Journal of Hydrology, 42 (1979) 39--62 39

© Elsevier Scientific Publishing Company, Amsterdam -- Printed in The Netherlands

[1]

AN ANALYSIS OF THE PROCESSES OF RIVER BANK EROSION

J.M. HOOKE
Geography Division, Manchester Polytechnic, Manchester M1 5GD (Great Britain)
(Received July 15, 1978; revised and accepted December 10, 1978)

ABSTRACT

Hooke, J.M., 1979. An analysis of the processes of river bank erosion. J. Hydrol., 42:
39--62.

Measurements of erosion have been made on sections of river banks in Devon, England
over a period of 2.5 yr. Erosion pins were used and found to be satisfactory. Data of the
hydrological and meteorological conditions during each erosion period were collected;
these were incorporated into a stepwise multiple regression analysis of the conditions
controlling the amount and distribution of river bank erosion. Field observations are
shown to aid considerably in understanding the processes of removal of material. Two
main methods of bank erosion are identified, corrasion and slumping, and these appear to
be associated with the influence of river flow levels and antecedent precipitation condi-
tions, respectively. Antecedent precipitation index emerges as the variable providing
highest explanation of the erosion characteristics but the diversity of results from
different sites illustrates the complex combinations of conditions related to erosion
events. A tentative grouping of the sites according to the types and amounts of erosion
and the variables explaining their distribution is suggested.

INTRODUCTION

River b a n k e r o s i o n is i m p o r t a n t g e o m o r p h o l o g i c a l l y in e f f e c t i n g c h a n g e s
in t h e river c h a n n e l c o u r s e a n d in d e v e l o p m e n t o f t h e f l o o d plain. T h e s e are
amongst the most dynamic elements of the landscape and thus an under-
s t a n d i n g o f t h e p r o c e s s e s is f u n d a m e n t a l t o o u r e x p l a n a t i o n o f t h e d e v e l o p -
m e n t o f fluvial features. R i v e r e r o s i o n is also i m p o r t a n t e c o n o m i c a l l y d u e t o
t h e loss o f f a r m l a n d a n d t h e u n d e r m i n i n g o f s t r u c t u r e s a d j a c e n t t o t h e river
c h a n n e l ; this p r o b l e m is o f t e n u n d e r e s t i m a t e d in Britain. R e l a t i v e l y f e w
d e t a i l e d studies o f t h e p r o c e s s e s o f b a n k e r o s i o n h a v e b e e n m a d e u n t i l
r e c e n t l y , b u t t h e s e h a v e s h o w n t h e c o n s i d e r a b l e a c t i v i t y o f rivers i n c l u d i n g
several in B r i t a i n ( W o l m a n , 1 9 5 9 ; T w i d a l e , 1 9 6 4 ; W a l k e r a n d A r n b o r g , 1 9 6 6 ;
L e w i n , 1 9 7 2 ; Hill, 1 9 7 3 ; K n i g h t o n , 1 9 7 3 ; M o s l e y , 1 9 7 5 ) .
T h e a i m o f this s t u d y is t o e l u c i d a t e t h e c o n d i t i o n s u n d e r w h i c h e r o s i o n
t a k e s p l a c e a n d t h e f r e q u e n c y o f t h e s e events, a n d t o investigate t h e distri-
b u t i o n a n d v a r i a t i o n in t y p e s o f b a n k e r o s i o n . F o r this p u r p o s e a n u m b e r o f
s e c t i o n s o f river b a n k s in s o u t h e a s t D e v o n w e r e i n s t r u m e n t e d a n d d e t a i l e d
40

measurements o f removal of material were made over the period December

1973 to June 1976.
River bank erosion is only one process in the total channel system in
which erosion is closely linked to other processes such as sediment trans-
port and deposition; the implications o f the results presented here for other
processes should be borne in mind although the study could not encompass
measurement of them.

METHOD OF EROSION MEASUREMENT

The m e t h o d used for monitoring erosion was that of erosion pins. These
were made from 6.35-mm diameter mild steel cut into lengths of 40 and 80
cm. The pins were knocked into the banks at equal intervals of 2 or 3 m
along a section and with the pins spaced vertically, about 30 cm apart. The
positions of the pins were referred to stable wooden pegs established on the
flood plain several metres from the bank edge and mapped in an initial plane
table survey. Resurvey of the bank line using such pegs could have been used
alone but, as Wolman (1959) found, the combination of pins and resurvey is
much more satisfactory. The erosion pins in this case functioned well,
especially on the larger rivers of 30 to 700 km 2 drainage area, where most of
the banks are composed of silty alluvial material. The pins are much less
suitable for gravel material. On these larger rivers no problem was
encountered with bank creep or pin m o v e m e n t and measurement errors were
negligible compared with the a m o u n t s of erosion. It appeared from observa-
tion that the pins gave a representative sample of the erosion in a section
except on the smaller streams o f 20 km 2 or less drainage area where erosion
was much more confined spatially and more sporadic in time o f occurrence.
A major problem arose from loss of pins when large amounts of bank
material collapsed, so it is r e c o m m e n d e d t h a t pins o f at least 80 cm length
should be used on all actively eroding banks. The presence of the pins them-
selves did n o t appear to encourage or inhibit the erosion processes. The
removal of material was assessed by measuring, either at frequent intervals or
after every storm, the length of pin exposed and then knocking the pin back
into the bank to a k n o w n distance.

