BED LOAD SIZE DISTRIBUTION AND FLOW CONDITIONS IN A HIGH MOUNTAIN CATCHMENT OF CENTRAL PYRENEES '

The bed load size distribution caused by different types of flow are compared in a high mountain catchment located in the upper Gallego river basin (Central Spanish'Pyrenees). Three kinds of hydrologie events could be defined: those triggered by heavy autumn rainfalls, those originated by Isolated summer rainstorms and those promoted by snowmelting. Each one Is characterized by a peculiar bed load size distribution. Thus, It could be demonstrated that the coarser fractions, above 30 mm in diameter, are up to six times more abundant. In percentage of total weight. In transports caused by heavy rainfalls than in the material collected after snowmelt flows. In its turn, bed load mobilized by snowmelt flows Is mainly composed by medium and fine gravel, from 2 to 8 mm. These may amount up to 60 % of total weight of bed load. The reasons for these so different size distributions are discussed.


BED LOAD SIZE DISTRIBUTION AND FLOW CONDITIONS IN A HIGH MOUNTAIN CATCHMENT OF CENTRAL PYRENEES'
RICARDO MARTÍNEZ-CASTROVIEJO" SUMMARY.-The bed load size distribution caused by different types of flow are compared in a high mountain catchment located in the upper Gallego river basin (Central Spanish'Pyrenees).
Three kinds of hydrologie events could be defined: those triggered by heavy autumn rainfalls, those originated by Isolated summer rainstorms and those promoted by snowmelting.Each one Is characterized by a peculiar bed load size distribution.Thus, It could be demonstrated that the coarser fractions, above 30 mm in diameter, are up to six times more abundant.In percentage of total weight.In transports caused by heavy rainfalls than in the material collected after snowmelt flows.In its turn, bed load mobilized by snowmelt flows Is mainly composed by medium and fine gravel, from 2 to 8 mm.These may amount up to 60 % of total weight of bed load.The reasons for these so different size distributions are discussed.

PIRINEOS 135
RÉSUMÉ.-Dans un bassin de haute montagne placé sur le Haut Gallego (Pyrénées aragonaises) on mis en rapport la distribution por tailles de la charge de  From a geoecological point of view, rivers are a reliable basis source for understanding the overall functioning of ecosystems incorporating geologic, géomorphologie, biogeographic and human activity information (MAR-TINEZ- CASTROVIEJO, 1989).Generally, the connection between river state and morphology and its environment is quite clear, particularly in mountain basins where events follow one another swiftly and where the drainage channels have available excess of energy to adjust themselves quickly.A great part of the information supplied by a river system can be extracted from the volumetric and temporal analysis of one of the consequences of such a functioning, that is, the sediment transport.
Most of authors agree that bed load transport processes are complex (KLINGEMAN & EMMET, 1982;BILLI &TACCONI, 1987;BATHURST, 1987).The recent literature on fluvial geomorphology devotes many pages to the improvement of sample techniques and analysis of the transported material, especially of the bed load (e.g.LEOPOLD & EMMET, 1976;KLINGEMAN etal, 1979;EMMET, 1979;KLINGEMAN & EMMET, 1982;TAGOONI & BILLI, 1987;CHURCH etal, 1987;HUBBEL, 1987 etc.) The development of these techniques has been the main source of advances known in the theory of fluvial transport throughout the last ten years.Thus, some of the mechanisms that govern the sediment transport can be described (JACKSON & BESCHTA, 1982;CARLING, 1987) as well as the implications of bed state and bed morphology for sediment mobility (WHITE & DAY, 1982;HAYWARD, 1980) and other aspects of the transport itself, such as the sediment movement in pulses (REID etal. 1985;SCHICK etal, 1987).Nevertheless, the most common application of this new knowledge is the establishment of sediment budgets at the catchment scale (e.g.TRIMBLE, 1983;ROBERTS & CHURCH, 1986;DUIJSINGS, 1987).
Although most of efforts have been centred on getting the quantity of material mobilized under certain channel and basin conditions, little attention has been paid to sediment quality.The assessment of bed load composition and their relation to flow conditions must complete present knowledge on transport processes and mechanics.The main problem for assessment is

