) model is used to demonstrate significant differences in
coarse sediment transfer and alluviation in medium sized catchments when responding to identical Holocene environmental
changes. Simulations for four U.K. basins (the Rivers Swale, Ure, Nidd and Wharfe) shows that catchment response, driven by
climate and conditioned by land cover changes, is synchronous but varies in magnitude. There are bursts of sediment transfer
activity, generally of rapid removal but with some sediment accumulation dspikesT, with longer periods of slow removal or
accumulation of sediment in different valley reaches. Within catchments, reach sensitivity to environmental change varies
considerably: some periods are only recorded in some reaches, whilst higher potential sensitivity typically occurs in the
piedmont areas of the catchments modelled here. These differential responses appear to be highly non-linear and may relate to
the passage of sediment waves, by variable local sediment storage and availability, and by large- and small-scale thresholds for
sediment transfer within each catchment. Differential response has major implications for modelling fluvial systems and the
interpretation of field data. Model results are compared with the record of dated alluvial deposits in the modelled catchments.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Erosion; Deposition; Environmental change; Modelling; Catchment; Non-linear; Sediment transfer
Modelling differential catchment
T.J. Coulthard
T
, J.
Institute of Geography and Earth Sciences, University
0169-555X/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2005.01.008
T Correspondin
E-mail addresses: T.Coulthard@aber.ac.uk (T.J. Coulthard),
mvm@aber.ac.uk (M.G. Macklin).
to environmental change
M.G. Macklin
Wales, Aberystwyth, Ceredigion, SY23 3DB, UK
222?241
www.elsevier.com/locate/geomorph
in relation toalent to higher surfaces downstream
g author. Tel.: +44 1970 622583.
longitudinal profiles (Germanoski and Harvey,
1993).
Field conditions, particularly in medium sized and
large river basins may be more difficult to interpret,
given a requirement to date alluvial bodies over an
extended time period, and the likelihood that forcing
stimuli such as climate change and human activity
may themselves have varied in a complex manner.
Nevertheless, a number of studies have pointed to the
differential behaviour of catchments and sub-catch-
ments, with periods of aggradation and incision being
contrasted in what may be nearby reaches of trunk
and tributary streams (Boison and Patton, 1985;
Rains and Welch, 1988). Considerations related to
network topology and storage basins within catch-
ments have recently been emphasised in a review by
Richards (2002), who concludes that catchment
modelling is required in order to examine the linkage
between dexternalT change (for example climate and
land-use) and responding alluviation which may be
delayed and differentiated within and between catch-
ments. Paola also describes how the timing of the
sedimentological record can be dfinickyT,being
sensitive to external controls as well as bthe internal
dynamics of the sediment?deposition systemQ (Paola,
2003, p. 459).
For fluvial modellers, this makes any general
forecasting of fluvial system response especially
difficult, as relationships established for one river
system can rarely be applied to others. Collecting site
specific data to calibrate and validate geomorphic
models renders them costly and time-consuming. An
ideal solution would be to develop a generic
modelling approach that could be applied with a
minimum of alteration to any river system. However,
in order to reach such a goal, it is first necessary to
establish whether catchments do actually respond in a
similar fashion to imposed environmental changes.
Investigating causes of catchment change, previous
authors have established three main groups of factors
(e.g. Schumm, 1979).
1. Changes to the external environment, as in tectonic
activity, sea level change or climate fluctuations.
Some of these may be rapid as in the case of local
response to faulting, but for the most part changes
occur over timescales of centuries to millennia.
2. Anthropogenic factors, as in land-use changes,
T.J. Coulthard et al. / Geomorphology
sediment extraction or river channel engineering.
These are also imposed on or are external to the
geomorphological system in their origin. Again,
immediate effects on catchments may be evident,
as in the case of impacts on sediment yields, but
broadly speaking such factors relate to periods of
human occupation which vary from decades to
millennia in different parts of the world.
3. Geomorphological factors, which include catch-
ment morphology and internal sediment transfer,
storage and availability as conditioned also by
prior erosional and depositional history. These can
be rapid but local, but extending over long time-
scales where bedrock exposure and catchment
transformation are involved.
It should be noted that sediment transfer processes
within any catchment may be affected by all of these
factors simultaneously.
Computer modelling would appear to be an ideal
tool for exploring the impact of these factors, as
modellers have total control over their simulated
environments, and thus the capability to compare
how landscape development is altered. Adopting a
reach based modelling approach would at first seem
logical, as a reach may exhibit an immediate response
to a forcing event such as an extreme flood, and
produce deposits recording its occurrence. However, it
is the differential triggering of sediment movement
and different rates of sediment routing down the
length of the fluvial system that is likely to effect large
scale deposition or removal of alluvial units over
extended time periods. Therefore, fluvial processes
operating throughout the catchment have to be
considered. To allow this, some workers have used
models that simulate the development of long profiles
(e.g. Tebbens and Veldkamp, 2001) and landscape
evolution models have been developed that model
whole drainage basins with a grid of square cells (e.g.
Willgoose et al., 1991; Howard, 1994; Tucker and
Slingerland, 1994) or with an irregular mesh of nodes
and links (Tucker et al., 2001; Braun and Sambridge,
1997). These studies have explored the impacts of the
first two groups of factors, including tectonics (Will-
goose et al., 1991; Braun and Sambridge, 1997;
Howard, 1994) and the effects of climate and land
cover changes (Tucker and Slingerland, 1997; Coulth-
ard and Macklin, 2001; Coulthard et al., 2002). Some
69 (2005) 222?241 223
models have also partially addressed the third group
of geomorphological factors, reproducing internal
ies and generating non-linear sediment dis-
from an early application of the CAESAR
(Coulthard et al., 1998), and from a braided
model (Murray and Paola, 1994). These have
attributed to the irregular input of material from
movement and slope processes (Coulthard et al.,
) and to the passage of mobile bar structures
Murray and Paola, 1994).