Field sites

The study was designed to relate the a m o u n t s o f erosion measured to the

meteorological and hydrological conditions in the intervening period. Thus
the choice of field sites to be monitored was restricted to streams with a
flow gauge and a rainfall gauge nearby. Sites of apparent m a x i m u m erosion
on the rivers were selected since these sites were likely to exhibit the
processes and conditions more clearly and frequently; t h e y are also the most
important sites geomorphologically. The channel sections and the measure-
m e n t sites are shown in Fig. 1. Measurement sites on a number o f rivers were
 41

iI
/,
S \ C

, \ a
/.//
/// 5
O

3~ ~, o 50m

a R. E X E
Erosion measurement section
b R. A X E
c R. C R E E D Y \;~,~
Steep slope
d R. C U L M (lower site)
e R. C U L M (upper s i t e ) /f / Railway
f R YARTY

Fig. 1. River channel planforms at erosion measurement sites.

established because a range o f sizes and types o f site would demonstrate the
variation in rates and modes o f erosion, and would also ensure that sufficient
data could be collected. On each river several sections within one reach were
instrumented so that variation under similar meteorological and hydrological
42

conditions could be tested. Altogether sixteen separate sections on seven

different streams were checked regularly, comprising 342 pins in total.
Details o f th e sites are given in Table I. T he River Exe is t he largest river
o n which meas u re m ent was made, with a drainage area o f 620 km 2 at the
erosion sites and an average channel width o f a p p r o x i m a t e l y 30 m. Site A is
a long straight bank section on a wide part o f t he channel with a low bank
and a gravel layer at t h e base. Site B was a rapidly eroding section which had
t o be a b a n d o n e d when bank p r o t e c t i o n structures were built. It was replaced
b y site E which is unusual in being on t he inside o f a sharp bend where
erosion takes place at high flows. T he bank line at site C is m uch m o r e irre-

TABLE I

Details of e r o s i o n m e a s u r e m e n t sites

Stream Site National Catchment Geology Bank

Grid area material
Reference (km 2 ) ' (% silt-clay)

Exe A SX 930989 620 a l l u v i u m over 46.1

B SX 929990 Lower Permian --
C SX 927988 Sandstone 48.0
D SX 928984 53.6
E SX 927990 51.1

Creedy 1 SX 880985 235 a l l u v i u m over 65.4

P e r m i a n Breccia
and Sandstone

Culm lower SX 946988 270 a l l u v i u m over 58.5

upper SX 950995 (268) Lower Permian
(divided c h a n n e l ) Sandstone

Axe 1 SY 266962 288 a l l u v i u m over 81.1

2 SY 266960 K e u p e r Marl 67.7
3 SY 266960 61.6
4 SY 266960 59.1
5 SY 269961 68.0

Yarty 1 ST 2 5 4 0 5 9 51 a l l u v i u m over 50.0

K e u p e r Marl

Coly 1 SY 2 5 0 9 3 6 74 a l l u v i u m over 68.8

K e u p e r Marl

Hookmoor G o r h u i s h SS 5 3 1 9 8 5 9.6 Crackington 69.5

Brook Formation of
C u l m Measures
 43

gular in plan profile than the other sites due to e m b a y m e n t s gouged o u t by

the river. Site D is on a long straight section of bank which exhibited a
remarkably high rate o f erosion during the measurement period.
The erosion site on the River Creedy is rather different in character from
the other sites being much more incised and generally more stable in plan-
form, as substantiated by the historical map evidence (Hooke, 1977). One
large e m b a y m e n t was monitored and showed some erosion activity during
the period. Two sections were established on the River Culm, the upper one
comprising a complete meander bend on a branch of the river flowing
through an open flood plain; the lower site is a similar section on the main

Vegetation Discharge gauge Rainfall gauge

and land
use location National type
Grid
Reference

pasture and Thorverton SS 936016 recorder Brampford Speke

scattered and Yendacott
trees

pasture and Cowley S X 901967 recorder Newton St. Cyres

woodland and Yendacott

pasture Woodmill ST 021058 recorder Stoke Canon and

Yendacott

pasture Whitford SY 262953 recorder Seaton Junction

pasture and Court Place ST 256070 observer Yarcombe

scattered board
trees

pasture and Colyton SY 250940 observer Colyton

scattered board
trees

rough pasture Gorhuish SS 531985 recorder

and wood-
land
44

river. As a check on the m e t h o d o f sampling and measurement this lower

reach was originally m o n u m e n t e d in a regularly spaced series of short
sections, some occurring on actively eroding banks and others on vegetated
banks. It was soon f o u n d t h a t the vegetated sections remained completely
stable and were m u c h less suitable for measurement by pins. The frequency
of bankfull flow is greater on the lower Culm than on the other streams
monitored though the time o f rise often lags t h a t of the Exe and Creedy.
The other sites were established on streams in east Devon because map
evidence indicated t h a t these were changing rapidly in planform (Hooke,
1977). Five sections were instrumented on the River Axe near Whitford,
where the gravel bed channel is wide and flows through an open floodplain.
A site on the River Yarty, a tributary o f the Axe, was included since erosion
appeared to be rapid, although the stream lacked a continuous flow record.
Sites were also established on the River Coly in the Axe basin and on the
H o o k m o o r Brook in central Devon but neither of these sites provided suffici-
ent data for detailed analysis.
All the sections monitored are in floodplain sites with most banks com-
posed of silty alluvial material overlying coarse gravel deposits, similar in
composition to the present river bed. The alluvial material varies in c o n t e n t
from about 40% to 90% silt-clay, with most samples yielding 50--70% silt-
clay from laboratory analyses. The stream catchments are almost entirely
rural mainly used as pasture for grazing cattle. The streams have a seasonal
flow regime with most of the high flows occurring between September and
March.

METHOD OF ANALYSIS AND CHOICE OF VARIABLES

The data were collected for use in a multivariate statistical analysis which
would incorporate discharge and precipitation parameters. Stepwise multiple
regression analysis is considered appropriate because, as Hauser (1974)
states:

"Stepwise regression may be regarded essentially as a search procedure to identify which

independent variables previously thought to be of some importance, actually appear to
have the strongest relationship with the dependent variable."