Study area
The Izas catchment is located in the upper Gallego river valley, in the Spanish central Pyrenees (Fig. 1 ).It has an area of 0.22 km^; its lowest point Is sited at 2,060 m. a.s.l., and the highest point is sited at 2,280 m a.s.l.Bedrock consists of densely fractured Carboniferous slates.In its upper part, the basin is drained by a dense and very steep gully system.Mass movements are the dominant processes: solifluction, formation of terracettes and soil creep.Terracettes are common on degraded soils of sunny slopes whose gradients range from 25 to 30°, while solifluction lobes are present on deeper soil areas in the middle and lower sections of shady slopes (GARCIA- RUIZ & PUIGDEFABREGAS, 1982;DEL BARRIO & PUIGDEFABREGAS, 1987).Although these landforms are still very active, they do not appear to contribute to the sediment output because they are disconnected from the fluvial system.The network has a high density (16 km km•^ DIEZ etal, 1988).Along the channels, bedrock reaches alternate with alluvial reaches, where the channel erodes older fan deposits.
The mean annual temperature of the basin is around 4°G and total rainfall is over 1600 mm.In winter, the whole basin is enclosed by the 0° C winter isotherm, so precipitation in that season, which is the most important of the year, falls as snow (GARCIA-RUIZ ei al, 1985).Above the tree line, alpine grasslands cover most of the slopes.Festuca eskya occurs on drier slopes and Nardus stricta in the wetter areas.Low density summer pasturing of sheep and cattle is the only human activity.

Equipment and nnethods
Izas catchment is equipped with a gauging station consisting of a pressure transducer connected to a data logger, several rain gauges, a propane heated snow gauge, an autographic rain recorder and a lot of thermometers connected to the data logger.Sediment transport is monitored by means a slot-trap for bed load and an automatic water sampler for suspended and dissolved solids concentration.The bed load trap is simply a cube dug in the bed of the stream and lined with concrete.The approximate trap capacity is 1,000 kg which normally is enough to trap the total bed load movement during individual flood events.Its capacity was exceeded during one exceptional event.On that occasion, bed load was.deposited in the stilling area of the gauging station which is just downstream of the trap.Periodically, mainly after heavy rainstorms and during the snowmelt (May-June), the trap was emptied and the volume of sediments weighed.If the volume obtained was not very large (less than 100 kg), all of it was taken to the laboratory for analysis.For more than 100 kg a proportional sample was obtained.In the laboratory, we sieve the bed load material to obtain the size distribution.In some cases, all particles with diameter below 0.5 mm have been rejected because it was believed that they were related to suspended sediment transport rather than to bed load; their deposition in the trap may be caused by energy losses created by the trap itself.We have available information from October 1987 to June 1989, although during the winter, with the basin completely covered by a snow mantle up to 4 m thick, there was virtually no water discharge and it was not possible to empty the trap.Thus, all the material transported during the winter is grouped in the first sample obtained in spring (in the middle of May).
Table 1 summarizes the most outstanding climatic and hidrological aspects of the sixteen recorded events throughout the study period.Disturbances to the gauging station data logging were associated with electrical storms, causing a break in the water discharge data series, but rainfall characteristics and distribution help us to interpret the hydrologie regime.When the shear stress supplied by discharge exceeds largely the critical value, particle may reach a new threshold and it leaves off contacting frequently with bed to be carried in suspension.Only between these two thresholds material behaves as bed load and falls into the trap.Table 2 shows the size distributions obtained in all transport events.Figure 2 shows the evolution of the most significant fractions together with hydrographs, total bed load and daily rainfall throughout the study period.
A simple statistical approach reveals very Interesting aspects.A coefficient of variation for each fraction has been used to describe its susceptibility to vary in different events.The results can be seen in Table 3. Lowest values are registered for fine and very fine gravel (2-4 and 4.8 mm) while coefficients of variation increase towards both extremes of the series.The most unstable sizes on flow conditions appear to be the extreme ones.Differences between typical size distributions according to magnitude of events will be reflected by these size fractions.

Mean
St. dev.

