A "design flood" is an extreme flood with a certain small probability of occurrence. The acceptable probability of occurrence depends on the category of construction and varies between 1 on 500 years to 1 on 10,000 years. An estimation of the design flood is based on observed discharge values. The available discharge series are extremely short compared to the design probabilities. The longest available observation series in Europe are seldom longer than 100 years. For other continents, and especially Central America, they are much shorter, 30 years or so.
It is obvious that the shortness of the observation series infers uncertainty in the design flood calculations. An ultimate goal for design flood calculations must be to minimise this uncertainty. A traditional practice for design flood calculations has been so-called "single site analysis". In this approach only information from the nearest observation site, with respect to the designed construction, is used in the calculations. Single-site analysis is the method applied in Costa Rica now. The accuracy of the method has, however, been questioned when ICE has applied for loans. This fact has actualised reconsideration of the presently applied methodology.
The results of the calculations according to this method are totally dependent on the length, quality and representativity of the single observation series used. For example, in tropical areas, tropical cyclones can strike a certain river basin so rarely, that they are not reflected in the available observations, though they may well be the main generating factor for the extreme floods.
An alternative, developed and widely applied in Europe, is to consider all available observation series in the region and/or to include also data on extreme precipitation. A recent European study on this subject is the one in the frame of the "Flow Regimes from International Experimental and Network Data (FRIEND)" project, which is a part of UNESCO IHP-IV. Within this project, in which two of the project members (Irina Krasovskaia and Lars Gottschalk) participate, there has been made a review of the existing new practices for design flood calculations. The most recent development is to also include scaling issues of extreme precipitation and floods, i.e. how the statistical characteristics depend on the size of the considered study area (drainage basin). During 1996 Irina Krasovskaia and Lars Gottschalk have been guest researches in Bern, Switzerland and in Lyon, France to apply and develop such flood design approaches to Swiss and French flood data sets.
These new practices, based on regional analysis (Index-Station approach) and joint rainfall/runoff analysis with consideration of scaling issues, can be most interesting for design purposes also in Costa Rica. The problem of calculating design floods has gained a special actuality in Costa Rica in connection with the launching of the Boruca project. A water-resources project of this dimension needs to be based on the best possible estimate of design floods.
Two computer programs resulted from the project. Colibri is a program for regional analysis of the maxima, and Tucan is a program for applying the results from the regional analysis in order to find the return period for any point of interest.