Here we aim to investigate these geomorphological
in greater detail by using a cellular model to
differences in behaviour between river catch-
Instead of focussing on how external factors
catchment evolution and sediment delivery as
onsidered in a previous paper (Coulthard and
2001), we explore how subtle changes in
morphology and internal instabilities may
to produce a variable erosion/sedimentation
se and thus a variable long-term simulated
tal history as revealed by site alluviation
erosion. For this purpose, the CAESAR model is
to four similar catchments from the Yorkshire
from catchment morphology), climate, and land cover
drivers.
2. Methods
A brief description of the CAESAR model is
provided here, but for more detailed information,
readers are referred to Coulthard et al. (2002).
CAESAR is a cellular model that uses a regular
mesh of grid cells to represent the river catchment
studied. Every cell has properties of elevation, water
discharge and depth, vegetation cover, depth to
bedrock and grain size. The model uses an hourly
Table 1
Characteristics of study catchments
River Swale Ure Nidd Wharfe
Modelled drainage area (km
2
) 383 646 281 697
Max. relief of modelled area (m) 514 564 560 546
Main channel length (km) 62 82 49 97
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241
Dales, northern England to simulate their evolution
over the last 9000 years (Figs. 1?4). To exclude the
influence of other factors, these simulations were
carried out using identical initial conditions (aside
instabilit
charges
model
river
been
mass
1998
(
factors
examine
ments.
affect
c
Macklin,
catchment
combine
respon
environmen
and
applied
224
Fig. 1. The study area, indicating the
rainfall record as the input for a hydrological model
(based upon TOPMODEL, Beven and Kirkby, 1979),
which may be altered to represent the hydrological
effects of different vegetation covers. The output
four catchments modelled.
from the hydrological model is then routed through
the catchment using a scanning multiple flow
algorithm that sweeps across the catchment in four
directions (from north to south, east to west, west to
east, and south to north). In each scan, flow is routed
to the three downslope neighbours (as per Murray
and Paola, 1994), but if the total flow is greater than
the subsurface flow, the excess is treated as surface
runoff and a flow depth is calculated using an
adaptation of Manning?s equation. The maximum
depth calculated for the cells over all four scans is
then recorded. Any flow that is not removed from the
basin remains for the following iteration, allowing
hollows to fill up and also any flow dtrappedT in
meanders remains, enabling complex channel pat-
terns (such as braids and meanders) to be simulated.
For all cells with a flow depth, fluvial erosion and
deposition are calculated using the Einstein Brown
equation (Einstein, 1950). This is applied to 11 grain
size fractions (from 1 to 256 mm) that are integrated
within a series of active layers (Hoey and Ferguson,
1994) that allows surface armouring to develop as
well as a limited stratigraphy. Over the course of
long simulations, this importantly allows previously
deposited finer sediment (as on a floodplain) to
become future erosion sources. Limited slope pro-
cesses are also included, with mass movement when
a critical slope threshold is exceeded, together with
Fig. 3. Catchmen
0
10
20
30
40
50
60
70
80
90
100
0 100 200 300 400 500 600 700
Swale
Ure
Nidd
Wharfe
Elevation (m)
P
e
r
c
entage of 
c
a
t
c
hment 
above elevati
o
n
Fig. 2. Hyposmetric curves for the modelled sections of the Rivers
Swale, Ure, Nidd, and Wharfe.
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241 225
t long profiles.
creep. These allow material from slopes to be
into the fluvial system as well as the input from
es (both large- and small-scale?e.g. bank
After the fluvial erosion/deposition and
process amounts are calculated, the elevations
grain size properties of the cells are updated
. A variable time step is utilised
ting between 10
C06
s and 10
4
s) that restricts
to 10% of the local slope, preventing
tational instability. Therefore, despite being
in operation, CAESAR only requires the
simple inputs of topography (a DEM), an hourly
rainfall record and a land cover record to drive a
sequence of erosion, deposition and landscape
evolution. This is then used to simulate individual
floods, responding to both local hydraulic responses
from runoff events, as well as cumulative inputs (or
deficits) arriving from up-catchment that may them-
selves have been triggered by previous conditions.
To test only the effects of different catchment
morphologies, identical initial and simulation con-
ditions were chosen for four basins. The climate input
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241
soil
fed
landslid
collapse).
slope
and
simultaneously
(opera
erosion
compu
complex
226
Fig. 4. Valley floor morphology and alluvial sediments on the River Swale at Catterick, Yorkshire (after Taylor and Macklin, 1997).
is derived from a combination of two proxy surface
wetness indices derived from peat bogs in Northern
England (Barber et al., 1994; from 6300 cal. BP to
present) and Scotland (Anderson et al., 1998; from
6300 to 9000 cal. BP). These records were combined,
interpolated and re-sampled at 50-year intervals. This
was then normalised to values between 0.75 and 2.25
to create a rainfall or wetness index (Fig. 5). The
model is then driven using a 10-year hourly rainfall
record that is duplicated 5 times to cover 50 years and
then multiplied by the rainfall index, creating a proxy
rainfall record for the last 9000 years. There are few
records of land cover for the regions modelled, so
local palynological records (Tinsley, 1975; Smith,
1986) are used to develop a land cover index ranging
from 2 (forested) to 0.5 (grassland; Fig. 5). This index
is then used to alter a key parameter in the hydro-
logical model that controls the magnitude and
duration of the flood peak.
3. Study sites
Study sites comprise the four major northern and
upland tributaries of the Yorkshire Ouse, U.K. (Fig.
1). The rivers Swale, Ure, Nidd and Wharfe drain
the eastern slope of the Pennine Hills which form a
dissected plateau rising to over 700 m developed on
sandstone, gritstones, and limestone of Carboniferous
age. The area underwent Quaternary glaciation,
leaving valleys which are trough shaped in cross
section, with exposed limestone scars on many
valley sides, but generally without thick glacial
deposits in the uplands. Valley moraines and filled
moraine dammed lakes have been suggested for
several valleys (King, 1960). The four valleys vary
in size (Table 1), but with broadly similar hypso-
metric characteristics (Fig. 2). Their long profiles are
contrasted so that for example, the Ure has a
markedly stepped profile that was largely shaped
1
1.5
2
2.5
4000
Land-Cover
Precipitation
4000
Years 
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241 227
0.5
0 1000 2000 3000
Land cover/precipitation index 
1
10
100
1000
10000
100000
1000000
10000000
100000000
0 1000 2000 3000
Log Sediment discharge (m
3
/50 years) 
Fig. 5. The top frame shows the land cover and rain indices used to drive the
four simulations.
