• A report on previous studies that delineated homogeneous hydrometeorological areas across Europe with respect to flood frequency.
• A catalogue of relevant information (e.g., catchment descriptors) recommended for use in flood frequency studies along with an indication of availability/unavailability across Europe.
• A report presenting analogies and differences between proposed methodologies and reporting a detailed description of selected methodologies.
• A report on the application of selected methodologies to the FloodFreq dataset (WG1) detailing: reliability, uncertainty and transferability for different hydrometeorological areas.
• A summary illustrating the recommended alternative methodologies or additional hydrological information to complement the traditional approach for a specific area.
• A collection of open-source computer codes for at-site and regional flood frequency estimation.
Flood frequency analysis is broadly used for estimating the design-flood at a given location (i.e., flood magnitude associated with the recurrence interval T, T-year flood). The approach can be implemented locally (At-Site Flood Frequency Analysis, SFFA), or regionally (Regional Flood Frequency Analysis, RFFA). RFFA is used to limit unreliable extrapolation when available data record lengths are short as compared to the recurrence interval of interest, or for predicting the flooding potential at locations where no observed data are available. Despite the maturity of SFFA and RFFA and the existence of consolidated methodologies available for many European regions, scientific progress continues and there is a constant need to evaluate the gains obtained from new methods (e.g., copula based multivariate flood discharge and volume analysis, Bayesian methods using Monte Carlo Markov Chains) over existing knowledge.
Different European countries (and sometimes even different regions within a country) adopt different methodologies. Also, methodologies are often selected on the basis of traditional choices or what is known to work well for a limited number of local catchments, rather than an objective assessment of available methods. FloodFreq will address this disconnected picture, promoting a pan-European and synergic approach to flood frequency analysis as requested by European Flood Directive. FloodFreq will homogenize and harmonize the current level of knowledge on the approach to flood frequency analysis through a comprehensive application and verification across Europe of methodologies proposed by the scientific community.
WG2 Reports from Vienna meeting June 25th, 2012
The PDF document that can be downloaded from this page by clicking on the title illustrates how to apply Top-kriging by means of the R-package rtop (see below where to download the package)
Jon Skøien (email to Attilio Castellarin, dated 30/11/2011): The FloodFreq action sounds very interesting, and you are of course free to share rtop package. Actually it is already available for anyone who is interested, through the R-forge repository. In an R-session, it is therefore now possible to download and install with:
> install.packages("rtop", repos="http://R-Forge.R-project.org")
HESS paper: Top-kriging - geostatistics on stream networks
J. O. Skøien, R. Merz, and G. Blöschl
Abstract. We present Top-kriging, or topological kriging, as a method for estimating streamflow-related variables in ungauged catchments. It takes both the area and the nested nature of catchments into account. The main appeal of the method is that it is a best linear unbiased estimator (BLUE) adapted for the case of stream networks without any additional assumptions. The concept is built on the work of Sauquet et al. (2000) and extends it in a number of ways. We test the method for the case of the specific 100-year flood for two Austrian regions. The method provides more plausible and, indeed, more accurate estimates than Ordinary Kriging. For the variable of interest, Top-kriging also provides estimates of the uncertainty. On the main stream the estimated uncertainties are smallest and they gradually increase as one moves towards the headwaters. The method as presented here is able to exploit the information contained in short records by accounting for the uncertainty of each gauge. We suggest that Top-kriging can be used for spatially interpolating a range of streamflow-related variables including mean annual discharge, flood characteristics, low flow characteristics, concentrations, turbidity and stream temperature.
R-package: rtop (R-Project CRAN)
WG2 Reports from Bratislava meeting November 25th, 2011
The ZIP archive contains the manuals for three different R-packages associated with the use of L moments. R is a free software environment for statistical computing and graphics. It compiles and runs on a wide variety of UNIX platforms, Windows and MacOS. To download R, please choose your preferred CRAN mirror.
The manuals describe the utilization of the following packages:
Functions related to L-moments: computation of L-moments of distributions and data samples; parameter estimation; L-moment ratio diagram; plot vs. quantiles of an extreme-value distribution.
(author: J. R. M. Hosking) DOWNLOAD
Functions for regional frequency analysis using the methods of J. R. M. Hosking and J. R. Wallis (1997), "Regional frequency analysis: an approach based on L-moments".lmomRFA: Regional frequency analysis using L-moments.
(author: J. R. M. Hosking) DOWNLOAD
The package implements the statistical theory of L-moments in R including L-moment estimation, probability-weighted moment estimation, parameter estimation for numerous familiar and not-so-familiar distributions, and L-moment estimation for the same distributions from the parameters. L-moments are derived from the expectations of order statistics and are linear with respect to the probability-weighted moments; choice of either can be made by mathematical convenience. L-moments are directly analogous to the well-known product moments; however, L-moments have many advantages including unbiasedness, robustness, and consistency with respect to the product moments. The method of L-moments can out perform the method of maximum likelihood. The lmomco package historically is oriented around canonical FORTRAN algorithms of J.R.M. Hosking, and the nomenclature for many of the functions parallels that of the Hosking library, which later became available in the lmom package. However, vast arrays of various extensions and curiosities are added by the author to aid and expand of the breadth of L-moment application. Such extensions include venerable statistics as Sen weighted mean, Gini mean difference, plotting positions, and conditional probability adjustment. Much extension of L-moment theory has occurred in recent years, including extension of L-moments into right-tail and left-tail censoring by known or unknown censoring threshold and also by indicator variable. E.A.H. Elamir and A.H. Seheult have developed the trimmed L-moments, which are implemented in this package. Further, Robert Serfling and Peng Xiao have extended L-moments into multivariate space; the so-called sample L-comoments are implemented here and might have considerable application in copula theory because they measure asymmetric correlation and higher co-moments. The supported distributions with moment type shown as L (L-moments) or TL (trimmed L-moments) and additional support for right-tail censoring ([RC]) include: Cauchy (TL), Exponential (L), Gamma (L), Generalized Extreme Value (L), Generalized Lambda (L & TL), Generalized Logistic (L), Generalized Normal (L), Generalized Pareto (L[RC] & TL), Gumbel (L), Kappa (L), Kumaraswamy (L), Normal (L), 3-parameter log-Normal (L), Pearson Type III (L), Rayleigh (L), Reverse Gumbel (L[RC]), Rice/Rician (L), Truncated Exponential (L), Wakeby (L), and Weibull (L).
(author: William H. Asquith) DOWNLOAD
WG2 Reports from Budapest meeting June 17th, 2011
WG2 Reports from Prague meeting October 29th, 2010