Escassez e abundância - entropia e estatística como ferramentas para hidrologia: uma ciência com distribuição irregular de dados e eventos

Abstract

During the past decades, the number of extreme weather events has increased at the same pace as the growing frequency of both unusual wet and dry weather periods. Extreme weather causes multiple natural hazards, such as floods, debris flow and oversiltation, as well as crop losses and forest fires. While hydrologists have to deal with complex processes under the urgency of the climate crisis, data to power their models can be both scarce, as in regions of the Global South, or too numerous and noisy. This scenario of unequally distributed data and (water) events was the motivation for the present dissertation. The text s driven by five questions regarding (1) gully erosion modelling, (2) sediment yield assessment in ungauged basins, (3) the key factors of flash drought, its definition and (4) how to identify such events, and (5) the tools available for hydrologists to tackle problems related to extreme weather and their impacts. These five questions led to the preparation of four scientific papers and an interactive application for flash drought identification. The first two papers deal with gully erosion modelling. In the first paper, we explore the key variables to successfully model long term erosion by gullies. We also present a physicallybased model for small permanent gullies, a common erosion process in the Brazilian Semiarid Region. In the second paper, we propose a novel equation for shear stress distribution in channel beds and walls. This equation, together with the model framework submitted in the first paper, leads to a novel entropy-based gully erosion model that allows simulation of gully erosion and assessment of total erosion for contribution areas up to 8 hectares. In the third paper, we introduce a novel methodology to assess sediment yield and delivery ratio in an event-based timeframe. The technique couples a temporal downscaling method and a well-published sediment transport model, both based on the principle of maximum entropy, a tool from Information Theory which allows the modeller to identify the best probability density function without admitting any unproven hypotheses. The proposed method presented promising results: it assessed sediment yield at a Nash-Sutcliffe Efficiency of 0.96. In the last paper, we study the definition and key factors of flash drought occurrence in Central European croplands. We compared the identification of flash droughts using five different well-published methods and one proposed by the authors under the light of our definition of flash droughts. A large degree of synchronicity of individual flash drought events was identified, but also some divergence in drought periods. So as to balance out the strengths and weaknesses of those methods, we suggested using an ensemble approach for event identification. All methods were implemented in an R package and are also available in a Shiny app for the public.