5000 6000 7000 8000 9000
5000 6000 7000 8000 9000
(cal BP)
Swale
Ure
Nidd
Wharfe
model; the lower frame shows the log of sediment yield from the
to the Holocene (Fig. 3). The upland plateau
are dominated by moorland and rough grazing,
peats and stagnohumic and stagnogle soils,
lst pastureland dominates the flatter, terraced
floors.
Fig. 4 shows valley floor morphology and the
of valley fill sediments for a studied site near
Yorkshire (Taylor and Macklin, 1997). This
evidence for episodes of incision and alluvia-
in a series of progressively incising river terraces.
details of the catchments and their hydrol-
are available in Jarvie et al. (1997) and Law et al.
. The nature of valley floor morphology and
is broadly representative of the alluvial
modelled in this paper.
The known, site-specific history of coarse sedi-
accumulation, erosion, and transfer in these
upland valleys strongly suggests that inter- and
intra-catchment modelling using CAESAR would be
appropriate as a means of assessing likely differ-
ential response within and between catchments.
During the Holocene they have responded to similar
environmental changes affecting sediment transfers
within bedrock valleys which have been inherited
from prior Quaternary (or earlier) conditions. The
catchments are also upstream of the influence of
Holocene sea level changes which have affected
sedimentation in the Ouse basin. The precise
morphology of the river catchments at the beginning
of the simulation (9000 cal. BP) is unknown, and to
reconstruct this would involve considerable and
largely speculative work. Therefore, to allow a
systematic approach to be taken for all four basins,
the present day surface is taken as an analogue.
0.5
1
1.5
2
2.5
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Land cover/precipitation index
Land-Cover
Precipitation
100
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241
0
10
20
30
40
50
60
70
80
90
0 1000 2000 3000 4000
Years (cal BP)
Percentage sediment yield
prior
areas
with
whi
valley
nature
Catterick,
shows
tion
Additional
ogy
(1997)
sediment
systems
ment
228
Fig. 6. The cumulative percentage of sediment yield from the four simulated
in the driving climate record and significant changes in sediment discharge.
5000 6000 7000 8000 9000
Swale
Ure
Nidd
Wharfe
catchments. Solid arrows indicate a linkage between wetness peaks
Dashed arrows indicate a weak/nonexistent linkage.
This assumption is not wholly unreasonable,
because following substantial valley floor incision
into gravels over the Late Glacial?early Holocene
transition, there is evidence that river bed elevation
in the Yorkshire Dales has only varied by F3m
during the Holocene (Macklin et al., 2000). The
model surface is therefore derived from a 50-m
resolution present day DEM, with the bedrock
surface defined 3 m below. This provides a uniform
3-m layer of sediment and soil capped with a turf
mat. Where stream powers are great enough, this
allows channels to incise to bedrock and become
supply limited. This parallels conditions found in
the field, where earlier in the Holocene there was a
plentiful supply of post glaciation sediment but later
incision to bedrock has limited this supply (Merrett
and Macklin, 1999). During the simulations, con-
tinuous sediment discharge data were recorded, and
catchment elevations and grain-size data saved at
50-year intervals. By subtracting elevations from
consecutive 50-year intervals, sediment budgets for
the four catchments were calculated (Fig. 5).
4. Results
The modelled sediment discharges (Fig. 5) indi-
cate that all four catchments behave in a broadly
similar manner, with several short periods of high
sediment discharges lasting 50?200 years. These
peaks are synchronous with wetter climate spells,
indicating that climate is the main driver of simulated
sediment yields. However, after 2000 cal. BP, the
peaks in sediment discharge are noticeably larger for
similar-size climate peaks?showing how simulated
deforestation can increase sediment discharge for a
given storm event. This can also be clearly seen in
Fig. 6, where there are much smaller jumps in
cumulative sediment discharge to the climate peaks
at 3200 cal. BP and 2500?2750 cal. BP (under
forested cover) than to those at 1800 and 1000 cal.
BP (deforested). Indeed, the Ure records a 3%
increase from 4000 to 1900 cal. BP but a 25%
increase from 1900 to 400 cal. BP after tree
clearance. This strongly suggests that the removal
of tree cover influences catchment response to
Years cal BP
2
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241 229
500000
400000
300000
200000
100000
Erosion/deposition (M
3
 per 50 years)
0
-100000
-200000
0 1000 2000 3000 4000
1
Fig. 7. Simulated cumulative sediment yields
reach 1
5000 6000 7000 8000 9000
reach 2
Erosion
Deposition
reach 3r reach 4
reach 5r reach 6
reach 7r reach 8
3
4
5
6
7
8
for the 8 reaches of the River Swale.
change by increasing the amplitude of the
peak for a given storm event.
Whilst the timing of the sediment peaks is
onous across all four catchments, Figs. 5 and
show that there are notable differences in the size of
peaks between basins. The smaller Rivers Swale
Nidd both have a far dflashierT, spiky Holocene
discharge compared to the larger Wharfe and
basins (Fig. 5), which have less variable sediment
As expected, the larger basins produce greater
of sediment output, but to compare the
mics of sediment delivery, i.e. the relative
e of their response, the totals are plotted as
ve percentages (Fig. 6). This normalises the
yield for all four basins to the same scale and
emonstrates significant differences in catchment
se. From 4200 to 400 cal. BP, the Nidd delivers
of its total sediment load, but during the same
the far larger Ure only 28%. Interestingly, the
Wharfe and Swale have very similar shaped
, despite the Wharfe draining nearly twice the
This firmly indicates that whilst total simulated
sediment yield may reflect the catchment area, the
timing and dynamics of sediment delivery are not
controlled by size alone.