CV (%)
> Another analysis consisting of affinities and incompatibilities between fractions helps to relate them to flow conditions.To obtain such behaviours, a matrix correlation of percentages of total for every fraction has been carried out (Table 4).Correlation coefficients are not as important as sign.A direct correlation Implies affinity between fractions, a tendency to be mobilize with the same type of flow.On the contrary, inverse correlations mean incompatibility, so increasing percentages in one fraction imply decrease in the other.Table 4 shows that all fractions correlate positively with the nearest in the series, but they lose their affinity towards more and more unlike sizes.More specifically, coarsest particles, above 8 mm, behave in a rather similar way to the smallest ones, while medium sizes correlate negatively both with coarse and small material.This can be graphically followed in Figure 2. The affinity of some of the size distributions is not inconsistent with flow conditions.Such affinities allow us to gather the distributions in three groups: that of bed load triggered by major hydrological events, that tiggered by the minor ones and that by snowmelt flows (Table 5).Oniy one event remains outside of this classification: that of May 16th, 1989, whose size distribution Is very far away from the group in which it should be.No doubt, such a bed load is related to more than one type of flow, rainfall peaks in the last days of autumn 1988 and snowmelt.dischargesduring spring 1989.(1) Mean of the period excluding the extraordinary event of 13 October, 1987.
Table 5 shows mean size distributions for the three groups.It has been carried out with absolute amounts of each fraction.Annual mean has been obtained in the same way.The annual mean excluding the flash flood of October 13th, 1987 is also shown in Table 5 and it proves how an extraordinary event may affect the whole study period.Both averages include the bed load collected on May 16th, 1989.
Figure 3 presents the cumulative curves corresponding to the mean size distributions.Those corresponding to bed load transport caused by heavy autumn rainfalls, on the one hand, and by isolated summer storms, on the other, are a substantial distance apart at all times.Coarser materials are less important in the minor events, as shown by a displacement to the left.The bed load caused by snowmelt discharges follows an intermediate way between the former two.In the coarser fractions it is very close to that of the isolated summer rainstorms, but reaching the fines it leaves this curve to behave similarly to the major events.The evolution of these three patterns confirm the existence of three kinds of bed load closely related to the type of flow that originated them.

Water discharge and bed load size distribution
Results obtained with the coefficient of variation of each size fraction and the analysis of the affinities and incompatibilities between them may be applied to the size distributions.So, patterns of mobility for different sizes in every kind of event are achieved and related to the type of hydrograph.
Coarser fractions, and more specifically that >30 mm, account for low percentages of the total transported material throughout the whole study period.They are markedly more abundant in those events caused by heavy rainfall.In that cases, sizes above 8 mm may amount up to 44 % by weight of total, 13,3 % for the >30 mm fraction.On the contrary, particles reaching 30 mm in diameter are absent in isolated summer events, with values below 4 % of total.Similar amounts are recorded for snowmelt flows, never reaching 3 %, while sizes above 8 mm only average 26 % of total by weight.
The movement of the coarsest particles appears to be related to the same mechanisms that trigger the greatest bed load transports.Figure 2 shows how the evolution of the >30 mm size fraction is almost parallel to that of the total bed load, at least throughout most of the study period.
Shields critical shear stress (eq. 1 ) is a direct relationship of the particle diameter.So, the shear stress required to put in motion the coarse particles is the highest one.Such a threshold can be surpassed only by the energy gradient created by flash floods and the peak discharge that accompanies them.The same explanation applies to the low percentages of coarse PIRINEOS 135 mateial in snowmelt flows and minor summer events: the rise in discharge is too small to reach the shear stress required by the largest sizes.
On the other hand, medium sizes (from 2 to 8 mm) account for most of bed load in all samples.It can be deduced that their movement is guaranteed with any kind of flow; these size fractions exhibit the minimum coefficients of variation throughout the study period.Nevertheless, the massive entrainment of fine and very fine gravel seems to be related to snowmelt flows, when they may represent up to 65 % of total, with and average value of 58 %.Events caused by summer and autumn rainfalls amount for 45 % and 51 % on average respectively.Minor pulses of meltwater during snowmelt, which are unable to mobilize great quantities of coarse and medium gravel, can entrain smaller materials in a massive manner because of relatively high discharges.The recurrence of these minor peaks throughout the snowmelt period, some of them promoted by rainfall with peak discharges lowered by snow mantle absorption, provides the energy gradient enough to put in motion that sizes.
In rainfall, events, the greater presence of other size fractions, mainly coarser, reduces the percentages of medium sizes, though they are still the majority.
The smallest sizes, less than Ú mm in diameter, are the minor part in all the distributions.They define events caused by isolated summer rainstorms because of their increased presence in bed load, which may be up to 24 % of total by weight.Those less than 1 mm amount 7 % as contasted with 2.9 % in the other two kind of flow conditions.
It is raher amazing to see how particles below 0.25 mm are more abundant in bed load associated with heavy autumn rainfalls than in any other type of event.Moreover, they are almost absent during snowmelt discharges.In both cases, flow conditions appear to be the responsible.In the rising-limb of a hydrograph, most of these particles are carried by flow in suspension.However, such a peak implies a falling-limb and, in consequence, the vertical components of flow turbulence disappear.So, suspended material begins to become bed load.This moment may be dalayed because particles keep some inertia in the manner that the stress required to keep them in motion is less than that to initialize it (LEOPOLD etai, 1964).Deposition related to the decrease in discharge may take days.
On the other hand, the former mechanics does not occur in snowmelt flows.Once in motion, the smallest sizes are suspended load.The lag of sharp falling-limbs in the hydrograph together with high discharges maintained through several weeks avoid their deposition into the trap.

