Flood frequency analysis (FFA) has been used for the design of civil engineering works for over a century. A multitude of methods for frequency analysis exist, and they can be broadly classified into four groups: (1) empirical, (2) phenomenological, (3) dynamic, and (4) stochastic watershed modelling in conjunction with Monte Carlo simulation. Some of the methods have been developed for both at-site and regional frequency analyses. Of these methods, empirical methods are more commonly employed in engineering design and planning, and among empirical methods the at-site frequency analysis is by far the most commonly used method.
This commentary addresses the following questions. Why are we still using empirical methods of FFA in hydrologic practice when we have superior methods available? What have we accomplished to date and where are we heading in FFA research? To answer these questions, let us revisit the at-site frequency analysis, since it is the most popular method used in practice.
The steps involved in this type of analysis are: (1) acquisition of data on flood extremes; (2) selection of a plotting position formula and computation of observed probabilities of occurrence; (3) fitting of an appropriate probability distribution with use of an appropriate parameter estimation technique; (4) testing the goodness of fit and acceptance of the selected probability distribution; (5) construction of confidence intervals; and (6) estimation of flood quantiles.
When examining the methods of at-site frequency analysis, it is clear that we have made great strides in hydrologic data collection and conceptualization of the processes surrounding floods. However, that is where the consideration of the hydrologic processes in FFA ends. One cannot help but notice that much of the progress achieved in FFA so far has been in computational aspects, whereas comparatively little progress has been in the hydrology of FFA. That may be because little effort has been devoted to integrating hydrologic laws (if such things exist) with computational aspects of FFA. Except in the first step involved in FFA listed above, the role of hydrology seems minor at best and the role of statistics seems to be the lead one, whereas it should be the other way round. This means that a statistician with little training in hydrology can do as well as a hydrologist with limited training in statistics. With this kind of approach it is doubtful if we will be able to advance the state of the art of hydrologic frequency analysis. It seems to us that after nearly a century of experience we should do better than we