Investigation of hydrological design practices based on historical flood events in an experimental basin of Greece (Lykorema, Penteli)

Typically, the hydrological design procedure in ungauged basins comprises three computational steps: (a) the formulation of the design storm; (b) the estimation of the “effective” rainfall (direct runoff); and (c) the derivation of the flood hydrograph at the basin outlet. In particular, the most widespread approaches with regard to (b) and (c) are the Soil Conservation Service Curve Number (SCS-CN) method and the unit hydrograph (UH), respectively. The SCS-CN method extracts the effective from the total rainfall through an elementary model that uses two parameters, i.e. the curve number (CN), which determines the potential maximum soil moisture retention of the basin, and the initial abstraction, which is in general assumed as 20% of the later. Next, the effective rainfall is propagated through the UH, which is a linear response function employing the spatiotemporal transformation of the direct runoff across the basin. In the absence of flow data, synthetic UHs are employed, for which various empirical formulas exist, derived from hydrological investigations in experimental basins worldwide. Yet, the suitability of such regionalization approaches is questionable, when aiming to apply them in areas with substantially different hydroclimatic and geomorphological characteristics. This issue certainly involves small-scale Greek basins of ephemeral runoff, which are affected by relatively short yet intense storm events causing flash floods. The objective of our study is the evaluation of the aforementioned methods, on the basis of historical flood data from the experimental basin of Lykorema. The basin is located in Penteli Mountain and covers an area of 15.2 km2. It is equipped with three meteorological stations and two flow gauges, from which we selected 35 rainfall and flood events to analyze. In all events was shown that the use of the SCS-CN method, with typical parameter values, in conjunction with two well-known synthetic UHs (Snyder and British Hydrological Institute) provided unrealistic predictions. The key reasons were the significant overestimation of both the CN value and the initial abstraction rate, as well as the improper representation of the shape of the UHs (particularly their rising branch). In this respect, we attempted to adjust the SCS-CN method, given that the CN is not a constant but a variable that actually depends on the soil moisture conditions, while the initial abstraction ratio is rather minor. In addition, we developed a synthetic parametric UH, described by a linear rising branch and a logarithmic falling branch. This uses as inputs the time of concentration, estimated by the Giandotti formula, and another duration parameter, estimated via calibration. Following a multi-criteria optimization approach, we represented with high accuracy all the important aspects of the flood hydrographs, in terms of runoff volume, magnitude and location of the peak. Although the implementation of the proposed framework in the specific basin was quite satisfactory, there is much more work to be done for establishing consistent design practices and guidelines of general use. An ultimately important step is the development of pilot basins and the collection of reliable flood data, which will allow providing much more accurate models and formulas.