See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/248744514

El Nin?o and annual floods in coastal Peru

Article

CITATIONS READS

3 124

2 authors, including:

Peter R. Waylen
University of Florida
116 PUBLICATIONS   3,541 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Savanna Science View project

Changes in land-cover and vegetation and their implications in Ecuador View project

All content following this page was uploaded by Peter R. Waylen on 26 March 2015.

The user has requested enhancement of the downloaded file.

Journal of Hydrology, 89 (1986) 141-156 141
Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

EL NII~O AND ANNUAL FLOODS ON THE NORTH PERUVIAN

LITTORAL

P.R. WAYLEN and C.N. CAVIEDES

Department of Geography, University of Florida, Gainesville, Florida (USA)
(Accepted for publication August 4, 1986)

ABSTRACT

Waylen, P.R. and Caviedes, C.N., 1986. E1 Nifio and annual floods on the north Peruvian littoral.
J. Hydrol., 89: 141-156.

A three-component mixed Gumbel distribution satisfactorily models the observed annual flood
frequencies of rivers in northern Peru which display highly variable annual peak-flood characteris-
tics corresponding to three sets of ocean-atmosphere conditions. Physical support for the "a priori"
subdivision of the series is provided by consideration of the regional offshore oceanic-atmospheric
conditions. Exceptionally warm waters (El Nifio) cause extensive heavy rains; unusually cold
waters (anti-E1 Nifio) restrict both the quantity and distribution of precipitation. Model pa-
rameters reveal marked spatial trends in the severity and extent of flooding during any set of
offshore conditions and geographic shifts in the regional flood-frequency characteristics between
such conditions.

INTRODUCTION

The oceanic-atmospheric conditions that characterize E1 Nifio Southern

Oscillation events have been extensively documented over recent years. Due to
its subequatorial and westerly position, coastal northern Peru experiences
heavy rains (Mugica, 1984) and extensive flooding (Caviedes, 1984) during
years associated with the eastward incursion of warm waters (El Nifio) from
the Equatorial Counter current. In others years normal coastal conditions are
influenced by an unusual strengthening of the upwelling centers of the Peru-
vian Current (anti-E1 Nifio) producing stabilized weather conditions, little rain
and limited stream flow. Annual flood series of thirteen rivers in the area were
investigated in order to identify and model the influence of oceanic-atmospher-
ic conditions on the regional flood hydrology of northern Peru.
The objectives of this paper are: (1) to obtain a method of characterizing
flood probabilities in a region experiencing highly variable annual hydrome-
teorologic conditions; (2) to describe the spatial variability of the parameters
of the selected probability distribution in such a way as to provide an indica-
tion of the nature and extent of the influence of offshore oceanic conditions on
regional flood hydrology.

0022-1694/86/$03.50 © 1986 Elsevier Science Publishers B.V.

142
METHODS

Annual flood is defined as the largest daily discharge, Q', observed in a year:
Q" = maxlQil i -- 1,2 . . . . . 365 (I)
where Q is daily discharge and i denotes dates in Julian days. The magnitude
of annual floods has traditionally been modeled by the Gumbel distribution,
with a probability function:

FT(Q" ~> X) = 1 -- e x p { - - e x p i - - ~ ( x - - fl)]} (2)

where, by the method of moments
= 1.281/a (3)
and
fl = p - 0.45a (4)
and # and a 2 are the population mean and variance, respectively (Lowery and
Nash, 1970). The Gumbel distribution generally fits the observed data satisfac-
torily when the annual floods are produced by a single hydrologic process
(Todorovic, 1978). It performs less well when two or more processes are respon-
sible for the generation of annual floods. Under such circumstances, distribu-
tions which incorporate the differing characteristics of floods generated by
each process have been employed (Rossi et al., 1984; Woo and Waylen, 1984).
Cunnane (1985) points out that a generalized mixture model of the form:
FT( Q >1 x) = Q1FI(Q >1 x ) + Q2F2(Q >~ x ) + . . . + QnFn(Q >~ x) (5)
where each of n separate processes, P, . . . . . Pn, possesses distributions
F, . . . . . Fn and occur with relative frequencies Q,. . . . . Qn, respectively, is
practical when there is some physical basis for the identification and analysis
of distinct processes.
To derive annual flood series in such cases it is therefore necessary to
estimate the parameters of each sub-population (i.e., ai, fli and ~i)" Within a
region these parameters might be expected to vary smoothly~ and convey
information concerning the changing nature of each generating process and its
relative spatial dominance.
Equation (2) may be substituted into (5) to describe the distribution of the
exceedance probabilities of annual floods due to three distinct processes:
FT( Q" >/ X) = QI[1- exp{-exp[-~,(x- fl,)]}]
+ e2[1 - e x p { - e x p [ - ~ 2 ( x - f12)]}]
+ Q3[1 - exp { - exp [ - a3(x - f13)]}] (6)
where the subscripts refer to the three, distinct, flood-generating conditions
and the parameters a and fl are estimated from the appropriate sub-sample
using eqns. (3) and (4). The annual variability for any given station may then
 143

