Modelling rainfall statistics in a changing climate : comparing and improving existing approaches

• Κύριος συγγραφέας: Μαμαλάκης, Αντώνιος
• Άλλοι συγγραφείς: Λαγγούσης, Ανδρέας
• Μορφή: Thesis
• Γλώσσα: English
• Έκδοση: 2016
• Θέματα: Human adaptation in the environment; Climatology; Προσαρμογή ανθρώπου στο περιβάλλον; Κλιματολογία; 551.577 015 1

Rainfall is one of the most representative, significant and dominant climatic variables, since it considerably affects hydrologic equilibrium, as well as constituting a control parameter in water management. However, due to its intermittent nature and highly variable character, its statistical modeling and forecasting under changing climate conditions, is one of the most pressing and crucial problems in modern Hydrology. In this thesis, we present, critically assess and compare the most popular approaches which have been used in the literature, in the fields of: a) extreme rainfall estimation, and b) reproducing rainfall statistics based on climate models results. In extreme excess modelling, one fits a generalized Pareto (GP) distribution to rainfall excesses above a properly selected threshold u. The latter is generally determined using various approaches, such as non-parametric methods that are intended to locate the changing point between extreme and non-extreme regions of the data, graphical methods where one studies the dependence of GP related metrics on the threshold level u, and Goodness of Fit (GoF) metrics that, for a certain level of significance, locate the lowest threshold u that a GP distribution model is applicable. In Chapter 1, we review representative methods for GP threshold detection, discuss fundamental differences in their theoretical bases, and apply them to 1714 over-centennial daily rainfall records from the NOAA-NCDC database. We find that non-parametric methods are generally not reliable, while methods that are based on GP asymptotic properties lead to unrealistically high threshold and shape parameter estimates. The latter is justified by theoretical arguments, and it is especially the case in rainfall applications, where the shape parameter of the GP distribution is low; i.e. on the order of 0.1 ÷ 0.2. Better performance is demonstrated by graphical methods and GoF metrics that rely on pre-asymptotic properties of the GP distribution. For daily rainfall, we find that GP threshold estimates range between 2÷12 mm/d with a mean value of 6.5 mm/d, while the existence of quantization in the empirical records, as well as variations in their size, constitute the two most important factors that may significantly affect the accuracy of the obtained results. Concerning reproduction of the statistical structure of daily rainfall at a basin level based on climate model (CM) results, two types of statistical approaches have been suggested. One is statistical correction of CM rainfall outputs based on historical series of precipitation. The other, usually referred to as statistical rainfall downscaling, is the use of stochastic models to conditionally simulate rainfall series, based on large-scale atmospheric forcing from CMs. While promising, the latter approach attracted reduced attention in recent years, since the developed downscaling schemes involved complex weather identification procedures, while demonstrating limited success in reproducing several statistical features of rainfall. In a recent effort, Langousis and Kaleris (2014) developed a statistical framework for simulation of daily rainfall intensities conditional on upper-air variables, which is simpler to implement and more accurately reproduces several statistical properties of actual rainfall records. In Chapter 2, we study the relative performance of: i) direct statistical correction of CM rainfall outputs using non-parametric distribution mapping, and ii) the statistical downscaling scheme of Langousis and Kaleris (2014), in reproducing the historical rainfall statistics, including rainfall extremes, at a regional level. This is done for an intermediate-sized catchment in Italy, i.e. the Flumendosa catchment, using rainfall and atmospheric data from 4 CMs of the ENSEMBLES project. The obtained results are promising, since the proposed downscaling scheme is more accurate and robust in reproducing a number of historical rainfall statistics, independent of the CM used and the characteristics of the calibration period. This is particularly the case for yearly rainfall maxima.