The deposition of sediment in reservoirs can variously impact their performance through storagecapacity losses, damage to valves and conduits, reduced flood attenuation and changes in waterquality. Sediment yield is the sediment load normalized for the catchment area and is the netresult of erosion and depositional processes within a basin. Thus, it is controlled by those factors that control erosion and sediment delivery, including local topography, soil properties, climate,vegetation cover, catchment morphology, drainage network characteristics and land use. Predicting the sediment discharge of a river section or the sediment yield of an upstreamcatchment has always been an ambitious goal for a number of different earth scientists, such asengineers, hydrologists, geomorphologists and others. In particular, estimation of sedimentdischarge is a vital key point for the assessment and design of major hydraulic systems, such asirrigation dams, hydroelectric projects and flood attenuation structures. A key element of the proposed method is to construct the Digital Elevation Models (DEM) fortwo periods of interest, one prior to the dam construction (1964) and the other during the hydrographic survey (1998-99). The hydrographic survey has been carried out using a differential Global Positioning System (GPS) technique and a typical fathometer operating at the frequency of 130 kHz for depth determination. Therefore the method is subject to the usual errors e.g. GPS limited availability and the definition of the water-mud interface. The DEM at the time prior to the dam completion was constructed from digitising the original survey maps (scale 1:5000). The corresponding DEM from the hydrographic survey resulted from an irregular network of points in three dimensions (position and elevation). The associated grids were interpolated from triangulation with linear interpolation procedures available in the SURFER mapping package. The difference in elevation results in the volume of deposited sediments. The spatial distribution of accumulated sediment in the reservoir shows profoundly that the total incoming sediment remains in the reservoir and particularly at the uppermost parts (deltaic deposits). The total sediment deposits volume was calculated equal to 69.6 hm3. To convertvolumetric changes to sediment yield in mass units the material properties of the deposited sediment were also investigated by collecting two core samples from the reservoir invert using appropriate instrumentation (LONGYEAR 36 hydraulic corer). Direct measurement of deposits density was not possible mainly because it was impossible to collect undisturbed samples. However, density was estimated from the proportion of sand, silt and clay in the samples using the Lane and Koelzer (1943) formula. The total sediment mass accumulated in the reservoir for the whole period of dam operation was estimated at 112.5 Mt. Therefore the mean annual sediment yield is estimated equal to 1005 t/km2 and the corresponding mean annual sediment dischargeequal to 106.4 kg/s. Agrafiotis River basin, which is the smallest one, contributes the mostconsiderable sediment load per unit catchment area. The corresponding value is one of the highestmean annual sediment yield found in international literature and is a result of rainfall intensity, geology, morphology and the small extent of its area.The hydrographic survey of a reservoir is a quite satisfactory procedure for reconstructing sediment yield records of a drainage basin. An apparent weakness of the method is that it givesonly an over year average of the sediment yield and not its temporal evolution. However, iffrequent hydrographic surveying of the reservoir is permitted (e.g. every 5 years) then sediment yield can be computed in finer time scales. Alternatively, this method can be combined with 28 hydrological models as well as sediment discharge measurements in upstream locations to reconstruct the temporal evolution of reservoir sedimentation. Its strongest merit, however, remains the illustration of the spatial distribution of accumulated sediments with in the reservoir. Dead storage remains almost free of deposited sediments whilstparts of the nominal useful storage are occupied from accumulated sediments. This obviously means that the total loss of stored water is significantly greater than it was originally assumed and it certainly becomes a waste of a valuable natural resource. In this specific case, the depositional pattern inside the reservoir reveals the apparent necessity of reconsidering the dead volume principle, in terms of a thorough investigation and modelling of sediment yield in the water resources management context. The main scope of the research is to compute the mean annual sediment yield and discharge for 14 river cross-sections in NW Greece. The basic data were the sediment discharge measurements and the mean daily stream discharges, all available from the PPC. The original suspended sediment measurements have been collected from the Public Power Corporation (PPC) andnormally cover a time span from 1965 to 1980. The number of measurements varies from site to site (e.g. 36 measurements at Poros Riganiou to 121 at Plaka Bridge) and the frequency of measurement is highly variable. Then mean daily sediment discharges were computed from the mean daily stream discharges and the mean annual sediment discharge is finally estimated. The cornerstone of this approach is the correspondence of the mean annual sediment discharge of Acheloos River at Avlaki gauging station computed by the broken line rating curve with the corresponding value computed by the sediment deposits in the Kremasta Reservoir located just downstream of the Avlaki gauging station. Sediment discharge measurements were taken during 1966-1970 where as daily river stages were recorded with frequent intervals without measurements. A physicall-based, distributed hydrologic model (the MIKE SHE model) was applied to fill in the periods with missing mean daily discharges from 1966 to 1998. Two alternative rating curves were deduced from the sediment discharge measurements, the first one with a unique power law expression for the whole set of discharges and the second with different power relations for two discharge classes above and below a threshold roughly corresponding to the bankfull discharge. It is assumed that the rating relationships are valid for the whole time span of the simulation since the catchment has undergone insignificant land use changes. The application of the first rating curve to the mean daily discharge yields mean annual sediment discharge equal to 13.5 kg/s, where as the application of the different power relations for two discharge classes yields a corresponding value of 73.3 kg/s. The first equation seriously underestimates the sediment discharge where as the second one results in an estimate close to that of hydrographic survey. This indicates that sediment rating curves can give good estimates if applied carefully, otherwise can result in serious inaccuracies. The broken line interpolation was introduced by Koutsoyiannis (2000) as a simple alternative to numerical smoothing and interpolating methods and is treated here as a surrogate for the ordinary single rating curve. The main concept is to approximate a smooth curve that may be drawn for the data points with a broken line, which can be numerically estimated by means of a least squares fitting procedure. If the only objective used for fitting the broken line is the minimization of total square error then there sult might be a very rough broken line, depending on the arrangement of the data points. However, the roughness of the broken line can be controlled by introducing as a second objective the minimization of the roughness. The broken line is a concatenation of straight-line segments, where the number of the straight-line segments is numerically the outcome of the compromise between the two objectives of minimizing the fitting error and the roughness of the broken line. Considering that the prevailing fluvial form in upstream Greek rivers is the gravel-bed form, we 29 assume a broken line with two segments. In such a fluvial form, there is a distinct threshold discharge for sediment motion, which is attributed to the development of the well-known armourlayer. Below this threshold there is no exchange of the suspended sediment with the riverbed. Once the surface, coarse material, armour layer fully breaks up beyond the threshold discharge and exposes a larger range of particle sizes underneath, the transport rate significantly increases. Additionally, bank erosion during high discharges will enhance the sediment availability in the river bed. Certain geomorphologic parameters have been computed from the catchments’ DTMsand their values were introduced in non-linear correlation analyses with the mean annual sediment yield and discharge. Two equations with high values of the coefficient of determination have been calculated in order to, at least, qualitatively describe the phenomena in terms of the relation between sediment yield and catchment geomorphology.