be summarized by the mean and the variance of floods generated by each

process and the relative frequency with which each process is observed.
This paper will proceed by the following steps:
(1) Identify years in which E1 Nifio, normal and anti-E1 Nifio conditions
prevailed in the southeastern Pacific.
(2) Extract flood sub-samples for each of the appropriate conditions from the
annual flood series of rivers in the region.
(3) Estimate statistics and parameters of each sub-population.
(4) Derive a mixture distribution using parameter estimates from (4) and
relative frequencies from (1) and compare with the observed data,
(5) Examine the spatial variability of the model parameters.

STUDY AREA AND DATA

Rivers draining over 600,000 km 2 of the littoral region of Peru (Fig. 1) were
selected for this study because of (1) their proximity to the location where the
major fluctuations in oceanic conditions have their maximum intensity and (2)
the extensive catastrophic flooding revealed by both the historic and arch-
aeological records (Moseley, 1983). Recently (1982-1983) the region was devas-
tated by some of the largest flows ever recorded, resulting in loss of life, crops
and property and damage to the hydraulic structures essential to agriculture
and communications (Caviedes, 1985). Flooding is highly seasonal and almost
entirely the result of summer (December-May) rainfall, which may vary con-
siderably from year to year (Fig. 2), augmented by varying small contributions
from snowmelt in only a few small basins.
Rivers in the area drain sub-parallel, west to southwest towards the Pacific
Ocean. The headwaters of all basins lie in the Andes at over 3000 m above m.s.1.
and locally may reach 5000 m. All rivers descend rapidly to the coastal plain,
much of which lies below 100 m. The plain attains its greatest width in the
north (200km) and diminishes to approximately 30 km in the south. Annual
precipitation is strongly elevation-dependent, ranging from less than 15 mm on
the coast to over 1500mm in the Andes (Servicio Agrometeorologica e Hi-
drologica (SENHAMI), 1962). In the extreme north of the study area, which
borders southern Ecuador, this pattern gives way to a more north-south trend
as the influence of equatorial conditions increases. Rivers bring water to the
intensively populated river oases located in an essentially arid region.
The annual flood series of thirteen rivers draining areas of between 800 and
13,000km 2 and located between 3°S and 10°S are considered (Table 1). The
selected stations provide more than 500 station-years of historic record, which
are available from the Peruvian Government (SENHAMI, 1962) and from local
water-management offices. There may be errors in flood measures, particularly
during the largest events, but such errors are unavoidable even with the most
modern technology (Potter and Walker, 1981). The majority of the stations are
well established at major urban locations and data were published under the
auspices of the United Nations until 1960. Records affected by recent flow-
regulation schemes have been excluded from the analysis.
144

8i1 o 7,,"

4"- • 4*

Pacific
Ocean 1~ '

Sechura

Desert • 6°

Casa
Grande

Elevation Puerto

m > 3000 rn
8" ,8 °
~ 1000-3000 m

1. I < 1000 m

6
• etreamflow guage

IIIIIIIIIDIIII Pacific drainage

divide

0 50 100
i i I
km

8'1" 7'9"

Fig. 1. Major physiography and rivers in the study area, n o r t h e r n Peru. Also shown are streamflow-
gauging stations which provided data for the study. For legend see Table 1.
 145

P u e r t o Chicama, Peru
d

is 2

~'~ , i = i = , i i , i = J i , , , = i i i i , i i ,

O,anOe, ,e u
E
30 [~ Annual total I-~
20-

10-

a.
0
Chicama River at Salinar, Peru
700
600
~)~ ~ Annual flood
500
I 400
E
E 300

200 60
*o
2 lOO 50
0 40
w
3O ¢
=
c 20

10

0 J
1934 1930 1938 1940 1942 1944 1940 1948 1950 1952 1954 1956 1958 1960

Fig. 2. Examples of the relationship between streamflow, precipitation and sea-surface tem-
peratures in and near the Chicama River basin.