Explanation and conclussions
Differences in bed load size distributions depending on the type of flow are explained according to inherent characteristics of transport processes and sediment supply.This is clear analyzing attributes and consequences of (c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://pirineos.revistas.csic.essnowmelt and rainfall flows in a high mountain catchment.Thus, major flows are commonly caused by snowmelt, but larger bed load transport events, with higher percentages of coarse gravel, occur when autumn rainfalls promote sudden peak flows in a few hours or even minutes.KLINGEMAN and EMMET (1982) argued that such an inconsistent relation between water discharge and bed load transport lies in great spatial variabilities in the availability of transportable bed material along a river's longitudinal axis.No doubt, bed material composition and organization disturb sediment transport processes in ways still under study.Nevertheless, other reasons must also be taken into account to explain such great variations in size distribution depending on flow origin.The similar distributions within each of the three groups defined above allow us to state that major differences are due to hydrologie conditions rather to changes in bed state, this being in part responsible for the within group variation.Coarse gravel mobilization is related not only to absolute water discharge, but also to the energy gradient created in the rising-limb of a flood peak.Thus, the initial motion of the coarsest fraction coincides with the shock wave of the flood.
On the other hand, sediment supply plays an important role in the total bed load transport.Some authors have pointed out the seasonal and interstorm variations of bed load, especially in rivers dominated by snowmelt regimes (BATHURST, 1987).According with that, erosion by overland flow during autumn rainstorms supplies an additional amount of sediment to the channel, which does not occur during a snowmelt flood.However, once into the main channel, stream power selects the material liable to be transported, remaining the rest stored in variable size and morphology depositional features.In fact, an increase in the supply of a given size fraction (e.g.coarser than 30 mm) from the slopes does not imply per se an increase in its percentage on total bed load.
Rainfall peak in the middle of the snowmelt flow falling-limb.First day of the period: 63.1 l/s.* Estimated data.(c) Consejo Superior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://pirineos.revistas.csic.esSuperior de Investigaciones Científicas Licencia Creative Commons 3.0 España (by-nc) http://pirineos.revistas.csic.esBED LOAD IN HIGH MOUNTAIN3.Bed load size distributionsBed load size distributions are a very good reference to assess the flow conditions that promoted their transport.Every particle of bed material responds to different initial motion thresholds clearly realted to particle size.This relationship is expressed in the Shields formula(SHIELDS, 1936stress (N.m"^)   Y^: specific weight of sediment particle (N.m"^) Y: specific weight fo water (N.m'^) d: particle diameter (mm)

Fig. 2 .
Fig. 2. Total bed load, water discharge and size distribution of bed load throughout the study period (winters excluded).Long periods in which the trap remained empty are not plotted.Dashed line: estimated data.

( c )Fig. 3 .
Fig. 3. Comparison of cumulative curves corresponding to the three kinds of typified events.