Peter's River in central Newfoundland supplies water to the town of Bo
twood from an intake located at a point just upstream of a right-angle
d bend. The pumphouse is situated on the outer bank of the river and i
s constructed on a foundation of sandy silt. In 1983, a large flood er
oded part of the bank and removed vegetation which had provided a prot
ective cover. Progressive erosion then threatened the pumphouse and fo
rced an engineering assessment of the situation. This was conducted in
the winter of 1984-1985 and included the prediction of the 50- and 10
0-year floods from the minimal data (4 years of records) which were th
en available. Estimates were made of velocities and flood levels, and
various protection schemes were considered. The installation of a reve
tment in 1986 stabilized the bank, but icing on the intake had always
caused problems and these continued unabated. In winter, frazil ice ad
hered to the submerged intake structure and was drawn into the pumps c
ausing frequent shutdowns. Various solutions, including electrical hea
ting, were considered. However, in 1991, a decision was made to instal
l infiltration pipes below the stream bed, to raise water levels with
a small broad crested weir and to disconnect the existing intake. Flow
estimates were again required, both for construction purposes and for
responding to the Department of Fisheries requirement that there woul
d be no immediate or long-term harm to fish habitat. This paper provid
es a brief description of the general problems, but focuses primarily
on the methods used, 8 years apart, to estimate flood data. These meth
ods were intended to be relatively unsophisticated, but were consisten
t with the magnitude of the river and the relatively small scale of th
e proposed works. In the study conducted in 1984-1985, only 4 years of
records were available and the estimation of the 50- and 100-year flo
ods relied strongly on the use of regional equations, the development
of unit hydrographs, and one large recorded flood. The more extensive
record available in 1991 permitted some more sophisticated statistical
techniques. However, the large flood which occurred in 1983 still cau
sed considerable difficulty, and the results of the analysis varied ac
cording to how this flood was included. Despite these difficulties, it
was shown that the estimates made in 1984 were not unreasonable.