Assessment of model error in regression estimates of floods at an ungauged site

摘要

Flood frequency analysis at ungauged drainage basins uses regression analysis to obtain flood estimates. Usually the model error occurring because of the deviation of the underlying relationship from that of the population is included in the noise term. However, it is shown in this paper that the regression estimates are affected by the model error arising by the exclusion of variables, using a hypothetical but hydrologically realistic data set. Further, it is shown that it is the first additional variable after log(area) that achieves the highest improvement in the estimates.