To investigate how sediment was moving within
the catchments, the four study basins were divided
into reaches approximately 5 km long. The positions
of these reaches were carefully selected so inter-reach
boundaries did not span major tributaries and, where
possible, wider sections of valley floor were captured
within a single reach. CAESAR saves the elevations
of all cells at 50-year intervals and, by comparing
elevations within reaches, simulated Holocene sedi-
ment budgets can be calculated. We have plotted
modelled sediment volumes accumulated or removed
on a cumulative basis for individual reaches for all
four catchments over the Holocene (Figs. 7?13).
Results for the Swale (Figs. 7?10) show consid-
erable variations between reaches, although most
respond to the major climate peaks, similar to the
catchment response as a whole (Figs. 5 and 6). This
provides clear evidence of reaches being controlled by
erosional thresholds, with most erosion occurring
Years cal BP
2
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241
500000
400000
300000
200000
100000
Cumulative erosion/deposition (M
3
)
0
-100000
-200000
0 1000 2000 3000 4000
1
climate
sediment
synchr
6
the
and
sediment
Ure
yields.
totals
dyna
magnitud
cumulati
sediment
d
respon
50%
period
Rivers
curves
area.
230
Fig. 8. Simulated cumulative sediment yields for
reach 1
reach 2
reach 3
5000 6000 7000 8000 9000
Erosion
Deposition
3
4
5
6
7
8
the top three reaches of the River Swale.
during short peaks triggered by wetter episodes of
climate, contrasting with little activity in intervening
periods. Viewed together, they show that the valley
floor has experienced progressive sediment loss
during the Holocene.
Fig. 8 shows sediment deposition/erosion patterns
in the top three reaches. Steady deposition is caused by
input from the valley walls adding material to the
edges of the floodplain. These three reaches are
geomorphologically relatively simple, as reach 1
drains headwaters of the Swale, and reaches 2 and 3
only have one major tributary. Reach 1 shows
considerable erosion at 4200 cal. BP, then steady
accumulation until 400 cal. BP. This may be caused by
material from upstream being transported clean
through the reach, and there is only significant removal
of valley floor material at 4200 and 400 cal. BP., when
the floods are of significant magnitude to breach local
erosional thresholds. Reaches 2 and 3 respond to more
major climate episodes than reach 1, but reach 2 also
responds at 2700 cal. BP unlike reach 3 where the
response may be dampened by the volumes of sedi-
ment arriving from reach 2 upstream. Indeed, reach 3
has a smaller response to the 1800 and 1000 cal. BP
peaks which again may be caused by the input of
material from upstream. It is interesting to note that
reach 1 has a tendency to accumulate sediment for
most of the Holocene, whereas reaches 2 and 3 pass
into a state of net removal after c. 2000 cal. BP.
In the middle two modelled reaches of the Swale
(reaches 4 and 5, Fig. 9), the situation is more
complex. Not only are reach responses triggered by
local erosional thresholds, but these reaches are also
receiving significant volumes of sediment from
upstream, as well as from the major tributaries of
Arkle Beck (upstream and north of reach 4) and
Marske Beck (upstream and north west of reach 5).
Reaches 4 and 5 behave similarly to 2 and 3 at 4200
and 1800 cal. BP, with substantial volumes of sedi-
ment removed in brief periods. At 800?1000 cal. BP
both reaches erode, but 50 years later, reach 4
aggrades, apparently receiving a delayed dslugT of
sediment (cf. Macklin and Lewin, 1989; Nicholas et
al., 1995) from the tributary Arkle Beck. Overall two
Years cal BP
2
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241 231
500000
400000
300000
200000
100000
0
-100000
-200000
0 1000 2000 3000 4000
Cumulative erosion/deposition (M
3
)
1
Fig. 9. Simulated cumulative sediment yields for
5000 6000 7000 8000 9000
Erosion
Deposition
reach 4
reach 5
3
4
5
6
7
8
the middle two reaches of the River Swale.
,
of slow sediment accumulation are indicated
4000 cal. BP and c. 2000 to 4000 cal. BP)
ted by a major removal event. After c. 2000 cal.
, more complexity of response is indicated.
Fig. 10 shows the lower three reaches located in
piedmont (reaches 6, 7, and 8) to lowland
and unlike the upper reaches, they receive
or no sediment from tributaries. Reach 6, at the
gin of the Pennine uplands, appears especially
ve to climate change, responding to all the
periods, including one at 6000 cal. BP. that no
reach shows. This may be due to the limited
storage capacity within the reach and/or it
relate to high local stream powers at a point in
catchment where the product of discharge and
is maximised. This piedmont reach may be the
interesting from a geomorphological perspective
it is the best drecorderT of climate changes. Reach 7
ds most weakly to the 4200 cal. BP peak, yet
at 1800 cal. BP and it is the only reach not to
major change during the Little Ice Age (LIA, c.
cal. BP). Possibly, this is caused by lower stream
powers or by receiving significant inputs from
upstream. The valley sides adjacent to reach 8 are of
low relief and consequently receive little or no
material from creep. However, it is far more active
than reach 7 and responds to all the main peaks,
including those at 2800 cal. BP. Interestingly, there is
a 100 year lag behind reach 6 for this peak?again
possibly reflecting a slug of sediment passing through
from upstream. Reach 8 responds to erosion loss
events in a similar manner to reaches 6 and 7, but it
shows little sign of accumulating sediment between
events. This reach thus appears to respond to
Holocene events as an erosion or transport zone
without dfillT periods, as what occurs further upstream.
Examining the other catchments, the Nidd (Fig.
11) shows a similar response to the Swale, with
most reaches reacting at the same time but with a
different magnitude of response. There is also
evidence of a lagged response at c. 1700 cal. BP
(reach 5) that may be caused by the influx of
sediment from upstream reaches. However, com-
pared to the Swale, the response to the LIA climate
Years cal BP
2
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241
500000
400000
300000
200000
100000
0
-100000
-200000
0
1000 2000 3000 4000
Cumulative erosion/deposition (M
3
)
1
periods
(before
separa
BP
the
transition,
little
mar
sensiti
wetter
other
sediment
could
the
slope
most
as
respon
strongly
show
400
232
Fig. 10. Simulated cumulative sediment yields for
5000 6000 7000 8000 9000
Erosion
Deposition
reach 6
reach 7
reach 8
3
4
5
6
7
8
the lower three reaches of the River Swale.