Certain of the original variables were excluded because of high multicoUine-

arity (correlation values of > 0.8). The assumption of a linear relationship is
complied with more closely for some sites than for others but this is used as
a general model at this stage. Data in both arithmetic and logarithmic form
have been analysed but the results of the latter are presented here. The
logarithmic values approach more closely a normal distribution and are also
the important cases for analysis since t h e y include only those events in
which erosion was recorded.
Surprisingly few systematic studies of bank erosion have been made in
 45

which a comprehensive set of factors likely to control erosion have been

analysed. Wolman (1959) considered, qualitatively, the influence of peak
discharge, rise in stage, moisture conditions and precipitation, and the effect
of temperature, particularly the occurrence of frost. Likewise, Twidale
(1964) provided a valuable discussion o f such factors b u t with no quantifi-
cation of the relationships. He re-iterated the importance o f moisture con-
ditions of the bank and demonstrated the significance of "falling-stage"
phenomena. Walker and Arnborg (1966) and Harrison (1970) discussed the
influences on bank erosion in environments where frost action has a pro-
found effect. No discharge data were available to Hill (1973) in his study of
erosion on river banks in Northern Ireland, b u t he produced correlation
values of the relationships between precipitation amounts and erosion and
between frost incidence and erosion, finding these factors to be of varying
importance at different sites. Knighton (1973) examined the mechanics of
erosion more closely, relating amount of erosion to bank conditions, velocity
distributions in the flow and to discharge magnitude and frequency.
Thus a number o f factors have been identified which can affect erosion of
the stream banks over time. These can be divided into five groups: (1) param-
eters of the flow conditions and hydrograph characteristics; (2) the rain-
fall characteristics of a storm; (3) the factor of time b e t w e e n peak flows; (4)
parameters of the soil moisture conditions; and (5) temperature conditions,
primarily the incidence of frost. Other factors such as stream size and sedi-
ment characteristics influence the spatial distribution of the rates of erosion
(Hooke, in press) though generally n o t the relative amount of erosion in
each event.
The variables selected for this analysis are listed in Table II. Discharge data
were available from continuous recorders for sites on the River Exe, Culm,
Creedy, Axe and H o o k m o o r Brook. Daily flow readings were available from
observer boards on the Coly and Yarty. Precipitation data were obtained
from daffy readings of South West Water Authority rain gauges located close
to the stream gauging stations. Additional rainfall data were also used for the
sites on the Exe, Culm and Creedy from a continuous recorder at Yendacott
(SS 912017).
Peak discharge is used as a direct measurement of the maximum force
exerted on the bank b y the flow during a storm event. Rate of rise of hydro-
graph and amount of rise were included because they may affect hydraulic
conditions; also, Walling (1974) found a parameter of flood intensity or rate
of rise to be useful in a model o f controls on suspended sediment concentra-
tion. This was measured for a hydrograph as:
[(peak flow) -- (preceding flow level)]/(time of rise)
This does not allow for the varying cross-sectional shape of the channel but
it does provide some indication of the relative intensity of the storm. Separa-
tion of the hydrographs was occasionally difficult b u t the basic criterion
used was that small subsidiary peaks occurring within 6 h o f a large peak
46

TABLE II

Variables used in stepwise multiple regression analysis

Factor Variables

Independent variables:
Discharge characteristics maximum peak discharge
rate of rise of discharge
amount of rise of discharge (later excluded)
Time and preceding conditions hours since last peak
discharge of last peak
m i n i m u m discharge prior to peak
Rainfall characteristics amount of rainfall in storm
duration of rainfall
max. 0.5-h intensity in storm
Soil moisture conditions antecedent Precipitation Index (API)

Temperature (not used for this study period)

Dependent variables:
Mean erosion for section at each measured event
Maximum erosion in section at each measured event
Proportion of pins registering some erosion in each event

should be incorporated into it but all other peaks should be measured indi-
vidually. A m o u n t of rise was later excluded because o f high multicollineari-
ty.
Problems were encountered in measuring rate of fall of hydrograph;
several alternatives were considered including fall in a given time period, fall
to previous level and rate of fall until further increase. None of these param-
eters could be measured on all hydrographs and thus it was t h o u g h t better
to exclude this parameter than incorporate an inconsistent measure.
The variable, hours since last peak, is used as a measure o f the time
between storms which would allow other processes to act and thus might
influence the state of the material and its susceptibility to fluvial erosion at
the next peak flow. Discharge of the last peak is included because this could
affect the a m o u n t o f material which was available for erosion in the next
storm. The minimum discharge prior to the hydrograph may m o d i f y the
effect of the hydrograph, mainly through its influence on soil moisture
conditions.
The other variables incorporated in the analyses are derived from the
precipitation data. Unfortunately, a m o u n t of rainfall in discrete storms
could not be obtained from the daffy gauge readings; 24-h totals are
considered misleading. From the Yendacott data storm totals were cal-
culated and also m a x i m u m 0.5-h intensity since this has proved important in
other soil erosion studies (Ellison, 1945; Wischmeier and Smith, 1958).
 47

Duration of the storm may influence the degree of wetting of the bank
material.
Soil moisture conditions are thought to be of importance in bank erosion.
The parameter used in this analysis is Antecedent Precipitation Index (API).
There are several forms of the index (Gregory and Walling, 1973) b u t most
incorporate an exponentially decreasing influence of previous rainfall. A
formula using cumulated daily totals multiplied by a constant, in this case 0.9,
was used:
APId = (APId-1 X 0.9) + Pd
where APId is that day's index, APId-1 is the previous days' index and Pd is
that day's precipitation.
Temperature data were to be collected and analysed b u t other studies
have demonstrated that the most important effect is at the freezing level.
Since the incidence of frost in this area o f Devon was very low in the period
of study and, since preliminary analysis excluding this factor produced
quite high levels of explanation, temperature data have not been utilized.
Drying and heating of the banks in summer does have some effect on the
material but causes little actual erosion; growth of vegetation in summer m a y
afford some protection. Inclusion o f a seasonal factor such as a simple sine
wave was contemplated to allow for these effects b u t lack of any significant
storms or loss of material between April and September restrict the range of
conditions which could be tested.
Three dependent variables have been used in the analyses: mean erosion at
each section, maximum erosion and proportion of pins registering erosion in
each event. The amounts of erosion have been recorded in linear distances
since this is most meaningful in terms o f bank line recession and channel
movement b u t volumes can be calculated b y multiplying b y bank height and
the spacing of the pins. The proportion of pins registering some erosion indi-
cates the spatial extent o f erosion.

