A wide range of temporal scales of variability exists in suspended
sediment and turbidity records from proglacial meltwater streams. These
variations and their causes are illustrated in this paper using data
from the stream draining the Haut Glacier d'Arolla, Valais,
Switzerland. The data include hourly pump-sampler suspended sediment
data from June to September 1990; a quasi-continuous 24 hour record of
turbidity and velocity sampled at 5 second intervals in July 1992; and
3-minute series of turbidity and velocity at various heights in the
boundary layer sampled at 10Hz, also obtained in July 1992. The
high-frequency measurements made in 1992 employed an active head,
infra-red suspended solids monitor deployed in close proximity to a
twin-axis, discoidal electromagnetic current meter. Data were analysed
using a combination of statistical and deterministic ('event
structure') modelling. Particular emphasis is placed on the parallel
variation of turbidity and velocity on time scales of less than 1
second. Variability at this scale has been relatively neglected, but
must be understood before turbidity monitoring programmes can be
designed that involve sampling from the continuous variation of
turbidity by averaging over 10-60 seconds every 30-120 minutes, logging
the data automatically, and calibrating the record using pump or hand
held samplers. Fluctuations in suspended sediment concentration and
turbidity are accounted for in terms of glacial-hydrological,
channel-marginal, and flow-related controls whose significance varies
with timescale. from the seasonal through the diurnal to the
instantaneous. The results have implications for research designs
employing turbidity meters as indicators of suspended sediment
variation. Representative turbidities, and comparisons between meters,
must consider time-averages scaled with respect to turbulence
characteristics at particular locations. The possibility of interaction
between more or less discrete transport events results in turbidity
variation which cannot, at present, be satisfactorily predicted.