-1300000
-1100000
-900000
-700000
-500000
-300000
-100000
100000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Years cal. BP
Cumulative erosion/deposition (m
3
)
reach 4
reach 5
reach 6
reach 7
-1300000
-1100000
-900000
-700000
-500000
-300000
-100000
100000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Cumulative erosion/deposition (m
3
)
reach 1
reach 2
reach 3
1
2
3
4
5
6
7
20km
Years cal. BP
Fig. 11. Simulated cumulative reach sediment yields for the River Nidd.
T
.J.
Coulthar
d
e
t
al.
/
Geomorphology
69
(2005)
222?241
233
-500000
-400000
-300000
-200000
-100000
0
100000
reach 1
reach 2
reach 3
reach 4
-1500000
-1300000
-1100000
-900000
-700000
-500000
-300000
-100000
100000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Years cal. BP
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Years cal. BP
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Years cal. BP
Cumulative erosion/deposition (m
3
)
reach 9
reach 10
reach 11
-1000000
-800000
-600000
-400000
-200000
0
Cumulative erosion/deposition (m
3
)
reach 5 
reach 6
reach 7
reach 8
1
2
3
4
5
6
7
8
9
10
11
20km
Cumulative erosion/deposition (m
3
)
Fig. 12. Simulated cumulative reach sediment yields for the River Ure.
T
.J.
Coulthar
d
e
t
al.
/
Geomorphol
ogy
69
(2005)
222?241
234
100000
200000
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
Years cal. BP
00
0
(
m
3
)
-500000
-400000
-300000
-200000
-100000
0
100000
200000
Cumulative erosion/deposition (m
3
)
-1000000
Reach 13
1
2
3
4
5
6
7
8
9
10
11
12
13
20km
Fig. 13. Simulated cumulative reach sediment yields for the River Wharfe.
T
.J.
Coulthar
d
e
t
al.
/
Geomorphology
69
(2005)
222?241
235
erioration is far smaller. There is also little
of net sediment accumulation during
between erosion events.
The upper reaches of the two largest modelled
s, the Ure and Wharfe (Figs. 12 and 13),
in a broadly similar manner to the Swale and
However, reaches 7, 8, and 11 in the Ure, and
11 and especially 12 and 13 in the Wharfe all
rd several periods of substantial deposition
5500 and 1000 cal. BP. In contrast to other
, these are storing the sediment brought down
upstream and then exporting it at a later date. On
Wharfe, they are all low-relief shallow gradient
T.J. Coulthard et al. / Geomorphol
reaches, and on the Ure in low gradient sub-basins?
both ideal depositional environments. Some of these
phases of deposition are significant, with in excess of
1000000 m
3
of sediment deposited from c.1800 to
1600 cal. BP on reach 7 in the Ure. The upper reaches
of both the Ure and Wharfe show net accumulation of
sediment, the Ure until c. 4200 cal. BP, and the
Wharfe for almost the whole of the Holocene. The
lower reaches of the Wharfe show a sensitivity to
early and late Holocene climate change. Reach 13 (the
most downstream) appears to record environment
fluctuations in a highly sensitive manner with marked
depositional spikes followed by slow sediment
removal.
When comparing the reaches in all four catch-
ments, three styles of reach response become appa-
rent, as shown in Fig. 14.
1. Rapid erosion, sometimes followed by gradual
accumulation over many hundreds of years. This is
Time (500 years)
1
2a
3
2b
Magnitude
det
evidence
periods
catchment
behave
Nidd.
10,
reco
between
reaches
from
the
236
Fig. 14. Different patterns of erosion and deposition found in the
simulations.
typical of most of the upland reaches (e.g. Swale 1,
Wharfe 2). Successive erosion lows may be higher
(Wharfe 2) or lower (Wharfe 9) than previous ones,
depending on the degree of depletion in the erosion
phase and the degree of build up between such
phases.
2. A short sedimentation spike, material from which
is then rapidly eroded. This appears to represent
rapid deposition following large amounts of
erosion from upstream (e.g. Ure 7). A more
common variant involves no sedimentation spike
(2b) but merely erosion events lowering storage
from one plateau to another (e.g. Swale most
reaches, Nidd 4). These plateaus follow the gradual
accumulation periods of style 1.
3. A rapid sedimentation spike which is then gradu-
ally depleted in a succession of episodes over
several hundred years. This is found in the lower
reaches of the Wharfe (13) where the volumes are
sufficient to cause floodplain accretion of c. 0.5m.
In some reaches (Wharfe 10?13, Ure 7?8) the
response is much dspikierT than above.
5. Discussion
Our detailed analysis of simulated reach reactions
to environmental change reveals a response that is
complex both within and between catchments.
Although there are similarities between timings of
erosion and deposition phases (largely corresponding
to climate changes), there are considerable differences
in the magnitude of river response at both the basin
and reach scale. Locally, the magnitude of response is
probably controlled by the local discharge, sediment
availability and the channel and valley morphology of
the reach (e.g. gradient and valley floor width). But
the magnitude of geomorphic response to climate
change also varies downstream, with upstream rea-
ches showing a smaller response than downstream
ones, which may be because stream powers, sediment
availability, and sediment throughput are greater in the
lower parts of the modelled river basins. Relationships
between available sediment sizes and local stream
power are likely to be important as much as their
absolute magnitudes, whilst the available volumes of
ogy 69 (2005) 222?241
alluvial material must constrain potential response to,
and recording of sediment-transfer events. Addition-
ally, reaches 8 and 11 on the Ure show a marked
increase in erosion and sediment delivery after 2000
cal. BP. This follows simulated catchment deforesta-
tion and shows how certain reaches may be especially
sensitive to changes in land use. Interestingly, this
reveals that only a small length of the River Ure (c. 10
km) may be responsible for generating over 60% of
the total catchment sediment discharge from 2000 cal.
BP to present (reaches 9?11). This has important
implications for forecasting how rivers may respond
to future climate changes, particularly if relatively
small regions have the potential to generate large
volumes of sediment.
