This study deals with the fundamental problems of hydrological design. Specifically, it explores the boundary conditions for design flood analysis. The problem of extrapolation of design relationships has been investigated by the systematic analysis of important design parameters namely unit hydrograph time to peak (tp), catchment antecedent conditions (CWI), runoff losses (PR) and the relationship between rainfall and flood return periods. In particular, this thesis investigates the combination of these variables representative of design analysis. A review of the hydrological design tools of response identification along with the physical evidence of catchment response is presented. The results of the investigation regarding time to peak (tp) show that it varies significantly between events, and the relationships developed between tp and flood magnitude (Qp) show the non-linear catchment behaviour which conform with most of the physical and field investigations. The relationships suggest that the unit hydrograph (UH) parameters derived from moderate events should be adjusted for extreme events and therefore a correction in UH tp has been developed which depends on the flood return period (Ashfaq and Webster, 2000a). The analysis of catchment wetness index (CWI) from a large number of observed events showed that antecedent conditions observed in the flood season are reasonably representative of the major events. This contrasts with the existing design recommendations which suggest consistently lower values. An alternative relationship of CWI therefore has been developed for design purposes (Webster and Ashfaq, submitted manuscript). The investigation showed that the percentage runoff (PR) characteristics of large events are consistently different than available from the existing design PR-method. The design method underestimates for the standardised conditions especially for large events because of its limited range of estimates for a wide range of return periods (e.g. 11% range in PR for 2 to 1000 year return periods). This problem is related with the PR-method itself for catchments having higher mean annual rainfall (SAAR> 800 mm) whereas for lower SAAR areas « 800 mm), it is related to the selection of design CWI values. The analysis of observed large events also revealed that a flood is generally associated with the storm of less return period than that of the flood. This contrasts with the suggested design rainfall-flood return periods relationship in the FSR (NERC, 1975; [H, 1999), but conforms with the curves presented by Webster (1998, 1999). This observation is further established by a detailed investigation of the characteristics of extreme events through a continuous model as well as the hydrological analysis of observed extreme floods of Easter-1998 (Ashfaq and Webster,2000b). The study demonstrates that the characteristics of extreme floods are different from those of small and moderate events. Relationships based on moderate events should be adjusted for the design of major events. The aggregated and integrated findings based on a comprehensive series of analysis led to the proposal of an alternative combination of design parameters. The performance of this combination showed improved flood estimates without any prior calibration in comparison to the FSR as well as FEH. A revised design methodology has therefore been proposed on the philosophy of 'independent treatment of input variables'. The application of this methodology on new catchments also provides encouraging estimates of flood quantiles. It is suggested that the methodology is equally applicable for both gauged and ungauged catchments especially where the observed data are limited or no data are available.