Congress Resources: Papers, posters and presentations

< Return to abstract list

How reliable can we detect changes in extreme precipitation events with statistical methods?

Congress: 2008
Author(s):
State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing, 210098, China

Keyword(s): extreme precipitation, climate change detection
AbstractExtreme hydro-meteorological events have become the focus of more and more studies in the last decade. Due to the complexity of the spatial pattern of changes in precipitation processes, it is still hard to establish a clear view of how precipitation has changed and how it will change in the future. Several approaches to the detection of changes in extreme hydro-meteorological events are assessed based on simulated daily precipitation series which are generated with two-part models, i.e., Markov chain models for precipitation occurrence and Gamma-distribution probability models for daily precipitation amounts. The approaches include: (1) detecting trend (with Mann-Kendall trend test method) in annual extremal series; (2) comparing probability distribution parameters for data observed for different periods of time; and (3) detecting changes in distribution of observed data for different periods of time with non-parametric methods (Kolmogorov–Smirnov test, Levene’s test and quantile test). Kolmogorov–Smirnov test, Levene’s test and quantile test detect changes in the overall distribution, variance and the shift of tails of different groups of data, respectively. The results show that, by using trend test, we often fail to find changes in overall statistical distribution property, while by comparing distribution parameters, the result is subject to parameter estimation uncertainties. Three non- parametric methods for evaluating changes in distribution work well for detecting changes in two groups of data with large data size (e.g., over 200 points in each group) and big difference in distribution parameters (e.g., over 100% increase of scale parameter in Gamma distribution), none of them are powerful enough for small data sets (e.g., less than 100) and small distribution parameter difference (e.g., 50% increase of scale parameter in Gamma distribution). Unfortunately, small dataset sizes and small distribution parameter changes are common in real world applications. Therefore, statistical testing methods available so far in detecting changes in extreme hydro-meteorological events are not reliable. Besides reasoning of possible changes based on the knowledge of the underlying physical mechanisms, graphical exploratory methods, such as Quantile-Qantile plots, are recommended to be used in combination with present statistical test methods.
© 2011 IWRA - International Water Resources Association office@iwra.org - http://www.iwra.org - Admin