ANALYSIS OF RESULTS

During the period of field measurement a range of hydrological conditions

was experienced including above average rainfall in the winter 1973--1974
and 1974--1975, the highest flow levels for 4 yr. or more, and the dry winter
of 1975--1976 which led to the drought situation of summer 1976. Erosion
was registered on 15--30 occasions at each site, almost all of these occurring
between September and March o f each year. Nearly all the erosion occurred
in association with high river levels and negligible amounts resulted from
drying of the banks, from frost incidence or from cattle trampling and
similar phenomena. (The effects o f frost were, however, observed during the
winter of 1976--1977 and appear to cause the removal of particles of soil,
leaving a dusting of material at the base of the bank.) Examples of the dis-
48

tribution of events and the mean a m o u n t s of erosion at several sites can be

seen in Fig. 2.
The detailed results o f all the stepwise regression results are not repro-
duced here but a summary table of the order of significance of the indepen-
dent variables is provided (Table III) and an example of one analysis is given

T A B L E III

S u m m a r y o f r a n k o r d e r o f significance o f i n d e p e n d e n t variables in s t e p w i s e m u l t i p l e
regression analyses

Variable Exe A Exe C Exe D Exe E Creedy Culm-

lower

Mean erosion:
Peak discharge 3 2 1 7 1 5
R a t e o f increase o f
discharge 7 5 8 4 4 3
H o u r s since last p e a k 6 8 7 5 9 4
Discharge o f last p e a k 5 1 6 6 2 6
M i n i m u m discharge 1 4 9 9 3 7
Rainfall a m o u n t 8 9 4 3 6 1
Rainfall d u r a t i o n 9 3 5 8 7 2
Rainfall i n t e n s i t y 4 7 2 2 9 4
A n t e c e d e n t Precipita-
tion Index (API) 2 6 3 1 5 8

Maximum erosion:
Peak discharge 4 1 1 5 4 7
R a t e o f increase o f
discharge 6 5 2 3 6 1
H o u r s since last p e a k 2 7 7 8 7 8
Discharge of last p e a k 5 2 9 6 8 3
M i n i m u m discharge 1 4 8 7 5 5
Rainfall a m o u n t 7 3 5 4 1 2
Rainfall d u r a t i o n 9 6 3 9 2 4
Rainfall i n t e n s i t y 8 8 6 2 9 6
A n t e c e d e n t Precipita-
tion Index (API) 3 9 4 1 3 9

Per cent pins registering erosion:

Peak discharge 9 6 2 9 1 9
R a t e o f increase o f
discharge 2 7 5 6 2 7
H o u r s since last p e a k 4 3 4 3 7 4
Discharge o f last p e a k 7 1 8 8 6 5
M i n i m u m discharge 1 9 9 5 5 8
Rainfall a m o u n t 5 5 3 4 3 2
Rainfall d u r a t i o n 6 8 6 7 8 3
Rainfall i n t e n s i t y 8 4 7 2 4 6
A n t e c e d e n t Precipita-
tion Index (API) 3 2 1 1 9 1
 49

in Table IV. In most cases only the first one or two variables are significant
at the 95% level as tested by the F criterion. A summary table o f number of
sites at which each variable is included either first, second or third is also
provided (Table V) and the total r 2 scores from these placings are given.
At Exe site A m i n i m u m prior discharge and API emerge as the most signif-

Culm- Axe 1 Axe 2 Axe 3 Axe 4 Axe 5 Yarty Total

upper ranking
score

6 6 5 2 6 6 5 56]98

8 2 6 1 2 4 54]93
5 5 4 6 4 1 4 68]98
1 3 2 5 5 3 1 46]98
8 4 1 4 3 2 3 58/98
2 33]63
8 42]63
4 36]63

3 1 3 3 1 5 2 43/98

9 5 6 4 6 6 2 60]98

2 4 5 1 4 2 41/93
7 6 2 6 2 1 5 68/98
1 3 4 5 3 3 3 55]98
8 2 1 3 5 5 4 58/98
3 25/63
6 30/63
4 41163

5 1 3 2 1 4 1 46/98

9 1 6 3 2 3 4 64/98

4 5 4 1 1 2 46/93
3 4 2 5 4 6 3 52]98
5 6 1 6 6 1 5 65/98
8 2 5 4 3 4 1 64/98
7 29/63
6 44/63
2 33]63

1 3 3 2 5 5 2 38/98
50

all
2o 1 ExEs,tec
Ool~l" F " M " A ' M ' J 'J 'A'S'O"
J,I N'J olJH J,
l J'F'M'A'M'J
,,.
' j S'O'
~, ~)llj' ! ' M * A" M '
1974 1975 1976
140.

12C

100.
EXE Site D

80-

60'

40'

20"

0
1974
F" M'J I . . . . o. "IN D',. JJF. M.,,.
~975
. ~.J.
. ,,,,. ~'. o ~ .I
D J F'M' A'M'
1976

2o
1
lO
I CREEDYSite J

0 el ' 'M'A'M'
,, j'~I'A
,,I, I
" S ' O " N' DI J ' F ' M ' A
,
7 M " J ' J ' A' S ' O ' N' D I J ' F ' M ' A ' M '
1974 1975 1976

10t J I F M IA CULM Lower Site

0 F"M'A'M'J'J'A'S'O ' ' ' J J A S O N D J F M A'M'

1974 1975 1976

AXE S i t e 4

10

I,II, I ,
M ' AI'M" j " j • A" S • O" N ' DI j ' F ' ! ' A ' " M ' j
i , ,
' j ' A" S " O ' N ' D I d ' ~' M' A ' M '
1974 1975 1976

Fig. 2. Mean a m o u n t s o f erosion at each event and their distribution over t i m e for
selected sites.
 51

TABLE IV

Stepwise multiple regression analysis of erosion -- EXE site D

Step No. Variable entered Multiple r r2 F value

Mean erosion:
I peak discharge 0.6808 0.4635 21.6010
2 antecedent precipitation index (API) 0.7217 0.5208 2.8673
3 rainfall intensity 0.7704 0.5936 4.1186
4 rainfall amount 0.7980 0.6368 2.6172
5 rainfall duration 0.8177 0.6687 2.0211
6 discharge of last peak 0.8279 0.6855 1.0683
7 hours since last peak 0.8311 0.6907 0.3208
8 rate of increase of discharge 0.8332 0.6942 0.2086
9 minimum discharge 0.8341 0.6957 0.0819

Maximum erosion:
1 peak discharge 0.5903 0.3485 13.3706
2 rate of increase of discharge 0.6195 0.3837 1.3740
3 antecedent precipitation index (API) 0.6329 0.4005 0.6435
4 minimum discharge 0.6384 0.4076 0.2621
5 rainfall intensity 0.6472 0.4189 0.4083
6 rainfall amount 0.6574 0.4322 0.4705
7 rainfall duration 0.6750 0.4556 0.8145
8 discharge of last peak 0.6774 0.4589 0.1115
9 hours since last peak 0.6837 0.4675 0.2736