However, it is important to appreciate that reach
response cannot be predicted solely on the basis of
internal characteristics, as they are also receiving
substantial volumes of sediment from upstream
reaches. Delivery of this sediment may be nearly
instantaneous in some reaches, but in other reaches
delayed, which may be attributable to the passage of
sediment dslugsT or waves (e.g. reaches 4 and 8 on the
Swale). Furthermore, over the duration of the Hol-
ocene, most reaches were static or slowly changing,
but were punctuated by infrequent often short-lived
periods of erosion, together with smaller spikes of
deposition in some reaches.
It is perhaps that for the modelled rivers, the net
tendency is for reaches to lose/yield sediment during
Holocene, especially during relatively short episodes.
Valley floor morphology in such circumstances should
be characterised by incision episodes separated by
extended periods of non-deposition or strath forma-
tion where rivers are laterally mobile and cut across
earlier deposits. A staircase of largely erosional
terraces (with a veneer only of lateral accretion
deposits) could be anticipated, with a greater degree
of incision as rivers occupy progressively lower levels
of entrenchment in prior sediments. This may be
contrasted with the picture elsewhere, in, for example,
more lowland environments, where dfillT events are
more dominant in alluvial sequences.
The differential reach response indicates that some
reaches are more sensitive to environmental change
than others. This means also that some areas will be
better recorders of environmental change, but impor-
tantly, this may not necessarily be related to gross
T.J. Coulthard et al. / Geomorphology
volumes of catchment sediment output. All the
catchments and reaches are responsive to the LIA
episode, but earlier episodes are variably recorded
(e.g. compare reaches 1 and 2 on the River Swale).
Erosional loss commonly wipes out any tendency to
gain sediment overall?although in practice, sed-
imentation niches may remain at the margins of
floodplain environments, which could record short
periods of sedimentation. It should be noted that the
reach values simulated here represent an average of
that reach, and it is likely that within some of these
reaches depositional zones will exist. However,
sedimentation peaks are found in a number of middle
and lowland reaches (Ure reaches 4, 7, and 8; Nidd
reach 3; Wharfe reaches 10?13). Therefore, the
reaches most sensitive to precipitation variations
seem to be located in the mid or lower catchment
(piedmont to lowland) environments, with intriguing
early Holocene simulations on the lower Wharfe
providing a highly complex sedimentation signature
(see Fig. 13).
To this extent, field geomorphologists may find
simulations using CAESAR useful for determining
which areas of a river basin are most worthy of
investigation in terms of the environmental sensitivity
that alluvial deposits can provide. However, this study
is a clear example of the caution necessary in basing
geomorphic interpretation on only one or a few sites
within a river basin (cf. Macklin et al., 1992). This
study simulates changes in response to climate and
land cover uniformly over the whole catchment?yet
produces significantly different response in individual
reaches solely as a result of internal controls. In light
of this, the confidence of interpretations of previous
studies of basin evolution based on one or two sites
must be questioned, especially considering the greater
degrees of freedom for variation within a real river
basin.
These results also have important implications for
modelling fluvial system change. As there is high
variability between catchments, as well as differences
in the style and degree of response between reaches,
results cannot simply be copied from one catchment to
another. To include the high levels of conditioning
morphological variability, river basins will have to be
modelled individually, and the whole catchment
simulated, to effectively incorporate downstream
sediment routing and storage.
69 (2005) 222?241 237
The high levels of inter-catchment and inter-reach
variability simulated here have been induced with
and human interactions. This raises the question as to
whether variations in all these parameters have to be
integrated in order to accurately model reach and
catchment response. This is presently impossible, so a
more prudent question might be which processes are
most important, and based on that, which do we need
to integrate and which are irrelevant?
When modelling large river systems over long
timescales compromises have to be made, usually to
aid simplicity and thus computational efficiency.
CAESAR has several limitations that could influence
the results presented here. In this application the grid
cell size is relatively coarse (50 m) and, as previously
mentioned, precipitation changes are implemented
uniformly over the entire catchment. However, by
comparing results from identical simulations (except
topography) we can show that intra-catchment climate
or land cover variations are not required to provide
this differential response. It could be argued that
variable responses are simply a facet of the model?
indeed they must be! But we would argue that much
of the behaviour shown by these simulations is
consistent with field evidence in so far as it is
available.
6. Validation
Validating these model runs is difficult as the
model generates output topographies every 50 years
and continuous hourly water and sediment discharge
data for 9000 years of simulation. There are no field
data sets that are comparable and it is not possible to
track spatial aspects of sediment transfer with
contemporary data. Previous long-term large-scale
studies have compared model simulations to catch-
relatively well constrained driving parameters. For
example, we have used simplified and uniform inter-
(climate) and intra-catchment (morphology) initial
conditions and drivers. Actual river basins have many
more drivers and far greater degrees of freedom for
variability. For example, rainfall is not spatially
uniform, as was the case in these simulations; it is
highly spatially and temporally variable. In reality,
further examples of spatial variability can be found in
vegetation cover and patterns, bedrock distributions,
T.J. Coulthard et al. / Geomorphol238
ment hypsometric curves (Willgoose et al., 1991) and
long profiles (Willgoose et al., 1991; Tucker and
Slingerland, 1994). More recently, Hancock et al.
(2002) have parameterised slope process and sediment
transport rates to hillslope plot experiments. However,
here we are modelling river systems where the overall
form and long profile has remained essentially similar
since the Late Pleistocene. Therefore, we need an
alternative methodology for validation. In a previous
paper (Coulthard and Macklin, 2001), we compared
the results of modelled sediment yields from the River
Swale to a histogram of
14
C dated alluvial units from
all of Great Britain. This showed a good correlation
between the number, magnitude, and approximate
timing of sediment peaks generated by the model. In
that paper, we argued that we were using the River
Swale as an exemplar for river system response to
Holocene environmental change in GB. The reasoning
behind this argument was that there were insufficient
14
C dates within the Swale to provide a suitable
regional validation. In this paper we can now make a
further step forward.
