Authors: Baez Villanueva, Oscar Manuel
Title: Streamflow simulation in data-scarce regions using remote sensing data in combination with ground-based measurements
Language (ISO): en
Abstract: Global water resources are currently under unprecedented stress, which is projected to increase due to the influence of multiple factors. Therefore, changes in governance are urgently required to improve water management and water use efficiency while maintaining the health of river systems and their water quantity and quality. Data is crucial in this process; however, most rivers in the world remain ungauged, and in data-scarce regions, the hydrometric and hydrometeorological networks of stations have been decreasing during the last decades. This hinders the implementation of proactive water management approaches that strive towards informed-based decision-making. This cumulative thesis shows how open access global precipitation products can be evaluated, corrected, and used to predict streamflow at the daily temporal scale in data-scarce regions in combination with ground-based measurements by following a three-step approach: i) performance evaluation of different precipitation products over regions with different climates and at multiple temporal scales; ii) development of a novel merging method to improve the representation of precipitation at the daily scale; and iii) assessment of the ability of the novel merged product altogether with state-of-the-art precipitation products to predict daily streamflow over ungauged catchments through the implementation of regionalisation approaches. This thesis showed that the precipitation products perform differently depending on the temporal scale, elevation, and climate; and that these products still have errors in detecting particular precipitation events. These insights served as a basis to develop a novel merging procedure named RF-MEP, which combines data from precipitation products, ground-based measurements, and topographical features to improve the characterisation of precipitation. RF-MEP proved to be a powerful method as the precipitation errors at different temporal scales were substantially reduced, outperforming state-of-the-art precipitation products and merging procedures. The precipitation product derived with RF-MEP has been included in a Chilean precipitation monitor platform from the Center for Climate and Resilience Research (Mawün) and users can apply this method in a friendly manner using the R package RFmerge. This merged product altogether with three state-of-the-art precipitation products was used to implement three regionalisation approaches by calibrating an HBV-like hydrological model over 100 near-natural catchments in Chile. The results showed that although these methods yielded relatively good performances, the precipitation products corrected with daily gauge observations did not necessarily yield the best hydrological and regionalisation performance. Additionally, the hydrological regime of the catchments influenced the performance of the evaluated regionalisation techniques, with the pluvio-nival and raindominated catchments yielding the best and worst performance, respectively. This cumulative dissertation shows that precipitation datasets can help to strive towards informed-based decision-making in data-scarce regions. However, these regions often lack the infrastructure and human capacity to use this type of information efficiently. Therefore, an informed-based decision-making process requires institutional transitions and changes that help address water resources management’s present and future challenges. In this sense, there is a need to move towards data-driven water resources management by implementing strategic approaches that systematically build the capacities and infrastructure of such regions.
Subject Headings: Data scarcity
Hydrological modelling
Machine learning
Merging procedures
PUB
Precipitation
Precipitation products
Random Forest
Regionalisation
RF-MEP
Subject Headings (RSWK): Maschinelles Lernen
Hydrologische Vorhersage
Niederschlag
Klassifikations- und Regressionsbaum
Abflussmenge
URI: http://hdl.handle.net/2003/41028
http://dx.doi.org/10.17877/DE290R-22876
Issue Date: 2022
Appears in Collections:Raumbezogene Informationsverarbeitung und Modellbildung

Files in This Item:
File Description SizeFormat 
Thesis.pdfDNB25.76 MBAdobe PDFView/Open


This item is protected by original copyright



Items in Eldorado are protected by copyright, with all rights reserved, unless otherwise indicated.