Per cent registering erosion

1 antecedent precipitation index (API) 0.6460 0.4174 17.9081
2 peak discharge 0.6966 0.4852 3.1624
3 rainfall intensity 0.7090 0.5026 0.8071
4 rainfall amount 0.7335 0.5380 1.6825
5 rainfall duration 0.7463 0.5569 0.8982
6 rate of increase of discharge 0.7592 0.5764 0.9174
7 minimum discharge 0.7645 0.5845 0.3704
8 hours since last peak 0.7804 0.6091 1.1323
9 discharge of last peak 0.7867 0.6189 0.4372

icant variables, explaining 78% o f the variation in mean erosion. These two
parameters reflect the influence of soil moisture conditions at this site. At
Exe site E the large amounts o f erosion recorded also appear to be closely
related to API and this is consistent with the t y p e of erosion observed. Inten-
sity o f rainfall is the only other significant variable. The analysis of Exe site
C showed preceding peak discharge to be important and this could be due to
the alteration of the bank configuration and the weakening of the bank
material which t o o k place. The variable is followed in order o f inclusion by
peak discharge in the case o f mean and m a x i m u m erosion and API for the
proportion o f bank eroded. This is a pattern which recurs at other sites
including Exe site D where peak discharge appears to be the most significant
control on amounts o f erosion and soil moisture, as measured by API,
influences the spatial extent of erosion.
TABLE V

S u m m a r y o f h i g h e s t placings o f e a c h variable in regression analyses a n d t o t a l r 2 values o f t h e s e placings

Variable Mean erosion Maximum erosion N u m b e r o f pins

1st 2nd 3rd total 1st 2nd 3rd total 1st 2nd 3rd total
r 2 r2 r2

Peak discharge 2 2 1 1.18 2 1 0 0.91 2 2 0.89

R a t e o f increase o f
discharge 1 2 1 0.87 2 3 1 1.03 2 3 1.11
H o u r s since last p e a k 1 0 0 0.24 1 3 0 0.55 0 1 0.58
Discharge o f last p e a k 3 2 2 1.23 1 1 5 0.64 3 0 0.90
M i n i m u m discharge 2 1 3 1.23 2 1 1 0.93 2 1 1.00
Rainfall a m o u n t 1 1 1 0.31 1 1 2 0.50 0 1 0.24
Rainfall d u r a t i o n 0 1 1 0.16 0 1 1 0.40 0 0 0.06
Rainfall i n t e n s i t y 0 1 0 0.07 0 1 0 0.08 0 2 0.17
Antecedent precipitation
index (API) 3 2 4 2.43 4 1 3 2.06 4 3 2.32
 53

Erosion occurred on m a n y fewer occasions at the Creedy site and mostly

comprised the removal o f prominent lobes of material. The levels of explana-
tion in the regression analyses are high, 83% or more, with peak discharge
emerging as most significant in relation to mean erosion and proportion of
bank eroded but m a x i m u m erosion most dependent on precipitation factors.
In contrast, explanations for the upper Culm site are extremely low with
discharge of the last peak emerging as one o f the more important variables
and API explaining 32% of the spatial variation. Amounts of erosion and dis-
tribution were observed to be rather erratic at this site and no obvious
controls were identified in the field. At the lower Culm site the orders of
inclusion are rather different for each o f the three independent variables but
no single variable is highly significant. Rainfall a m o u n t explains 21% of the
variation in mean erosion, rate of rise explains 23% of the m a x i m u m erosion
and API accounts for 35% of the spatial extent o f erosion. At Axe site 1 the
most significant variable in relation to mean and m a x i m u m erosion is API;
this is probably due to the higher clay c o n t e n t of the material and the
greater bank height of the site. The spatial extent of erosion appears to be
controlled by the flow stage due to bank configuration. Rate of rise of
discharge is included first in all the analyses for Axe site 2/3, explaining
about 50% of the mean and m a x i m u m erosion and 31% of the proportion of
bank eroded. API appears to be the d o m i n a n t control on erosion at Axe site
4 and produces an r 2 value of 0.51 or more for all three independent
variables. At Axe site 5, hours since last peak is the most significant variable
in relation to mean and m a x i m u m erosion but all the independent variables
together only explain 35% of the spatial variation in erosion. This is partly
because erosion at this site was spatially rather constricted.
A more restricted set o f variables was analysed for the Yarty site and the
derivation of the discharge data means that less reliance can be placed on
these results, but it does appear t h a t soil moisture conditions are of
importance.

OBSERVATIONS OF THE PROCESSES OF BANK REMOVAL

The processes o f bank erosion were observed, noted and photographed at

various flow stages during the 2.5-yr. period of field measurement. Two main
processes have been identified as acting on these river banks in Devon.
Firstly, d o m i n a n t at m a n y of the sites is the direct shearing o f f of material
or corrasion of the bank at high flows. This is indicated by smoothed banks
remaining after the flood passage, with little undercutting and no superflu-
ous material on the bank face or slumped material at the base. Often over-
hangs are left, bound by roots in the upper vegetation layer, particularly
when flow has approached bankfull; in time, these overhangs are weakened
and drop o f f or are swept away in the succeeding flood. The other type of
bank erosion process occurs when material is in a thoroughly moistened
condition (as measured crudely by API), and bank hydraulic conditions are
54

such that the main removal takes place b y collapse or slumping of large
blocks of material after the peak has passed. This process occurs in associ-
ation with b o t h high and moderate peak flows and m a y take place on the
same bank sections as the direct corrasion but usually at different times. It
has been suggested that this slumping may be related to the rate of fall of
the hydrograph (Twidale, 1964) b u t this is difficult to observe or measure.
The process of slumping involves large blocks which topple forward b u t usu-
ally remain intact with the grass continuing to grow on the side. The blocks
are often of similar size even on streams of differing width, depth and bank
height, the blocks averaging about 1--1.5 metres in length along the bank,
about 0.75 m in t o p surface width and a b o u t 1 m in height. There is some
tendency for this process to be more c o m m o n on the lower banks and this
may be partly due to a relation b e t w e e n bank height and t y p e of material

41¸¸:¸¸¸i~¸¸~!: ~

Fig. 3. Slumped bank material, River Exe.

