Examination of Flood Estimation Techniques in the Irish Context

Abstract

This thesis deals with the determination of the most suitable methods of finding the food magnitude-return period relationship for Irish flood data using the annual maximum statistical model. Data from approximately 200 gauging stations in the Republic of Ireland, from the archives of OPW, EPA and ESB, were available. The research work is primarily based on the data of 115 of these gauging sites.

As a preliminary step to building statistical models for frequency analysis, flood flow characteristics are examined with a view to determining the regional behaviour of flood flow statistics, selecting appropriate statistical distributions to describe the flood data and examining the seasonal aspect of flooding. Most floods occur during the winter half of the year with some notable floods occurring during summer, especially in August. Because of Ireland's humid climate the range of variation of flow values from year to year, as measured by the coefficient of variation, is quite small by international standards. Likewise, the skewness of flow series is modest. While no single statistical distribution can be considered to be "best" at all locations in the context of at-site analysis it has been found that both the Extreme Value Type 1 (Gumbel) and the lognormal distributions provide reasonable models for the majority of stations.

Flood estimation (QT) by the Index Flood method using the Region of Influence (ROI) approach is investigated. The members of the pooling group in the ROI scheme are chosen with the help of a Euclidean distance or similarity measure, dij. Tests have been carried out on the effect on the estimated value of growth factor (XT) of catchment area, peat coverage, lakiness (as measured by FARL), geographical location and period of record, where the periods fall between the 1950s and the 2000s. None of these attributes were judged to be of sufficient influence to warrant special provision. Also, tests were carried out on the effectiveness of different combinations of catchment descriptors in the definition of dij and the most effective descriptors were found to be AREA, SAAR and BFI. The GEV distribution is recommended for use in the Index Flood method, except in cases where it implies an upper bound, where then EV1 is recommended.

Homogeneity is examined in the context of the estimation of XT because a homogeneous pooling group of sites is required to minimise the error in estimating XT. Tests based on Monte Carlo simulation were conducted to assess how successful a region of influence method of identifying pooling group membership is in selecting groups that qualify as being homogeneous. The results show that even with a carefully considered selection procedure, it is not certain the pooling groups identified are perfectly homogeneous. As a compromise it is recommended that a group containing more than 2 values of L-coefficient of variation outside the 95% confidence limits for that variable should not be considered as being homogeneous.

The standard errors (se) associated with estimates of XT and QT are investigated. The standard error of QT estimated by the index flood method is dominated by se(Qmed). The se(QT), expressed as a percentage, is found to vary only slightly with T. When Qmed is estimated from at-site data and XT is estimated from a pooling group containing approximately 500 station years of data then se(QT) is of the order of 5% to 10% QT regardless of the magnitude of the return period. If Qmed is estimated from a catchment descriptor based formula alone and XT is estimated from a pooling group containing approximately 500 station years of data then se(QT) is of the order of 36% QT .

The performance of pooling group based estimation is also investigated. Experiments using Monte Carlo simulation show that the size of a pooling group is not a significant factor for estimating flood quantiles provided the number of station years is more than 350. An increased heterogeneity decreases the advantage of pooled estimation over that of at-site estimation. A heterogeneity measure (H1) less than 4.0 can render the pooled estimation to be preferable to that of at-site estimation for estimating extreme quantiles. The at-site estimation is preferred when the record lengths at the site concerned exceed 50.

Guidance is provided on the estimation of the design flood of required annual exceedance probability at both gauged and ungauged locations. A number of examples is presented in Chapter 7 which cover a range of difficult cases that can arise in practice. In some of these cases, the user may be recommended to consider using an at-site based estimate in preference to the generally recommended regional pooling based method.