Instream flooding in coastal areas becomes more severe when heavy precipitation and high tide occur simultaneously. Flood modelling under such conditions requires the joint design of precipitation and tide. Data of 24‐h heavy precipitation and the corresponding daily high tidal level in the Xixiang basin, south China were used. Annual maxima and peaks over threshold sampling methods were applied to construct the precipitation and tidal level series. The joint distribution of precipitation and tide was derived by Archimedean copulas and their joint return period was calculated by the Kendall measure function. Then, equalized frequency and most likely weight function methods were employed for the joint design of precipitation and tide. Results showed that for the peaks over threshold sampling heavy precipitation coincided with more and greater high tidal levels. The general normal distribution was found to fit the marginals of precipitation and tide well, but the location, scale and shape parameters of two sampling methods differed markedly for heavy precipitation and slightly for high tidal level. Although the dependence of heavy precipitation and high tidal level was quite small and positive, Archimedean copulas effectively modelled the joint distribution of precipitation and tide. Based on the equalized frequency method, the joint design precipitation and tide using Gumbel–Hougaard copula for the peaks over threshold sampling was safer than that for the annual maxima sampling. However, for the most likely weight method design pairs remarkably tended to the lower heavy precipitation and the higher high tidal level. If the sampling method and the joint distribution were determined, a reciprocal situation for the design pairs of precipitation and tide occurred for a given joint return period, that is, a greater design value of heavy precipitation corresponded to a smaller design value of high tidal level, and vice versa. Therefore, which design method was safer was underdetermined.

No abstract

Spatial information of climatological frequency distribution of daily precipitation is highly valuable for a wide range of applications. Accurate estimation of climatology can be made for gauged locations where quality and lengthy observations are available. For ungauged or poorly gauged locations, however, indirect estimation is needed. One approach is to use a gridded daily precipitation dataset derived from interpolating observations. However, gridded daily precipitation data can be subject to large errors when gauge density is low. In addition, most interpolation methods tend to smooth the extreme values and increase the low ones, leading to unrealistic statistical properties and therefore poor estimation of daily climatology. Another approach is to first derive climatology at gauged locations and then interpolate climatology to ungauged locations. While this approach is likely to be more robust than the first approach, low gauge density can still cause significant errors especially in areas of complex terrain. In this study, we develop a method that postprocesses spatially consistent and rich reanalysis data using accurate observations at gauged locations. At an ungauged location, daily precipitation amounts from the reanalysis are bias‐corrected using quantile‐mapping guided by frequency distributions of reanalysis data and observations at a nearby gauged location (reference location). The bias‐corrected precipitation amounts are then used to estimate the climatology for the ungauged location. This method eliminates the need for interpolation and therefore its adverse effects. Special care is taken in quantile‐mapping when extrapolating beyond the range of reanalysis data at the reference location. We evaluate the method at 50 locations in Australia, using the Bureau of Meteorology Atmospheric high‐resolution Regional Reanalysis for Australia (BARRA) and precipitation observation network across Australia. These locations are chosen to represent different climate regions in Australia and have observations to validate the postprocessed reanalysis climatology of daily precipitation. Results show that the postprocessed climatology is consistent with observations, in terms of frequency distribution, high quantiles, probabilities of wet and dry days and their transitions.