 55

but also because a lower bank is more rapidly and frequently wetted
throughout. It did appear that the coarser, sandier bank material was more
liable to slumping. The slumped material usually remains at the base of the
bank until the blocks disintegrate and the material is gradually removed by
the river flow (Fig. 3).
Little is k n o w n of the details of hydraulic conditions in river banks and
h o w the m o v e m e n t o f water m a y change with passage of a flood peak and
thus affect the stability of the bank. Some studies have been made on bank
seepage b u t these mostly involve mathematical or laboratory modelling of
simplified situations and with specified conditions of the water table; for
example, the viscous flow model and numerical solution m e t h o d of Desai
(1973) and the work of Freeze (1971). The effects of seepage inflow and
o u t f l o w on erosion o f bed particles has been assessed b y Watters and Rao
(1971). One o f the most valuable studies o f bank seepage, however, is by
Burgi and Karaki (1971) who performed empirical tests under certain condi-
tions and concluded that erosion is related to the hydraulic gradient across
the bank and the velocity of flow in the channel. They suggested the former
is dominant at low flows and the latter at high flows.
The t w o processes of erosion identified on the river banks in Devon
concur with other observations reported in the literature (e.g., Coleman,
1969; Gupta and Fox, 1974). However, it is possible that a third process is
important in other areas; this is rotational sliding which produces multi-
stepped bank profiles. Turnbull et al. (1966), Laury (1971) and Coleman
(1969) have shown h o w the complex stratigraphy of river banks such as the
Mississippi and Brahamaputra can affect the bank processes. On the Devonian
streams there was little influence from structure, the only differentiation
present mostly being a layer o f gravel below the silty alluvium. This gravel
tends to be more easily eroded than the overlying material b u t is seldom
undercut more than 10--20 cm and so does not cause oversteepening.
The structure o f the currents in the stream flow appears to be another im-
portant determinant of the distribution, t y p e and extent of erosion. Circular
upwellings and spiral vortices were observed frequently at many of the
sections (Fig. 4), though the pattern of currents alters with stage and
turbulent features become more pronounced at high flows. These turbulent
features often recur in the same positions and they are obviously related
to irregularities in the bed and bank, b u t it has n o t been possible to
relate the distribution and a m o u n t o f erosion to the flow patterns or turbu-
lence features. The effectiveness o f eddies and vortices as erosive agents still
depends much on the nature and condition o f the material though t h e y are
thought to be important in the corrasion process.

SUMMARY OF E F F E C T S OF I N D I V I D U A L F A C T O R S

The effects o f the individual factors in the statistical analyses can n o w be

reassessed in the light o f these observations of the operation o f the erosion
processes. Peak discharge appears fourth or fifth in explanation ranking o f
56

Fig. 4. T u r b u l e n c e flow s t r u c t u r e s during a f l o o d o n t h e River Culm.

the variables and is the primary variable in relation to mean erosion at only
two sites. It is difficult to identify a definite pattern but these are sites where
the direct corrasion processes dominate. At sites where peak discharge is the
primary control over the spatial e x t e n t of erosion the amounts of erosion are
low and intermittent in occurrence with the flow reaching parts of the bank
only at high discharges. It is clear that at other sites a m o u n t of erosion is not
directly related to the forces of water at high flow.
The variable, rate of rise of discharge, appears high in the total ranking of
variables, particularly in relation to m a x i m u m erosion and proportion of
bank eroded. It is t h o u g h t to affect the rate o f wetting of the bank material
and thus the hydraulics and to determine to some extent the development of
turbulence structures and the a m o u n t of undercutting. A problem arises
with the interpretation of the statistical analysis because of some high corre-
 57

lations of rate of rise with peak discharge but both are an indication of the
size of the flood; any site at which either of these variables emerges as
d o m i n a n t is primarily influenced by discharge conditions rather than by
other factors, and tend to be where corrasion predominates.
Hours since last peak appears low in the order of inclusion of variables for
most sites. At only one site, Axe 5, is it the primary variable and at this site
the distribution and nature o f the erosion varies considerably and shows
little systematic pattern. The lack of relation to amounts of erosion may also
be caused by the variable measuring effects which act in opposite directions;
increased length of time will tend to enhance preparation and erodibility
but, depending on the season, the greater time may also allow vegetation to
gain a firmer hold and thus increase resistance. It could also be t h o u g h t to
represent the time available for drying out of the banks.
Discharge of the last peak is a difficult variable to interpret but was
included to test whether there is any systematic lag-over effect. However,
there are t w o processes which this parameter could be measuring indirectly
and these may have occurred at the same site at different times. The previous
flow could be such t h a t it undercuts and weakens the bank so much t h a t
erosion is encouraged at the next high flow. Secondly, the previous flow
could remove all loose material and so smooth the banks as to make them
highly resistant to erosion in the subsequent flood. It emerges from the
analyses that this variable is related to the peak discharge and/or minimum
discharge prior to the peak because m a n y of the high flows succeed each
other in the winter months. Thus it represents a cumulative effect of the
high flows and high moisture content. It appears moderately high in rank
orders of inclusion, especially in relation to mean erosion, but is particularly
significant at sites such as Exe site C where the bank line is irregular and
alteration of the configuration takes place.
The discharge values of m i n i m u m flow prior to the peak arise from a com-
bination of factors but this variable is probably best considered as a second
parameter of soil moisture status and hydraulic conditions in the bank. This
is generally more adequately estimated by API but minimum flow appears
high in total ranking of variables.
The a m o u n t of rainfall in a storm is revealed as of primary importance at
only one site. This is not entirely unexpected since this factor should have
no direct influence which is not already measured by other variables such as
rate of rise of hydrograph, peak discharge and moisture levels. However, it
yields only a low correlation with these variables. Likewise duration of a
storm does not emerge as important and probably only slightly influences
the degree of wetting o f the bank material. Intensity of rainfall may have
some effect in c o m m i n u t i o n of the bank material but the intensities are not
sufficiently high for this to be important and the vertical nature of the banks
mostly precludes this direct action.
API emerges from the analyses as an extremely important variable,
particularly in relation to mean erosion and proportion of bank eroded. Most
58

of the sites where API is the primary variable are ones where the amounts o f
erosion are large and where erosion often occurs by large blocks collapsing
during the falling-stage of the hydrograph. These are also sites with a rather
lower percentage of silt-clay in the bank material and a greater sand propor-
tion than at other sites. It appears t h a t a moist condition of the bank
material encourages more widespread erosion with a particular effect on the
lower banks where the whole bank is wetted through more easily.

