Improving hydrological modelling in cold regions using satellite remote sensing and machine learning techniques

Forskningsoutput: AvhandlingDoktorsavhandling (sammanläggning)

41 Nedladdningar (Pure)

Sammanfattning

Water resources are fundamental to life, ecosystems, and human development. In an era of climate change and increasing water scarcity, effective management of these resources is crucial, particularly in cold regions where glaciers and snow play vital roles in hydrological cycles. Understanding and accurately modelling these complex systems is essential for sustainable water resource management, climate change adaptation, and ecological preservation. The aim of this PhD research was to enhance hydrological modelling in cold regions by developing and applying advanced techniques that address the complex interactions between snow, glaciers, and runoff processes. This research was divided into four subprojects. Subproject-1 focused on implementing and evaluating the FLEXG model for glacio-hydrological modelling in the Torne River basin, northern Sweden, for the first time. The results demonstrated that the FLEXG model performed very well in runoff simulation. Subproject-2 developed the Adjusted Normalized Difference Snow Index (ANDSI) for improved glacier mapping using Sentinel-2 imagery. This index, combined with machine learning (ML) algorithms, showed enhanced accuracy in differentiating glaciers from water bodies compared to existing methods across various glacierized regions. Subproject-3 aimed to develop and evaluate a multi-variable calibration approach by using satellite-derived snow cover area data to enhance hydrological simulation accuracy and realism. Subproject-4 investigated hybrid modelling frameworks that combine ML algorithms with the FLEXG model for improving hydrological simulations in cold regions. The results indicated that the hybrid modelling framework, which incorporated precipitation, temperature, evapotranspiration, glacier mass balance, snow cover area, relative humidity, sunshine hours, solar radiation, and wind speed as inputs of the ML model, shows excellent performance in runoff simulation, especially in detecting peak flows. The proposed methods demonstrate the potential for improving our understanding of the complex interactions within glacierized catchments and enhancing the accuracy of hydrological predictions in these sensitive environments. This research supports the use of advanced modelling techniques and reliable data integration for the sustainable management of water resources and environmental protection in cold regions.
Originalspråkengelska
KvalifikationDoktor
Tilldelande institution
  • Institutionen för naturgeografi och ekosystemvetenskap
Handledare
  • Duan, Zheng, handledare
  • Pilesjö, Petter, Biträdande handledare
Tilldelningsdatum2024 okt. 23
UtgivningsortLund
Förlag
ISBN (tryckt)978-91-89187-45-0
ISBN (elektroniskt)978-91-89187-46-7
StatusPublished - 2024

Bibliografisk information

Defence details
Date: 2024-10-23
Time: 09:00
Place: Department of Physical Geography and Ecosystem Science, Pangea (Hörsal 229), Lund University, Sölvegatan 12, 223 62
External reviewer(s)
Name: Nolin Anne
Title: Professor
Affiliation: Department of Geography, University of Nevada-Reno, USA.
---

Ämnesklassifikation (UKÄ)

  • Naturgeografi
  • Vattenteknik
  • Fjärranalysteknik

Fingeravtryck

Utforska forskningsämnen för ”Improving hydrological modelling in cold regions using satellite remote sensing and machine learning techniques”. Tillsammans bildar de ett unikt fingeravtryck.

Citera det här