TABLE 1

Locational details of drainage basins in northern coastal Peru

No. River Gauge Lat. Long. Area Years of

location (S) (W) (km 2) record

1 Chancay Carhuaqero 06.38 79.29 2860 59

2 Chicama Salinar 07.40 78.58 4087 62
3 Chira Sullana 04.53 80.42 12712 22
4 Jequetepeque Ventanillas 07.17 79.17 4809 52
5 Leche Puchaca 06.24 79.30 914 51
6 Moche Quirihuac 08.02 78.50 1873 60
7 Nepefia San Jacinto 09.10 78.15 1335 25
8 Piura Piura 05.13 80.38 7836 53
9 Quiroz Paraje Grande 04.26 80.15 1792 25
10 Sechin Puente Casma 09.29 78.17 776 15
11 Tumbes Tumbes 03.34 80.28 4565 12
12 Viru Huacapongo 08.25 78.40 803 34
13 Zafia E1 Batan 06.50 79.18 845 54
146

RESULTS

Identification of processes

Conceptually, there should be a strong link between sea-surface tem-

peratures, annual precipitation and flood discharges in this region. Examina-
tion of the synchronous period of deviations of sea-surface temperatures at
Puerto Chicama and hydrometeorological data at Casa Grande and the Chica-
ma River (Fig. 2) indicates that the relationship is not as consistent as might
be expected. Because the interactions at the ocean/atmosphere/land interfaces
are not fully understood or detailed in this region, the sea-surface temperatures
do not provide a good ~'~ priori" basis for the classification of annual floods by
the generating process (Cunnane, 1985). Not even the discharges themselves
can be used as indicators of the generating processes involved. Figure 3 shows
the empirical estimates of the return periods of annual floods in the study area
for four years which are widely considered to be influenced by E1 Nifio con-
ditions. It can be seen that there is considerable spatial variability between
contiguous basins.
In the absence of any acceptable, single, classificatory hydrologic variable,
the annual series are divided on the basis of oceanic-atmospheric anomalies
derived from a search of the pertinent literature. Schweigger (1959) singles out
E1 Nifio years as those which concurrently produce heavy summer rains in
Northern Peru, abnormally high sea-surface temperatures at Puerto Chicama
and Callao, the depletion of fish stocks and the starvation of sea birds. Quinn
and Burt (1970) utilize barometric and pluviometric criteria to distinguish E1
Nifio years. If summer pressures at Darwin, Australia, rise and rainfall on
several equatorial Pacific islands (Canton, Malden, Fanning and Christmas)
increase, then anomalous oceanic-atmospheric conditions are induced. Bellido
Delgado (1983) made similar classifications on the basis of air and sea-surface
temperatures along the Peruvian coast.
Covey and Hastenrath (1978) were among the first to attempt an identifica-
tion of anti-E1 Nifio years - - as opposed to E1 Nifio years - - on the basis of
sea-surface temperatures and atmospheric pressures. Subsequently Barnett
(1984) and Van Loon and Shea (1985) have identified "cold episodes" in the
sea-surface temperatures at Talara, Peru, and related them to zonal winds and
the fluctuations of the Southern Oscillation. Complete accordance between
these classifications is impossible given the variation in the natural criteria
selected, the geographic locations of the studies and the time periods under
consideration. However, an objective classification in terms of flooding may be
extracted and is listed in Table 2.

Probability distributions

The simple Gumbel distribution is considered to be significantly different to

the observed data using a modified Kolmogorov~Smirnov test at the 0.20 level
 147

/ ~ 1983 1 1953
I

Fig. 3. Observed r e t u r n periods (years) of a n n u a l floods over the study area for four years influen-
ced by E1 Nifio.
148

TABLE 2

Annual classification of prevailing oceanic-atmospheric conditions (1911 1983)