As a result of a series of studies undertaken as
part of the UK Natural Environment Research
Council?s Land Ocean Interaction Study (LOIS)
community research programme between 1994 and
2000 in the Yorkshire Ouse basin (Taylor and
Macklin, 1997; Hudson-Edwards et al., 1999;
Howard et al., 2000; Macklin et al., 2000), the
chronology of Holocene river development in the
region is now underpinned by 60
14
C dates. This
includes several sites on the Swale, Ure, Nidd and
Wharfe at which the sequences of fluvial incision
and fill have been established (as in Fig. 4 above). In
Fig. 15, following the methodology recently set out
in Macklin and Lewin (2003), we have plotted the
cumulative of
14
C dates (16 in total) that mark
modification in sedimentation style or rate, allowing
the timing of geomorphologically significant changes
in river activity to be picked out. These correspond
well with the changes in the cumulative simulated
sediment yields from all four catchments evident at
4200, c. 3000, 1800, and 1000 cal. BP. Of equal
interest, however, are phases where the Holocene
alluvial record indicates lower rates of geomorphic
activity that match with periods of low sediment
discharge from the model. This importantly indicates
that the model is not only simulating the high
ogy 69 (2005) 222?241
sediment yield events, but also periods of quiescence.
Nevertheless, despite these new
14
C dates, more are
7. Conclusion
5000
for
This study has modelled the response of four
catchments to spatially generalised Holocene fluctua-
still required to fully validate the performance of the
model.
0
10
20
30
40
50
60
70
80
90
100
0 1000 2000 3000 4000
Years (cal BP)
Percentage sediment yield
Fig. 15. Cumulative percentage sediment yields
T.J. Coulthard et al. / Geomorphology
tions in precipitation and land cover. A whole-catch-
ment approach has allowed sediment volumes to be
tracked downstream, including an analysis of the
differential response of individual river reaches down-
valley. Whilst there is some evidence that simulated
activity bears comparison with the actual record of
Holocene river erosion and alluviation, it should be
appreciated that at this stage modelling involves some
simplification both of the environmental record and
the likely processes of sediment transfer involved.
Further field evidence is needed for some of the
simulated responses that CAESAR produces. Never-
theless, the exercise has important implications for
both interpretation of the fluvial sedimentary record
and for the forecasting of future impacts of environ-
mental change:
1. River catchments process environmental changes
in relatively complex ways, and simulations of
continuous fluctuations in climate and land-use
appear to produce bursts of sediment transfer
activity (rapid removal together with some accu-
mulation dspikesT) in relation to transport thresh-
olds. At other times, there may be slow removal or
slow accumulation of sediment in different reaches.
2. The four catchments modelled responded to envi-
ronmental change to different degrees, not only
with respect to overall sediment discharge but also
in the variable scale of erosion and/or alluviation
6000 7000 8000 9000
0
2
4
6
8
10
12
14
16
18
Frequency of 
14
C dated units
Swale
Ure
Nidd
Wharfe
Cumulative change
dates
all 4 catchments with the
14
C change dates.
69 (2005) 222?241 239
during particular periods of the Holocene.
3. Within a single catchment, reach sensitivity to
environmental change varies considerably. Some
periods are recorded only in some reaches, whilst
higher potential sensitivity appears to occur in the
piedmont area of the catchment analysed here. It
appears that inter-reach transfers of sediment may
have a considerable effect on delaying or even
blanketing out-phased alluvial response, whilst
erosion/sedimentation thresholds may be crossed
significantly only in certain reaches.
In the future it will clearly be possible to improve
model sophistication as environmental records are
augmented and possibly differentiated within catch-
ments, and as process modelling is improved. At this
stage, however, the conclusions above seem robust,
and the implications that need to be considered are
clear enough. Catchment morphology in relation to
sediment transfer processes and thresholds does make
environmental responses complex and differential
Coulthard, T.J., Macklin, M.G., 2001. How sensitive are river
Germanoski, D., Harvey, M.D., 1993. Asynchronous terrace
ogy 69 (2005) 222?241
development in degrading braided channels. Physical Geogra-
phy 14, 16?38.
Hancock, G.R., Willgoose, G.R., Evans, K.G., 2002. Testing of the
SIBERIA landscape evolution model using the Tin Camp Creek,
Northern Territory, Australia, field catchment. Earth Surface
systems to climate and land-use changes? A model-based
evaluation. Journal of Quaternary Science 16, 347?351.
Coulthard, T.J., Kirkby, M.J., Macklin, M.G., 1998. Non-linearity
and spatial resolution in a cellular automaton model of a
small upland basin. Hydrology and Earth System Sciences 2,
257?264.
Coulthard, T.J., Macklin, M.G., Kirkby, M.J., 2002. A cellular
model of Holocene upland river basin and alluvial fan evolution.
Earth Surface Processes and Landforms 27, 269?288.
Einstein, H.A., 1950. The bed-load function for sediment transport
on open channel flows. Tech. Bull. No. 1026, USDA, Soil
Conservation Service, 71.
according to the individual catchment or reach being
considered. This has to be considered when attempt-
ing to predict the effects of environmental change.
Acknowledgements
We are most grateful to NERC for supporting
investigations in the Yorkshire Ouse catchment
through a recent grant (GST/02/0758) to MGM. We
also wish to thank Jonathon Phillips and Trever Hoey
for their insightful comments on the manuscript. The
CAESAR model may be downloaded from TJC?s
website http://www.coulthard.org.uk.
References
Anderson, D.E., Binney, H.A., Smith, M.A., 1998. Evidence for
abrupt climatic change in northern Scotland between 3900 and
3500 calendar years BP. The Holocene 8, 97?103.
Barber, K.E., Chambers, F.M., Maddy, D., Stoneman, R., Brew,
J.S., 1994. A sensitive high resolution record of late Holocene
climatic change from a raised bog in northern England. The
Holocene 4, 198?205.
Beven, K.J., Kirkby, M.J., 1979. A physically based variable
contributing-area model of catchment hydrology. Hydrological
Sciences Bulletin 24, 43?69.