FREQUENCY OF EVENTS AND THRESHOLDS

It can be seen from Fig. 2 and other observations that erosion events
occurred several t i m e s a year at all sites. None of the peak flows during the
period of measurement had an estimated recurrence interval of more than
2.5 yr. (Institute o f Hydrology, 1975); thus the amounts and characteristics
of the erosion are not t h o u g h t to be abnormal and could be expected to
occur under similar conditions and on similar streams in the same area. The
highest peaks and their estimated recurrence intervals on each river are listed
in Table V. Another important implication of the analyses is that, at m a n y
sites threshold values o f flow and moisture levels are indicated, below which
erosion rarely takes place. Such thresholds are more distinct at some sites
than at others as can be seen from Fig. 5. Values corresponding to the graphs
are suggested in Table VI. The thresholds are particularly clear at sites Exe
A, Exe E and Axe 1, where a certain level of water is needed to reach the
bank. The thresholds for erosion may depend on a combination of condi-

TABLE VI

R e c u r r e n c e intervals o f p e a k discharges and estimated values o f t h r e s h o l d s for e r o s i o n

Site Peak Recurrence Threshold Recurrence Threshold

discharge interval discharge interval API
(mS/s) (yr.) (m~/s) (yr.)

Exe A 223.4 2.51 50 0.079 45

C 25 0.048 50
D 30 0.053 40
E 50 0.079 45

Creedy 1 105.0 0.90 15 0.130 20

Culm lower 94.3 2.44 -- -- 15

upper 5 0.103 10

Axe 1 110.3 1.34 12 0.060 25

2 15 0.O65 20
3 12 0.060 20
4 - - - - _

5 10 0.056 15
 59

80iXE Site A 80iXE Site C

%•
•o •o • I ,
0 %0 0
~c
I 1(]0

14CEXE Site D 14CEXE Site E

• .:,"
u 1(~0 0 ~o

2oCREEDY CULM Lower Site

151 •

.., o "." i
0

E
0 1(~0
,o
AXE Site 2/3
Z AXE Sute 1
,,o-O~o15'I •

IX: •
• OO • • |
J...% o.• "0 • .'~ ~ o.
%Z 0 I n u~ ~ u • n I0"-~]0 0 -'2 ©l __ 0 n
100
Z 0

Z
AXE Site 4 15AXE Site 5
w •

• • •

. o%_ ~, = , , ~ o ",'~'~ ,qc -- i I~[Oo

0 100 0

ANTECEDENT PRECIPITATION INDEX

Fig. 5. Mean distance of bank erosion (cm) plotted against antecedent precipitation index
for erosion sites.
60

tions and in reality m a y n o t be represented simply by the single variables.

At these sites most of the flow and moisture levels which have been identi-
fied as thresholds occur on eight or more days per year and therefore are
unlikely to be important restrictions on the a m o u n t of erosion taking place.
Fig. 4 also indicates the mean amounts of bank erosion which occurred in
each storm and the extent o f variation within and b e t w e e n sites.

CONCLUSIONS

Two main processes of bank erosion have been identified, direct corrasion
and slumping. The former appears to be more directly controlled by river
flow conditions and the latter mainly b y soil moisture conditions. More
widespread erosion takes place with high moisture levels; low banks of silty
material also appear to enhance erosion because the whole height of the
bank is more frequently wetted. Although soil moisture emerges as providing
highest explanation o f the erosion, from observation and measurement it
appears that little erosion occurs which is not associated with peak flows. Of
the other five major factors originally identified, the time factor is difficult
to distinguish because of the conflicting effects of increased length of time
between storms. The rain factors do not emerge as important, most of the
effects being subsumed in other variables. The effect of temperature through
frost incidence has already been discounted in the period of study. Thus it
would seem that certain of the variables could be eliminated from the model
but multiple regression analyses for each site using only four variables --
peak discharge, API, hours since last peak and rainfall intensity -- yielded
much lower levels o f explanation, decreasing the average total r 2 for each
regression from 0.657 to 0.408. It is concluded that the amount of erosion
in each event is controlled b y a complex combination o f conditions and that
no single model o f controls emerges. The analyses and observations have
shown that there is considerable variation in types of erosion and factors
which influence the processes even within a reach experiencing similar rainfall,
discharge and temperature conditions. It is likely that sediment composition
is a major factor affecting the spatial distribution of erosion and its relation-
ship to the erosion processes though, in general, banks with higher clay com-
position appear to be more resistant and experience less slumping. The sites
may nevertheless be divdided into four types on the basis of the analyses and
observations and these could prove useful in further work.
(1) Sites where erosion is mostly b y corrasion at high flow and the mag-
nitude of the peak discharge is of prime importance, e.g., Exe D, Exe C,
Axe 3 and Axe 4.
(2) Sites, possibly of coarser, sandy material, where soil moisture condi-
tions are of greatest significance and erosion is usually b y collapse of
material, e.g., Exe A, Exe E, Yarty.
(3) Sites where again soil moisture conditions (API) are important b u t the
amounts of erosion are low, the material is highly resistant and high flows
are also necessary for erosion to occur, e.g., Axe 1, Culm-lower, H o o k m o o r
Brook.
 61