E1 Nifio Normal Anti-E1 Nifio

1911 1915 1940 1912 1967

1925 1916 1942 1913 1968
1926 1917 1944 1914 1970
1932 1918 1945 1921 1974
1933 1919 1946 1924 1979
1939 1920 1949 1930 1980
1941 1922 1952 1937
1943 1923 1955 1947
1953 1927 1956 1948
1957 1928 1959 1950
1965 1929 1969 1951
1972 1931 1971 1954
1973 1934 1975 1960
1977 1935 1976 1963
1978 1936 1981 1964
1983 1938 1982 1966

on three rivers. In all remaining cases, the observed data show varying degrees
o f s y s t e m a t i c d e v i a t i o n f r o m t h e s i n g l e G u m b e l f u n c t i o n ( F i g . 4). S u b s e q u e n t l y ,
the annual series were subdivided and separate estimates of the mean and
s t a n d a r d d e v i a t i o n m a d e ( T a b l e 3). A s o c e a n i c - a t m o s p h e r i c changes are on
s u c h a l a r g e s p a t i a l s c a l e i t i s r e a s o n a b l e t o a p p l y t h e s a m e v a l u e s o f Qi a c r o s s
the whole study area.

TABLE 3

Estimates of the mean and S.D. of a n n u a l floods under differing oceanic-atmospheric conditions a
I

No. River E1 Nifio Normal Anti-E1 Nifio

mean S.D. mean S.D. mean S.D.

1 Chancay 0.1131 0.1253 0.0635 0.0263 0.0397 0.0169

2 Chicama 0.1162 0.0833 0.0654 0.0409 0.0311 0.0144
3 Chira 0.2828 0.1667 0.0719 0.0172 0.0271 0.0074
4 Jequetepeque 0.0984 0.0591 0.0570 0.0319 0.0327 0.0142
5 Leche 0.0886 0.0311 0.0629 0.0484 0.0373 0.0082
6 Moche 0.0866 0.0686 0.0636 0.0366 0.0607 0.0793
7 Nepefia 0.0238 0.0131 0.0133 0.0074 0.0090 0.0047
8 Piura 0.1672 0.0819 0.0458 0.0340 0.0053 0.0029
9 Quiroz 0.3825 0.2256 0.1608 0.1435 0.1112 0.0529
10 Sechin 0.0187 0.0146 0.0195 0.0083 0.0093 0.0084
11 Tumbes 0.4624 0.1754 0.1139 0.0489 0.0899 0.0226
12 Viru 0.0933 0.0671 0.0704 0.0426 0.0452 0.0432
13 Zafia 0.1153 0.0916 0.0469 0.0256 0.0261 0.0154

"All data are in units of specific runoff (m s s 1 km-2).

 149

0.7(~ o Tumbes

• Piura o

• Jequetepeque
0.60-
• Nepei~a

g"
~E
~< o.5o-

IJ.
i, 0.40-
O
Z
D
nr
Q
O o.3o-
0
u.
.J o • • •
<¢
Z 0.20-
Z
<

0.10-

=jBilII~/li~I~- .A & _ & _ A ~ & _ - . A - - ---L--'-'--'-~- -

0.00 i r I I 1 .410 I .210 I J I J

.999 .919 .90 ,Be .70.~0.50 .30 .10 .05 .03 .02 .01

EXCEEDANCE PROBABILITY

Fig. 4. Examples of the observed annual flood frequencies of four rivers and the estimated Gumbel
function. Data are standardized to units of specific discharge (m3 s -1 km -2) to allow spatial
comparison.

T h e m i x e d G u m b e l d i s t r i b u t i o n (Fig. 5) is n o t s i g n i f i c a n t l y different f r o m t h e
o b s e r v e d d a t a in a n y o f t h e s a m p l e d basins, a n d is successful in r e p r o d u c i n g t h e
v a r y i n g d e g r e e s of c u r v a t u r e in t h e a n n u a l series plots in Fig. 4. In all cases it
c a n be s e e n t h a t t h e s u b - s a m p l e s a r e a d e q u a t e l y described by t h e single G u m b e l
d i s t r i b u t i o n . T h e l a r g e s t a n n u a l floods a r e d o m i n a t e d , a l t h o u g h n o t exclusive-
ly so, by e v e n t s in E1 Nifio years. C o n v e r s e l y , the l o w e s t floods a r e d o m i n a t e d
by anti-E1 Nifio years, a n d reflect t h e i r s t a t i s t i c a l c h a r a c t e r i s t i c s on t h e l o w e r
end of t h e plot of t h e m i x e d G u m b e l d i s t r i b u t i o n .
 Jequetepeque River Nepe~a River Piura River