Boison, P.J., Patton, P.C., 1985. Sediment storage and terrace
formation in coyote gulch basin, South-Central Utah. Geology
13, 31?34.
Braun, J., Sambridge, M., 1997. Modelling landscape evolution on
geological time scales: a new method based on irregular spatial
discretization. Basin Research 9, 27?52.
T.J. Coulthard et al. / Geomorphol240
Processes and Landforms 27, 125?143.
Hoey,T.B.,Ferguson,R.,1994.Numericalsimulationofdownstream
fining by selective transport in gravel bed rivers: model develop-
mentandillustration.WaterResourcesResearch30,2251?2260.
Howard, A.D., 1994. A detachment-limited model of drainage basin
evolution. Water Resources Research 30, 2261?2285.
Howard, A.J., Macklin, M.G., Black, S., Hudson-Edwards, K.A.,
2000. Holocene river development and environmental change in
Upper Wharfedale, Yorkshire Dales, England. Journal of
Quaternary Science 15, 239?252.
Hudson-Edwards, K.A., Macklin, M.G., Taylor, M.P., 1999. 2000
years of sediment-borne heavy metal storage in the Yorkshire
Ouse basin, NE England, UK. Hydrological Processes 13,
1087?1102.
Jarvie, H.P., Neal, C., Robson, A.J., 1997. The geography of the
Humber catchment. Science of the Total Environment 194/195,
87?99.
King, C.A.M., 1960. The Yorkshire Dales Geography. Sheffield,
24 pp.
Law, M., Wass, P., Grimshaw, D., 1997. The hydrology of the
Humber catchment. Science of the Total Environment 194/195,
119?128.
Macklin, M.G., Lewin, J., 1989. Sediment transfer and trans-
formation of an alluvial valley floor: the River South Tyne,
Northumbria, UK. Earth Surface Processes and Landforms 14,
233?246.
Macklin, M.G., Lewin, J., 2003. River sediments, great floods and
centennial scale Holocene climate change. Journal of Quater-
nary Science 18, 101?105.
Macklin, M.G., Rumsby, B.T., Passmore, D.G., 1992. Climate and
cultural signals in Holocene alluvial sequences: the Tyne basin,
Northern England. In: Needham, S., Macklin, M.G. (Eds.),
Alluvial Archaeology in Britain, Oxbow Monograph, vol. 27.
Oxbow Press, Oxford, pp. 123?139.
Macklin, M.G., Taylor, M.P., Hudson-Edwards, K.A., Howard, A.J.,
2000. Holocene environmental change in the Yorkshire Ouse
basin and its influence on river dynamics and sediment fluxes on
the coastal zone. In: Shennan, I., Andrews, J. (Eds.), Holocene
Land?Ocean Interaction and Environmental Change Around the
North Sea. Special Publication, vol. 166. The Geological
Society, London, pp. 87?96.
Merrett, S.P., Macklin, M.G., 1999. Historic river response to
extreme flooding in the Yorkshire Dales, Northern England. In:
Brown, A.G., Quine, T.M. (Eds.), Fluvial Processes and
Environmental Change. Wiley, Chichester, pp. 345?361.
Murray, A.B., Paola, C., 1994. A cellular model of braided rivers.
Nature 371, 54?57.
Nicholas, A.P., Ashworth, P.J., Kirkby, M.J., Macklin, M.G.,
Murray, T., 1995. Sediment dslugsT: large-scale fluctuations in
fluvial transport rates and storage volumes. Progress in Physical
Geography 19, 500?519.
Paola, C., 2003. Floods of record. Nature 425, 459.
Rains, B., Welch, J., 1988. Out-of-phase Holocene terraces in part
of the North Saskatchewan river basin, Alberta. Canadian
Journal of Earth Sciences 25, 454?464.
Richards, K.S., 2002. Drainage basin structure, sediment delivery
and the response to environmental change. In: Jones, S.J.,
Frostick, L.E. (Eds.), Sediment Flux to Basins: Causes, Controls
and Consequences. Special Publications, vol. 191. Geological
Society, London, pp. 149?160.
Schumm, S.A., 1979. Geomorphic thresholds, the concept and its
application. Transactions-Institute of British Geographers 4,
485?515.
Smith, R.T., 1986. Aspects of the soil and vegetation history
of the Craven District of Yorkshire. In: Manby, T.G.,
Turnbull, P. (Eds.), Archaeology in the Pennines, B.A.R. British
Series vol. 158. British Archaeological Research, Oxford,
pp. 3?28.
Taylor, M.P., Macklin, M.G., 1997. Holocene alluvial sedimentation
and valley floor development: the River Swale, Catterick, North
Yorkshire, UK. Proceedings of the Yorkshire Geological Society
51, 317?327.
Tebbens, L.A., Veldkamp, A., 2001. Exploring the possibilities and
limitations of modelling Quaternary fluvial dynamics: a case
study of the River Meuse. In: Maddy, D., Woodward, J.,
Macklin, M.G. (Eds.), Archives of Fluvial Change. Balkema,
Rotterdam, pp. 469?484.
Tinsley, H.M., 1975. The former woodland of the Nidderdale Moors
(Yorkshire) and the role of early man in its decline. Journal of
Ecology 6, 1?26.
Tucker, G.E., Slingerland, R.L., 1994. Erosional dynamics, flexural
isostasy, and long-lived escarpments: a numerical modelling
study. Journal of Geophysical Research 99, 12229?12243.
Tucker, G.E., Slingerland, R., 1997. Drainage basin responses to
climate change. Water Resources Research 31, 2047?3021.
Tucker, G.E., Lancaster, S.T., Gasparini, N.M., Bras, R.L., Rybarc-
zyk, S.M., 2001. An object-oriented framework for distributed
hydrologic and geomorphic modeling using triangulated irreg-
ular networks. Computers & Geosciences 27, 959?973.
Willgoose, G.R., Bras, R.L., Rodriguez-Iturbe, I., 1991. A coupled
channel network growth and hillslope evolution model: 1.
Theory. Water Resources Research 27, 1671?1684.
T.J. Coulthard et al. / Geomorphology 69 (2005) 222?241 241