(4) Sites where erosion is infrequent and takes place by sudden excavation
of lobes at high flows, e.g., Creedy.
Some sites are difficult to classify, for example Culm-upper and Axe 5,
but amounts of erosion were low at both these sites.
There are several reasons for the lack of a more systematic explanation of
the erosion events at these sites, some of which may be overcome in
designing further study but m a n y of which present practical and analytical
problems. Some of the parameters used prove not to be suitable because
of the diverging influences which t h e y measure. This could be overcome in
some cases by qualifying the values from observations o f the processes but
this introduces a subjective element which is not wholly desirable. Incorpora-
tion of a seasonal factor m a y possibly increase the level o f explanation but
this also includes the opposing effects of change in moisture level and vegeta-
tion growth. A parameter of rate of fall o f the hydrograph would be
desirable but a constant such as m a x i m u m rate o f fall is of different signifi-
cance when translated into discharge terms and the individual cross-sections.
Another major difficulty is that of separating the effects of multiple or
closely-spaced storms. This may be a practical difficulty in the field if the
water level falls insufficiently to expose the pins or if the storms are too
frequent for visits to the sites between each peak. Duration o f flow para-
meters instead of instantaneous flow parameters would overcome this to
some extent and should be tested, although t h e y depend on the availability
of suitable data. The other restriction from the data used here is that the
numbers of samples may be insufficient to distinguish all the combinations of
conditions which occur and the effect of individual factors. Therefore, this
type of study is better suited to a much longer time period, probably of the
order of 5--10 yr. The use of several sites has also illustrated the dangers of
basing all inferences on measurements from only one or two sites.
More detailed work is needed on the mechanics of the processes to identi-
fy the effects of soil moisture, the pattern of forces on the bank and the
changes in shear strength of the bank material. However, this type of study
does provide some information on the effects of environmental factors and
it has demonstrated the complexity of the processes when analysed at this
scale. The m e t h o d o f using erosion pins was found to work well on these
streams of southeast Devon but is not recommended for streams of less than
about 20 km 2 drainage in area in such an environment. This t y p e of
measurement programme also provides valuable information on the spatial
and temporal distribution of erosion and on the rates of erosion (Hooke, in
press). The prospects for prediction of erosion under different conditions
still seem distant without a priori data from each individual site but under-
standing of the processes is gradually increasing.

ACKNOWLEDGEMENTS

This work was carried out during tenure of a NERC award at the Univer-
sity of Exeter, for which the a u t h o r is grateful. The author is indebted to
62

P r o f . K . J . G r e g o r y f o r his a d v i c e , t o D r . I . D . L . F o s t e r f o r p r o v i s i o n o f r a i n -
fall f i g u r e s a n d t o t h e S o u t h W e s t W a t e r A u t h o r i t y f o r d i s c h a r g e a n d r a i n f a l l
data.

REFERENCES

Burgi, P.H. and Karaki, S., 1971. Seepage effect on channel bank stability. J. Irrig. Div.,
Proc. Am. Soc. Cir. Eng., 97: 59--72.
Coleman, J.M., 1969. Brahmaputra River: channel processes and sedimentation. Sedi-
ment. Geol., 3: 129--239.
Desai, C.S., 1973. Seepage in Mississippi River banks. U.S. Army Waterways Expt. Stn.,
Tech. Rep. TRS-73-5-1.
Ellison, W.D., 1945. Some effects of raindrops and surface flow on soil erosion and infil-
tration. Trans. Am. Geophys. Union, 26: 415--429.
Freeze, R.A., 1971. Three-dimensional, transient, saturated--unsaturated flow in a
groundwater basin. Water Resour. Res., 7 : 347--366.
Gregory, K.J. and Walling, D.E., 1973. Drainage Basin Form and Process. Arnold,
London, 456 pp.
Gupta, A. and Fox, H., 1974. Effects of high-magnitude floods on channel forms -- a case
study in Maryland Piedmont. Water Resour. Res., 10: 499--509.
Harrison, S.S., 1970. Note on the importance of frost weathering in the disintegration
and erosion of till in east central Wyoming. Bull. Geol. Soc. Am., 81 : 3407--3410.
Hauser, D.P., 1974. Some problems in the use of stepwise regression techniques in geo-
graphical research. Can. Geogr., 18: 148--158.
Hill, A.R., 1973. Erosion of river banks composed of glacial till near Belfast, Northern
Ireland. Z. Geomorphol., 17: 428--442.
Hooke, J.M., 1977. The distribution and nature of changse in river channel patterns. In:
K.J. Gregory (Editor), River Channer Changes. Wiley, London, pp. 265--280.
Hooke, J.M., in press. Magnitude and spatial distribution of rates of river bank erosion.
Earth Surf. Processes.
Institute of Hydrology, 1975. Flood Studies Report. N.E.R.C., Wallingford.
Knighton, A.D., 1973. Riverbank erosion in relation to streamflow conditions, River
Bollin-Dean, Cheshire. East Midland Geogr., 5: 416--426.
Laury, R.L., 1971. Stream bank failure and rotational slumping. Bull. Geol. Soc. Am.,
82: 1251--1266.
Lewin, J., 1972. Late-stage meander growth. Nature (London), 240: 116.
Mosley, M.P., 1975. Channel changes on the River Bollin, Cheshire, 1872--1973. East
Midland Geogr., 42: 185--199.
Turnbull, W.J., Krinitsky, M. and Weaver, F.J., 1966. Bank erosion in soils of the Lower
Mississippi Valley. J. Soil Moisture Div., Proc. Am. Soc. Civ. Eng., 92: 121--136.
Twidale, C.R., 1964. Erosion of an alluvial bank at Birdwood, South Australia. Z. Geo-
morphol., 8: 189--211.
Walker, H.J. and Arnborg, L., 1966. Permafrost and ice-wedge effect on river bank erosion.
Proc. Int. Conf. on Permafrost, Washington, D.C., pp. 164--171.
Walling, D.E., 1974. Suspended sediment and solute yields from a small catchment prior
to urbanisation. In: K.J. Gregory and D.E. Walling (Editors), Instrumented Water-
sheds. Inst. Br. Geogr., Spec. Publ., 6: 169--193.
Watters, C.G. and Rao, M., 1971. Hydrodynamic effects of seepage on bed particles. J.
Hydraul. Div., Proc. Am. Soc. Cir. Eng., 97: 421--439.
Wischmeier, W.H. and Smith, D.D. 1958. Rainfall energy and its relationship to soil loss.
Trans. Am. Geophys. Union, 39: 285--291.
Wolman, M.G., 1959. Factors influencing erosion of a cohesive river bank. Am. J. Sci.,
257: 204--216.