I S 5 0 " - - e~.6 7OO

1200- O eo- O
40O I

I060- 52.s 100

0
-1 SO0
u. CO0- 4s
..i
< O OO
~60- 37.6 - o isoo -

z
< 1200 -
S00" 0 0 30-

o
m 460- ooo -
22.s -

300- ls- eoo -

0 7s- 300 -

o o- 0

1360 6Ts- 7oo ,

~200 eo- 400 ~

52.6 loo -4
1oso

eoo aoo ~

75o 37.s -
/ soo -I
0

eoo .o~ o-/

<

45o 22.s
z
<

3oo.
../ ,oJ eoo

16o 7.6] 3oo

oi o
.gg~ .ol .99g .g0 .T0 .50 .~0 .2o .lO .0s . 0 3 . 0 2 .01
°;s 9
J ,90 .70 .50 .30 .20 .10 .05 .03 ,02

EXCEEDENCE PROBABILITY

Fig. 5. Probability distributions of annual floods in E1 Nifio (closed circles), normal (squares) and anti-E1 Nifio (open circles) years on t h r e e
exemplifying rivers. The corresponding mixture models are shown on the bottom row.
 151

Spatial variability
When fitted with the mixed Gumbel distribution, the annual flood series in
this region show some strong spatial trends. In order to make spatial com-
parisons it is necessary to standardize the annual flood parameters. The flow
data are expressed as specific runoff (m3s -1 km -2) and are listed in Table 3. In
general, flood discharges increase towards the north and east (Fig. 6), corres-
ponding to the increased influence of equatorial conditions and orographic
enhancement of precipitation. As might be expected, the pattern of annual
flood size is dominated by elevation in normal and anti-E1 Nifio years. In E1
Nifio years the pattern becomes more strongly influenced by latitude, as the
equatorial air masses related to the Intertropical Convergence Zone and the
Equatorial Countercurrent move further south.
The varying degree of curvature of the annual flood-series plots (Fig. 5) is a
measure of the proportional differences in the flood statistics between each set
of generating conditions. Figure 7 shows the ratio of the mean annual flood in
E1 Nifio years relative to those of anti-E1 Nifio years. The greatest change in
conditions appears to be associated with those basins that have centroids in
northern coastal locations, diminishing towards the south and into the moun-
tains. The diminishing influence with elevation is consistent with Mugica's
(1984) observations of precipitation during the 1982-1983 E1 Nifio event. A
comparison of precipitation totals in the first six months of 1983 with mean
annual precipitation at stations in the Piura and Chira River basins revealed
a 40- to 60-fold increase in coastal regions but only a two-fold increase in the
mountains. Orographic precipitation will be induced at lower elevations in the
warmer air masses,while the southerly movement of the Intertropical Conver-
gence Zone will greatly enhance convective activity. The differences in the
oceanic-atmospheric conditions for various years may also be a function of
groundwater-transmission losses along the arid coastal plain. The larger nor-
thern basins (notably the Piura and Chira) have a greater proportion of their
basin areas on the coastal plain (Fig. 1). During normal and anti-E1 Nifio years,
when precipitation on these regions is low, transmission losses will be high,
while during E1 Nifio years coastal rainfall will reduce or reverse the direction
of groundwater flow.
Figure 8 illustrates a fairly consistent log-log relationship between mean
annual flood and drainage-basin area in E1 Nifio years. Four basins stand out
as residuals: the Tumbes (11) and Quiroz (9) to the north of the study area and
the Nepefia (7) and Sechin (10) to the south. The influence of equatorial and
mountainous conditions on the two northern basins has already been men-
tioned; in this case they reveal higher flood runoffs than their more southerly
counterparts. The southernmost residuals produce less flood runoff t h a n might
be expected. A possible explanation may be found in an examination of Fig. 1,
which reveals t h a t these basins do not extend as far into the Andes as the other
basins and t h a t this westernmost branch of the cordillera is the driest of all
because of the stabilizing influence of the Peru Current. Much of the Andes in
3°~
t,O

4°$ ,.4824 e.1139

2828 .0718•
,.3825 .1608 .1112
5°:S
?
8°8
.0888• 326
+.1131
• .1153
.0884
.0884 •.0570
.0%%, • .028t 0

