Flood Frequency Analysis: Exploring the Role of Statistical Models in Engineering Education and Practice

Flood presents a significant risk to communities globally as flood damage is increasing due to climate change. To effectively reduce flood damage, scientists use a term 'design flood', which refers to a flood discharge linked to annual exceedance probability. This study investigates the role of several statistical models adopted in design flood estimation using flood and catchment data from 88 catchments in eastern Australia. The first objective of this study is to compare the effectiveness of two statistical modeling approaches, generalised additive models (GAM) and log-log regression in estimating flood quantiles. Generalised Extreme Value (GEV) distribution with L-moments was employed where the mean, the coefficient of variation and the coefficient of skewness were adopted as the response variables and the catchment characteristics as predictor variables. This study reveals that GAM performs better than log-log regression technique in capturing the variability of flood quantile estimates. The second objective of this study is to illustrate the learning aspects of the adopted statistical models. It is found that most of the students do not understand the fundamentals of these statistical modelling techniques and often reach inappropriate conclusions. The findings of this study will assist students and junior researchers to understand the assumptions related to statistical flood modelling approaches and how this affects decision making in sustainable flood plain management.