,.0654
*.0888
8°8
e"0e'0933 075 ¶O452
~ i 0 5 0

8°8
EI Nh~o Mean
\
025
i~238
~~g~"°:gg' +.0133
A n t i - E l Nitro Mean ~ --.°2"J

e.lO187 i ~ ~ ,[0185 + , \ .:::::

3°$ r , , T

4° S 4 g

*~435
5°8

8°8

• ":::8: *.016g
*.0258 *.0154
0591 *.0319 *.01
*.0409

s.0366 / ~.0783
..°:o%, + ~ ! 4 2 6

Normal ~--- ~02~

9oE Standard Deviation ~ ..o131 +.0074
, \ .;o848 I J ~ *i 0083 I I ~, I
81°W 80°W 79rW 78ow 81°W 80°W 7g°W 78°W 81oW 80°W 79°W 78°W

Fig. 6. A r e a l v a r i a t i o n of m e a n a n d S.D. of a n n u a l floods (m 3 s-1 k m 2) in r e l a t i o n to t h e t h r e e s e t s of o c e a n i c - a t m o s p h e r i c c o n d i t i o n s in n o r t h e r n

Peru.
 153
3°S

4 °:

f 1044
e344
S° I

3126

6°S
\ \ .263
7°S

N (o143
8°5 \ e206

Mean .....
El__Nino x 100~
Anti ~
9°S e265
~201
81°W 80°W 79°W 78°W
Fig. 7. Areal variation of the ratio of mean annual floodsin E1 Nifio and anti-E1 Nifio years.

this region is drained by the Santa River which flows parallel (southeast-north-
west) with the maj or mountain ridges, orthogonal to the regional drainage. The
relative importance of orographic precipitation during E1 Nifio years is
therefore considerably reduced in the southern basins, and nonexistent during
normal or anti-E1 Nifio years.
When mean annual floods in anti-E1 Nifio years are plotted on the same
graph a very marked and consistent regional shift in the basin-area/runoff
relationship is observed. Tumbes and Quiroz fall along the line described by the
majority of basins in E1 Nifio years. These in turn appear to adopt the E1 Nifio
characteristics of the dry southerly basins. Such an observation suggests
extensive shifts in regional flood hydrology dependent on upset oceanic-atmo-
spheric conditions. The Piura River (8) shows the most dramatic change in
conditions and is the major exception to this generalization. In anti-E1 Nifio
years its mean annual flood is more in keeping with t h a t of the dry southern
basins. However, the Piura basin occupies a very large area of arid coastal
plain and the headwaters correspond to one of the lowest segments of the
Andes, seldom exceeding 3000 m in elevation. Within each of the broad bands
154

50001

11

I
1OOO~
cO
E
LU
L~
n"
.<
"I"
O
5
0
1OO 13
0
1 2q~
LI-
..J

Z
Z
,,<

• EL NINO

C ANTI-EL NII;,IO
10

6 , r i , t,f ;~r , i ~ i w r i1", I r , , r f ,,

100 1000 10000 1O O , 0 0 0

B A S I N AREA (kin 2)

Fig. 8. Relationship between basin area and mean annual flood size in E1Nifio (closed circles) and
anti-E1 Nifio (open circles) years.

described in Fig. 8 individual variability may be observed, such as the Chira (3)
which, in E1 Nifio years, approaches the characteristics of its northern neigh-
bours but during anti-E1 Nifio years shows a considerably diminished mean due
to the large proportion of low-lying land. In contrast, the Moche (6) shows
minimal change between conditions. Figure 1 reveals that almost all of the
Moche lies above 1000 m and that approximately 50% of the basin may be over
3000m. Mugica's (1984) observation of a negative correlation between eleva-
tion and increased precipitation in E1 Nifio years appears to be borne out by the
examination of flood characteristics.

CONCLUSIONS

The proposed mixture model provides a good fit to the observed distributions
of annual floods on rivers in an area that historically has experienced highly
variable flood conditions and extremely large "outliers". Information concern-
ing flood frequency in the region may be extracted cheaply from the existing
data and used for practical purposes such as flood-plain management and
 155

hydraulic design. The improvement in fit of a nine-parameter distribution over

the standard Gumbel distribution is not surprising. However, there is very
strong independent physical evidence for subdividing the annual flood series
into three sub-samples. The element of subj ectivity in classification is removed
and the need for parameter estimation at each station considerably reduced by
the reasonable assumption that changes in atmosphere/ocean conditions are
regional and that the proportions of the mixtures are therefore constant.
Examination of floods at the aggregate level reveals temporal and spatial
patterns of flooding that are otherwise obscured on an individual basis. The
majority of basins reveal marked differences between flooding in each of the
sub-samples; however, the probabilistic approach allows the incorporation of
such considerable variation within each sample. The changing spatial nature
of the between-sample characteristics reflects the physical atmosphere/ocean
processes dominant at the time. Annually, the effects of E1 Nifio reveal a very
patchy distribution from one basin to the next. However, when considered
together each process reveals very pronounced spatial trends. It has been
postulated that E1 Nifio is not the result of massive southward invasions of
warm water in Peruvian coastal waters but rather the consequence of region-
ally restricted intrusions of warm waters with isolated pockets of various
numbers and size depending on the severity of the E1 Nifio event (Smith, 1983).
This hypothesis appears to be in keeping with the observations of the mag~
nitude of flooding and supports the adoption of a stochastic approach.
The observations on the role of elevation in controlling the degree of varia-
bility in floods experienced between E1 Nifio and anti-E1 Nifio conditions
warrant further, more detailed, hypsometric investigation. Such a dependence
would have immediate implications for future damage-reduction programs,
land-use planning and construction projects. It may also have greater ar-
chaeological implications for the pattern of abandoned coastal settlements and
migration to the highlands (Moseley, 1983).

REFERENCES

Barnett, T.M., 1984. Prediction of E1 Nifio 1982-83. Mon. Weather Rev., 112: 1403-1407.
Bellido Delgado, R., 1983. Metodo empirico para pronosticar la ocurencia del fenomeno del Nifio.
Documenta (Lima), 11: 23-28.
Caviedes, C.N., 1984. E1 Nifio 1982-83. Geogr. Rev., 74: 267-290.
Caviedes, C.N., 1985. Emergency and institutional crisis in Peru during E1 Nifio 198~1983. Disas-
ters, 9: 70-74.
Covey, D.L. and Hastenrath, S., 1978. The Pacific E1 Nifio phenomenon and the Atlantic circula-
tion. Mon. Weather Rev., 106: 1280-1287.
Cunnane, C., 1985. Factors affecting choice of distribution for flood series. Hydrol. Sci. J., 30: 23-37.
Lowery, M.D. and Nash, J.E., 1970. A comparison of methods of fitting the double exponential
distribution. J. Hydrol., 10: 259-275.
Moseley, M.E., 1983. The good old days were better: Agrarian collapse and tectonics. Am. An.
thropol., 85: 773-799. Mugica, R., 1984. Departamento de Piura rainfall in 1983. Trop. Ocean--
Atmos. Newslett., 28: 7.
156

Potter, K.W. and Walker, J.W., 1981. A model of discontinuous measurement error and its effect
upon the probability distribution of flood discharge measurements. Water Resour. Res., 17:
1505-1509.
Quinn, W.H. and Burt, W.V., 1970. Prediction of abnormally heavy precipitation over the Equato-
rial Pacific dry zone. J. Appl. Meteorol., 6: 20-27.
Rossi, F., Fiorentino, W. and Versace, P., 1984. Two component extreme value distribution for flood
frequency analysis. Water Resour. Res., 20: 847-857.
Servicio Agrometeorologica e Hidrologica, 1962. Boletino de Estadistica Meteorologica e Hi-
drologica, Nos. 2, 3, 4, 7, 8, 9, 11, 12, 13, 14, 18 and 19. Ministerio de Agricultura, Republica del
Peru.
Schweigger, E., 1959. Die Westkiiste Siidamerikas im Bereich des Peru-Stroms. Keysersche Ver-
lagsbuchhandlung, Heidelberg-Miinchen, West Germany.
Smith, R.L., 1983. Peru coastal currents during E1 Nifio 1976 and 1982. Science, 221: 1397-1398.
Todorovic, P., 1978. Stochastic models of floods. Water Resour. Res., 14: 345-356.
Van Loon, H. and Shea, D., 1985. The southern oscillation part IV: the precursors south of 15°S to
the extremes of the oscillation. Mon. Weather Rev., 113: 2063-2074.
Woo, M.-K. and Waylen, P.R., 1984. Areal prediction of annual floods generated by mixed processes.
Hydrol. Sci. J., 29: 75-87.
Wyrkti, K. et al. (1976) Predicting and observing E1 Nifio. Science, 191: 343-346.

View publication stats